First Results from LIGO Keith Riles University of Michigan High Energy Physics Seminar Michigan State University – Feb. 24, 2004

First Results from LIGO

Keith Riles University of Michigan

High Energy Physics Seminar

Michigan State University – Feb. 24, 2004

K. Riles - First Results from LIGO -2/24/04

2

Outline

Nature & Generation of Gravitational Waves

Interferometric Gravitational Wave Detection

The Initial LIGO Detector

The LIGO Scientific Collaboration

Science Run S1: Sensitivity & Data Quality

First analysis results and future prospects

Preparing for Advanced LIGO


3

Gravitational Waves = “Ripples in space-time”

Perturbation propagation similar to light (obeys same wave equation!) Propagation speed = c Two transverse polarizations - quadrupolar: + and x

Amplitude parameterized by (tiny) dimensionless strain h: L ~ h(t) x L

Nature of Gravitational Waves

Example:

Ring of test masses

responding to wave

propagating along z


4

Why look for Gravitational Radiation?

Because it’s there! (presumably)

Test General Relativity: Quadrupolar radiation? Travels at speed of light? Unique probe of strong-field gravity

Gain different view of Universe: Sources cannot be obscured by dust Detectable sources some of the most interesting,

least understood in the Universe Opens up entirely new non-electromagnetic spectrum


5

What will the sky look like?


6

Generation of Gravitational Waves

Radiation generated by quadrupolar mass movements:

(with I = quadrupole tensor, r = source distance)

Example: Pair of 1.4 Msolar neutron stars in circular orbit of radius 20 km (imminent coalescence) at orbital frequency 400 Hz gives 800 Hz radiation of amplitude:


7


Major expected sources in 10-1000 Hz band:

Coalescences of binary compact star systems

(NS-NS, NS-BH, BH-BH)

Supernovae

(requires asymmetry in explosion)

Spinning neutron stars, e.g., pulsars

(requires axial asymmetry or wobbling spin axis)

Also expected (but probably exceedingly weak):

Stochastic background – Big Bang remnant


8


Sources well below LIGO bandwidth: Binaries well before coalescence Inspiral of stars into massive black holes Coalescence of massive black holes Stochastic background (e.g, big bang remnant, superposition of binaries)

Irreducible seismic noise argues for space-based system at low frequencies - LISA

Three satellites forming interferometers with 5 x 106 km baselines (launch 2012?)


9


Strong indirect evidence for GW generation:

Taylor-Hulse Pulsar System (PSR1913+16)

Two neutron stars (one=pulsar)

in elliptical 8-hour orbit

Measured periastron advance

quadratic in time in agreement with

absolute GR prediction

Orbital decay due to energy loss


10


Can we detect this radiation directly?

NO - freq too low

Must wait ~300 My for characteristic “chirp”:


11


Coalescence rate estimates based on two methods: Use known NS/NS binaries in our galaxy (two!) A priori calculation from stellar and binary system evolution

Large uncertainties!

For initial LIGO design “seeing distance” (~20 Mpc):

Expect 1/(3000 y) to 1/(4 y)

Will need Advanced LIGO to ensure detection


12


Most promising periodic source: Rotating Neutron Stars (e.g., pulsar)

But axisymmetric object rotating about symmetry axis

Generates NO radiation

Radiation requires an asymmetry or perturbation:

Equatorial ellipticity (e.g., – mm-high “mountain”): h α εequat

Poloidal ellipticity (natural) + wobble angle: h α εpol x Θwobble

(precession due to different L and Ω axes)

Notation henceforth: ε = ( εequat or εpol x θwobble )


13

Periodic Sources of GW

Serious technical difficulty: Doppler frequency shifts Frequency modulation from earth’s rotation (v/c ~ 10-6) Frequency modulation from earth’s orbital motion (v/c ~ 10-4)

Additional, related complications: Daily amplitude modulation of antenna pattern Spin-down of source Orbital motion of sources in binary systems

Modulations / drifts complicate analysis enormously: Simple Fourier transform inadequate Every sky direction requires different demodulation

All-sky survey at full sensitivity = Formidable challenge


14

Periodic Sources of GW

But two substantial benefits from modulations: Reality of signal confirmed by need for corrections Corrections give precise direction of source

Difficult to detect neutron stars!

