Components based on optical fibers with internal electrodes

Niklas Myrén

Royal Institute of Technology Department of Physics and Quantum Optics

and

Acreo AB

Stockholm, Sweden, November 2003

Thesis for the Degree of Licentiate of Technology

Royal Institute of Technology Department of Physics and Quantum Optics Stockholm Center for Physics, Astronomy and Biotechnology Roslagstullsbacken 21 SE-106 91 Stockholm, Sweden Acreo AB Electrum 236 SE-164 40 Kista, Sweden Akademisk avhandling som med tillstånd av Kungliga Tekniska Högskolan framlägges till offentlig granskning för avläggande av teknisk licenciatexamen i fysik, fredagen den 19 december 2003. Avhandlingen kommer att försvaras på engelska. TRITA-FYS 2003:60 ISSN 0280-316X ISRN KTH/FYS/--03:60--SE ISBN 91-7283-660-1 Cover: HeNe laser coupled into twinhole fiber with indiffused silver. The silver region

starts in the middle of the picture. Components based on optical fibers with internal electrodes Niklas Myrén, 2003

Abstract The topic of this thesis is development of devices for telecom applications based on poled optical fibers. The focus is on two different specific functions, wavelength conversion and optical switching. Optical switching is demonstrated in a poled optical fiber at telecom wavelengths (~1.55 µm). The fiber has two holes running along the core in which electrodes are inserted. The fiber device is made electro-optically active with a poling process in which a strong electric field is recorded in the fiber at a temperature of 270 o C. The fiber is then put in one arm of a Mach-Zehnder interferometer and by applying a voltage across the two electrodes in the fiber the refractive index is modulated and the optical signal switched from one output port to the other. So far the lowest switching voltage achieved is ~1600 V which is too high for a commercial device, but by optimizing the design of the fiber and the poling process a switching voltage as low as 50 V is aimed for. A method to deposit a thin silver electrode inside the holes of an optical fiber is also demonstrated. A new way of creating periodic electrodes by periodically ablating the silver film electrode inside the holes of an optical fiber is also shown. The periodic electrodes can be used to create a quasi-phase matched (QPM) nonlinearity in the fiber which is useful for increasing the efficiency of a nonlinear process such as wavelength conversion. Poling of a fiber with silver electrodes showed a huge increase in the nonlinearity. This could be due to a resonant enhancement caused by silver nanoclusters. Keywords: Poling, twinhole fiber, fiber electrodes, silver film electrodes, silver diffusion, quasi-phase matching, optical switching, frequency conversion, optical modulation

Preface The work described in this thesis was performed at Acreo AB in Kista, Sweden. The work has been partially founded by the European GLAMOROUS project. The finishing of this thesis was possible because of partial financial support from Fredrik Laurell’s group at KTH. The thesis consists of a theoretical introduction to give a background to the work performed, which is followed by a description of the performed experiments and the reprints of the publications listed below.

List of publications I. ”Widely tunable fiber-coupled single-frequency Er-Yb:glass laser” G. Karlsson, N. Myrén, W. Margulis, S. Tacheo, F. Laurell. Applied Optics, Vol. 42 Issue 21 Page 4327 (July 2003) II. “Comparison of characterization techniques and the effect of surface condition in poled silica” C. S. Franco, G. A. Quintero, N. Myrén, A. Kudlinski, H. Zeghlache,H. R. Carvalho, A. L. C. Triques, D. M. González, P. M. P. Gouvêa, G. Martinelli, Y. Quiquempois, B. Lesche, I. C. S. Carvalho, W. Margulis. Submitted to Journal of Applied Physics, November 2003 III. “In-fiber electrode lithography” N. Myrén, W. Margulis, M. Fokine, L. Kellberg, O. Tarasenko. Submitted to Optics Letters, November 2003 Additional publications by the author: IV. “Periodic electrodes fabricated by laser ablation” N. Myrén, W. Margulis, M. Fokine, L. Kellberg, O. Tarasenko. Conference on Bragg grating, poling and photosensitivity (BGPP), Monterey, California, September 2003. V. “Comparison of characterization techniques and the effect of surface condition in poled silica” C. S. Franco , G. A. Quintero, H. R. Carvalho, I. C. S. Carvalho, D. M. González, P. M. P. Gouvêa, A. L. C. Triques, B. Lesche, N. Myrén, W. Margulis, G. Martinelli, Y. Quiquempois, A. Kudlinski, H. Zeghlache. Conference on Bragg grating, poling and photosensitivity (BGPP), Monterey, California, September 2003. VI. “Tuning of an Er-Yb:glass mini-laser using external feedback from fiber Bragg grating” N. Myrén, W. Margulis, G. Karlsson and F. Laurell, S. Taccheo. Conference on Lasers and Electro-Optics (CLEO) Los Angeles, California, May 2002

