178
Coherent Production of Light Pseudoscalars (Axions) Inside a Magnetic Field with a Polarized Laser Beam by Yannis K. Semertzidis Submitted in Partial Fulfillment of the Requirements of the Degree of DOCTOR OF PHILOSOPHY Supervised by Professor A. C. Melissinos Department of Physics and Astronomy University of Rochester Rochester, New York 1989 * Work supported in part by U.S. Dept. of Energy Contract DE-AC02-76ER13065.

Coherent Production of Light Pseudoscalars (Axions)

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Coherent Production of Light Pseudoscalars (Axions)

Coherent Production of Light Pseudoscalars (Axions)

Inside a Magnetic Field with a Polarized Laser Beam

by

Yannis K Semertzidis

Submitted in Partial Fulfillment of the

Requirements of the Degree of

DOCTOR OF PHILOSOPHY

Supervised by Professor A C Melissinos

Department of Physics and Astronomy

University of Rochester

Rochester New York

1989

Work supported in part by US Dept of Energy Contract DE-AC02-76ER13065

Abstract

We have constructed a highly sensitive ellipsometer to

measure the effect of a transverse magnetic field on the

propagation of polarized light The total path length is

of order 7 Km in a field of 22 T and rotation angles as

low as 10-10 rad can be measured

We have used this instrument to set limits on the coupl ing

of light pseudoscalar and scalar particles to two photons

thus placing constraints on some recent theories of the

elementary particles (certain classes of supersymmetric

theories) We have also used the apparatus to measure for

the first time the Cotton-Mouton coefficients of the noble

gases Neon and Helium

In addition the apparatus has been designed so as to

measure photon-photon (Delbruck) scattering in the visible

as predicted by QED We have demonstrated that the apparatus

has adequate sensitivity to reach this goal and have

identified its present limitations

ii

Acknowledqements

The experiment was first approved (unanimously) by the

BNL committee in November 1987 It is encouraging to see

that big laboratories still support small (though elegant)

experiments Brookhaven National Lab has supported us with

facilities and personnel so there are numerous people to

thank Charlie Anderson Audrey Blake Frank Cullen Norman

Franklin Robert Gottschalk Ron Hauser Arnie Kreisberg

Jim Licari John Mazzeo Vera Mott Cynthia Murphy The

Riggers Ritchie Savoy and Diana Votruba A separate thanks

to Jim Briggs David Cattaneo Bill De Jong George Ganetis

Herb Hildebrand Rich Howard Paul Ribaudo Joe Skatudo

and Dan Wilson who were generous both with their time and

skills

This support was matched by the people at the University

of Rochester Pat Borelli Ernie Buchanan Betty Cook

Thomas Haelen Connie Jones and Judy Mack My appreciation

goes to the members of our research team Ruth Cameron

Giovanni Cantatore Henry Halama George Kostoulas Don

Lazarus Bruce Moskowitz Frank Nezrick Al Prodell Carlo

Rizzo and Joe Rogers for their remarks and contributions

Special thanks to Prof Emilio Zavattini for his

contagious enthusiasm and to my supervisor Prof Adrian

Melissinos for all his support encouragement and guidance

iii

Finally my special thanks go to my wife Georgia

Afxendiou-semertzidis for her support and understanding

but most of all for sharing with me the excitement of thesis

writing

This work was supported by the us Department of Energy

under contracts DE-AC02-76CH00016 and DE-AC02-76ER13065

iv

curriculum vitae

Yannis Semertzidis was born on September 16 1961 He

attended the Aristoteleio Panepistimio Thessalonikis

(Aristotle University of Thessaloniki) in Greece from 1979

to 1984 where he received a Bachelor of Science Degree in

Physics May 1985 he began graduate studies at the University

of Rochester In 1987 he was a recipient of the Furth

Fellowship administered by the University His research

work has included the CP Violation experiment at LEARCERN

during the summer of 1983 and from spring 1984 to summer

1984 Fermilab experiment 723 a search for anomalous forces

at highly relativistic velocities from May 1985 to December

1985i BNL experiment 805 the search for Galactic axions

from January 1986 to spring 1989 Subsequently he has worked

on BNL experiment 840 a coherent production of any

pseudoscalar (or scalar) that couples to two photons this

was from summer 1987 to the present This work was supervised

by Dr A C Melissinos

v

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 2: Coherent Production of Light Pseudoscalars (Axions)

Abstract

We have constructed a highly sensitive ellipsometer to

measure the effect of a transverse magnetic field on the

propagation of polarized light The total path length is

of order 7 Km in a field of 22 T and rotation angles as

low as 10-10 rad can be measured

We have used this instrument to set limits on the coupl ing

of light pseudoscalar and scalar particles to two photons

thus placing constraints on some recent theories of the

elementary particles (certain classes of supersymmetric

theories) We have also used the apparatus to measure for

the first time the Cotton-Mouton coefficients of the noble

gases Neon and Helium

In addition the apparatus has been designed so as to

measure photon-photon (Delbruck) scattering in the visible

as predicted by QED We have demonstrated that the apparatus

has adequate sensitivity to reach this goal and have

identified its present limitations

ii

Acknowledqements

The experiment was first approved (unanimously) by the

BNL committee in November 1987 It is encouraging to see

that big laboratories still support small (though elegant)

experiments Brookhaven National Lab has supported us with

facilities and personnel so there are numerous people to

thank Charlie Anderson Audrey Blake Frank Cullen Norman

Franklin Robert Gottschalk Ron Hauser Arnie Kreisberg

Jim Licari John Mazzeo Vera Mott Cynthia Murphy The

Riggers Ritchie Savoy and Diana Votruba A separate thanks

to Jim Briggs David Cattaneo Bill De Jong George Ganetis

Herb Hildebrand Rich Howard Paul Ribaudo Joe Skatudo

and Dan Wilson who were generous both with their time and

skills

This support was matched by the people at the University

of Rochester Pat Borelli Ernie Buchanan Betty Cook

Thomas Haelen Connie Jones and Judy Mack My appreciation

goes to the members of our research team Ruth Cameron

Giovanni Cantatore Henry Halama George Kostoulas Don

Lazarus Bruce Moskowitz Frank Nezrick Al Prodell Carlo

Rizzo and Joe Rogers for their remarks and contributions

Special thanks to Prof Emilio Zavattini for his

contagious enthusiasm and to my supervisor Prof Adrian

Melissinos for all his support encouragement and guidance

iii

Finally my special thanks go to my wife Georgia

Afxendiou-semertzidis for her support and understanding

but most of all for sharing with me the excitement of thesis

writing

This work was supported by the us Department of Energy

under contracts DE-AC02-76CH00016 and DE-AC02-76ER13065

iv

curriculum vitae

Yannis Semertzidis was born on September 16 1961 He

attended the Aristoteleio Panepistimio Thessalonikis

(Aristotle University of Thessaloniki) in Greece from 1979

to 1984 where he received a Bachelor of Science Degree in

Physics May 1985 he began graduate studies at the University

of Rochester In 1987 he was a recipient of the Furth

Fellowship administered by the University His research

work has included the CP Violation experiment at LEARCERN

during the summer of 1983 and from spring 1984 to summer

1984 Fermilab experiment 723 a search for anomalous forces

at highly relativistic velocities from May 1985 to December

1985i BNL experiment 805 the search for Galactic axions

from January 1986 to spring 1989 Subsequently he has worked

on BNL experiment 840 a coherent production of any

pseudoscalar (or scalar) that couples to two photons this

was from summer 1987 to the present This work was supervised

by Dr A C Melissinos

v

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 3: Coherent Production of Light Pseudoscalars (Axions)

Acknowledqements

The experiment was first approved (unanimously) by the

BNL committee in November 1987 It is encouraging to see

that big laboratories still support small (though elegant)

experiments Brookhaven National Lab has supported us with

facilities and personnel so there are numerous people to

thank Charlie Anderson Audrey Blake Frank Cullen Norman

Franklin Robert Gottschalk Ron Hauser Arnie Kreisberg

Jim Licari John Mazzeo Vera Mott Cynthia Murphy The

Riggers Ritchie Savoy and Diana Votruba A separate thanks

to Jim Briggs David Cattaneo Bill De Jong George Ganetis

Herb Hildebrand Rich Howard Paul Ribaudo Joe Skatudo

and Dan Wilson who were generous both with their time and

skills

This support was matched by the people at the University

of Rochester Pat Borelli Ernie Buchanan Betty Cook

Thomas Haelen Connie Jones and Judy Mack My appreciation

goes to the members of our research team Ruth Cameron

Giovanni Cantatore Henry Halama George Kostoulas Don

Lazarus Bruce Moskowitz Frank Nezrick Al Prodell Carlo

Rizzo and Joe Rogers for their remarks and contributions

Special thanks to Prof Emilio Zavattini for his

contagious enthusiasm and to my supervisor Prof Adrian

Melissinos for all his support encouragement and guidance

iii

Finally my special thanks go to my wife Georgia

Afxendiou-semertzidis for her support and understanding

but most of all for sharing with me the excitement of thesis

writing

This work was supported by the us Department of Energy

under contracts DE-AC02-76CH00016 and DE-AC02-76ER13065

iv

curriculum vitae

Yannis Semertzidis was born on September 16 1961 He

attended the Aristoteleio Panepistimio Thessalonikis

(Aristotle University of Thessaloniki) in Greece from 1979

to 1984 where he received a Bachelor of Science Degree in

Physics May 1985 he began graduate studies at the University

of Rochester In 1987 he was a recipient of the Furth

Fellowship administered by the University His research

work has included the CP Violation experiment at LEARCERN

during the summer of 1983 and from spring 1984 to summer

1984 Fermilab experiment 723 a search for anomalous forces

at highly relativistic velocities from May 1985 to December

1985i BNL experiment 805 the search for Galactic axions

from January 1986 to spring 1989 Subsequently he has worked

on BNL experiment 840 a coherent production of any

pseudoscalar (or scalar) that couples to two photons this

was from summer 1987 to the present This work was supervised

by Dr A C Melissinos

v

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 4: Coherent Production of Light Pseudoscalars (Axions)

Finally my special thanks go to my wife Georgia

Afxendiou-semertzidis for her support and understanding

but most of all for sharing with me the excitement of thesis

writing

This work was supported by the us Department of Energy

under contracts DE-AC02-76CH00016 and DE-AC02-76ER13065

iv

curriculum vitae

Yannis Semertzidis was born on September 16 1961 He

attended the Aristoteleio Panepistimio Thessalonikis

(Aristotle University of Thessaloniki) in Greece from 1979

to 1984 where he received a Bachelor of Science Degree in

Physics May 1985 he began graduate studies at the University

of Rochester In 1987 he was a recipient of the Furth

Fellowship administered by the University His research

work has included the CP Violation experiment at LEARCERN

during the summer of 1983 and from spring 1984 to summer

1984 Fermilab experiment 723 a search for anomalous forces

at highly relativistic velocities from May 1985 to December

1985i BNL experiment 805 the search for Galactic axions

from January 1986 to spring 1989 Subsequently he has worked

on BNL experiment 840 a coherent production of any

pseudoscalar (or scalar) that couples to two photons this

was from summer 1987 to the present This work was supervised

by Dr A C Melissinos

v

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 5: Coherent Production of Light Pseudoscalars (Axions)

curriculum vitae

Yannis Semertzidis was born on September 16 1961 He

attended the Aristoteleio Panepistimio Thessalonikis

(Aristotle University of Thessaloniki) in Greece from 1979

to 1984 where he received a Bachelor of Science Degree in

Physics May 1985 he began graduate studies at the University

of Rochester In 1987 he was a recipient of the Furth

Fellowship administered by the University His research

work has included the CP Violation experiment at LEARCERN

during the summer of 1983 and from spring 1984 to summer

1984 Fermilab experiment 723 a search for anomalous forces

at highly relativistic velocities from May 1985 to December

1985i BNL experiment 805 the search for Galactic axions

from January 1986 to spring 1989 Subsequently he has worked

on BNL experiment 840 a coherent production of any

pseudoscalar (or scalar) that couples to two photons this

was from summer 1987 to the present This work was supervised

by Dr A C Melissinos

v

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 6: Coherent Production of Light Pseudoscalars (Axions)

Table of contents

Abstract ii Acknowledgments iii Curriculum vitae v

List of Figures vii List of Tables ix

Chapter 1 Theoretical Motivation 1 11 Introduction 1 12 QED Vacuum Polarization 10 13 Axion 16

Chapter 2 Apparatus 34 21 Magnets 34 22 Laser 45 23 optics 54 24 Cavity 63 25 Ray Transfer 73 26 Telescope 78 27 Jones Vectors and Matrices 84

Chapter 3 Data Acquisition 93

31 Electronics 93

32 Misalignment Correction 101 33 Laser Power Stabilizer 104

Chapter 4 Analysis of Results 113 41 Noise Sources 113 42 Data Analysis 123

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases 144

Chapter 6 Conclusions 160

Index 168

vi

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 7: Coherent Production of Light Pseudoscalars (Axions)

List of Fiqures

Chapter 1 11 Primakoff effect bull bull bull bull bull bull bull bull bull bull bull bull 3 12 Lowest order QED Vacuum Polarization bullbull 6 13 Axions mix with pi zero and eta bullbullbull 20 14 Axions decay through a triangle anomaly bull 22 15 Axion induced rotation bullbullbullbullbullbullbull 27 16 Axion induced ellipticity bullbullbullbullbullbull 28 17 Axion limits from assumed rotation bullbullbullbullbull 30 18 Axion limits from assumed ellipticity bullbullbull 31

Chapter 2 21 Experiment layout bull bull bull bull bull bull bull bull bull bull 35 22 optics inside the enclosure bullbullbull 36 23 Cross section of a CBA magnet bullbullbull 37 24 Lead pot end cap and the two magnets bullbullbull 38 25 Magnet quench current vs temperature bullbullbull 39 26 LM0014 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 42 27 LM0018 transfer function bullbullbullbullbullbullbullbullbullbullbullbullbull 43 28 stray magnetic field bullbullbullbullbullbullbullbull 44 29 Atomic Ar levels 47 210 A typical Ar ion laser resonator bullbullbullbullbull 52 211 QWP and elliptically polarized light bull 57 212 Support tubes for the Faraday cell 60 213 Lissajous pattern bullbullbullbullbullbullbullbullbullbull 65 214 Time delay measurements bullbullbullbullbullbullbullbullbullbullbull 67 215 Birefringence axis setup bullbullbullbullbullbullbullbullbullbullbullbullbullbullbullbull 69 216 Circle for birefringence axis bullbullbullbullbullbullbullbullbullbull 70 217 First readout with one mirror rotated bullbullbull 72 218 Birefringence axes of a mirror bullbullbullbullbullbull 74 219 Parameters of a paraxial ray bullbullbullbullbullbullbullbullbullbullbullbullbull 75 220 Stable regions of mirrors resonator bullbullbullbullbull 77 221 Telescope lens setup bullbullbullbullbullbullbullbullbullbullbull 81

vii

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 8: Coherent Production of Light Pseudoscalars (Axions)

Chapter 3 31 Electronics setup of the experiment 95 32 EGampG low noise preamplifier characteristics 98 33 Laser light spectral density bull 105 34 Nominal noise reduction vs frequency bull 108 35 Typical data with LPC OFF bullbullbullbullbull 109 36 Typical data with LPC ON bull 110 37 lf noise reduction with LPC ON bullbullbull 111

Chapter 4 41 lgain squared vs frequency bullbullbullbullbull 118 42 Noise sources vs amount of light bull 119 43 Single record rotation data bullbullbull 125 44 RMS average of 26 files 126 45 vector average of 26 files bullbullbullbullbull 127 46 Typical rotation data with magnets ON 130 47 Shunt mirror data with magnets ON bull 132 48 Cavity mirrors move due to the magnet motion 134 49 RotationEllipticity limits from E840 bullbullbullbull 142

Chapter 5 51 Magnet modulation at 78 mHz 147 52 Fourier analysis of the magnet current 148 53 Typical Nitrogen gas data 149 54 Vacuum ellipticity data 150 55 Ellipticity vs Nitrogen gas pressure 152 56 Ellipticity vs Ar gas pressure middot 153 57 Ellipticity vs Ne gas pressure middot 154 58 Ellipticity vs He gas pressure middot 155

viii

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 9: Coherent Production of Light Pseudoscalars (Axions)

List of Tables

Chapter 2 21 Effective magnetic field length 22 Laser performance parameter specifications 23 Laser specifications (Innova 9o_5) bullbullbullbullbullbullbullbullbullbull

24 Laser output power specifications bullbullbull 25 REP and LEP light through a QWP bullbullbullbull 26 Ray transfer matrices 27 Jones vectors 28 Jones matrices

Chapter 3

31 Silicon photodiode 32 Data taking sheet

Chapter 4

41 Rotation data 42 Ellipticity data bullbullbullbullbullbull

Chapter 5 51 Cotton-Mouton constants 52 Various parameters for the gases 53 Radius of first excited state of some gases bull

41 49

50

51 58 82 85 86

96

112

137 141

157 158 158

ix

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 10: Coherent Production of Light Pseudoscalars (Axions)

1012

Chapter 1 Theoretical Motivation

11 Introduction

One of the most important problems in physics today

is the behavior of the interactions of the elementary

particles at very high energies in the range 10 2 to 10 19

GeV However present and even future particle

accelerators will be able to explore only the first few

decades of this range Thus we will have to rely on

indirect information by studying the relics of the very

early universe where such energies existed for a very

brief time One such particle created at an energy of

GeV is the axion which is a pseudoscalar with very

light mass ma~10-5 eVe This makes it possible to search

for the axion either in the cosmic radiation1 or to attempt

to directly produce it in the laboratory This thesis

is an experimental search for axions and for analogous

light scalar particles We have not observed axions but

have set a limit on their coupling to two photons or

equivalent to two quarks

The search for axions was based on the effect that

they would have on the propagation of light in a magnetic

field in vacuum This can be understood by considering

1

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 11: Coherent Production of Light Pseudoscalars (Axions)

the coupling of the axion to two photons as shown in figure

1lb Since a magnetic field is equivalent to a cloud

of virtual photons a real photon from a 1ase- bea can

scatter off the virtual photons to produce an axion 2 The

production of the axion will affect the polarization of

the incident beam as explained in detail in section 13

We have therefore constructed a very sensitive

polarimeter or ellipsometer and have achieved a

sensitivity in the change of polarization of order 10-10

rad The polarimeter is described in chapters 2 and 3

and I give here only a brief sketch

The photon source is a 2W (single line power) Argon-ion

laser operating at ~ = 514 nm High quality polarizers

are used to establish and to analyze the polarization and

are set for best extinction The magnetic field is as

high as possible 22 kgauss in the present investigation

and extends over -10m To enhance the effect the light

is made to travel repeatedly through the field region

This is achieved by forming an optical cavity between two

mirrors and deforming one of the mirrors we have reached

790 traversals

Since the signal is so weak it is important to measure

the change in the photon field amplitude rather than

intensity This can be done by interfering the effect

from the cavity with a deliberately introduced rotation

in our case this is done with a Faraday cell For

2

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 12: Coherent Production of Light Pseudoscalars (Axions)

a 11M Photon

(a)