But search is nonetheless attractive & intriguing:

Unknown number of electromagnetically quiet, undiscovered

neutron stars in our galactic neighborhood

Realistic values for ε unknown

A nearby source could be buried in the data, waiting for just the

right algorithm to tease it into view


15

Gravitational Wave Detection

Suspended Interferometers (IFO’s)

Suspended mirrors in “free-fall”

Broad-band response

(~50 Hz to few kHz)

Waveform information

(e.g., chirp reconstruction)

Michelson IFO is

“natural” GW detector


16


Major Interferometers coming on line world-wide

LIGO (NSF-$300M)

Livingston, Louisiana & Hanford, Washington

2 x 4000-m

1 x 2000-m

Advanced

Commissioning

VIRGO

Near Pisa, Italy 1 x 3000-mEarly

Commissioning

GEO

Near Hannover, Germany 1 x 600-mAdvanced

Commissioning

TAMA

Tokyo, Japan 1 x 300-mAdvanced

Commissioning


17

LIGO Interferometer Optical Scheme

end test mass

LASER/MC

6W

recyclingmirror

•Recycling mirror matches losses, enhances effective power by ~ 50x

150 W

20000 W(~0.5W)

Michelson interferometer

4 km Fabry-Perot cavity

With Fabry-Perot arm cavities


18

“Locking” the Inteferometer

Sensing gravitational waves requires sustained resonance in the Fabry-Perot arms and in the recycling cavity

Need to maintain half-integer # of laser wavelengths between mirrors

Feedback control servo uses error signals from imposed RF sidebands

Four primary coupled degrees of freedom to control

Highly non-linear system with 5-6 orders of magnitude in light intensity

Also need to control mirror rotation (“pitch” & “yaw”)

Ten more DOF’s (but less coupled)

And need to stabilize laser (intensity & frequency), keep the beam pointed, damp out seismic noise, correct for tides, etc.,…


19


Initial LIGO Design Sensitivity

Dominant noise sources:

•Seismic below 50 Hz

•Suspensions in 50-150 Hz

•Shot noise above 150 Hz

Best design sensitivity:

~3 x 10-23 Hz-1/2 @ 150 Hz


20

LIGO Observatories

Livingston

Hanford

Observation of nearly simultaneous signals 3000 km apart rules out terrestrial artifacts


21

LIGO Detector Facilities

Vacuum System

•Stainless-steel tubes

(1.24 m diameter, ~10-8 torr)

•Gate valves for optics isolation

•Protected by concrete enclosure


22


LASER Infrared (1064 nm, 10-W) Nd-YAG laser from Lightwave (now commercial product!) Elaborate intensity & frequency stabilization system, including feedback from main

interferometer

Optics Fused silica (high-Q, low-absorption, 1 nm surface rms, 25-cm diameter) Suspended by single steel wire Actuation of alignment / position via magnets & coils


23


Seismic Isolation Multi-stage (mass & springs) optical table support gives 106 suppression Pendulum suspension gives additional 1 / f 2 suppression above ~1 Hz

102

100

10-2

10-4

10-6

10-8

10-10

Horizontal

Vertical

10-6


24

Livingston Problem -- Logging

Livingston Observatory located in pine forest popular with pulp wood cutters

Spiky noise (e.g. falling trees) in 1-3 Hz band creates dynamic range problem for arm cavity control

Solution:Retrofit with active feed-forward isolation system (using technology developed for Advanced LIGO)

Work started January 2004


25

Correlation between seismic noise and lock livetime (4-day weekend)


26

Sampling of Milestones

December 1999 First light in Hanford 2-km cavity & brief arm lock

April 2000 Engineering run “E1” (1 day, one 2-km arm)