Acknowledgements

A big thanks goes to my supervisor and inspiring mentor Walter Margulis. No matter how busy he was he always answered my numerous questions. I thank him also for listening to my ideas and for feeding me with new ones. I also thank my main supervisor Fredrik Laurell for having me in his group and saving me when the times got tough at Acreo. I wish to thank Leif Kjellberg for giving me support when building all the gadgets and gizmos used in this work and Oleksandr Tarasenko for helping me preparing fibers. I thank all the people at our fiberlab in Hudiksvall for supplying me with fibers, especially Håkan Olsson, Niclas Sjödin, Per Helander and Lars Norin. Thank you also Ola Gunnarsson, Åsa Claesson and Michael Fokine for taking the time to answer all my questions and for proofreading my thesis. I thank Ingemar Petermann for the interesting discussions on real life matters and Fredrik Carlsson for producing the special gratings needed in some experiments. I would like to thank Mikhail Popov for theoretical discussions on different aspects of fiber optics and Gunnar Karlsson for fruitful collaborations. Finally, I really appreciate all the support I have been given from all the people at the photonics department and the people working within the GLAMOROUS project.

Thank you all!!!

Content

1 Introduction____________________________________________________________ 1

2 Nonlinear Optics ________________________________________________________ 3 2.1 Background________________________________________________________________ 3

2.2 Wave equation _____________________________________________________________ 4

2.3 Second Harmonic Generation in silica _________________________________________ 7

2.4 Phase-matching ____________________________________________________________ 8

2.5 Electro-Optic effect _________________________________________________________ 9

3 Silica Glass fibers ______________________________________________________ 13

4 Thermal Poling ________________________________________________________ 15 4.1 Thermal Poling in Bulk _____________________________________________________ 15

4.1.2 Creation of the depletion region ____________________________________________ 15

4.3 Poling of fibers ____________________________________________________________ 17

5 Experiments___________________________________________________________ 23 5.1 EFISH ___________________________________________________________________ 23