Photon 11M a

(b)80

Figure 11 Primakoff effect a) Decay ofaxions inside a magnetic field

b) Production ofaxions inside a magnetic field

3

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 13: Coherent Production of Light Pseudoscalars (Axions)

ellipticity measurements the sought effect is transformed

to a rotation by a suitable A4 plate Standard practice

in such small signal measurements is the modulation of

the signal at a well known fixed frequency this allows

detection in a very narrow frequency band thus reducing

noise and eliminating static effects Unfortunately the

high magnetic fields cannot be modulated rapidly we

reached 78 mHz in this investigation whereas the Faraday

cell was driven at 260 Hz

The whole apparatus and cavity were placed in a high

vacuum ranging from 10-4 Torr to 10-7 Torr and in a

constant temperature enclosure These precautions as well

as computer feedback on the key optical components are

absolutely necessary to reach the previously mentioned

sensitivity of 10-10 rad Furthermore the signal must

be Fourier analyzed the line width being of order 1 - 2

mHz This in turn sets the stability requirements of the

corresponding driving oscillators and receiver

electronics A typical Fourier spectrum had 400 channels

for a total width of -1 Hz For our instruments this

implied an acquisition time of order of 17 minutes per

record The records were written to disk and then further

analyzed (ie combined) off line on a VAX 11780 It

was thus possible to use very long effective integration

time Since we had phase information available we formed

vector averages this reduces the noise level as 1N

4

(where N is the number of records averaged) while the

signal remains constant

An important aspect of experiments that search for

small effects is the ability to calibrate them

Fortunately there is an atomic effect that induces an

ellipticity in light traversing gases subject to a

transverse magnetic field the Cotton-Mouton effect this

is not a Faraday rotation which involves a longitudinal

magnetic field By introducing a suitable gas in the

cavity and varying the pressure in the range of a few

Torr we are able to generate a known small ellipticity

and thus calibrate our device We used Nitrogen gas and

found that our apparatus was properly calibrated In view

of our improved sensitivity we were able to measure for

the first time the C-M coefficient of Neon and Helium and

to improve on the value for Argon These measurements

and their interpretation are discussed in chapter 5

The apparatus had been designed so that when its

ultimate sensitivity is reached it would be possible to

detect an important effect predicted by Quantum Electro

Dynamics (QED) This is photon-photon scattering as shown

in figure 12 where the incident photon scatters (twice)

from the magnetic field due to a virtual electron loop

This process known as Delbruck scattering has been

predicted 50 years ago but has never been observed with

visible photons where its interpretation is unambiguous

5

k k

B

Figure 12 The lowest order graph giving rise to dispersive effects

6

For the design parameters of the apparatus the ellipticity

induced by Delbruck scattering is 5x10-12 rad This

necessitates a 5T field replacement of the Faraday cell

by an EO modulator and further refinements in noise

reduction and data acquisition The theoretical analysis

of Delbruck scattering and of its detection are given in

chapter 1

The experiment is installed at Brookhaven National

Laboratory because of the availability of high field

magnets We are using two dipoles which were built as

prototypes for the Colliding Beam Accelerator (CBA) and

have superconducting windings Thus the necessity for a

sizeable Helium refrigeration plant (200 W) and power

supply (I= 5 kA) which were provided by Brookhaven High

vacuum technology is also necessary since as already

discussed residual gas can give rise to ellipticity

induced by atomic effects In contrast the optics was

produced by the University of Rochester shops as well as

was some of the mechanical construction Such an

experiment would not have been possible without a laser

for a light source and this explains why no Delbruck

measurements have been attempted previously The laser

has been purchased commercially as were the Fast Fourier

Analyzers The experiment is a collaboration between the

University of Rochester Brookhaven Fermilab and the

University of Trieste It was approved in the Fall of

7

1987 and the installation was completed two years later

As soon as the first data were obtained we observed

a signal which exhibited the characteristic of vacuum

optical rotation as expected for an axion However by

rotating the polarization of the light from 45middot with

respect to the magnetic field to 0middot the signal remained

present indicating that it was not due to selective

absorption along the two field orientations (parallel and

orthogonal to the field) but due to a pure rotation This

background signal sets the present limit in our

experimental measurement which corresponds to ES 4 X 10-8

rad From the measured value of E we set a limit on the

axion coupling to two photons

1 1 gayylt M = 4x 10 5 GeV

This is the first laboratory measurement of the coupling

of a light pseudoscalar or scalar particle (ma s 10-3 eV)

to two photons It el iminates several models of

supersymmetric theories3 and places important constrains

on any future theory

To set our result in a broader context we note that

accelerator experiments that have searched for axions

through their decay a ~ e+e- or a -+ yy and in k -+ na set

limits at the level M c 102 GeV These limits are much

weaker than our results but are not restricted to the low

mass region On the other hand astrophysical arguments

8

from the evolution of the sun predict that M ~ 108 GeV

This level of sensitivity corresponds to a rotation of

10-11 rad which is well within reach of the apparatus

when full field will be available and its final

configuration reached

We have searched for the source of the spurious rotation

that we observe and quickly established that it originates

in the cavity and disappears when the shunt mirror is

used We also established that it is not induced by

residual gas it is not pressure dependent We then

observed that the endcaps holding the mirrors are moving

by -6 nm in phase with the magnetic field since a

multipass cavity induces a static rotation of the

polarization plane mirror motion can induce a

corresponding time-dependent signal Incidentally the

static rotation is a direct manifestation of Berrys

phase26 which is present for the classical EM field A

new version of the apparatus will provide better isolation

of the mirrors thereby eliminating this signal We are

also examining the possibility of halo scattering from

the magnetized walls of the vacuum chamber which could

induce an ellipticity and rotation

If we exclude the spurious signal the sensitivity of

the apparatus as obtained from a rotation measurement is

E = 6 x 10-10 rad which corresponds to 2 x 10-8 radJHz

Thus a measurement of 106s should yield e-2x 10- 11 rad

9

bull

Similar sensitivity is obtained for the ellipticity

Further reductions in the noise level by using an EO

modulator and a larger dynamic range in the detector will

yield the remaining factor of ten in sensitivity These

modifications of the apparatus are part of a future long

range program

12 QED Vacuum Polarization

When a photon beam is traveling in vacuum the two

polarization states have the same speed as a consequence

of the isotropy of space This is not true any more when

a magnetic field is present because of photon-photon

interactions (fig 12) The photon beam according to

Quantum Electrodynamics (QED) creates virtual e+e- pairs

which it reabsorbs in the course of traveling When the

magnetic field is present the time evolution of the e+eshy

pair depends on whether they lie in a plane parallel or

orthogonal to the magnetic field That is the vacuum

is not isotropic any more but there is an axis defined

by the external field The same is true for an electric

field but a strength of 3x l01V em would be required to

match the effect of a 10 Tesla field

This problem was first discussed in the 1930s and I

10

shall follow the formalism of the Euler-Heisenberg4

effective electromagnetic Lagrangian which contains the

classical fields and the quantum corrections due to vacuum

polarization This Lagrangian in the low frequency limit

(hwlaquo 2mec2) is

( 11 )

in unrationalized Gaussian units valid to fourth order

E and B are respectively the total eleotric and magnetic a2 (ftlmc)3

fields and A=- 2 133 x l0- 32 em 3 erg is a constant90n mc

where me is the electron mass Any fermion pair can be

used instead of the e+e- pair with the proper replacement

of the charge and mass values in the calculation of the

constant A Equation 11 holds for fields that are well

below the critical ones Le E cr 10 15 V em and

Bcr 441 x lOl3gauss

Since the average polarizations of matter are

represented by the electric displacement D and the magnetic

field H we will use them to derive the indices of

refraction in vacuum 4

oL oLD=4nshy H=-4nshy ( 12)

oE oB

For a polarized laser beam (with Eph and Bph being the

electric and magnetic vectors of the laser beam) of a few

watts and beam diameter of a few millimeters 5 propagating

11

along z inside a magnetic field Bext and assuming Eph is

in the x direction we have

E = E e ei(kz-wt) B -+- B A i(kz-wt) - B -+- B ph ph x B -

-ext pheye - Qxt ph ( 13)

Then utilizing only linear terms of Eph and Bph

D -= 4 n ( iJ ~~h ex) = [E ph - 4 A B XI E ph -+- 14 A ( E B ext) ( B axt bull ex) ]ex (14 )

and using the formula D=E E=

( 1 c- E = 1 - 4 A B xt -+- 14 A ( ex B ex ) 2 )

Similarly

iJL 2H -4n-=B-4AB B=

iJB

B - 4 A [ B xc B ext -+- B h B QXC -+- 2 ( B ext B ph ) B xl -+shy

(16)

where in the above we have used the relation H=Hext+Hph

Using the relation Hph - ~ -1 Bph we obtain

(17 )

Now let us assume that the external magnetic field is

in the plane defined by the electric and magnetic vectors

of the laser beam Then we can distinguish two cases

1) Eph parallel to Bext From (15) and (17) we have

E = 1 - 4 A B xt -+- 1 4 A B xt = 1 -+- lOA B xt 2 - J 2

~ = ( 1 - 4 A B ext) =lt 1 -+- 4 A B ext

12

The parallel index of refraction is

~ 2 2 4 12n p = V Ell = ( 1 + 14 A B Qxt + 40 A B ext ) ~

n p 1 + 7 A B xt ( 18)

2) Eph orthogonal to Bext Again from (15) and (17)

we have

E == 1 - 4 A B ~xt and

The orthogonal index of refraction is

( 19)

Therefore we have

n p = 1 + 7 A B xt

(l10)

that is the vacuum becomes birefringent for B oxt t O

In case the magnetic field does not lie in the Eph

Bph plane Bext is the projection of the external magnetic

field on that plane4 and the more general formulas

n p == 1 + 7 A B XI si n 2 e

(l11)

apply where (

k)2 2 e 1 B axt bull sin = - bull B extk

and k is the photon wave-vector

Equation 111 is accurate to better than 01 for

13

B ex SO 1B cr 6 For magnetic fields of 01 SBwBuS 10 a

detailed calculation is given in reference 6 The

polarized laser beam entering the magnetic field region

with polarization (ie electric vector) at a 45deg angle

with respect to the magnetic field will acquire an

ellipticity given by

( 112)

where L is the length of the path of the laser light inside

the magnetic field region and A the laser wavelength The

phase retardation between the parallel and orthogonal

components is cent = 21jJ (ellipticity ljJ is defined as the ratio

of the semiminor to the semimajor axis of the ellipse

traced by the electric vector) Here we have used sine =

1 In these units (natural unrationalized Gaussian) 1

gauss can be expressed as 691-10-2 eV2 or 104 (ergcm3 ) 12

and 1 cm as 5104 eV-1 For our experimental parameters

A = 5145 nm (green light from an argon ion laser)

L = 104 m

and Bext = 5 Tesla (= 50 kgauss)

we get

( 113)

As we see the induced ellipticity is very small and

this effect can safely be ignored in the usual treatment

of electromagnetism in other words it is correct to use

14

the linear Maxwell equations in everyday life) This

is not true any more when regions with high magnetic fields

are considered as for the surface of neutron stars where

magnetic fields of the order of 1012 gauss are present

In fact Novick et al 7 proposed to use the degree of

ellipticity of the X-rays emitted from a neutron star

in order to distinguish between X-rays coming from the

surface and those coming from regions well within the

surface

The QED vacuum polarization has never been observed

directly but there are indirect observations where the

=-Il-=(1165924plusmn85)X 10- 9

effect is a small part of other processes 8 One such

measurement is the 9 j1 - 2 experiment where the muon

anomaly is found to be 9

9 -2 U

exp 2

while the theoretical value is

= (1165921 plusmn 83) x 10- 9 U 1h = U QED + U hadr + U weak

where

U QED = (11658520 plusmn 19) x 10- 9

=(667plusmn81)X 10-9u hadr

=(21 plusmn06)X 10-9U weak

The photon-photon contribution to this anomaly is

15

which contributes about two parts in 104 to the whole

process and would be good to check in a more direct fashion

Another precision measurement involves forward

scattering10 of v-rays in the electric field of a nucleus

Here again the Delbruck scattering (ie photon-photon

interaction) contribution is a small part of the dominant

processes ie Thomson Rayleigh and nuclear resonant

scattering

13 AXion

The Quantum-Chromo-Dynamics (QCD) Lagrangian contains

the term

(114 )

where F QIlV is the gluon field and e is an angular parameter

This term violates parity (P) but conserves charge

conjugation (C) and therefore violates the combined CP

symmetry There is a series of vacua distinguished by

a topological number n which are the classical solutions

of the gauge equations in QCD but are not invariant under

all possible gauge transformations The state of the true

vacuum (ie the one that is gauge invariant) is

16

constructed by a linear combination of Ingt vacua and e

is usually defined as

lllG19gt = I e 1 ngt (115)

n--oo

9 is therefore a periodic variable with a period of 2~

The angle e is related to the electric dipole moment (EDM)

of neutronll which if finite would violate CP symmetry

We can show this as follows

Any electric dipole moment of the neutron must be

aligned with the spin vector (the only preferred axis)

If we apply the T operator (time reversal symmetry) the

EDM does not change sign whereas the spin does Therefore

the presence of an EDM would violate T symmetry since

the physical state of the neutron would change under T

reflection Because of the CPT theorem which requires

that every reasonable quantum field theory must be CPT

invariant the presence of an EDM implies that CP is also

violated Actually if the neutron has an EDM P symmetry

is violated as well as can be shown by arguments similar

as used for T We discuss the EDM of a neutron because

a strict experimental upper limit on it of 4x 10-25 e middot em

already exists12 this corresponds to a limit on 9 of

191lt 10-9 or 19-nllt 10-9

bull

Because the notion of fine tuning of theoretical

parameters is not popular it is believed that e must be

exactly 0 or ~ but in principle it could be anywhere

17

between 0 and 2~ Different values of e would correspond

to different theories with different coupling constants

and there would be no problem with the term of equation

114 except for the presence of strong CP violation

Peccei and Quinn13 first introduced the idea that 0 is

not a parameter of the theory but rather a dynamical

variable and that different values of e describe different

energy states of the vacua of the same theory Below the

QeD energy scale e naturally gets the value 0 because

the Lagrangian has an effective potential with minimum

at that value The Peccei-Quinn mechanism is an extension

of the standard SU(3)XSU(2)LXU(l) theory14 with two Higgs

doublets

(116)

The U(l)pQ axial symmetry corresponds to invariance

under the following transformations

iT)Vs -2iT) ltIgt ~u i ~ e u i centu e u

IT)VS e -2iT) ltIgtd i ~ e d i ltlgtd ~ d (117)

Le the Lagrangian is invariant under these phase

transformations In the above ui and di refer to the

SU(2) quarks and n is a rotation angle (some arbitrary

phase)

Weinberg15 and Wilczek16 realized that when the axial

U(l)pQ symmetry is spontaneously broken it would give

18

rise to a Goldstone boson named the axion The axion

is massless in the zero-mass limit for light quarks but

because the light quark masses are not zero the axion

mixes with the ~O and n and becomes massive (fig 13)

The axion is therefore a pseudo-Goldstone boson Since

the breaking occurs at the electroweak scale the mass

of the axion is supposed to be greater than 100 KeV and

lt B ) 7 (lOOKeV)51ts l1fet1me -c(a~e e-)_10- s -c(a~yy =0 s --- bull Suchmiddot

an axion model is excluded by beam-dump17 and

branching-ratio experiments For example theory14

predicts 1 6 x 10-S for the product of the yen an d Y branching

ratios into y+a whereas the experimental limit for this

product is 06 x 10-9 lS Another discrepancy between

theory and experiment is the branching fraction in the

K+ decay n+a theory19 predicts BR(K+~n+a)85x 10- 6

and experimentally20 this branching ratio is found to be

less than 4S x 10-S bull

But this was not the end of the axion because the mass f2f m bull

of the axion is ma = fa where f n = 94MeV ffin = 135 MeV

for the pion mass and fa is the symmetry-breaking scale

of the PQ symmetry First it was thought that fa should

coincide with the weak symmetry-breaking scale

constant) However if we assume that fa is very large

(fagt lOS GeV) then the axion mass becomes very small

19

q

a

Pi zero Eta zero

q

Figure 13 Mixing with pi zero and eta axions become massive in the non zero light quark limit

20

and its coupling very weak such axions have been named

invisible 21 In this latter axion model an extra complex

scalar Higgs field ltP is required that has a vacuum

expectation value

so that

where x is the ratio of the vacuum expectation values of

the two Higgs doublets ex = U dIU 11) can be large In that

case the axion become the preferred candidate for

accounting for the dark matter in the universe

Axions couple to two photons through a mechanism

referred to as the triangle anomaly (fig 14) the axion

lifetime is predicted to be of the order of 1050 sec for

a mass of the order 10-5 eV However I the Primakoff effect

could be used (fig lla) to enhance axion decay Two

experiments are currently using this technique to look

for cosmic axions22 23 on the assumption that 100 of

the dark matter in the local halo is made up of axions

We can flip (T-symmetry) the graph of figure lla and

also produce axions via the Primakoff effect by shining

(laser) light through a magnetic field Here a photon

21

Photon

a

Photon

Figure 14 The axions couple to two photons through the triangle anomaly

22

from the light beam with the correct polarization can

be combined with a virtual photon from the magnetic field

and produce an axion24 as shown in figure 1lb

A suitable extension of the electromagnetic Lagrargian

of equation 11 including the axion field is then (in

natural Heaviside-Lorentz units)

(118)

where a is the axion field rna its mass me the electron

mass the electromagnetic field tensor

(F IV-~IlAV_~VAI) P =~ FPC t d 1-0 Cl r llv 2EvPC 1 S ua and a 1 I 137 is

the fine structure constant M = Igayy where gayy is the

coupling constant of the axion to two photons

Now we consider a polarized laser beam entering at a

45deg angle with respect to the external magnetic field

Bext and propagating in the z direction (normal to BeXI) bull

Since the axion field is pseudoscalar the coupling term

is of the form (8 eXI EPh) namely only the polarization

component parallel to Bex can mix with the pseudoscalar

axion field 25

From the above Lagrangian using the Euler-Lagrange

equations for a (~ - ~I[~(~a) ] = 0) and for A II

23

we get the following equations of motion (written here

in matrix form)

o Qp == o (119)

BeX1WA1

where

2 ~_ a Bexwith ( 120)

( )4Sn B cr

A o A pare respectively the amplitudes of the orthogonal

and parallel photon components and a is the amplitude

of the axion field The gauge A 0 == 0 has been used and

the longitudinal photon component due to QED vacuum

polarization is neglected in the above calculations In

our case (in vacuum where the index of refraction is

such that In-lllaquol and w+k=2w because k=nw with k

the photon wavevector) we have

and then we can approximate

0 6 p 0 (121 )lW-io+l1 ~]] l]6 M

where

-m 24 7 a B oXI

6 = -w~ t =-w~ ta = 2w and t - shyo 2 p 2 M 2M

24

Because of the mixing of the A p and a fields we can

diagonalize the two lowest parts of the above equation

by rotating the original fields through a mixing angle cent

(as in any mixing problem)

[ coscent sincentJ[A p ] ( 122)

- si n cent cos cent a

AMwith tan2cent=2~ the new ts are now expressed as

p bull

tp+t a tp-t a = ( 123)

2 2cos2cent

Then by rotating back we can get the original components

Ap(Z)] [Ap(O)]= R(z)[ a(z) a(O)

with R(z) being the operator that propagates the photon

axion fields along z

COS cent si n cent ] R(z)= [

sml cos cent

Because cent is small in our case it suffices to expand R(z)

to second order in cent R(z) = RO(z) + centRl(z) + cent2R2(Z) with

Ro(~)=[~ e~tJ Rl(~)=(l-eit)[~ ~l and ( 124)

The phase shift is then given by

cent II (z) = - pound (R 11 (z)) = cent 2 ( t p - t a) Z - si n [( t p - t II) Z] ( 1 25)

and the attenuation by

25

( 126)

both of order cp2 (figures 1 5 and 1 6) bull For cent small

tan2cp2cp and for m a = 10- 5 eV (126) becomes

( 127)