October 2000 Hanford 2-km “First-Lock” (whole kit & caboodle)

March 2001 Engineering run “E3” (3 days, one Livingston arm)

January 2002 Eng. run “E7” (17 days, all IFO’s + GEO + ALLEGRO)

September 2002 Science run “S1” (17 days, all IFO’s + GEO + TAMA)

Results reported here

Feb. – April 2003 Science run “S2” (59 days, all IFO’s + TAMA)

October 2003 Eng. run “E10” (7 days, all IFO’s)

Nov 2003 – Jan 2004 Science run “S3” (66 days, all IFO’s + GEO + TAMA)

Just ended!


27

The LIGO Scientific Collaboration (LSC)

LSC formed to

• Determine the scientific needs of the project and set R&D priorities

• Present the scientific case for the program

• Carry out the scientific and technical research program

• Carry out the data analysis and validate the scientific results

• Maximize scientific returns in the operations of LIGO Laboratory facilities

• Determine the relative distribution of observing and development time

• Establish the long term needs of the field & set priorities for improvements

Modeled in part on the successful high energy physics arrangement between national facilities and user group collaborations:

LIGO Laboratory = Caltech + MIT + Hanford + Livingston


28

LSC Membership (~400 scientists)

University of Adelaide ACIGA

Australian National University ACIGA

Balearic Islands University

California State Dominquez Hills

Caltech LIGO

Caltech Experimental Gravitation CEGG

Caltech Theory CART

University of Cardiff GEO

Carleton College

Cornell University

Fermi National Laboratory

University of Florida @ Gainesville

Glasgow University GEO

NASA-Goddard Spaceflight Center

University of Hannover GEO

Hobart – Williams University

India-IUCAA

IAP Nizhny Novgorod

IUCCA India

Iowa State University

Joint Institute of Laboratory Astrophysics

Salish Kootenai College

LIGO Livingston Observatory (LLO)LIGO Hanford Observatory (LLO)

Loyola New Orleans

Louisiana State University

Louisiana Tech University

MIT LIGO

Max Planck (Garching) GEO

Max Planck (Potsdam) GEO

University of Michigan

Moscow State University

NAOJ - TAMA

Northwestern University

University of Oregon

Pennsylvania State University

Southeastern Louisiana University

Southern University

Stanford University

Syracuse University

University of Texas@Brownsville

Washington State University@ Pullman

University of Western Australia ACIGA

University of Wisconsin@Milwaukee


29

Michigan LIGO Group MembersOld fogeys:

Dick Gustafson, Keith Riles

Graduate students:

Dave Chin, Vladimir Dergachev

Research assistant:

Jennifer Lindahl

Undergraduates:

Tim Bodiya, John Lucido, Jake Slutsky, Phil Szepietowski

Former undergraduates:

Jamie Rollins (now at MIT)

Justin Dombrowski (now at seminary)

Joseph Marsano (now at Harvard)

Michael La Marca (now at Arizona State)


30

Michigan Group – Main Efforts

Initial work -- R&D at Caltech 40-Meter Prototype Interferometer

Demonstration of “recycling” with suspended interferometer

Last major R&D milestone for LIGO project

Apprenticeship for Gustafson and Riles

Detector Characterization (leadership, instrumentation, software)

Riles, Gustafson, Chin, Dergachev, Bodiya, Lucido

Commissioning & Noise Reduction

Gustafson (in residence at Hanford), Chin

Control System Development

Gustafson, Lindahl, Szepietowski, Riles

Search for Periodic Sources (rotating neutron stars)

Chin, Dergachev, Riles, Gustafson, Slutsky


31

Data Analysis

Four “upper limits groups” organized: (to become “search groups”)

Inspiraling binary systems

Unmodelled bursts

Periodic sources (e.g. pulsars)

Stochastic background (e.g., big-bang remnant)

First steps:

Limits not yet astrophysically interesting

Understanding detector artifacts

Defining analysis algorithms and strategy (e.g., acceptable single-IFO signal rate prior to coincidence searching)