Equation 127 holds for a single pass of the photon beam

through the region of the magnetic field

If a pair of mirrors is used to bounce the light back

and forth equation 127 has to be modified Instead of

Z2 we must use N [2 where N is the number of reflections

and l is the length of the magnetic-field region This

is because axions go through the mirrors while photons

are reflected and the coherence of the two fields is

therefore destroyed after every pass We then find

Similarly the phase shift is given by

cP =N ( B ext W ) 2 m l _ sin ( m l ) ( 129) a Mm~ 2w 2w

In equations 128 and 129 QED vacuum polarization is

neglected The rotation is e=(e2)sin26 where e is the

angle between the initial polarization and the direction

of the external magnetic field and the ellipticity is

1jJ=cp2 (see section 28) In these units (natural

26

x

t 2epSilOmiddot

y

Figure 15 Only the parallel component produces axions resulting to the rotation of the original vector E

27

Axion

Figure 16 The parallel component creates and reabsorbs axions and therefore lags behind the other component (ellipticity)

28

Heaviside - Lorentz) a magnetic field of 1T can be expressed

as 195 eV2 and a length of 1cm as 5 x 104 eV-1 Figures

17 and 18 show the limits that can be obtained on the

axion coupling constant as a function of its r~ss for

rotation and ellipticity limits of 10-12 rad respectively

for N=600 reflections l=880 cm ()j=241 eV and B QxF5T

These are the kinds of sensitivities we expect from our

experiment in the final configuration For an axion mass

ma~8X10-4 eV the oscillation length of the axionphoton

system equals the magnetic field length l and the

probability of creating an axion with that mass in our

system is exactly zero The same is true when the magnetic

field length is a multiple of the above oscillation length

and is demonstrated in figures 1 7 and 18 with the

oscillating feature in the limits of the coupling constant

29

1E9

0 L-J 1E8

-+- c 0

-gt- (fJ 1[7c 0 0

CJl C -

w 0

-g- -1 E6 0 u

I1E5+1-shy1E-5 1E 4 1E 3

Axion rnass [eV]

Figure 17 Limits (the lower part of the area is excluded) on axion coupling constant vs mass assuming

-1a rotation of 10 rad

1E9x------------------------------- shy

------ gt(])

C) 1E8LJ

gtshy+- c 0

+-

1E7(f)

c 0 u en c

w ~ Q lE6

J 0

U

1E5+----+--~~-middot-~-r~~H------middot-+---~--~~+-rH-~--

1E-5 1 E 4 1E 3

Axion IllOSS leV]

Figure 18 limits (the lower part of the area is exch ded) on axion coupling constant VS mass asslJming an ellipticity of 1(j 12 rad

References

1) S De Panfilis et al Phys Rev Lett 59 839 (1987)

2) H Primakoff Phys Rev 81 899 (1951)

3) T Bhattacharya and P Roy Phys Rev Lett 59 1517 (1987) the authors discuss a model with a chiralscalar s that couples to two photons with a coupling constant 9s yy 2 x IO- 3 Cey-l Our experiment excludes this coupling constant by 3 orders of magnitude

4) H Euler and Heisenberg Z Phys 98 718 (1936) V s Weisskopf Mat-Fys Medd Dan Vidensk selsk 14 6 (1936) S L Adler Ann of Phys (NY) 87 599 (1971) J Schwinger Phys Rev 82 664 (1951)

5) The equivalent magnetic field from a laser light of 1 watt and 2 rom diameter is B ph 052 gausslaquo B extshy

6) Wu-yang Tsai and Thomas Erber Phys Rev D12 1132 (1975)

7) R Novick M C Weisskopf J R P Angel and P G Sutherland Astroph J 215 Ll17 (1977)

8) E Iacopini CERN-EP84-60 1984

9) J Calmet et al Rev of Mod Phys 49 21 (1977)

10) M Delbruck z Phys 84 144 (1933) C Jarlskog Phys Rev D8 3813 (1973)

11) J E Kim Phys Rep 150 1 (1987)

12) Baluni Phys Rev D19 2227 (1979)

13) R D Peccei and H R Quinn Phys Rev Lett 38

1440 (1977) Phys Rev D16 1791 (1977)

32

14) I Antoniadis The axion problem in 1st Hellenic School on Elementary Particles Corfu 1982 ed World Scientific

15) S Weinberg Phys Rev Lett 40 223 (1978)

16) F Wilczek Phys Rev Lett 40 279 (1978)

17) E M Riordan et al Phys Rev Lett 59755 (1987)

18) C Edwards et al Phys Rev Lett 48 903 (1982) Sivertz et al Phys Rev D26 717 (1982)

19) I Antoniadis and T N Truong Phys Lett 109B 67 (1982)

20) c M Hoffmann Phys Rev D34 2167 (1986) N J Baker et al Phys Rev Lett 59 2832 (1987) Y Nagashima Proceedings of Neutrino 81 Hawaii July 1981

21) J Kim Phys Rev Lett 43 103 (1979) M Dine W Fischler and M Srednicki Phys Lett 104B 199 (1981)

22) W U Wuensch et al Phys Rev D40 3153 (1989)

23) Talk given by Chris Hagmann Proceedings of the Workshop on Cosmic Axions C Jones ed World Sientific (1989)

24) L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

25) Georg Raffelt and Leo Stodolsky Phys Rev D37 1237 (1988)

26) M V Berry Proc Roy Soc London A392 45 (1984) R Y Chiao and Yong-shi Wu Phys Rev Lett 57 933 (1986) A Tomita and R Y Chiao Phys Rev Lett 57 937 (1986)

33

Chapter 2 Apparatus

21 Magnets

An apparatus (fig 21 22) capable of measuring

optical rotation or ellipticity at the level of 10-12 rad

must be very carefully designed and must utilize state

of the art optics and electronics equipment 1 To

appreciate the smallness of the angle note that it is

the angle seen by an observer in Rochester when a point

at Brookhaven (approximately 400 miles away) moves by 1

micron On the other hand even to reach such an angle

a large magnetic field over a long effective length is

required (in terms of light travel in the region of the

field) Therefore our task is to combine the delicate

optics with the massive superconducting magnets and the

cryogenics in such a way that they can work without causing

problems to the measurable effects

In order to be able to observe the QED effect we need

to have a dipole magnetic field on the order of 5 Tesla

and be able to modulate the whole field as fast as possible

In our experiment we are using two CBA (Colliding Beam

Accelerator) magnets (fig 23 24) each 44m long

They are maintained at 47degK by supercritical helium gas

circuit Figure 25 shows the quench current versus

34

~

() U1

r---- ~-~~~~~-~~i ~ A----------- w-----0 -~--0- ---~ ~idnetlc Optical table

bull bull W FG awP

A bull ~ Vacuum box An Analyzer1--( -u- t3 -0- -1 LASER EO Electrooptic crystal Per _ rei EO Pol Fe Faraday cell

CI HWP Half wave plate Per Periscope

Optical table Pol Polarizer QWP Quarter wave plate Tel Telescope W Window

Figure 21 Apparatus for measuring small optical rotationbirefringence

Figure 22 Apparatus in the enclosure The large box in the rear to the left is the vacuum box containing most of the optics

36

CORR[CTION

EPOXY

W J

FIBERGLASS-EPOXY

COl D BORE TUBE

WITH

SPACE fOR SurER INSUlATION-

WARM 80Rpound TUBE IVACUUM CUAMBERI

EPOXY SPACERS

FIBERGlASS-EPOXY WITU HELIUM COOLING CHANNELS

Figure 23 End section of a CBA magnet

bull

Q5

5

38

CBA MAGNET QUENCH LINE

5r-------~----------------------~

4

3

2

o~----~----~------~----~----~ 4 5 6 7 8 9

T [K]

Figure 25 Quench current vs temperature

39

temperature These magnets were prototypes for the CBA

ring (their code numbers are LM0014 and LM0018) a high

intensity p-p collider consisting of two strings of

magnets Special care was taken in design to mini~~z~

the stray magnetic field

The modulation of the magnets at a high speed presents

a problem since the specifications were for a ramping rate

of 4 As We have operated the magnets at 115 As which

for full modulation (0 H 3500 A) results in a frequency

of -165 mHz

The measured magnetic field as a function of current2

is given in table 21 The magnetic field is B[TJ = TF

x I[Amps] x 10-4 where TF is the transfer function (figs

26 27) The field configuration is that of a dipole

with vertical direction and homogeneity (field variation)

better than 10-4

The stray magnetic field (fig 28) is small a fact

which made our task of shielding the optics considerably

easier The most sensitive elements are the two big

mirrors at the cavity ends the polarizeranalyzer the

Faraday glass and the A4 Plate (QWP) If we want the

background to be less than 10 at a 10-12 rad angle then

the limitations are very stringent According to reference

3 the dominant effect on the optical properties of a

dielectric mirror is the Faraday rotation with

4gtF37 X lO-IOradIG This of course refers to a single

40

TABLE 21

MAGNET Curr [A] JBdl [Tm] Effective Length [m]

LMOO18 264 1 8190 44251

LMOO18 1200 82356 44051

LMOO18 2000 43878

LMOO18 3500 217335 43566

LMOO14 264 1 8183

LMOO14 1200 82349

LMOO14 2000 43865

LMOO14 3500 21 7443

41

--

________ __________

2 LM0014 Transfer function

- v)

I

0shyE 0

Vl VlO J~ 0 01

Z 0 1-

~ J L

a IoJ ~

~C lt Q I shy

~________~__________

degOQ _ (lt00(01)0 l J

~~ A ~

bull1 60

0

0 o~

omiddot omiddot6o middot6 of

0 o~

oa I)

0 6

0 tf6

01)

6

6 6

6 6

6 6

6 6

6 6

6

1pound shyN-o

o I N o - = o

~ I~__________

o 1000 2000 3000 4000 I (amper e)

Figure 26 LM0014 Transfer function

~ ~

42

--

I

2 LM0018 Transfer function

-J) I Ie

E

~

l n J _0 shy I shy ~-0

ill 11)0 111 o shy shy

N j CI

~z I - i

I- I -= e

lt z shy -~

=

~L-________ _________L________~~________~________~ ~

a 000 2000 OOO 000 I (ampere)

Figure 27 LM001 B Transfer function

43

I

STRAY FIEDS AT r=20

160

120

Vl Vl

o MEASUREMENT I j

r C~ Leu L to T 0 ~~ (M 0 P) ( I I

shy56 KG

l 100lshyltX lt-shyc I

I w i LI - 801-

I J

Iu I I ~

J I

IIJ i lt I I ltX 60 L

I I

40

20

0 0

0 00 00 0 0

1000 2000 3000 4000 I AMPS

Figure 28 Stray Magnetic Field

44

bounce If we have about 1000 bounces and want to have

a background of less than 10-13 rad then the permissible

modulated stray field in the normal direction to the

mirror is less than 027 ~gauss

Our Faraday glass has a Verdet constant of 325 radTm

for A = 5145nm The Faraday rotation is given by centF =

VBI where V is the Verdet constant in radTesla-meter and

is linear with the wavelength of the light B is the

magnetic field and I is the length in m (see section 24

OPTICS below) For less than 10-13 rad spurious signal

we would need to have a longitudinal stray field of less

than 10-8 gauss The stray field on the polarizeranalyzer

and QWP must be of the same order (10-8 gauss)

22 Laser

Our light source is an Innova 90 5 Argon Ion laser

manufactured by Coherent It can deliver 5 watts in all

lines (but the main ones are 488nm (blue) with 15 watts

and 5145nm (green) with 2 watts) Usually the 488nm is

the dominant line of Ar+ lasers but it has a lower

saturation level than the 514 5nm and therefore the latter

becomes the dominant one at high powers

45

Figure 29 shows the levels of Ar+4 The neutral Ar

gas must first be ionized by the discharge electrons and

then the upper level of the Ar+ must be populated by a

second collision with the electrons A current density

of 700 Ampscm2 is needed to maintain continuous lasing 5

In order to suppress high order modes of oscillations

(ie TEMnn with nraquol) and because a very large current

is needed the discharge is confined within a tube whose

diameter is a few millimeters When the electrons strike

and ionize the Ar atoms the Ar+ ions sputter the walls

at high speed A solenoid magnetic field is established

around the discharge tube and serves a twofold purpose

a) to prevent the atoms from sputtering the wall thus

decreasing gas consumption and wall damage and b) to

confine the discharge along the center part of the tube

consequently increasing the current density Because

consumption of Ar gas is unavoidable our laser has an

automatic refill system which keeps the pressure stable

within 10 mTorr which in turn means that the pressure is

within 3 of the optimum condition (-1 Torr)

The ions inside the tube have a temperature of 3000 0 K

because they have been accelerated by the discharge

electric field and water cooling is needed to remove the

generated heat A minimum of 22 gallonsminute is

46

- --plaa

25

i-20

-_______ 35 Amiddot

(Ground atate

Ar

Figure 29 Atomic Ar levels

47

required for our laser We tune the laser at the green

light (5145nm) which corresponds to

(21 )

with a line width of 35 GHz

The laser resonator (fig 210) consists of a plane

high reflectivity mirror (greater than 99) and a curved

mirror with curvature of several meters and reflectivity

of 95 The Brewster windows made of crystalline quartz

select a particular polarization (in our case vertical

ie the photon electric field vector is vertical) and

the net polarization ratio is 1001 The prism at the

far end (not shown in figure 210) is used for wavelength

selection

There are two modes of operation a) current regulation

mode which tries to keep the current in the tube constant

by feedback techniques and b) light modulation where the

sample light going through a beamsplitter at the output

coupler is utilized and compared to a stable voltage

When the light changes the feedback system corrects by

supplying the necessary current to the tube Tables 22

23 and 24 show the laser specifications (as given by

the manufacturer)

Vacuum

48

TABLE 22

PERFORMANCE PARAMETER SPECIFICATIONS

Beam Diameter (at the

output coupler)

15 rom at 1e2 points

Beam waist diameter

(located 140 cm behind the

output coupler)

12 mm at 1e2 points

Long Term Power stability

(over any 30 minute period

after 2 hour warm-up)

Current Regulation plusmn3

Light Regulation plusmn05

optical Noise Current Regulation 02

rms

Light Regulation 02 rms

49

TABLE 23

LASER SPECIFICATIONS (COHERENT INNOVA 9o_5)

Bore Configuration Tungsten disk with one piece

ceramic envelope

Plasma Tube Cooling Conductively water cooled

Resonator Construction Thermally compensated Invar

rod structure

Cavity configuration Flat high reflector long

radius output coupler

Output Polarization 1001 Electric Vector

Vertical

Cavity Length 1093 mm (4303 in)

Excitation Current Regulated DC

Input Voltage 208 plusmn 10 Vac 60 Hz 3 phase

Maximum Input Current 45 Amperes per phase

Maximum Tube Discharge

Current

40 Amperes

Cooling water

Flow Rate

Incoming Temperature

Pressure

85 liters (22 gallons) per

minute (minimum)

300 C (860 F) maximum

Minimum 1 76 kgcm2 (25 psi)

Maximum 352 kgcm2 (50 psi)

50

TABLB 24

OUTPUT POWBR (TEMOO) SPECIFICATIONS

Wavelength Output Power

(nro) (roW)

All lines 5000

5287 350

5145 2000

501 7 400

4965 600

4880 1500

4765 600

4727 200

4658 150

4579 350

4545 120

351 1 - 3638 200

51

Feedlhroughs Ceramic sleeve Gas fill port

ilI~ l t-1 tl VA

1middotdeg Mirror mount

Gaha ~ae lacke

Tube support s

Inv3r resonator rods o ring seal

Ul tJ

Figure 210 A typical Ar ion laser resonator

As can be seen from figure 21 the optical elements

are resting on two tables The near end table which

holds all the optics except the endcap containing the

return mirror is made of granite weighs 2 tons and its

dimensions are 96X48X14 It sits on an iron frame

which in turn sits on three jacks directly on the tunnel

floor The table at the far end is also a marble plate

with an iron frame

The optics box on the table is connected to the vacuum

and contains most of the optical elements It is made

of aluminum and weighs approximately 200 kg This large

mass helps to keep the optics at uniform and stable

temperature which would otherwise affect the index of

refraction A 20 litis ion pump backed by a turbo pump

maintains the vacuum at 10-4 Torr The turbo pump is

attached directly to one side of the box but currently

there is no bellows between it and the box This made

it necessary to turn the pump off while data were taken

Another turbo pump is used for the insulating vacuum

of the magnets it is occasionally used to pump down the

optics box as well The nominal speed is 120 litis but

because of long lines and small diameter constraints the

effective speed is significantly less

The principal vacuum tube where the magnetic field

is and where the light travels back and forth several

hundred times is pumped by three 20 litis ion pumps and

53

three titanium sublimators We monitor the vacuum with

three ion gauges (figure 24) operating in the range from

10-3 to 10-12 Torr

Inside the optics box there are several mounts and

over 25 remote control motors and cables which outgas

While taking data all the mechanical pumps are off and

the vacuum inside the box is around 10-4 Torr In the

tube the vacuum depends on whether the system was

previously baked and if the sublimators were fired In

that case and when the valve between the box and the tube

is closed the vacuum attained is in the 10-9 Torr range

and presumably 10-10 Torr could be reached for a clean

system With the valve open and the pressure in the box

around 10-5 Torr we had 33 x 10-6 Torr and 89 x 10-7

Torr at the two ends of the beam tube

23 optics

Immediately following the laser head there is a

polarizer followed by a power stabilizer with EO crystal

(see section 33) Following that we have installed a

54

telescope to match the cavity a V2 Plate (HWP) a

periscope and a 45deg mirror All these components are

outside the vacuum (see fig 21)

Polarizers

The light enters through a window in the vacuum box

and is incident on the first polarizer as shown in figure

21 The polarizers are air spaced for high laser power

vacuum compatible and the exit surfaces are tilted by

3deg to eliminate etalon effects They were manufactured

by Karl Lambrecht corporation in chicago6 and the nominal

extinction is better than 10-6 over two thirds of the

surface This is a very conservative specification because

10-7 is routinely obtained We have achieved extinctions

down to 10-8 several times Extinction is the ratio of

the light coming through two crossed polarizers to the

light just before the second polarizer Following that

there is a mirror with a hole through which the light is

allowed to pass and a shunt mirror which is used for

diagnostics After the ray returns from the cavity it

bounces off the mirror with the hole and then off another

spherical mirror with radius of curvature 4 meters It

then passes through a Quarter Wave Plate (QWP) a Faraday

cell and the second polarizer (analyzer)

55

QWP

The QWP mount sits on two translators with one inch

travel each and is used only in ellipticity measurements

A two inch total travel is needed in order to make

successive ellipticityrotation measurements The QWP is

a device that has two axes a fast axis and a slow axis

In birefringent materials where the ordinary and

extraordinary rays have different velocities (and

therefore the index of refraction is different) the phase

difference between the two components is

-2n64gt=64gt -64gt =-(n -n )d (22)

0 A 0

where - is the light wavelength ne and no are the

corresponding indices of refraction for the extraordinary

and ordinary rays and d is the thickness of the crystal

that the light crosses When 164gtI=n2 then jn -n o d=A4

hence the name of such a device The polarization of the

light that is parallel to the fast axis is gaining a 90 0

phase with respect to the polarization component parallel

to the slow axis From figure 211a we see how a QWP

converts an ellipticity to a rotation Let the fast axis

be along x and the light be REP (right elliptically

polarized ie Ex lags behind Ey by n(2) then after

the QWP the Ex component will gain a phase n2 with respect

to the Ey and therefore they will have the same phase

Now we can recombine the vectors and we see that the net

56

x

y

a)

x

y

b)

Figure 211 a) Fast axis of QWP along x axis b) Fast axis along y

57

result is linearly polarized light but with the

polarization vector rotated by an angle ~ with respect to Ey

the x axis tan 11 = E It is clear that the fast axis must x

be oriented along one of the principal axes of the ellipse

if the fast axis is along y (fig 211b) then the Ex and

-Ey components have the same phase yielding again linearly

polarized light but with the polarization vector rotated

in the opposite direction Table 25 shows the sense of

rotation for LEP and REP light going through a QWP

TABLE 25

position LEP REP

Fast axis along X Ey

tan ~ =shyEx

Eytan ~ =-shy

Ex

Fast axis along Y Ey

tan~=--Ex

Eytan ~ =shy

Ex

Half wave plate

This plate simply rotates the polarization vector

The component that is parallel to the fast axis gets ahead

of the components parallel to the slow axis by 180deg The

phase difference is

58

2nllcent=n=--(n -n )d (23)A e 0

We find the new polarization vector by replacing Ex ~

-Ex where Ex is the component parallel to the fast axis

Faraday cell

Our Faraday cell (fig 212) consists of a plate of

BK7 glass inside a coil (solenoid) A tube holds the

glass in place and there are through cuts in both mounts

which serve to reduce any heating effects due to eddy

currents The coil actually is wound in two coils one

on top of the other The length of the coil is 344-inch

and the diameter I-inch The wire used is gauge (AWG)

No 28 (32 mils in diameter) The first (inner) coil

consists of ten layers 1500 turns and the resistance

is Ri = 28D The second (outer) coil consists of eight

layers 1100 turns and the resistance is Ro = 29D The

inductance of the inner coil is Li = 12mH and that of the

outer Lo = 15mH The magnetic field produced is (taking

into account the fact that the length of the coil is not

infinite)

B i = 4nNI =206 x J[Amps] gauss (24 )

for the inner coil and

B =4nNl=1506 X I[Amps] gauss (25)o

The impedances of the two coils are

59

AlumInum tube Part I side view 22middot~ OOSmiddot --

193~

l

bull 06 toOOSmiddot OSSmiddot OOOS 04Smiddot0005

bull

42tO05middot OSOOS

0 5

2tO05StO05middot AlumInum Coil 2 Side vIew

L~===~3~44~~0~0~05~-==~J7750005

0063 0005- 750005middot

Sharo edges ---------- shy 04005 ~ 8l2taps ~

(aligned)

Figure 212 Support tubes for the Faraday cell

60

(26)

and

where we used w = 2nv and v = 260 Hz as an example with

the two coils connected in parallel the magnetic field

is Bt = Bi + Bo with il-VIZ (V is the voltage acrossI

the two terminals) and i 0 = V I Z 0 206 1506)

B E = -+ [= 10XV[Volts] gauss (28)( Zi Zo

When polarized light goes through a crystal with a

Verdet constant V and there is a magnetic field along

the axis of light propagation then the polarization plane

rotates by an angle

e == V Bl (29)

where B is the magnetic field component along the axis

of propagation and I the total length of the crystal Our

crystal ( W - 13P 1-inch length furnished by optics for

Research Co) has a Verdet constant V = 40 radTm at A

= 633nm Since V is proportional to the wavelength A we

can find the constant at 514 5nm (Green) V = 325 radTm

Thus

(210)

The advantage of using the Faraday effect is that a properly

aligned crystal does not distort the light polarization

61

(extinctions are excellent) The main disadvantage of the

Faraday cell is that the coil acts as a low pass filter

as is apparent from equations 26 and 27 This makes

it very difficult to drive it at very high frequencies

Mounts

Most of the optics inside the vacuum box reside on

ORIEL Mounts For most of the elements the x y (which

define the plane normal to the beam direction) and ecent

motions are standard In some of the elements (where the

orientation of the birefringence or polarization axis

matters) the 3600 mounts are used allowing the adjustment

of the rotational degree of freedom as well

The mounts are motor controlled and some have position

readout with resolution Ol~ (~ 361 x 10-6 rad) We

calibrated the readout microns to rotation in rad by

rotating one mount360deg and dividing by the motor total

travel in microns The motor controllers are connected

with the motors inside the vacuum box through vacuum sealed

feedthroughs

62

24 cavity

The cavity consists of two high reflectivity (998)

interferential mirrors ground and coated by the optics

shop at the University of Rochester The mirrors have a

diameter of 112Scm and a radius of curvature of 1903rn

One mirror has a hole of S7mm diameter at its center

From equations 112 and 1 28 it is apparent that to maximize

the effect to be measured the light should go through

the magnetic field region several times since the effect

is proportional to the number of bounces This is achieved

by reflecting the light on the mirrors several hundred

times The necessary expressions for calculating the spot

position on the mirrors is given in reference 7 Usually

the light passes through a hole in the first mirror and

with the system al igned the successive spots form an

ellipse the number of bounces depends on the distance D

of the mirrors and the radius of curvature The number

of reflections achieved in this manner is limited by the

spot diameter and the diameter of the mirrors (in reality

the limitation is more stringent because of the reentrance

condition)7 For that reason we deform one mirror (so

that its focal length is different in one direction than

in the other) and when the ellipse is closed for the

first time the spot misses the hole and traces out more

ellipses Thus a Lissajous pattern8 is formed this is

63

shown in figure 213 where there are about 400 reflections

(a total of 800 for both mirrors) We use two methods

to measure the number of reflections

A) When the mirror reflectivity is known we can measure

the attenuation the light suffers inside the cavity The

intensity of the light emerging from the cavity is

- I I

o e Ao (211)

where nO is the number of reflections needed to attenuate

the light intensity to its lIe value n is the number of

reflections and IO is the initial intensity no is related

to the reflectivity

1 n =-- (212)

o 1 - R

After n reflections

and for (l-R)laquol we can approximate

and therefore no=ll-R)

The ratio IIIo is the ratio of the photodiode signal

when the light is traveling inside the cavity to that

when the shunt mirror is in place In such a measurement

the two polarizers must not be exactly crossed since the

ratio IIIo is usually around 05 and would be obscured

by extinction fluctuations if a good extinction was used

64

bull

Figure 213 Lissajous pattern on end mirror

65

If the reflectivity of the mirrors is not known then

nO can be found from equation 211 by counting a small

number of reflections This is however difficult for

reflectivities better than 998 because the attenuation

of the light intensity (for say 50 reflections) is very

small

B) The other method is based on measuring the time

delay for the traversal through the cavity We are using

a beamspl i tter at point A (fig 2 1) to feed one photodiode

and a chopper to chop the light (at -2 KHz) before point

A We feed this output to the first channel of an

oscilloscope (which we use at the alternate configuration)

and we trigger with this channel What we see on the

oscilloscope is a square wave NOw we feed to the second

channel of the oscilloscope the photodiode signal (fig

21) The oscilloscope screen shows two square waves

shifted relative to one another shifted (fig 214 a

b) by the time the light spends inside the cavity Figure

214b shows the null measurements that is the signal

photodiode sees the light before the cavity and therefore

the two square waves coincide The photodiode outputs

should be fed to the oscilloscope input directly (with a

500 termination for fast rise time) and one should beware

of introducing false phase delays (amplifiers) The slow

rise time in the figure is due to the chopper which limits

the accuracy of the method

66

a)

b)

---- Time (5x10-6 sDiv)

Figure 214 a) Time delay between the two paths b) Null measurement

67

Interferential mirrors introduce ellipticity because

they have a slow and a fast axis Therefore it is

important to align the mirrors so that the fast (or slow)

axis coincides with the polarization pl~ne of the laser

light In order to find this axis we setup a small test

system as shown in figure 215 The first polarizer

polarizes the light which is then reflected back and

forth between the two mirrors several times forming a

circular pattern Typical values are for the distance

between the two mirrors D = 50 cm and for the number of

reflections approximately 50 on each mirror (a total of

100 reflections) We need about this many reflections

because the total ellipticity which is cumulative is

100 times more than in a single bounce and the effect

becomes measurable If the axis of one of the mirrors

is known the measurement is straightforward If this is

not the case the procedure is much more time consuming

and the following measurements must be performed

a) Form a circle (fig 216) with several bounces on

each mirror and catch the exiting beam with the mirror

with the hole After this exiting beam goes through the

analyzer we rotate the latter to extinguish it We record

the intensity (as seen on the FFT at the chopper frequency) bull

b) We rotate one mirror by 5deg and try to extinguish

the beam recording at the end the minimum value We

repeat this until we complete a rotation through 360deg

68

a w

~

(I) 0 0IV shy

-NOshygt0 as 0

laquo 0 c s

c I 0

laquo 0

8bull 0

Q) ~~ 0 () Tmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddotmiddoty

(I)

8 c

(I) ~

(I) 0

~ (I)

0

(I) N ~

as 15 0

15 0

en C IV E IV (I) as (I)

E 25c (I) enc ~

Q5 zs E c (I)

C)

Iii N e en

u

~

~

69

bullbullbullbullbullbull bull bull bull bull bull bull bull bull bull bull bull bull bullbullbullbullbullbullbull

bull bull t bull

bull

Figure 216 Circle for birefringence axis

70

The total readout looks like figure 217

c) We find the minimum of the above minima and rotate

the mirror so that its axis (it is not the real axis yet)

coincides with the polarization plane of the light

d) We take the same measurements but now we rotate

the other mirror with the same increment of 5deg We find

again the minimum of the minima

e) Now we align the two axes of the mirrors found

from above and take exactly the same measurements by

rotating both mirrors together 5deg for each measurement

and find again the minimum of the minima The real axes

of both mirrors are along the new minima For data taking

in the experiment these axes should be along the direction

of the light polarization In addition when the light

is reflected off a mirror it acquires an ellipticity910

which is proportional to the square of the angle of

incidence and independent of the orientation of the fast

or slow axis of the mirror Therefore special care should

be taken to readjust the circles on the mirrors so that

they are always the same circles Another detail to be

concerned about is that different points on the mirrors

might have different reflectivity In order to exclude

this case we used a QWP in front of the analyzer and

tried to extinguish the beam In this way we can check

whether the non-extinction is due to ellipticity or due

to a local absorption

71

600t-2

500r2

400-2

300(gt-2

200-2

100-2

I r r

iA

~ j

0 to 20 30

Figure 217 Extinction (arbitrary units) vs rotation

Position 33 corresponds to 360 degrees rotation

72

40

The above procedure is tedious because the minimum in

the first stage (a) is not so obvious and because

readjustments must be made while rotating both mirrors

Fortunately this procedure needs to be performed only

once the final result shown in figure 218ab

25 Ray Transfer

A paraxial ray moving through an optical system will

change position and slope according to the following

equation (fig 219)11

(213)

where XIX I are the position and slope of the original

ray and X2X~ that of the transformed one Table 26

shows the ray transfer matrices for different optical

systems they have unit determinant (AD - Be = 1) assuming

no absorption The focal length of the system and the

location of the principal planes are related to the matrix

element by11

f = -1 IC

D-l h l =-shy (214)

C

A-I h =-shy

2 C

73

Figure 218 Extinction (arbitrary units) vs rotation

In the polar graph R is extinction in -dBV The two axes (fast amp slow) appear clearly

74

_----=_bull- - ~ - - shy ~ y p I I-----~~I ---shy TH 1 1 - ________

I I I I I I I I I I

X~

INPUT OUTPUT PLANE PLANE

Figure 219 Parameters of a paraxial ray

75

where the notation for h1 and h2 is given in figure 219

When the light bounces back and forth between two spherical

mirrors it undergoes a periodic focusing which is

equivalent (if we imagine the system unfolded) to the

light passing through a series of focusing lenses The

ray transfer matrix of this system is evaluated by means

of Sylvesters theorem 11

I~ BII1=_1_Asin8-Sin(n-l)8 Bsin n8 (2 IS) D sin8 Csinn8 Dsinn8-sin(n-l)8

where cos 8 = I 2 (A + D)

In order for periodic sequences to be stable

(periodically focused) they mus t obey - 1 lt 2 1

( A + D) lt 1 i

otherwise the sines and cosines become hyperbolic

functions and the spot size becomes bigger and bigger as

it passes through the elements In the case of a cavity

with spherical mirrors its dual a sequence of lenses

have focal lengths equal to those of the mirrors and the

distance between them is equal to the distance between

the two mirrors In that case the above inequality is

given by

(216)

Figure 220 shows the stable and unstable regions of the

cavity Our cavity (d = 1256m R = 1903m) corresponds

to point A far away from any unstable region

76

CONfOCAL (lIItalllt z d)