32

S1 Instrumental Sensitivities

LIGO S1 Run----------“First

Upper Limit Run”

23 Aug–9 Sept 200217 daysAll IFOs in power recycling config LLO 4Km

LHO 2Km

LHO 4Km

GEO in S1 RUN----------

Ran simultaneouslyIn power recycling Lesser sensitivity


33

Science Run S1 Data Analysis

As expected, S1 data much improved over earlier engineering run data in

• Sensitivity

• Gaussianity

• Stationarity

But significant residual artifacts remained and new problems emerged as the noise floor fell

Artifacts substantially complicated the data analysis


34

S1 Data Quality(three hours of Livingston 4-km)

Histograms vs time of filtered GW channel

CLEAN DATA

RAGGED DATA


35

S1 Data Quality – Inspiral Range (kpc)(“seeing distance” for inspiraling binary system)

Calibration not as stable as we wanted: (one day of S1 Hanford 4-km running)


36

S1 Data Quality

Despite commissioning improvements, occasional artifacts appeared:

Hanford 4-km “heartbeat”

Tracked down to undamped “side motion” of one end mirror kicked up by Oregon earthquake


37

Science Run S1 Data Analysis

Will summarize here results from searches for four types of gravitational wave sources:

• Inspiral coalescence of binary neutron star system

• Periodic source (pulsar)

• Stochastic gravitational wave background

• Bursts


38

S1 Results

Search for binary inspiral coalescence of two neutron stars

Reporting limit on inspiral rate / milky-way equivalent galaxy (per year)

Search method relies upon matched filtering (frequency domain) of bank of template waveforms against the data

Searching over grid of neutron star mass pairs with M1 < M2 < 3 MSUN

Efficiency depends on distance to binary system

Define “inspiral candle” range to be distance at which binary system of 2 × 1.4MSUN neutron stars can be detected with SNR=8 in optimal orientation


39

S1 RangesL1: 110 kpc < D < 210 kpc

H1: 40 kpc < D < 75 kpc

H2: 38 kpc < D < 70 kpc

Visible: Milky Way

LMC, SMC


40

Inspiral Results: [gr-qc/0308069 – submitted to PRD]

Use all candidates from Hanford and Livingston: 236 hours Cannot accurately assess background (be conservative, assume zero) Use maximum signal-to-noise ratio statistic to establish the rate limit. Monte Carlo simulation efficiency = 0.51 90% confidence limit = 2.3 / (efficiency * time) Express the rate as a rate per Milky-Way Equivalent Galaxies (MWEG)

R < 2.3 / (0.50 x 236 hr) = 1.7 x 102 / yr / (MWEG)

Previous observational limits Japanese TAMA R < 30,000 / yr / MWEG Caltech 40m R < 4,000 / yr / MWEG

Theoretical prediction R < 2 x 10-5 / yr / MWEG


41

Loudest Surviving Event Candidate

Not NS/NS inspiral event 1 Sep 2002, 00:38:33 UTC S/N = 15.9, 2/dof = 2.2 (m1,m2) = (1.3, 1.1) Msun

SNR vs time:

χ2 vs time:

Study of auxiliary channels reveals:

Photodiode Saturation!


42

S1 Results (cont.)

Search for periodic sources of gravitational radiation

Chose to focus on single (fastest) pulsar:

J1939+2134 fGW = 1.28 kHz

Two independent searches carried out:

Time Domain (heterodyne with modulation correction)

Frequency Domain (FFT stitching with correction)

Doppler modulations and pulsar spin-down explicitly taken into account in these coherent targeted analyses


43

Graphic by R. Dupuis, Glasgow

Detectable amplitudes with a 1% false alarm rate and 10% false dismissal rate by the IFOs during S1 (colored curves) and at design sensitivities (black curves)

Upper limits on <ho> from spin-down measurements of known radio pulsars (filled circles)

<ho>= 11.4 [Sh(fo)/T]1/2

Expectations for pulsar strain amplitude sensitivity

S1: NO DETECTIONEXPECTED

PSR J1939+2134P = 0.00155781 sfGW = 1283.86 Hz

P = 1.0511 10-19 s/sD = 3.6 kpc

.