I

N I

~~~

Figure 220 Stable (white) regions and unstable (shaded) regions of mirror resonators Point A is for the cavity used in this experiment

n

26 Telescope

When the light exits the laser head it has a diameter

of 2w = 15mm at the lie point of the electric vector

and divergence 05 mrad The beam waist is 2wo = 12 mm

and is located 140 cm behind the output coupler of the

laser head at the output the beam has a Gaussian profile

A complete treatment of the propagation of a laser bear

can be found in references 11 and 12 the main results

being

(217)w(z) = W[ I + (n~n and

[ (nw2)2]R (z) =z 1 + l A ZO bull (218)

Here 2wO (beam waist) is the minimum diameter of the

Gaussian beam and at this point the phase front is plane

A is the light wavelength 2w is the beam diameter at the

lie points after a distance of travel z measured from

the point where w = wO- R(z) is the radius of curvature

of the wavefront at z If we SUbstitute in equation 217

A = 5145 nm Wo = O6mm for a distance z = 18m we would

have a beam diameter 2w = 10mm This actually would be

the spot size on the far end mirror To avoid this we

are using a telescope to match the beam to the cavityl011

78

We only need one middotlens in order to transform the beam

characteristics Then

and

(220)

where fo=nwIW2f f must be larger than fO and K=plusmnl

d1 (d2) is the distance between the position of the waist

of the original (transformed) beam and the lens similarly

w1 (w2)is the waist of the original (transformed) beam

ButWO(=w2) is given by

W 4 = (~)2 D(R 1-D)(R2 - D)(R 1+ R2 - D) (221)

o n (R 1 +R 2-2D)2

where R1 R2 are the radii of curvature of the two mirrors

and Wo the cavity waist For R1 = R2 the waist Wo is

located at the center of the cavity due to the symmetry

otherwise the distances t1 and t2 from the first and

second mirror respectively are given by

(222)

In our case R1 = R2 = R = 1903 m and D = 1256 m Therefore

W 4 = (~)2 D(R - D)2(2R - D) = (~)2 D(2R - D) =9

o n 4(R-D)2 n 4

Wo= 121mm (223)

79

ie the beam waist in the cavity diameter 2wO = 24mm

The beam diameters on the mirrors are

and

(224)

In our case (AR)2 D w 4 =w 4 =w 4 =i- =w=148mm (225)

1m 2m n 2R - D

ie beam diameter 2w = 3 mm

From the above when R = D (confocal cavity) we have

2 AR 2 AR w =- and w =shy (226) o 2nn

That is if we had a confocal cavity (R = D = 1256 m)

then the beam waist diameter and the beam diameter at the

mirrors would be

2wo =203mm and 2w=287mm (227)

As is apparent now our telescope must match the beam

waist w1 = 06mm which is located 140cm behind the output

coupler (from laser specifications) to the beam waist

at the center of the cavity w2 = 12mm (from equation

223) We see from equations 219 220 that once d11

d2 are fixed then f is fixed too since fO = 44m In

our case d1 = 2m d2 = 9m and

80

--e

Je

Ie N N Q) I 2l

Ji lL o

Ibull

81

Table 26

Ray Transfer Matrices of six Elementary Optical

structure

No OPTICAL SYSTEM RAY TRANSFER MATRIX

1

-d - I I 1 I

I 1 I I

[~ ~J

2

f

4shy I I

r shy1 0shy

1 1-7

3

--d-shy ~f

I 2

r -1 d d1-shy I shy -

f f

4

d_-df-~

~f of I

bull I 2

[ d did1-shy d l +d 2-shy11 I

l-~- d2_~+ d l d 21 1 d 2-Tt- 12 + 112 11 12 12 1112

5

- ---- ~~~~ n t j

bull~

n n~-inrl I I

- If~ 1 n 2cosd - Sind -

no ~nOn2 no

- ~non2sin d~ ~cosd -no no

6

~-7~ ---~ gt)

Y (

~1 ~~ middoth [~ dn]

1

82

=9f=641m (228)

We can use the above technique or use a more practical

approach which employs a combination of two lenses one

divergent and the other convergent put in series (fig

221) In the above figure H HI are the positions of

the principal planes of the telescope11 with

(229)

where f1 is the focal length of the convergent lens f2

that of the divergent (f2 is negative) and

where d is the distance between the two lenses As is

shown in figure 221 d1 d2 of equations 219 and 220

are with respect to the principal planes It should be

noted that equations 229 and 230 can be found from table

26 and equations 214

Therefore our telescope must satisfy the following

equations

(231 )

83

middot ddl=l+[ +fshyfl

d =D -(l+(+d-f~)2 lot f 2

with 1 = 140cm d ~ 5cm I ~ 40cm and Dtot ~ 11m By

solving these equations for fl f2 we can obtain the

parameters of the telescope In reality we used a computer

program in which we chose fl with a value that is available

commercially and varied d (within reasonable limits) so

that f2 is also available

27 Jones vectors and Matrices

The treatment of monochromatic light going through

polarizers QWP Faraday cell etc is greatly simplified

with the use of the Jones vectors and matrices Table

27 shows the Jones vectors for different kinds of

polarizations13 and table 28 shows the Jones matrices

for different optical elements 13 When the laser light

goes through an optical element the final Jones vector

can be found by multiplying the original Jones vector by

the Jones matrix of that element As an example we are

going to use here this formalism to treat our system in

two configurations

84

Table 27

Jones vectors

Light NormalizedVector

Vector

Linearly Polarized light along the x axis [Ax~middotmiddotJ [~J Linearly Polarized light along the y axis [~J[A~middot] Linearly Polarized

Axe [COSSJ[ ]light in an angle 9 SineA 14gt~with respect to x axis ye

Linearly Polarized [ COse ]axis in an angle -9 [ Axmiddot] - sin eA Ifgtxwith respect to x axis - ye

Left Circularly Polarized light (LCP) [ Amiddotmiddot ] ~[~Ji~-n2)

Axe

Right Circularly Polarized light (RCP) [ Amiddot ] ~[~iJi~ n2)

Axe

Elliptically Polarized light [A ] [ cose -]

lltIgtyAye sin Se

85

2

Table 28

Jones Matrices

optical Element 8= deg qeneral 8

Ideal Polarizer at an angle 8 with

respect to the axis [~ ~J [ cose cosesin eJ cos8sin 8 sin 2 8

14 plate (QWP) with

fast axis at an angle 8 [~ ~J [ cos

2 e-isin

2 e cosesin9(1 +i)J

cosesin 9( 1 + i) -icos 2 e+sin 2 9

AJ2 plate (HWP) with

fast axis at an

angle e [~ _deg1 ] [ cos28 sin 28 ]

sin28 - cos29

Plate introducing phase delay ltP with fast axis at e [~ e~~ ]

[ cos 2e sin 2ee-middotmiddot cosesineCl-e-)] cos9sineCI -e-middotmiddot) cos 2 ee-middotmiddot sln 2e

0

Note e is the angle between the polarization direction

or the fast axis and the x axis The Jones matrices for

e degare obtained from those for 8 =degby using the rotation

matrices ie rotate first the x axis along the fast

axis (polarization axis) of the element perform the

multiplication with the Jones matrix of the element at

0 and then rotate the x axis back to the original

direction

86

a) Ellipticity Polarizer at 0middot phase shift ~ from the

cavity due to the magnetic field with fast axis at e (the

fast axis is along the direction orthogonal to the magnetic

field see equation 110) a QWP with fast axis at 0middot

Faraday cell (rotation) and an analyzer at i+a where a

is the misalignment from the perfect crossing between the

polarizer and the analyzer The final Jones vector is

(from tables 27 28)

0 0J[ cosa sinaJ[ cosTj sin TjJ[ 1 [ deg 1 -sina cosa_ - sin Tj cos Tj deg

cos2S+sin2Se-i4gt (232)[ cos8sin8( l-e- i

4raquo

The misalignment angle a is of the order of 10-6 rad

Tj bull Tj (t F) is the Faraday rotation with a frequency of 260

Hz and is of the order of 10-3 rad1 ~ contains the time

dependence of the magnetic field since it is the phase

shift introduced to the light inside the magnetic field

This is much smaller than any other phase shift introduced

into the system and we take it to be between 10-8 - 10-12

rad Therefore the above matrices can be rewritten

(approximately) as

87

cos 2e+Sin 2e(l-icent) [ tcentcosesine

(233)

Therefore the current detected by a photodiode after the

analyzer is

then

(234 )

where we have used T)=T)OCOSWFt 1p=cent2=ljJocosw M t cent is

the phase shift and 1p the ellipticity and 10= IAxl2 the

light intensity after the polarizer It should be noted

that the signal of interest is maximum when e= plusmnnl4

b) Rotation We use the same configuration but without

the QWP

(235)

where E = Eo COSW Mt is the rotation introduced by the magnetic

field Expanding equation 235 we have

88

(236)

and as before the photocurrent is

(237)

In the above case we have used E as an angle of optical

rotation In reality there is an attenuation E (axion

case see figure 15) of the photon component parallel

to the magnetic field For an angle -8 between the first

polarizer and the magnetic field the Jones matrices become

sin8J[1-e[0 0J[ 1 aJ[ 1 11J[ cosedeg 1 - a 1 - T] 1 - si n e cose deg

[cose -sin8][1 sin e cose deg

neglecting the small terms in the above we obtain

(238)

89

comparing equations 236 and 238 we conclude that

for 6=plusmnn4 or

E=(E2)sin26

in general where E is the rotation of the polarization

plane of the beam and E is the attenuation of the component

parallel to the magnetic field Therefore

(239)

90

References

1) This experiment was first proposed by E Iacopini and

E Zavattini Phys Lett ass 151 (1979) see also E

Iacopini B smith G Stefanini and E Zavattini II

Nuovo Cimento 61S 21 (1981)

2) Magnet Measurements Analysis Analysis amp Measuring

Group BNL Report numbers TMG-259 (1982) and TMG-270

(1983) E J Bleser et al Nucl Instrum Meth A23S

435 (1985)

3) E Iacopini G Stefanini and E Zavattini Effects

of a magnetic field on the optical properties of dielectric

mirrors CERN-EP83-55 1983

4) Orazio Svelto Principles of Lasers translated and

edited by David C Hanna second edition Plenum Press

(QC688 S913) 1982

5) Rudi Wiedemann Lasers and Optronics October 1987

p 55

6) The code name of our polarizers is MGLQD-12

manufactured by Karl Lambrecht corporation 4204 N

Lincoln Avenue Chicago Illinois 60618 USA

91

7) D Herriott H Kogelnik and R Kompfner Appl opt

3523 (1964)

8) D Herriott and Harry J Schulte Appl opt bull 883

(1965)

9) S Carusoto et al The Ellipticity Introduced by

Interferential Mirrors on a Linearly Polarized Light Beam

orthogonally Reflected1 CERN-EP88-114 1 1988

10) M A Bouchiat and L Pottier Appl Phys B29 1 43

(1982)

11) H Kogelnik and T Li Proc IEEE 5 1312 (1966)

12) Miles V Klein and Thomas E Furtak OPTICS second

edition John Wiley amp Sons Inc 1986

13) I Spyridelis et al Askiseis Optikis Tefhos ~

ARISTOTELEIO PANEPISTIMIO THESSALONIKIS 1980

92

Chapter 3 Data Acquisition

31 Electronics

In the previous chapters we described a system designed

to search for ellipticity and rotation introduced by a

magnetic field The sources of ellipticity are both the

QED vacuum polarization and axions (depending on the axion

mass and coupling constant) while the source of rotation

is only the axions As we saw in the last chapter (equations

232 235) the two effects require a different setup

To search for ellipticity we must have a QWP in the optical

path in which case any induced rotation gives no effect

to first order To search for a rotation we remove the

QWPi in the latter case it is the ellipticity that gives

no effect to first order The reason for this is that

ellipticity does not mix (interfere) with the Faraday

rotation and therefore the useful signal rather than being

linear in the field amplitude remains proportional to the

square of the effect Of course the signal of interest

is always proportional to the sought after effect times

the Faraday amplitude (see equations 234 and 239) If

we Fourier analyze IT we obtain an unwanted signal at the

Faraday cell frequency w F and two satellites containing

93

the signal of interest at frequencies W F plusmn W M where W M

is the magnet modulation frequency The magnet frequency

however is in the tens of millihertz region and it becomes

apparent that the width of the Faraday signal must be

small and noise free Also all the clocks must be locked

together to avoid any phase jitter

Figure 31 shows the electronics setup of our system

The Fast-Fourier Transform or FFT (HP 35660A Dynamic

Signal Analyzer) has an internal synthesizer with very

stable frequency output We use this synthesizer at 260

Hz and 35V zero-peak to feed a TECHRON 5515 voltage

amplifier which in turn drives the Faraday cell The

width of the Faraday frequency line is much less than 1

mHz as measured with the same FFT

The transmitted light is fed to a silicon photodiode

(table 31) The area of this photodiode is kept small

so as to minimize its capacitance which restrains the high

frequency response Because of this we use a focusing

lens before the photodiode with a focal length between

Bmm to 10mm We make sure the focal point is not exactly

on the photodiode surface to avoid surface current density

limitations In order to avoid Etalon effects we removed

the glass cover of the photodiode otherwise the incoming

ray would interfere with the multiple reflections in the

glass and the signal would become very sensitive to the

beam pointing stability

94

II W () () if t-

ATCOMPAllBLE

COMPUTER

HPI8

HP35660A

FFT

INTERNAL

INPUT

RS232C ORIEL

18011

ENCODER

MOTOR CONTROLLER

BAND PASS FILTER

MOTORS

E SYNTHESIZER CURRENT PREAMPLIFIER

260Hz 3SV

260 Hz 10V

FARADAY

CELL

OPTOCOUPLER

SCALAR

t3328

OPTOCOUPLER MAGNET TRIGGER

Figure 31 Electronics setup

95

Table 31

Silicon Photodiode

Type No S1336 - lSBQ 1 - TO - lS

11 x 11Size (rom)

Range (nm) 190 - 1100

Radiant Sensitivity (AW) 027 5145nm

Short Circuit Current Ishl 100 lux Min (jJA) 1 Typ (jJA) 12

Dark Current Id VR = 10mV Typ (pA) 2 Max (pA) 20

Temperature Dependence of Dark Current Typ 115 (Timesoc)

Shunt Resistance Rsh VR = 10mV Min (GD) 05 Typ (GD) 5

Junction Capacitance CjVR = OV Typ (pF) 20

Rise Time tr VR = OV RL == 1kD Typ (Ils) 01

4 x 10-15NEP Typ (WHz12)

3 x 1013D Typ (cm Hz12W)

5

Temperature Range Operating (OC)

Reverse Voltage VRmax (V)

-20-+60 Storage COC) -55-+S0

96

The photodiode output goes to a EGampG model 181 Current

sensitive preamplifier This amplifier has six different

feedback resistors selectable with a front panel switch

(10-4 VIA to 10-9 VIA in increments of factors of 10)

Figure 32 shows the noise gain and input impedance vs

frequency for the different settings A battery-powered

Hamamatsu C1837 amplifier with two feedback resistor

settings (R = 107 0 and R = 109 0) was used for the early

measurements (until October 10 1989) The amplifier is

mounted far away from any metal surface to avoid

capacitively coupled ground loops which could pickup

transient noise The output cable was two conductor

shielded (one is used as the signal carrier and the other

as the return) the outer shield was grounded on the

amplifier end only Similar cable is used between the

TECHRON 5515 voltage amplifier and the Faraday cell The

above configuration seems to have eliminated interference

from transient noise

It should be noted that the optics (vacuum box with

the table laser head and one of the endcaps) are in a

temperature controlled enclosure The only electronics

that are inside the enclosure are the preamplifier the

motor controller and the laser power stabilizer This

minimizes the need for accessing the enclosure the heat

load and electrical noise The output of the preamplifier

if fed to a bandpass filter (Frequency Devices 9002 used

97

FREOUENCY Hz

Figure 32 EGampG low noise preamplifier characteristics

Typicil noIbullbull CUFfnl It bull function 01 frqulncy nd nl shy11 lty

I)

t 0

~

-

Hl ~

u U Z

0 laquo

~ ~

i ~ i

FREQUENCY Hz

Tplcal inrpu1lmpednc bull bullbullbull function of tvlty nd I CIUM1C

FREQUENCY Hz

Tplcal fquency ponn a tunction Olnliivlly

shy~ Cl II 0 shy0 J

Z r c

FREQUENCY Hz

98

in differential input frequency band 220 Hz to 300 Hz

gain 1) used to eliminate the DC component and to attenuate

the signal at twice the Faraday frequency This allows

the FFT to detect lower level signals otherwise obscured

due to dynamic range limitations The output of the

bandpass is fed to the input of the FFT A typical FFT

setting is the following

Center 260 Hz (same as the Faraday frequency)

Full bandwidth (400 Channels) 15625 Hz

Window Hanning

Trigger External

Input AC Coupling Autorange

We use vector average since the noise (in volts) H z)

decreases as fN due to the randomness of the noise phase

where N is the number of averages whereas the signal

phase is of course fixed and therefore the signal vector

remains unaffected We take one measurement (one record)

at a time and calculate the vector average OFF line A

preliminary vector average is performed ON line for

diagnostic purposes

The output display of the FFT is read through an HPIB

(Hewlett-Packard Interface Bus) and a 37203A HPIB

extender by an AT compatible personal computer After

one measurement the computer commands the FFT to change

bandwidth (usually to 0 - 800 Hz) and take one average

from which it obtains the DC amplitude and the signal

99

at twice the Faraday frequency It stores the data on

the hard disk and updates the vector average on the screen

A note is need here about the HP 3660 FFT and its units

The setup for the units is dBVrms ie if the readout

is A dBVrms this corresponds to B Vrms

11

10 20B = V rms M A = 2010g B dBV rms

When we read a spectrum on the display through HPIB the

FFT always sends the values in linear units and in zero

to peak amplitudes independently of the displayed units

In case we read the marker as we do for the DC and the

signal at twice the Faraday frequency then the data

transferred through the HPIB is the same as the ones

displayed (ie dBVrms )

A more important detail concerns the DC readout from

the HP35660A FFT When the displayed values are zero

to peak amplitudes the DC readout is twice as much as the

actual value that is there is an offset of 6dB Now

if we change the displayed values from zero to peak to

rIDS (root mean square) the FFT just subtracts -3dB from

every channel (ie divides every displayed number by f2 ) even the channel with the DC Therefore if the

displayed values are in rIDS in order to obtain the correct

DC value we have to subtract only -3dB and not 6dB In

our case it is most convenient to use zero to peak

amplitudes and just subtract 6dB at the DC readouts

100

32 Misalignment Correction

From equations 234 and 239 we realize that our signals

are very close to the Faraday frequency where is present

the unwanted peak proportional to a 110 Because the

satellites are so close together there is no filter that

can remove the center peak If a is very large then a

problem arises because of the dynamic range limitations

(see section 41) Therefore a should be kept as low as

possible (below 10-5 rad) The task of keeping the

misalignment low is handled by the AT computer with the

help of the ORIEL 18011 motor controller (MC) The two

instruments communicate via an RS232 serial interface

The motor controller has three inputs and is connected

with the motor that rotates the analyzer another motor

that rotates the polarizer and the third one moves the

theta motion of the polarizer (theta is defined in the

polarization plane of the laser beam) Before the AT

commands the MC to move any motor it checks with the FFT

the amplitude at the Faraday frequency W F and at 2w F bull

Since at w F the signal is proportional to 2a110 and at 2WF

proportional to n~2 then the ratio is equal to 4ano

The Faraday amplitude no is kept between 10-4 and a few

times 10-3 rad Therefore our goal is to keep the above

ratio at about 100 (40 dB) which means that a is kept in

101

the 10-5 to 10-6 rad range If the misalignment is already

in this range the AT does not try to move any motor and

sets up the FFT for the next measurement Otherwise it

moves either the analyzer rotation or the polarizer theta

depending on the misalignment level In principle since

we know ~o very precisely we could rotate the analyzer

by a fixed amount (in linear dimensions 1~ of micrometer

motion corresponds to 361 x 10-5 rad) Furthermore if

we know the phase of the misalignment component we would

also know in which direction to move The position readout

of the motor controller has a resolution of Ol~m but the

motors move reliable only for distances in excess of O 5~m

(not to mention the backlash problem which is of the order

of a few microns) Thus in order to reach rotation

resolution of 10-6 rad we would have to have O 04~ readout

resolution on the micrometers Therefore the program

moves the analyzer rotation only if the misalignment is

within the range of the micrometers otherwise it moves

the polarizer theta By moving the theta motion of the

polarizer a few microns the extinction remains unchanged

whereas the projection of that motion on the rotation

plane gives us the necessary resolution to lower the

misalignment to the 10-6 rad range

There are two alternatives to the above scheme One

is to use a Faraday cell and apply a DC current through

the coil to rotate the polarization plane and to match

102

the extinction plane of the analyzer This could be done

with a HP 3325A synthesizer or a computer controlled DC

current source The other scheme is to use negative

feedback on the Faraday glass and keep the misalignment

continuously low

When the misalignment is at an acceptable level the

AT commands the FFT to change center frequency and

bandwidth After the FFT settles its digital filters

etc the AT enables the ARM command of the FFT which can

now accept a trigger to start the measurement The settl ing

time is a considerable part of our overhead because it

takes about 50 of the actual data taking time If the

time needed for one average is 512 sec then the settling

time is 256 sec however if the input of the FFT is

overloaded by a transient signal while it is settling a

new 256 sec period is initiated

Trigger

We have chosen to vector average so that we can have

both amplitude and phase information from our signals

To do that we must trigger the FFT and the magnets with

the same signal The output of the Faraday frequency

signal from the FFT I S internal synthesizer is split with

a tee connector One output is fed to the TECHRON 5515

Power Supply amplifier to amplify the voltage and drive

the Faraday coil and the other is fed to the a scaler

103

bull

which divides the original frequency by 3328 to obtain f

= 78125 mHz This is the magnet modulation frequency

(T = 128 sec) and care has been taken so that this

frequency falls in the center of an FFT channel

To eliminate any ground loops between the FFT the

scaler and the magnet electronics we use an optocoupler

to feed the scaler from the FFT and another one to feed

the magnet trigger from the scaler The optocoupler

consists of an LED (Light Emitting Diode) whose light is

picked up by a photodiode which in turn generates a current

with the same frequency This signal passes through a

currentvoltage amplifier with a peak value of about 15

volts which in turn is fed to a TTL converter

33 Laser Power Stabilizer

The laser amplitude noise at the lasers output is

much higher than the Shot Noise Limit (SNL) even though

the laser has a feedback system Figure 33 shows the

light spectral density versus power A major goal of the

experiment is to have very good extinction so that the

SNL dominates the amplitude noise

Nevertheless we can further reduce the light intensity

104

------

20 Laser sre_~ClI_density at 1400_H_z_____

r---1

N shy-

1 5 -shy1--

N I - L-J

-

(f) 1 0 ~ C - (J)

U ---shy~

0 ~ ~- L 0 +- O I) -ltshy

UI shyU (]) 0_

(f)

00 ---~--+__--+---_t_---+-___I

o 400 8 12 1600 2000 wer [mW]

Figure 33 Laser light spectral density (in units of 1E-S) VS laser output power

bull

fluctuations and any other thermal noise introduced by

the optics by using a Laser Power Controller The laser

light passes through many optics elements and bounces off

many mirrors All these elements introduce llf noise due

to thermal fluctuations (eg index of refraction etalon

effects position fluctuations etc) Furthermore the

cavity itself with many hundreds of bounces could move

the beam so that there are amplitude fluctuations caused

by different absorptions at the different positions of

the optics elements

A Laser Power Controller (LPC) consists of an

Electrooptic Crystal (EO) a polarizer and a photodiode

Polarized light goes through the Electrooptic Crystal

then through a polarizer and finally through a beamsplitter

which redirects 2 of the light to a temperature controlled

(33 0 C) photodiode with a Smm x Smm sensitive area The

output of the photodiode is fed to an electronics module

which monitors this readout and tries to keep it stable

by modulating the EO crystal The polarizer before the

beamsplitter allows only one component of the light to

pass through namely the one that is parallel to the

polarization axis By applying voltage to the EO crystal

the original vector is rotated (so that the output of the

polarizer is attenuated or amplified) by an amount equal

(negative in sign) to the change in amplitude of the laser

light that the photodiode observes Actually applying

106

voltage to an EO crystal introduces ellipticity to the

original polarized beam but effectively it is the same

as if the beam was rotated by an angle equal to the

ellipticity because the polarizer that follows rejects

one component no matter what its phase relative to the

other component

Of course we can use the photodiode at a different

position on the light path and stabilize the power there

by modulating the EO crystal In our experiment the light

returning from the cavity is split to two parts by the

analyzer one part is the transmitted 11 (or extraordinary)

ray which has all the signal information and the other

is the rejected (or ordinary) ray which has no signal

information but has the same amplitude noise as the

transmitted light Therefore we placed our photodiode

in the path of the rejected light Figure 34 shows

the nominal noise reduction versus frequency and figures

35 and 36 show our data with LPC OFF and ON respectively

Figure 37 shows the 11f and amplitude noise reduction

using the LPC The shot noise limit corresponds to -105

dBV (R = 1070 DC = 9 dBV) the first ten channels are

not a faithful representation of the intensity because

the AC coupling attenuates these very low frequencies

At the beginning of each run we complete a form (table

32) so that we keep track of the runs and files we write

on the hard disk

107

o~____________________________________________~LPC Noise Reduction IX)

0 to

C 0 0 0Jshyc Q)

-+oJ laquo

0 ~

0

0 0 000

0000 000

0 0

000

000 ltXgt

0 0 deg0 o~

101 2 102 2 103 2

Hz

Figure 34 Nominal noise reduction in dB vs frequency

108

-50shy

70

-90 gt m u

110 I bull I It fill II JIlfll ampfIn lflLI II

I- I0 I 11ft Vlni~

_ 1 3 0 l 1 r T~ ~~ ~ Irq I0

150+---~--~--~--

200 -120 l-----imiddot ---1--__1_---1

1 0 40 120 200

CHANNfl Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 35 Typical data with LPC OFF

bull

50

70

-middot90 ~shygt CD u

1 10 shy

~ ~L 1 lh a ~I~ ~J-~A ~r~ ~l ~V~ shyo

-130

-150 -tshy200 -120 L1 0

bull

--------------------------shy

~

40 120 200

CHANNEL

Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -80 dBV

Figure 36 Typical data with LPC ON The noise floor around 78 mHz is 10 times smaller than in previous figure

AVERACE COMiCETE s-c

A Mo~ka X 0 Hz VI -B6961 dBv~mc____---

-45 --1-- ------r---1 l-- -- f Ibull

dBVmc

LogMog ~~I+I~~I~I~--~----~----~----~----~------~-----------------10 dB

d i v 1uV It I All I J I I I I

~

~ ~

i

~ I yJ

I bull p amp - f I ~ I If IdWI I I I J I f Ii ~ fI- I- ----- I NYi ~ WChjJ ~~~~JcJ--1 I-

-05 i

I~ I 1 - I 1---------- shy

-1251 I I~ Stot C Hz Stop 1 e kHz Spctum ChClM 1 RMS 100

Figure 37 1f and amplitude noise reduction using the LPC The two

spectra shown are taken with the LPC OFFON

Table 32

DATE ITIME ILOGFILE

Laser Power FFT Trigger

LPC ON Constant Power Constant Transmission LPC OFF Transmission

EGampG Preamplifier Resistance n

MAGNET ON OFF Cavity Shunt mirror Shape

TM = s MIN Current of reflections MAX Current Polarization RAMP RATE UP As QWP IN OUT

DOWN As

Filter OUT DC (Source off) DC (Source on) 12wF

Filter IN DC

12wF

DC Offset 2W F Offset 12wFllwF

Voltage at the Faraday Cell 11 0 BW 0 2 1V

Cavity with Gas Vacuum TYPE PRESSURE Pressure ENCLOSURE END TEMPERATURE DOOR END 1 2 3 BOX

Notes

112

Chapter 4 Analysis of Results

41 Noise Sources

In equations 232 and 235 we used the matrices for

the ideal crossed polarizersanalyzers

[~ ~J [0deg 0J 1 and

In this case the light going through the pair of crossed

polarizers is zero which is the ideal case In reality

there is always light after the second polarizer equal

to the extinction (ie ITIO = 0 2 see chapter 2 under

optics) times the light intensity before the second

polarizer To accommodate for this the above matrices

can be rewritten

[~ ~J [~ ~J and

so that the light intensity after the analyzer is I =

1002 Taking this into account we can rewrite equation

234 to lowest order

(41 )

113

Equations 237 and 239 change accordingly We will call

our signal Wo and we will know that this corresponds to

Wosin26 for ellipticity Eo for rotation and ~sin26 for

attenuation

Shot Noise Limit

The DC current IDC = Io (0 2 + ~ + a 2 ) flowing through a

resistor fluctuates1 resulting to a rrns current noise

given by

JSNL(rms) = ~2eJ dcBW ( 42)

where e = 1 6 x 10-19 C and BW is the measurement bandwidth

This noise is white (same in amplitude at every frequency)

It is also the limit for most precision experiments today

even though new techniques that can beat this noise are

becoming available 2

In the case where the shot noise is the limit the

signal to noise ratio can be deduced from equation 41

( 43)

where q = 065 is the quantum efficiency of the photodiode

(027 AW radiant sensitivity at 5145 nmmiddot 241 eV = ftw

see table 31) and N is the total number of current

electrons collected In equation 43 we have used the

114

zero to peak shot noise and neglected the term a 2 bull

Therefore

SNR= ( 44)

where P is the laser power before the analyzer and T the

measurement time In practice we use 02=n~2 (see the

discussion of the amplitude noise below) Then for a

signal to noise ratio 1 laser power P = 05 watts before

the analyzer and Wo = 10-12 rad the integration time

needed is

2(SN R)]2rw)T= -=SSdays ( 4S)[ Wo pq

It should be noted that for Wo = 4 x 10-12 rad the integration

time becomes less than 35 days From equation 44 it

is apparent that SNR is maximum when n~raquo 0 2 but the

function n~(02+n~2) is slowly varying and it only changes

from 1 (for n~2 02) to 2 (for n~raquo02) It is however

difficult to reach the Shot Noise Limit for intense light

Therefore we keep the Faraday rotation angle at such a

2level that n~2 0 bull

Amplitude Noise

We can now include in the above calculation the laser

amplitude noise which is proportional to the light power

115

(46 )

where A is the amplitude noise or spectral density at

the signal frequency In the case where the amplitude

noise is dominant then

(ie the SNR is independent of the laser power) The

SNR given in equation 47 has a maximum when n~2=a2 and

becomes

( 48)

from which we conclude that the SNR improves for better

(smaller) extinction

From equation 46 we obtain the limit of the noise

spectral density if we are to be shot noise dominated

154 x 10-9 ~BW[Hz] ( 49)

~T~ a +shy2

that is

Alaquo 154 x 10-6~Hz=-1162dB~Hz for a2+n~2= 10- 6

Alaquo487x 10-6~Hz=-1062dB~Hz for a2+n~2= 10- 7

and (410)

Alaquo 2flw

116

for P = 05 watts q = 065 and hw = 241 eVe

Preamplifier noise dynamic range

Any resistor at finite temperature generates a white

noise voltage across its terminals known as Johnson noise

The rms voltage noise is given by

V rms ( R ) = J4 k T R B W (411)

where R is the resistance in n BW the bandwidth k the

Boltzmann constant and T the absolute temperature 4kT

= 162 x 10-20 V2HZD at room temperature (200 C = 68 0 F

= 293 0 K) The current noise is obtained from Vrms divided

by R

I R ) = ~ 4 k T B 11 (412)rms( R

Figure 32 shows the current noise for the different

settings of the EGampG amplifier versus frequency Even

though Johnson noise is white Irms depends on frequency

because R is the feedback impedance of the amplifier

Usually there is a capacitor in parallel with the feedback

resistor to prevent the amplifier from oscillating at high

frequency and therefore the impedance becomes smaller at

higher frequencies Also the photodiode capacitance (and

hence the sensitive area) should be kept as small as

possible for the same reason Figure 41 shows 1G2 where

G is the normalized gain versus frequency and R = 107n

117

(f)

gt

o --

N I

x Ql

Ie ~

r

118

----- N

-shy(f) (l

E laquo

L-J

(]) (f)

0 I- L_I- 0

1E-10

1 E - 1 1

1E-12

1 E 13

1E 14

1 E 15

E - 1 6 1 E

r------shy

Amplitude Noise

Initial laser power 05 watts

Shot Noise

[GampG Arnplifier Noise

J bull _l--LL l~tlL J~~-----LL J-LLd -LlJshybull

10 1E-9 1 E --8 1 [ 7 1E-6 1E-5

DCF

Figure 42 Noise sources In the experiment The amplitude noise drawn here corresponds to spectral density of 1E-5 per root Hz

using the Hamamatsu amplifier (C1837) Using this curve

we found the 3dB (ie 112 ) gain point to be at 1420

Hz At zero frequency G = 1 Figure 4 2 shows the

contribution from different noise sources in Amps~Hz as

a function of the DC factor (DCF) ie a2+~~2

When an analog signal is converted to digital with a

certain sampling rate an error is intr~d~ced because of

the finite spacing of the digital readout Assume that

Vs is the resolution (spacing) of the digital readout

Then the rms noise associated with one measurement is

1 JV12 V 2ltV2gt=_ V2dV =_s~

o V 5 -V2 5 5 12

I 2 V 5J ltV gt=-- (413)

o 23

and for Ns measurements the error becomes

Vs a --== (414)

q 2J3N s

If further the measurement consists of N

photoelectrons the measurement error is a SN =IN (shot

noise) Supposing that the 2N photoelectrons correspond

2n 2nto the full scale of the digital converter - 1 ~

where n is the number of bits then the resolution is

Vs = 2N2n and the quantization error becomes

N a =--== (415)

q 2 n J3N s

120

If we want the quantization error to be equal to the shot

noise the sampling rate must be

(416 )

Let us take as an example 05 watts of laser light

which corresponds to 85 x 1017 photoelectronss and

assume a typical extinction of 10-7 (ie N = 85 x 1010

photoelectronssec) and n = 12 bits The sampling rate

must be at least Ns = 1700 Hz If we require the

quantization error to be 10 times less than the shot noise

the sampling rate must be 170 KHz

Our FFT (HP35660A) has 12 bits 125 KHz sampling rate

and a dynamic range of 72 dB This means that when the

highest level signal is A volts the FFT is insensitive

to any signal that is more than 72 dB (- 3981 times) weaker

than A volts From equation 41 we see that in our spectrum

we have a DC component with relative strength of 10-6 to

10-8 (in units of IO) and a signal at 2w F also with

relative strength of 10-6 to 10-8 Our signal ~o~o is

supposed to have a relative strength of 10-14 to 10-16

Therefore we need to use a bandpass filter to eliminate

or attenuate the DC and the 2w F signal The problem

becomes more difficult when we consider the misalignment

peak at W F and our signals at W Ft W M where W M is in the

tens of mHz The ratio of the two peaks is 2aljJo with a

in the 10-5 to 10-6 rad range and Wo in the 10-12 range

121

then we would need a dynamic range of 120 dB it is obvious

that the dynamic range of the FFT is not adequate to

accommodate this range and there is no filter with such

narrow bandwidth as to separate the sidebands from the

W F peak

The way to overcome this difficulty is to introduce

random noise3 4 the level of which is within the dynaDic

range of the instrument Using vector averaging the

noise reduces as 1N a where Na is the number of averages

and the peaks can appear after an adequate number of

averages Actually we don I t have to introduce any random

noise because we have the laser shot noise (equation

42)

As an example we can look at laser power P = 05 watts

(which corresponds to 10 = 85 x 1017 photoelectronss) 2

2 noBW = 4 mHz and 0 =2 Then the ratio

2anoo a n 10 -= 0 ~ 72dB (417)2 el DCBW ~ el 0 ll~BW

determines that a must be less than 27 x 10-7 rad In

the above we have used the zero to peak shot noise

lf or flicker noise

Equation 234 shows that the signal is given by Is =

10 1l0Wo and the misalignment peak by I w =21 0a1l0 Because

the two signals are very close together in frequency 10

122

has to remain stable throughout the measurement In

reality however 10 is subject to lf noise the source

of which has been discussed earlier (see section 33)

42 Data Analysis

Using equation 128 we can set limits on the axion

coupling constant 5

(418)

The external magnetic field is Bxt = (BDC+BO)2 = B~c+Bt

+ 2B DC B obull where Boc is the DC magnetic field and Bo is

its amplitude modulation In our case the dominant term

is 2BocBO and we therefore use B~xl=2BDCBo I is the

magnetic field length and we use 1 = 2 x 439 m (see table

21) E is the rotation we measure or a limit on the

rotation angle In these units (natural Heaviside -

Lorentz) a magnetic field of 1T can be expressed as 195

eV2 and 1 cm as 5 x 104 eV-1 therefore for e = 45deg

I

M = 2 1 4 G eV x ~ 2 B DC B 0 [ T 2] 2 (419)

123

Rotation

A) Data with magnet off Figure 43 presents a single

record of data with the magnets off and no QWPi the

bandwidth is 390625 mHz (single channel bandwidth 0976

mHz) The number of reflections is 790 plusmn 35 (the time

delay is found to be 33 plusmn 1 5 ~s in a cavity 1256 m long)

FigurEs 44 and 45 show the rms (taking into account

only the amplitude of the data points) and vector (taking

into account both the amplitude and phase of the data

points) average of 26 records (files) The small peaks

that appear are due to the FFT external trigger (most

probably from ground loops) and are absent when the trigger

is removed The voltage at the Faraday coil is VFC = 79

Vo-p (from equation 210 we find TJomiddotOo- p = 826 x 10-5

(radV) x 79 V = 653 x 10-4 rad) and the peak at 2WF

is 12wF = -10 dBVO-p The Faraday frequency is WF = 260

Hz and the signal peaks are supposed to appear at (260 plusmn

001953) Hz (ie plusmn 20 channels away from the center peak)

The amplitude of these peaks is -106 dBVO-p and using

the following equations we can deduce the rotation angle

that this corresponds to The amplitude at 2WF is (using

equation 237)

( 420)

whereas at the signal frequency

124

---------------------~~----------~ 0

~

o

f ~ 1~

0 ~

- CO L -CO - 0

r --

0 c

CO I I ~

-0 0+--

I--z5~ i

I ~

8I

III IIII

6lI c

(j)

j

I I III+~ c

l

1 en u

0 0 N

0 0 0 0 0 0t1) f- v) - I) Lr)

I - - -shyI I I

8 0

125

50

-70 ~

--90 gtm D

-110

tIJ 0

-130

JIrIVH LJj~~~~

150+---+---+---~---~----~--

200 -120 t10 40 120 200

CHANNEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953mHz Amplitude of first harmonic -105 dBV

Figure 44 RMS average of 26 files

--

50

-70

-90 -

CD -0

-110

fJ

-130

_---------------------

~ 150+---~~------

200 -120 -40 40 120 200

CHAN~JEL

Center frequency 260 Hz Full BW 390 mHz Magnet frequency 1953 mHz Amplitude of first harmonic ~107 dBV

Figure 45 Vector average of 26 files (the same as in previous figure)

Is=oTJoEo (421)

From equations 420 and 421 we have

TJo Is E =--- (422)

o 2 12~

and substituting the values of I zw = -10 dBVO-PI Is =

-106 dBVO-PI we obtain

6 53 x 10-4 d 10-106120V ra o-p -9 Eo= 2 -)020 =517XIO rad (423)

10 V O-P

The above rotation limit corresponds to the false peak

(induced by the FFT trigger via ground loop) at 1953

mHz (the magnet period at the time was 512s but this

is data with the magnets off) Assuming the false peak

is removed Is = -113 dBVO-p and the rotation becomes 23

x 10-9 rad The noise around 80 mHz from the center peak

is -125 dBVo-p This corresponds to Eo = 58 x 10-10 rad

and comparing this with equation 423 the need to drive

the magnets as fast as possible becomes apparent

To acquire one record takes 1024 s (for a total

bandwidth of 390625 mHz) With the 26 measurements we

can place a limit of Eo = 58 x 10-10 rad therefore the

sensitivity of the experiment is

Es = EoX [T = 58 x 10- 10 rad x ~ 1024 26 =

=86X 10-sradlJHz (424 )

128

It should be noted that in this run the Laser Power

Controller (LPC) which can give us another factor of 10

in sensitivity was not in place During a later run

when the LPC was in operation we achieved a limit of Eo

= 54 x 10-10 rad corresponding to a sensitivity of Es =

18 x 10-8 radJHz

B) Data with magnet on Figure 46 shows the data

with the magnets on (log file name FNM30811ASC which

means it is the third run on 081189) The vacuum inside

the cavity is 65 x 10-6 Torr and 5 x 10-7 Torr at the

two ends and 10-5 Torr in the box The Faraday voltage

is VFC = 17 VO-P (ie no = 14 x 10-3 rad) 12wf = 004

dBVO-p total BW = 780 mHz and the magnet frequency is

3906 mHz (magnet period is 256s) The peaks are at Is

= -85 dBVO_p and from equation 422 we have

(425)

The number of reflections is the same as before 790 plusmn