Crab pulsar


44

Pulsar Results: [ gr-qc/0308050 –to appear in PRD]

No evidence of continuous wave emission from PSR J1939+2134 – As expected

Summary of 95% upper limits on h:

• ho < 1.4 x 10-22 constrains ellipticity < 3 x 10-4 (M=1.4Msun, r=10km, R=3.6kpc)

• Previous best result for PSR J1939+2134: ho < 10-20 (Glasgow, Hough et al., 1983)

IFO Frequentist FDS Bayesian TDS

GEO (1.9 0.1) x 10-21 (2.2 0.1) x 10-21

LLO (2.7 0.3) x 10-22 (1.4 0.1) x 10-22

LHO-2K (5.4 0.6) x 10-22 (3.3 0.3) x 10-22

LHO-4K (4.0 0.5) x 10-22 (2.4 0.2) x 10-22


45

S1 Results (cont.)

Search for stochastic background of gravitational waves

Sources: early universe, many weak unresolved sources emitting gravitational waves independently Random radiation described by its spectrum (isotropic, unpolarized, stationary and Gaussian)

Analysis goals: constrain contribution of stochastic radiation’s energy GW to the total energy required to close the universe critical :

0

(1/ ) ( ) GWGW

critical

f f df


46

Methods for the Stochastic Search

Optimally filtered cross-correlation of detector pairs: L1-H1, L1-H2 and H1-H2.

o

• Detector separation and orientation reduce correlations at high frequencies (GW > 2xBaseLine): overlap reduction function» H1-H2 best suited

» L1-H1(H2) significant <50Hz

• Achievable sensitivities to by detector pairs in S1 [assumes (f) = 0 ]


47

Results of Stochastic Search [ gr-qc/0312088 – submitted to PRD]

Non-negligible LHO 4km-2km (H1-H2) cross-correlation due to common

acoustic coupling (now much improved!)

51 hrs

64 hrs

Tobs

GW (40Hz - 314 Hz) < 23 5

GW (40Hz - 314 Hz) < 55 11

90% CL Upper Limit

LHO 2km-LLO 4km

LHO 4km-LLO 4km

Interferometer Pair

Previous best upper limits:

Measured: Garching-Glasgow interferometers :

Measured: EXPLORER-NAUTILUS (cryogenic bars):

GW ( f ) 3105

GW (907Hz) 60


48

S1 Results (cont.)

Search for gravitational wave bursts

Sources: Known and unknown phenomena emitting short transients of gravitational radiation of unknown waveform (e.g., supernovae, black hole mergers).

Analysis goals: Broad frequency band search to (a) establish a bound on their rate at the instruments, (b) interpret bound in terms of a source and population model on a rate vs. strength exclusion plot.

Search method: Time-Frequency domain algorithm identifies regions in the time-frequency plane with excess power (threshold on pixel power and cluster size – “tfcluster”).


49

Bursts Search Pipeline

basic assumption: multi-interferometer response consistent with a plane wave-front incident on network of detectors.

design the capability to veto data epochs and events based on quality criteria and auxiliary channels.

essential: use temporal coincidence of the 3 interferometer’s ‘best candidates’

correlate frequency features of candidates (time-frequency domain analysis).

Search code generated events

Epoch veto’ed

Can be veto’ed by auxiliary channels


50

Upper Limit on Rate of Bursts

End result of analysis pipeline: number of triple coincidence events. Use time-shift experiments to establish number of background events. Use Feldman-Cousins to set 90% confidence upper limits on rate of

foreground events:

< 1.6 events/day

Background estimation for TFCLUSTERS in S1

Zero-lag measurement

Poisson fit of timeshifted coincidencesbetween the LIGO sites


51

90% CL Rate limit vs. Strength Plots for a Burst Model

Optimally oriented(per IFO)

Average over direction,Polarization (per IFO)

Burst model:

Sine-Gaussians with varying central frequencies

Burst model:

1ms, 2.5 ms Gaussian impulses

Determine detection efficiency of the end-to-end analysis pipeline via signal injection of various morphologies.