35 whereas the magnetic field is modulated from 1100 to

1600 Amps with a ramp rate 100 Ampss and from 1600 Amps

to 1100 with a rate of 35 Ampss At these currents the

magnet transfer function is 1563 gaussAmp (see fig 26

and 27) that is Bl = 172 Tesla and Bh = 25 Tesla +8 8-8 1 8 2_8 28 11

Using B DC =-2- B O=-2- =+ 2BDCBOT we conclude

from equations 419 and 425 that

129

50

70 ~-

--g gt rn TJ hN~ vJ~Wyen

w o

130

-150+------4------~----middot+---

200 - 120 - 40 40 120 200

CHANNE Center frequency Full BW 780 mHz

260 Hz

Magnet frequency 39 mHz Amplitude of first harmonic -85 dBV

Figure 46 Typical rotation data with the magnets ON

790 M=214GeVx 8 392x 10shy

( 426)

assuming this limit to be correct within 10 due to

uncertainties on the number of reflections the Faraday

rotation (ie Faraday cell calibration) and the

polarization direction (angle e) bull

In order to find the source of these peaks we took

data with the shunt mirror (fig 21) blocking the light

from entering the cavity This way any electronic pickup

would appear again at the right frequency with the same

amplitude To account for the excess of light seen by

the signal photodiode (the cavity attenuates the light

which is now bypassed) we used neutral density filters

Figure 47 shows that under these conditions the peaks

appear at -105 dBVO_p This level is the same as the

peaks which appeared during the first run without the

magnet We have I zw = -85 dBVo-p VFC = 17 VO-p (~o = 14 x 10-3 rad) and therefore Eo = 105 x 10-8 rad Thus

the modulation of the light intensity increases by a factor

of 3 when the light traverses the cavity and we must

search for the source of this effect

The amount of gas inside the cavity cannot produce

such a large rotation and the endcaps were shielded against

any stray magnetic field below 1 mgauss without the

131

-50

7 () -shy

-90middotmiddotmiddot gt m TJ

-110 w I)

-130

I I ~~-150 -200 120 -- 4 0 40 120 200

C11AN r-j r-L Center frequency 260 Hz Full BW 780 mHz Magnet frequency 39 mHz Amplitude of first harmonic -1 05 dBV

Figure 47 Shunt mirror data with the magnets ON

magnetic shielding we measured a stray DC magnetic field

of 24 mgauss and a modulation of 3 mgauss at the far end

endcap where the lead pot is located (the above field

strengths are measured along the light propagation

direction only this component produces Faraday rotation

on the mirrors) Any Faraday effect on the mirrors is

linearly proportional to the magnetic field and the 3

mgauss could only cause a modulated rotation of less than

10-9 rad In any case further shielding the endcaps

with Il-metal reduced the stray field (even the DC component)

below the range of our magnet probe (1 mgauss) This had

no effect at the level of the peaks (actually the data

shown in figure 46 are with the Il-metal covering both

endcaps) In the next step we established that the effect

was proportional to both the BDC and Bo which implies a

quadratic dependence on the magnetic field

The ultimate test is rotating the polarization to 0deg

(ie parallel to the external magnetic field) Trying

this we found that the amplitude of the peaks was around

the same value (within 50) After that we positioned a

Mitutoyo displacement gauge on the back of the far end

end-cap We observed a signal at the magnet frequency

shown in figure 48 which corresponds to a 6 nm periodic

motion We also established the source of this signal

to be the magnet motion of about 04 JlID Table 41

133

bull

R~AL-TJM~ Ave COMPLETE

A Mark X 390625

-54 dBVm

LogMag 10 dB

div

w

-134

stct t 0 Hz Spct um Chctn I

---I-----middot~

____L-____

StOpl 15625 Hz OVLD RMSS

mHz

s c Tlk

YI -BlBB dBVrma

Figure 48 The cavity mirrors move due to magnetic motion

summarizes some of our rotation data with the magnetic

field on and off and for light polarization at 45middot or

omiddot with respect to the external magnetic field

Ellipticity

To search for ellipticity we include a QWP in front

of the Faraday cell to convert it to a rotation Data

taking is the same as previously except for making sure

that the QWP has reached thermal equilibrium Table 42

presents the results of the ellipticity data Several

trials with different elliptical orientations established

the fact that there was a strong correlation between

results and orientation The data shown in table 42 are

with different orientations of the cavity ellipse

Figure 49 shows the limits obtained from the rotation

data with N = 790 reflections Eo == 2x IO-Brad and B~xt

= 117 T2 assuming the signal to be due to axion

production Superimposed are the ellipticity limits for

induced ellipticity tlJo = 29 x IO-8 rad N = 38 reflections

and B~Xl = 1 36 T2 If both the rotation and ellipticity

peaks were due to axions then the axion mass and its

coupling constant could be found from the intersection

points of the two curves in figure 49 Assuming this

is the case we obtain an axion mass of about 10-3 eV and

a coupling constant somewhere between

2x I04GeVltMlt6x l04GeV due to many intersection points

135

Our noise floor during data taking was dominated by

amplitude noise which at best was a factor of 5 from the

shot noise We did not take particular care to lower

this noise because the spurious signals where at much

higher level

In order to find the absolute rotation of our signals

we have used the calibration given in equation 210 for

the Faraday cell An alternative approach is to use the

extinction to find the Faraday rotation It is easy to

show following equation 41 that

n J2W110 = 20 - shy

J DC (426)

where J~c is the DC light measured while the Faraday cell

is off

We saw in the various sets of data gathered with the

two methods a rate of agreement from 5 - 20 We decided

therefore to use the calibration of the Faraday cell for

consistency Applying the Faraday cell calibration to

the measurement of the Cotton-Mouton constant of N2 and

comparing our results with previous measurements (which

are very accurate) we found a 5 agreement

136

Table 41

Rotation data

Log file name FNM

00725 02728 02810 01811 03811

Magnetic field

(2BDC B O[T2]) 00 117 1 65 00 165

Polarizashytion

(degrees) 45 45 45 45 45

Faraday rotation

[rad] 65X10-4 12x10-3 21x10-3 21x10-3 14X10-3

Extinction 14X10-7 22x10-7

790

3x10-7 3X10-6

Number of Reflections 790 790 790

0 Shunt Mirror

Rotation Eo[rad]

52X10-9 2X10-8 43x10-8 45X10-9 10X10-8

Phase of the two

harmonics (-+) [ 0 ]

70 -70 No FFT trigger

79 -79 20 -20 86 -86

Frequency [mHz]

Visible peaks

1953

YES

1953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] NA 39 37 NA NA

Vacuum in the cavity ends [Torr]

65x10-6 5X10-7

2x10-6 2X10-6 NA

Number of records

26 4 5 36 49

Table continued on next page

137

Rotation data (Continued)

Log file name FNM

30S12 00S13 10S13 00S14

Magnetic field

( 2 B DC B0 [ T 2 ])

074 0S2 123 1 65

Polarization (degrees

45 45 45 45

Faraday rotation

[rad]

1 4X10-3 14X10-3 14X10-3 14X10-3

Extinction 3X10-6 3x10-6 3x10-6 3x10-6

Number of Reflections 790 790 790 790

Rotation Eo[radJ 17X10-S 92X10-9 16x10-S 35x10-S

Phase of the two

harmonics (-+) [ ]

No FFT trigger

No FFT trigger

No FFT trigger

No FFT trigger

Frequency [mHz]

Visible peaks

3953

YES

3953

YES

3953

YES

3953

YES

Coupling constant

M [105 GeV] 39 57 53 41

Vacuum in the cavity ends [Torr]

2x10-6 2x10-6 2x10-6 2x10-6

Number of records

45 35 6 3S

Table continued on next page

13S

Rotation data (continued)

Log file name FNM

30816 40816 10818 21005 31005

Magnetic field

(2B DC B 0[T2]) 078 082 165 1 36 1 36

Polarizashytion

(degrees) 45 45 0 0 45

Faraday rotation

[rad] 1 4x10-3 14X10-3 14X10-3 95X10-4 12X10-3

Extinction 3X10-6 3X10-6 1 5X10-5 3X10-7 14XIO-6

Number of Reflections 790 790 790 33 33

Rotation Eo[rad]

23X10-8 27x10-8 2X10-7 14X10-8 16x10-8

Phase of the two

harmonics (-+) [ 0 ]

No FFT trigger

No FFT trigger

No FFT trigger

2 -2 23 -23

Frequency [mHz]

Visible peaks

39

YES

39

YES

39 78

YES YES

78

YES

Coupling constant

M [105 GeV] 35 33 17 12 11

Vacuum in the cavityends [Torr]

2x10-6 2x10-6 1x10-5 1x10-4 1x10-4

Number of records

10 9 5 10 15

Table continued on next page

laquo

139

Rotation data (continued)

Log file name FNM 11006 21006

Magnetic field ( 2 B DC B 0 [ T 2 ]) 1361 36

Polarizashytion degrees 4545

Faraday rotation [rad] 11X10-362x10-4

3x10-7 3x10-7Extinction

Number of Reflections 38 38

47X10-8 59x10-8Rotation Eo[radJ

-31 31 harmonics (-+)

Phase of the two -36 36 [ 0 ]

78 Visible peaks Frequency [mHz) 78

YES

Coupling constant M (105 GeV]

YES

2022

5x10- 4

ends [Torr]

Number of records

5x10-4Vacuum in the cavity

59

140

Table 42

Ellipticity data

Log file name

FNM 01012 21015 31016 21016

Magnetic

field

( 2 B DC B 0 [ T 2 ])

00 1 36 1 36 1 36

Polarization

(degrees) 45 45 45 45

Faraday

rotation

[rad]

2X10-3 2X10- 3 2X10-3 2X10- 3

Extinction 1 5x10-7 15X10-7 15x10-7 15X10-7

Number of

Reflections 38 38 38 38

Ellipticity

V o[rad] 1lX10-8 29X10-8 11X10-7 22X10-7

Phase of the

two harmonics

(-+) [ 0 ]

-81 81 76 -76 -137 137 -128 128

Frequency

[mHz]

Visible

peaks

78

NO

78

YES

78

YES

78

YES

Vacuum in the

cavity ends

[Torr]

10-4 10-4 10-4 10-4

ltI

141

CJ

I

f

t I

-l-

I w

T)

I 0 w ----- ~ W E

gtQ

0(f) (f) shy0 -fIJ

sectE 2c 0

c 9shy

~ w I c -

W 0

- ctl 0a iri ori Q) ) 0)

u

L()

w ltD w shy

L() w shy

~ w shy

[18~ ] 1 +UO+SUO~ 6uldno)

w0shy0 0 shy

142

References

1) The art of electronics Paul Horowitz Winfield Hill Cambridge Univ Press TK7815H67 1980

2) Talk given at Brookhaven National Lab by Richard E Slusher ATampT Bell Labs NJ

3) J Butterworth et al J Sci Instrum 44 1029 (1967)

4) Gary Horlick Anal Chem 47 352 (1975)

5) This experiment was first proposed by L Maiani R Petronzio and E Zavattini Phys Lett 175B 359 (1986)

bull

143

Chapter 5 Cotton-Mouton Coefficients of the Noble Gases

Phenomenologically the origin of the cotton-Mouton

effect is similar to the QED vacuum polarization where the

role of the e+e- pairs is played by the electrons in gas

atoms In the latter case the electric field of the light

induces a polarization of the atoms

We can measure the Cotton-Mouton constant of different

gases by filling up the cavity with the particular gas while

including the QWP in our system When a polarized light is

traveling in a direction orthogonal to the external magnetic

field it is gaining an ellipticity1 equal to

lJ = nCM sin (28) f dx(i3 x k)2 (51 )

where CM is the Cotton-Mouton constant of the gas at given

pressure and temperature B is the external magnetic field

vector k is the unit vector along the light propagation

and e is the angle between the light polarization (ie the

electric field of the photon) and Bxk For our apparatus

we can safely assume the ellipticity as

lJ = nCM sin (28)B 2 IN (52)

where N is equal to the number of bounces and 1 is the

magnetic field length The eM constant is defined as

(53)

144

where np and no are the parallel and orthogonal indices of

refraction respectively B is the magnetic field and A the

light wavelength

We have measured the Cotton-Mouton constant of N21 He

Ar and Ne The CM constant of N2 and Ar had been previously

measured the former rather accurately23 and the latter

with less precision Therefore we checked our system

against any systematics by measuring the CM constant of N2

improved the measurement of Ar and measured the constants

for He and Ne for the first time

The gas pressure was monitored by a WALLACE amp TIERNAN

manometer mounted at the center of the cavity The manometer

had an offset of 12 mmHg and the precision of the readout

was 05 Torr Prior to every measurement we pumped down

to 4 x 10-4 Torr filled up with gas and pumped down a

second time again to 4 x 10-4 Torr (cavity pressure) he

purpose of this was to prevent any gas that was previously

trapped on the box walls to be flushed out by the incoming

gas and thus contaminating our system This is important

because the Cotton-Mouton constant of 02 is 10 times larger

than that of N2 whose constant is in turn more than 103

greater than that of He The cavity was opened at the

beginning of the run so that the sublimators were saturated

not absorbing any more gas All the pumps (including the

ion pumps) were off during the measurements and constant

pressure was maintained throughout

145

The temperature was monitored at the two cavity ends and

at the cavity center There was consistently a temperature

gradient the cavity end close to the enclosure being warmer

than the end near the tunnel exit

The number of reflections was kept small (between 30 and

40 the light spots formed one ellipse on each mirro) so

that the light dispersion due to the gas was slight The

Laser Power Controller was always in constant transmission

mode (ie effectively off) because the signals were well

above the noise a Hanning wiridow (the data is multiplied

in the time domain by a sine wave with a period equal to

the time record this way the frequency width of the

misalignment peak at W F is kept small) was used in the FFT

The magnetic field period was 128 sec with modulation from

1250 ~ 1600 Amps and ramp rate of 115 As both ways The

magnet current modulation is shown in figure 51 The

corresponding magnetic field (using figures 26 and 27

where we find the transfer function to be TF = 1563 gaussAmp

with an accuracy of 1) is BI = 1250 Amp x 1563 gaussAmp

~ BI = 1954 Tesla and Bh = 25 Tesla Therefore BDC = 22

Tesla and B~ = 0274 Tesla In order to find BO of the

first harmonic we used a diagnostics voltage from the

electronics driving the magnets This voltage is

proportional to the magnet current with calibration of 2

mVAmp Fourier analyzing (fig 52) the voltage we find

BO = 199 Amps x 1563 gaussAmp = 031 T that is

146

--

------

-----------------

---~-----

------ ---- shy-~------

en -CD N en - lcr Q)

C 0gt Q) c ta-Q)

E 1=

bull

=I--

Q) - c-c Q)0) ta l ~ o

147

N J E CD -ta-en c E laquo 0 0 ltD -0 If)

c E laquo 0 IJ) CI--2 ni 5 C 0 E E c 0) ta ~

Lri Q) l u 0)

-----J--shy

bull i----middotf --1----1

middoti It R-I--II--middotH illmiddot ----+---1------1

-B3 Lshy__--shy__-

REAL-TIME AVG COMPLETE Sr-c TIl-lt

A Mar-ke XI 7B 125 mHz V -11 026 dBVr-ma I -~-- -- ------------------------shy

-3 dBVr-m_

-----I------r-----~-----~---~LcgMag 10 dB

dlv -----+1----+----shy

~ 0)

Start 0 Hz Stopa 15625 Hz Spcactrum Chan 1 RMS5

Figure 52 Fourier analysis of the magnet current

o -------------------shy

-20

40

gt -60-shym u

-80 b It)

100

-120 200 -120- 110 40 120 200

CHI~JN EI Center frequency 260 Hz Full BW 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic ~22 dBV

Figure 53 Typical N2 data at 147 Torr

bull

- 40 - -----shy

60

-80 -shy

gtCD 0 -100

VI o -120

-140 I

-200 -120 -40 40 120 200

CHANNEL Center frequency 260 Hz Full 8W 156 Hz Magnet frequency 78 mHz Amplitude of first harmonic -85 d8V

Figure 54 Vacuum ellipticity drih

BO = 1135 XB~ Therefore 2BOBDC = 136 T2 Figure 53

shows typical N2 data whereas figure 54 shows data without

gas The peaks which appear without gas are attributed to

the motion of the magnet which moves the light ray to a

different spot on the QWP Figures 55 56 57 58 show

the ellipticity versus N2 Ar Ne and He pressure

respectively From the above figures it can be concluded

that the induced ellipticity is proportional to the gas

pressure We used this assumption to derive the Cotton-Mouton

(CM) constants because the slope of the lines (ellipticity

vs pressure) is proportional to the CM constants per Torr

with a proportionality factor given from equation 52

nSin(2B)S2lN = (135007) x 1013 gauss2-cm with N = 36

reflections B2 = 136 T2 and B = 45deg10deg Using this

method the CM constant becomes independent of the absolute

measured ellipticity (in case the Faraday cell is not

calibrated correctly) The two methods one using the slope

of the above graphs and the other using the absolute

calibration of the Faraday cell give the same values for

the CM constants within 5 for all the gases We therefore

conclude the Faraday cell calibration to be correct within

5 Table 51 shows the CM constants for the various gases

at 760 Torr and temperature of 255degC

In the case of Ar Ne He (monoatomic gases) a two-level

atom model is utilized4 in which the Cotton-Mouton constant

151

L()

G

1

agt~-1 l

q- () ()

e0 c c Q)r-- 0)

IshyIshy 0 0 -Z

L-l ui

gt

~

amp n 0

t

0 (f) Qen (J) Qi

Ishy 0 Q) ()

L l~~ ~J 0

- - E

Lri Lri

C-J Q) 0 0)= u

bull

shy

f-shyI

lJ1 qshy

f-shyI

W n

f-shyi

W C-J

f--O I

W

[pOJ] 4 I J14 dl ll-lbullbull bull J

l52

I I

j

1

shy

shy

o

Ci Al)qd3L

153

0 0 ~

0 f-shy i

shy

as I

0 cent -shy

--- I-lshy

0 i-

VJ VJ Q) e laquo

L-J ~ (lj ~ 0

-shyIshy

J U

u ~ (f) Q CD

0 rmiddot L

ID 0

0 Ishy

lt I -g

OJ cO ampri Q) I enu

0 11)

o N

I r

-

~ l-- -- 0

1 N I

f--o---

0

fshy

0 ~

--7-4

0

- j I 0

1 CO

0 L()

0 N

fshy fshy fshy ro CO I I I I

w w w w w 0 ill N 0 0 N -shy -shy ro ~

r ~

LpOJ J j(~J~d113

---- L L r-

f-L-J

C

~ fJ

(I)

Q) shy

r-L

Q) 7

e Ugt Ugt (l) ~

0 (l)

Z ui

2 -gt gtshy

poundshy(l)

0 (l) () 0 E 0 Q) 2gt u

154

16E-7

j r- U j(1 L

L-J L12E 7 gt

J

()

-1-

O

-W J

I- 80E 8 ~

111 111

40E - 8 I L __~_-L--L- ~_-L-_----l_--------

120 160 200 240 280 370 360 400 440

11 0 Fres-ure [T (lrr]

Figure 58 Induced ellipticity vs He pressurP

1

~

bull

is related to the linear dimensions of the excited states

of the atom

40nmc 2 5CAQ----+-- (54 )a(n-l) 2n

where Wo is the energy difference betvleen the ground state

10 14and the first exciter stat euro I W= 2nv whEre v is 583 x

Hz for the green light (see equation 21) me is the electron

mass Ae (-hmec) is 2n times the electron Compton wavelength

rl is the radius of the excited state c the speed of light

in vacuum a the fine structure constant and n is the index

of refraction of the gas Table 52 delineates the parameters

for the three gases and table 53 presents their

Cotton-Mouton constants the corresponding ~ltrTgt and the

atomic sizes of the closest atoms

bull

156

Table 51

Cotton-Mouton constants

Gas SlopeX1010 [radTorr] CMX1020 at 255degC and 760 Torr [1gauss2-cm]

N2 7550plusmn100 -4300plusmn100

Ar lllplusmn2 62plusmn2

Ne 97==004 54 01

He 34plusmn01 19plusmn01

CM = Slope(nlB 2 N) = slope x 737Xlo-14gauss2-cm at T

= 255degC assuming e = 45deg The assigned error above is

the statistical one the CM values could be as much as 5

higher due to the uncertainty in the polarization angle 9

which introduces a biased (systematic) error The signs of

the CM-constants are derived from the phase of the vector

signal and are relative to that of N2 whose sign is assumed

from references 2 and 3

bull

157

Table 52

Gas Wo

[cIs]

W (green)

[cIs]

n - 1

Ar 1 46x10 16 367x1015 296x10-5

Ne 221x1016 367X1015 685X10- 5

He 21xI0 16 367xI015 357xlO- 5

Table 53

Gas

Cotton-Mouton

Constant

[1gauss2-cm]

Jlt rf gt

[A]

Alkali

Atom

Ref 5

ro[A]

Ar 62x10-19 240 K 238

Ne 54x10-2O 187 Na 190

He 19x10-2O 185 Li 1 55

The reported values are at 255degC for the CM-constants and

Jltr~gt and 17degC for the rest

158

References

1) F Scuri et al J Chern Phys 85 1789 (1986)

2) A D Buckingham Tras Faraday Soc 63 1057 (1967)

3) S Carusotto et al Jour opt Soc Am Bl 635 (1984)

4) S Carusotto et al CERN-EP83-181 (1983)

5) American Institute of Physics Handbook McGraw Hill NY 3rd ed (1972) bull

bull

159

Chapter 6 Conclusions

bull

bull

In the above pages I have described an experiment

performed at Brookhaven Laboratory where we have searched

for the effect of a magnetic field on the propagation of

light in vacuum In view of the success of QED in other

physical processes in particular the Lamb shift and g-2 of

the electron and muon one expects that the vacuum will be

birefringent due to Delbruck scattering The present level

of sensitivity does not allow us to observe this effect as

yet

On the other hand our experiment has set a limit on the

coupling of light scalar or pseudoscalar particles to two

photons of

1 1 gayylt M = 4x lOsCeV

This can be translated to a quark coupling

1 1 lO-4 C V-Igaqq--- - e fa Nag ayy

with N the number of flavors This limit is an improvement

of two orders of magnitude as compared to laboratory

experiments searching for such particles More stringent

limits exist from astrophysical observations and in

particular the evolution of the sun Light pseudoscalars

160

coupling to two photons would be produced in the solar

interior but because of their weak coupl ing they could

escape thus increasing the cooling rate beyond their

observed values The limit on gavv from the sun is M ~ 108

GeV which will be also reached by the present apparatus in

the near future

We have also measured for the first time the Cotton-Mouton

coefficient for the noble gases Neon and Helium which was

previously unknown These results fit reasonably well within

existing calculations

The technical aspects of this work involved the

integration of two massive superconducting magnets with a

highly sensitive optical system We have demonstrated that

this is feasible and established the present limits of

sensitivity These arise from the coherent motion of the

cavity mirrors which induces a rotation of order

e 0 -6x 10-9 rad

for 33 reflections in the cavity The observed rotation is

proportional to the number of reflections The noise level bull ignoring the induced signal is at

e-6X 10- 10 rad

which corresponds to a sensitivity of

Es-2x 10-8rad~Hz

The analysis of the data is carried out in the frequency

domain the basic modulation frequency of the magnetic field

being -78 mHz and limited by eddy current heating Thus

161

bull

lf and phase noise become the dominant contributions

Nevertheless we were able to reach within a factor of five

of the limit imposed by the shot noise of the laser

By eliminating the spurious noise introduced by the

cavity mirror motion and by increasing the dynamic range of

our analyzer we should be a~le to rea=h a sensitivity of

Es-IO-9rad~Hz

Thus a measurement of 106s (approximately 20 days of data

taking) should allow the observation of the QED Delbruck

scattering with a signal to noise ratio of approximately

three

The present apparatus is l1mited to a full field of 5T

at a modulation rate of OlTsec However new magnets for

accelerator applications are being constructed with peak

fields of lOT The availability of such magnets in the

future will make possible the measurement of Delbruck

scattering with precision Our sensitive ellipsometer and

the further improvements in the apparatus are an essential

contribution to the achievement of this goal which has eluded

physicists for over half a century

It is also natural to ask whether the noise level can

be improved by increasing laser power Our experience is

that this is not necessarily true because increased power

introduces heating of the optical elements with resultant

noise as well as instabilities We believe that 1 W of

162

light exiting the cavity is optimal An alternate approach

to increasing the sensitivity is to increase the optical

path length with our present delay-line method we feel

that 1000 traversals are a good match to the mirror (and

also magnet aperture and stability) Higher reflectivity

mirrors will be available in the future 1 - R lt 10-4 but

in that case one should examine a Fabry-Perot resonator

This does not require a great aperture and can accommodate

a large finesse but for a 10m spacing its stability is a

serious problem necessitating highly sophisticated feedback

techniques

The future direction we will explore is to replace the

Faraday cell with an electrooptic modulator (Pockel s cell)

This removes any limit on the modulation of the beat frequency

but does not solve the magnet modulation problem It does

however eliminate the need of a Aj4 plate for the measurement

of ellipticity and is expected to yield better noise figures

Combined with increased modulation amplitude and larger

dynamic range in the detector (ADC) we should gain at least

a factor of 10 in sensitivity the principal limit being

the laser shot noise

With regards to elementary particle theories our

experiment has set limits on axion-photon-photon coupling

gayy ~ 25X IO-6 GeV -1

bull

163

by a purely laboratory experiment for ma lt 10-3 eV As

discussed this eliminates a certain class of theories We

also set limits for a neutrino magnetic moment by

substituting the electronpositron pair with the

neutrinoantineutrino pair in the figure 12 and assuming

an ellipticity limit of 2 x 10-8 rad

With m v10eV this corresponds to Ilvlt 10-5 1lB which is not

very stringent as compared to the GUT prediction of

Ilv = 10- 19 118lt

Our limit on gayy is weaker than the solar limit but the

improved apparatus should reach and exceed that limit We

are also not as sensitive as the cosmic axion searches which

establish

but over a very narrow mass range 45 x 10-6 eV ltmalt 16 x

10-5 eV Furthermore the cosmic axion search is based on bull

the existence of an axion halo which provides the closure

of the universe clearly our experiment is free of this

condition and is thus a direct test of models of elementary

particles independent of cosmological assumptions

Several proposals have been made for experiments to

search for light scalars and pseudoscalars in view of the

strong belief that these particles must exist However

164

none of these experiments has been mounted as yet and no

results other than ours have been so far obtained One

possible class of experiments is to produce and detect

axions as compared to our approach where we observe only

the production by its effect on the incident beam Such

experiments require higher sensitivity since they involve

g4 as compared to our g2 Van Bibber et ale proposed the

use of a very high power pulsed laser available at Livermore 1

Buckmuller and Hoogeveen2 propose the use of X-rays from a

synchrotron light source and axions are produced in the

Bragg scattering from a crystal In this case a larger mass

range is available for the search In contrast Vorobyev

et al 3 propose the use of microwaves incident on a ferrite

whereby axions are produced by their coupling to the aligned

spins 4 A limit on axion production by microwaves in a

magnetic field has been given by Rogers et ale 5 We mention

these proposals to indicate the current interest in the

subject and possible future directions

Another approach is to search for axions produced in the

sun 6 In that case the axions would have energies in the

KeV range (solar interior) and one would have to coherently

convert the axions to X-rays which would then be detected

Pointing toward the sun would improve the signal to noise

ratio

Finally we note that our experiment is also sensitive

to light spin-2 particles that couple to two photons as it

165

is the case for the graviton In that case of course M =

MPlanck =1019 GeV indicating a very weak coupling as compared

to the sensitivity of our experiment which can reach only

M = 108 GeV The common feature of all these processes is

that light particles propagate with the same speed as photons

and thus the two fields can remain coherent over long

distances This enhances the production rate and indeed we

have a mixing (or oscillations) between the two eigenstates

of the field one state is that of the incident photons

the other that of the sought after particle such processes

are thought to be occurring in stars with strong magnetic

fields 7

In conclusion our experiment is the first attempt to

exploit the coherent transformation of photons into other

weakly coupled light particles Whether these predicted

particles do actually exiampt or not can only be answered by

further experiments bull

bull

166

References

1) K van Bibber et al Phys Rev Lett 59 759 (1987)

2) W Buchmuller and F Hoogeveen Coherent Production of Light Scalar Particles In Bragg Scattering ITP-UH 989 Univ of Hannover FRG (1989)

3) P V Vorobyev H V Kolokolov and B F Fogel JETP Lett 50 58 (1989)

4) R Barbieri et al Phys Lett B226 357 (1989)

5) J Rogers et al Appl Phys Lett 52 2266 (1988)

6) K van Bibber et al Phys Rev D in press (1989)

7) S L Adler Ann of Phys (NY) 87 599 (1971)

bull

167

Index

Accelerator 7 8 34 162 Amplifier 94 97 Analysis 7 113 123 Argon 2 5 45 145

151 157 Astrophysical 8 160 Attenuation 25 64 89 Axial symmetry 18 Axion 8 16 19 23 25 89 123 135 163 165 Axion mass 19 29 135 Bandwidth 99 117 124 Berrys phase 9 Birefringence 13 56 160 BK7 glass 59 Bounces 45 63 144 Branching ratio 19 Brewster 48 Calibration 62 131 51 Capacitance 94 96 cavity 40 55 63 78 87 106 124 CBA 7 34 40 Chiral 32 Coherence 26 Coil 59 102 Compton 156 Cosmic I 21 164 Cotton-Mouton 5 136 144 158 Coupling constant 18

23 29 32 123 135 137 CP symmetry 16 CPT 17 critical fields 11 Cryogenics 34 Curvature 48 55 63 DCF 120 Delbruck 5 16 160 Electric 10 11 14 144 Electric dipole moment 17 Electric displacement 11 Electronics 93 94

Electrooptics 106 163 Electroweak scale 19 Ellipticity 14 87 135 141 151 Euler-Heisenberg 11 Extinction 55 12 136 137 Extraordinary ray 56 107 Fabry-Perot 163 Faraday cell 59 94 Faraday rotation 40 4587115133 Feedback 48 97 103 Feedthroughs 62 Filter 62 97 Fourier 4 93 146 Frequency 4 11 40 87 104 137 141 Gauge 16 54 133 Goldstone boson 19 Green light 14 48 156 Ground loops 97 104 124 Helium 5 7 145 151 157 Higgs 18 21 HWP 55 58 86 Index of refraction 11 53 145 Interferential 63 68 Isotropic 10 Jitter 94 Jones matrices 84 86 89 Jones vectors 84 85 Lagrangian 11 16 23 Lamb shift 160 Laser 45 48 78 104 165 Lifetime 19 21 Magnetic field 11 23 26 34 46 53 59 61 123 137 141 146 Magnetic field length 26 29 123 Maxwell 15 Microwaves 165 Mirror 9 26 45 48 63 68 163

168

Misalignment 87 101 122 Mixing 25 166 Muon 15 160 Neon 5 145 151 157 Neutron 17 Neutron stars 15 Nitrogen 5 145 151 157 Optocoupler 104 Ordinary ray 56 107 oscillation length 29 Parity 16 Peccei-Quinn mechanism 18 Phase 4 14 18 25 58 94 103 137 141 Photodiode 94 106 Photons 1 5 8 21 26 160 165 166 Pion 19 Planck mass 166 Pockels cell 163 Polarizer 45 55 86 Power controller 106 129 Power stabilizer 104 Primakoff 21 32 Pseudoscalar 8 23 160 164 Pump 53 QeD 16 18 QED 5 10 15 160 Quantization 120 Quarks 1 Quench current 34 QWP 56 58 86 135 Ray transfer 73 82 Rayleigh 16 Reflections 26 64 124 137 151 Reflectivity 48 63 64 163 Relics 1 Rotation 8 9 26 56 88 102 124 135 137 161 Satellites 93 101 Scalar 1 821 32 160 Sensitivity 2 9 128 161 Silicon 96

Spectral density 104 116 Spurious 9 45 Standard theory 18 Stray magnetic field 40 131 Sublimator 54 145 Sun 9 160 165 Supersymmetric theo~ies

8 Systematic 145 157 Telescope 78 Tensor 23 Tcffison 16 Topological 16 Transfer function 40 129 Triangle anomaly 21 Units 11 14 23 26 100 Vacuum 4 21 48 54 129 138 141 Vacuum polarization 10 11 Verdet constant 45 61 Virtual 2 5 10 23 Wavelength 14 51 61 X-rays 15 165

Noise 11f 106 107122 161 amplitude 104 107 115 flicker 122 shot 104 107 114 136 bull

169

Page 14: Coherent Production of Light Pseudoscalars (Axions)
Page 15: Coherent Production of Light Pseudoscalars (Axions)
Page 16: Coherent Production of Light Pseudoscalars (Axions)
Page 17: Coherent Production of Light Pseudoscalars (Axions)
Page 18: Coherent Production of Light Pseudoscalars (Axions)
Page 19: Coherent Production of Light Pseudoscalars (Axions)
Page 20: Coherent Production of Light Pseudoscalars (Axions)
Page 21: Coherent Production of Light Pseudoscalars (Axions)
Page 22: Coherent Production of Light Pseudoscalars (Axions)
Page 23: Coherent Production of Light Pseudoscalars (Axions)
Page 24: Coherent Production of Light Pseudoscalars (Axions)
Page 25: Coherent Production of Light Pseudoscalars (Axions)
Page 26: Coherent Production of Light Pseudoscalars (Axions)
Page 27: Coherent Production of Light Pseudoscalars (Axions)
Page 28: Coherent Production of Light Pseudoscalars (Axions)
Page 29: Coherent Production of Light Pseudoscalars (Axions)
Page 30: Coherent Production of Light Pseudoscalars (Axions)
Page 31: Coherent Production of Light Pseudoscalars (Axions)
Page 32: Coherent Production of Light Pseudoscalars (Axions)
Page 33: Coherent Production of Light Pseudoscalars (Axions)
Page 34: Coherent Production of Light Pseudoscalars (Axions)
Page 35: Coherent Production of Light Pseudoscalars (Axions)
Page 36: Coherent Production of Light Pseudoscalars (Axions)
Page 37: Coherent Production of Light Pseudoscalars (Axions)
Page 38: Coherent Production of Light Pseudoscalars (Axions)
Page 39: Coherent Production of Light Pseudoscalars (Axions)
Page 40: Coherent Production of Light Pseudoscalars (Axions)
Page 41: Coherent Production of Light Pseudoscalars (Axions)
Page 42: Coherent Production of Light Pseudoscalars (Axions)
Page 43: Coherent Production of Light Pseudoscalars (Axions)
Page 44: Coherent Production of Light Pseudoscalars (Axions)
Page 45: Coherent Production of Light Pseudoscalars (Axions)
Page 46: Coherent Production of Light Pseudoscalars (Axions)
Page 47: Coherent Production of Light Pseudoscalars (Axions)
Page 48: Coherent Production of Light Pseudoscalars (Axions)
Page 49: Coherent Production of Light Pseudoscalars (Axions)
Page 50: Coherent Production of Light Pseudoscalars (Axions)
Page 51: Coherent Production of Light Pseudoscalars (Axions)
Page 52: Coherent Production of Light Pseudoscalars (Axions)
Page 53: Coherent Production of Light Pseudoscalars (Axions)
Page 54: Coherent Production of Light Pseudoscalars (Axions)
Page 55: Coherent Production of Light Pseudoscalars (Axions)
Page 56: Coherent Production of Light Pseudoscalars (Axions)
Page 57: Coherent Production of Light Pseudoscalars (Axions)
Page 58: Coherent Production of Light Pseudoscalars (Axions)
Page 59: Coherent Production of Light Pseudoscalars (Axions)
Page 60: Coherent Production of Light Pseudoscalars (Axions)
Page 61: Coherent Production of Light Pseudoscalars (Axions)
Page 62: Coherent Production of Light Pseudoscalars (Axions)
Page 63: Coherent Production of Light Pseudoscalars (Axions)
Page 64: Coherent Production of Light Pseudoscalars (Axions)
Page 65: Coherent Production of Light Pseudoscalars (Axions)
Page 66: Coherent Production of Light Pseudoscalars (Axions)
Page 67: Coherent Production of Light Pseudoscalars (Axions)
Page 68: Coherent Production of Light Pseudoscalars (Axions)
Page 69: Coherent Production of Light Pseudoscalars (Axions)
Page 70: Coherent Production of Light Pseudoscalars (Axions)
Page 71: Coherent Production of Light Pseudoscalars (Axions)
Page 72: Coherent Production of Light Pseudoscalars (Axions)
Page 73: Coherent Production of Light Pseudoscalars (Axions)
Page 74: Coherent Production of Light Pseudoscalars (Axions)
Page 75: Coherent Production of Light Pseudoscalars (Axions)
Page 76: Coherent Production of Light Pseudoscalars (Axions)
Page 77: Coherent Production of Light Pseudoscalars (Axions)
Page 78: Coherent Production of Light Pseudoscalars (Axions)
Page 79: Coherent Production of Light Pseudoscalars (Axions)
Page 80: Coherent Production of Light Pseudoscalars (Axions)
Page 81: Coherent Production of Light Pseudoscalars (Axions)
Page 82: Coherent Production of Light Pseudoscalars (Axions)
Page 83: Coherent Production of Light Pseudoscalars (Axions)
Page 84: Coherent Production of Light Pseudoscalars (Axions)
Page 85: Coherent Production of Light Pseudoscalars (Axions)
Page 86: Coherent Production of Light Pseudoscalars (Axions)
Page 87: Coherent Production of Light Pseudoscalars (Axions)
Page 88: Coherent Production of Light Pseudoscalars (Axions)
Page 89: Coherent Production of Light Pseudoscalars (Axions)
Page 90: Coherent Production of Light Pseudoscalars (Axions)
Page 91: Coherent Production of Light Pseudoscalars (Axions)
Page 92: Coherent Production of Light Pseudoscalars (Axions)
Page 93: Coherent Production of Light Pseudoscalars (Axions)
Page 94: Coherent Production of Light Pseudoscalars (Axions)
Page 95: Coherent Production of Light Pseudoscalars (Axions)
Page 96: Coherent Production of Light Pseudoscalars (Axions)
Page 97: Coherent Production of Light Pseudoscalars (Axions)
Page 98: Coherent Production of Light Pseudoscalars (Axions)
Page 99: Coherent Production of Light Pseudoscalars (Axions)
Page 100: Coherent Production of Light Pseudoscalars (Axions)
Page 101: Coherent Production of Light Pseudoscalars (Axions)
Page 102: Coherent Production of Light Pseudoscalars (Axions)
Page 103: Coherent Production of Light Pseudoscalars (Axions)
Page 104: Coherent Production of Light Pseudoscalars (Axions)
Page 105: Coherent Production of Light Pseudoscalars (Axions)
Page 106: Coherent Production of Light Pseudoscalars (Axions)
Page 107: Coherent Production of Light Pseudoscalars (Axions)
Page 108: Coherent Production of Light Pseudoscalars (Axions)
Page 109: Coherent Production of Light Pseudoscalars (Axions)
Page 110: Coherent Production of Light Pseudoscalars (Axions)
Page 111: Coherent Production of Light Pseudoscalars (Axions)
Page 112: Coherent Production of Light Pseudoscalars (Axions)
Page 113: Coherent Production of Light Pseudoscalars (Axions)
Page 114: Coherent Production of Light Pseudoscalars (Axions)
Page 115: Coherent Production of Light Pseudoscalars (Axions)
Page 116: Coherent Production of Light Pseudoscalars (Axions)
Page 117: Coherent Production of Light Pseudoscalars (Axions)
Page 118: Coherent Production of Light Pseudoscalars (Axions)
Page 119: Coherent Production of Light Pseudoscalars (Axions)
Page 120: Coherent Production of Light Pseudoscalars (Axions)
Page 121: Coherent Production of Light Pseudoscalars (Axions)
Page 122: Coherent Production of Light Pseudoscalars (Axions)
Page 123: Coherent Production of Light Pseudoscalars (Axions)
Page 124: Coherent Production of Light Pseudoscalars (Axions)
Page 125: Coherent Production of Light Pseudoscalars (Axions)
Page 126: Coherent Production of Light Pseudoscalars (Axions)
Page 127: Coherent Production of Light Pseudoscalars (Axions)
Page 128: Coherent Production of Light Pseudoscalars (Axions)
Page 129: Coherent Production of Light Pseudoscalars (Axions)
Page 130: Coherent Production of Light Pseudoscalars (Axions)
Page 131: Coherent Production of Light Pseudoscalars (Axions)
Page 132: Coherent Production of Light Pseudoscalars (Axions)
Page 133: Coherent Production of Light Pseudoscalars (Axions)
Page 134: Coherent Production of Light Pseudoscalars (Axions)
Page 135: Coherent Production of Light Pseudoscalars (Axions)
Page 136: Coherent Production of Light Pseudoscalars (Axions)
Page 137: Coherent Production of Light Pseudoscalars (Axions)
Page 138: Coherent Production of Light Pseudoscalars (Axions)
Page 139: Coherent Production of Light Pseudoscalars (Axions)
Page 140: Coherent Production of Light Pseudoscalars (Axions)
Page 141: Coherent Production of Light Pseudoscalars (Axions)
Page 142: Coherent Production of Light Pseudoscalars (Axions)
Page 143: Coherent Production of Light Pseudoscalars (Axions)
Page 144: Coherent Production of Light Pseudoscalars (Axions)
Page 145: Coherent Production of Light Pseudoscalars (Axions)
Page 146: Coherent Production of Light Pseudoscalars (Axions)
Page 147: Coherent Production of Light Pseudoscalars (Axions)
Page 148: Coherent Production of Light Pseudoscalars (Axions)
Page 149: Coherent Production of Light Pseudoscalars (Axions)
Page 150: Coherent Production of Light Pseudoscalars (Axions)
Page 151: Coherent Production of Light Pseudoscalars (Axions)
Page 152: Coherent Production of Light Pseudoscalars (Axions)
Page 153: Coherent Production of Light Pseudoscalars (Axions)
Page 154: Coherent Production of Light Pseudoscalars (Axions)
Page 155: Coherent Production of Light Pseudoscalars (Axions)
Page 156: Coherent Production of Light Pseudoscalars (Axions)
Page 157: Coherent Production of Light Pseudoscalars (Axions)
Page 158: Coherent Production of Light Pseudoscalars (Axions)
Page 159: Coherent Production of Light Pseudoscalars (Axions)
Page 160: Coherent Production of Light Pseudoscalars (Axions)
Page 161: Coherent Production of Light Pseudoscalars (Axions)
Page 162: Coherent Production of Light Pseudoscalars (Axions)
Page 163: Coherent Production of Light Pseudoscalars (Axions)
Page 164: Coherent Production of Light Pseudoscalars (Axions)
Page 165: Coherent Production of Light Pseudoscalars (Axions)
Page 166: Coherent Production of Light Pseudoscalars (Axions)
Page 167: Coherent Production of Light Pseudoscalars (Axions)
Page 168: Coherent Production of Light Pseudoscalars (Axions)
Page 169: Coherent Production of Light Pseudoscalars (Axions)
Page 170: Coherent Production of Light Pseudoscalars (Axions)
Page 171: Coherent Production of Light Pseudoscalars (Axions)
Page 172: Coherent Production of Light Pseudoscalars (Axions)
Page 173: Coherent Production of Light Pseudoscalars (Axions)
Page 174: Coherent Production of Light Pseudoscalars (Axions)
Page 175: Coherent Production of Light Pseudoscalars (Axions)
Page 176: Coherent Production of Light Pseudoscalars (Axions)
Page 177: Coherent Production of Light Pseudoscalars (Axions)
Page 178: Coherent Production of Light Pseudoscalars (Axions)