Assume a population of such sources uniformly distributed on a sphere around us: establish upper limit on rate of bursts as a function of their strength.


52

Burst Search Results and the Future [ gr-qc/0312056 –to appear in PRD]

Search and raw results sensitive to a wide variety of waveform morphologies and broad frequency features (as long as signal has significant strain amplitude in LIGO’s frequency band).

90% CL upper limit of 1.6 events/day leads to limits on characteristic strain strength ~ 10-19 -- 10-17 / √Hz.

In the near future: Use multiple-interferometer information on amplitude of putative

signal and correlation statistic of their raw time-series. Improve time-resolution of event trigger generators. Pursue rigorously an externally triggered (by GRB’s, neutrinos)

search for bursts (excercised during S1).


53

Looking Ahead

S1 analysis was a good exercise for operations and data analysis

Learning how to confront instrumental artifacts in astrophysical searches (no more Gaussian noise modelling!)

Excellent preparation for more sensitive future Science Runs

Meanwhile, IFO’s ever improving with further commissioning …


54

Looking Ahead

May 01

Jan 03

Livingston 4-km IFO sensitivity history until S2:

Initial LIGO Design


55

Looking Ahead

Initial LIGO Design

S11st Science Runend Sept. 2002

17 daysS22nd Science Runend Apr. 2003

59 days

S33rd Science Runbegin Nov. 2003

70 days


56

Looking Ahead

The three LIGO interferometers are part of a forming global network.

Multiple signal detections will increase detection confidence and provide better precision on source locations and wave polarizations

LIGO GEO Virgo

TAMA

AIGO (proposed)


57

Looking Way Ahead

Despite their immense technical challenges, the initial LIGO IFO’s were designed conservatively, based on “tabletop” prototypes, but with expected sensitivity gain of ~1000.

Given the expected low rate of detectable GW events, it was always planned that in engineering, building and commissioning initial LIGO, one would learn how reliably to build Advanced LIGO with another factor of ~10 improved sensitivity.

Because LIGO measures GW amplitude, an increase in sensitivity by 10 gives an increase in sampling volume, i.e, rate by ~1000


58

Advanced LIGO

Detector Improvements:

Improved seismic isolation: Passive Active

Lowers seismic “wall” to ~12 Hz

New suspensions: Single Quadruple pendulum

Lower suspensions thermal noise in bandwidth

Increased test mass: 10 kg 30 kg

Compensates increased radiation pressure noise

New test mass material: Fused silica Sapphire

Lower internal thermal noise in bandwidth

Increased laser power: 10 W 180 W

Improved shot noise (high freq)


59

Advanced LIGO

Sampling of source strengths vis a vis Initial LIGO and Advanced LIGO

Lower hrms and wider bandwidth both important

“Signal recycling” offers potential for tuning shape of noise curve to improve sensitivity in target band (e.g., known pulsar range)


60

Advanced LIGO

Ambitious upgrade program:

• MRE proposal submitted to NSF in January

• Technical review went very well, but more layers to go

• Hope to begin detector upgrades in 2007

Begin observing in 2008

First 2-3 hours of Advanced LIGO is equivalent

to a “Snowmass year” of Initial LIGO


61

Summary

Initial LIGO commissioning well underway

Much instrumental noise to beat down, but no show-stoppers have appeared

Engineering runs giving way to sporadic science runs (with astrophysical measurements as primary purpose) interspersed with ongoing commissioning

Confronting dirty-data analysis First astrophysics results

Looking ahead to several years of high-duty-cycle data taking

Looking farther ahead to major detector upgrade with more than 1000-fold increase in event rate

Direct GW detection only a matter of time

Exciting years to come!

Documents

First Results from LIGO Keith Riles University of Michigan High Energy Physics Seminar Michigan State University – Feb. 24, 2004