5.2 Indiffusion of silver in fiber with thin film silver electrodes _______________________ 25

5.3 Measuring the temperature in the fiber _______________________________________ 28

5.4 Poling experiments. ________________________________________________________ 30

5.5 Ablation of silver electrodes _________________________________________________ 32

6 Conclusions ___________________________________________________________ 37

References _____________________________________________________________ 39

List of publications_______________________________________________________ 41

1 Introduction The ever-increasing need for higher capacity in the telecom networks calls for cheaper and more robust optical components. The cost of optical components is not only determined by the cost of the component itself but also by the cost of installing and maintaining it. Optical packet switching in today’s WDM systems is primarily done by Mach-Zehnder modulators containing an electrooptical crystal. Protection switching is mainly done by Micro Electronic Mechanical Systems (MEMS) switches. A function that is becoming more important as more wavelength channels are added to the systems is to optically change the wavelength of a single channel or a range of channels from one wavelength to another. The devices needed to perform the above mentioned tasks are costly and because of their free-space configuration they introduce loss that must be compensated for by adding optical amplifiers. Inexpensive and low loss optical silica fibers are used in almost all transport links of optical networks. The key characteristic of a transport fiber is to affect the guided radiation as little as possible. Since silica is a centrosymmetric material it has no second order nonlinearity and is therefore, in principle, a very poor material for active devices. By poling silica however, e.g. by heating up the material and applying a high voltage, one can break the centrosymmetry and create and electrooptically active material. This poling process can be applied not only to bulk silica glass, but also to optical fibers. By drawing a fiber with one hole on each side of the core and inserting electrodes into these holes one can pole the fiber and thus make it electrooptically active. This technique makes it possible to integrate optically active devices into optical fibers. The main advantages are the low splicing loss and the possible low cost since silica fibers can be produced in very long lengths. Most of the work done in this thesis was performed in the frame of the GLAMOROUS (GLass-based MOdulators, ROUters and Switches) project. The three-year project is funded by the European Union and started September 1st 2001. The project has nine participants, ranging from universities and research institutes to companies. The scientific and technological objective of GLAMOROUS is to demonstrate development of groundbreaking technology that allows fabrication of active glass-based components that are industrially viable. This is to be accomplished by increasing the nonlinearity in the glass material through poling. The value of the permanently induced electrooptic coefficient should be sufficiently high to allow poled fibers and waveguides to perform advanced optoelectronics functions by responding to electrical fields with ultrahigh speed. Since fibers have low cost and low loss, are light, small, easily spliced and modular, such devices could then be used as building blocks in a number of environments where logical decisions are to be taken. The more specific goals of projects are mainly to 1) to optimize the poling conditions (in terms of time, temperature and voltage) to increase the stability of poled samples (to > 10 years at room temperature). 2) A 2x2 electrooptical switch with a time response < 100 µs and driving voltage x poled length < 50 Vm. 3) A wavelength converter with conversion efficiency >-10 dB, number of simultaneously switched channels ≥ 8 at < 2W pump power.

2 Nonlinear Optics 2.1 Background In the optical phenomena that we encounter everyday, the interaction between light and matter is a linear process. It is linear in the sense that there is no interaction between light waves in the material. It is also linear in the sense that processes are independent of the intensity of illumination. An optical material can be thought of as a collection of positively charged ions surrounded by electron clouds. An optical wave entering the material creates a propagating perturbation in the position of the positive and negative charges. In a dielectric material there are no free charges and the displacement of the charges give rise to a polarization in the medium. In the case of linear optics the polarization of each dipole is a linear function of the electric field it is subjected to. At large amplitudes the induced polarization will no longer be linear, but rather a nonlinear function of the electric fields. One can make a mechanical analogy with two particles connected with a spring. If the particles are pulled apart only slightly the restoring force of the spring will be a linear function of the displacement. For larger forces the displacement will depend on higher order terms of the displacement (fig. 2.1). The polarization can be expressed as a power series in E on the form [1]:

Where PL is the linear contribution and PNL is the nonlinear contribution. χ(N) is a susceptibility tensor of order N and rank (N+1) e.g. χ(2)

ijk. In order to observe nonlinear phenomenon the incident optical field in the material must be comparable to the internal field that binds the electrons and ions together. This field is ~3x1010 V/m. An incident intensity of ~1014 W/cm2 would give a comparable field. In some cases the radiation from the individual dipole oscillation will add up constructively along the medium and give a much higher intensity. This is known as phase matching and makes it possible to compensate for weak nonlinear coefficients by increasing the interaction length. In waveguides such as optical fibers this is particularly true since the energy can propagate long distances without significant loss of energy or spreading of the beam.

...)( 3)3(2)2(0

)1(0 +++=+= EEEPPP NLL χχεχε 2.1 2.2.1

Fig. 2.1 Figure (a) shows the linear regime. The electric field is small and the induced polarization is to a good approximation a linear function of the electric field. Figure (b) shows the nonlinear regime. The polarization of the material depends on higher order terms of the electric field [1]. 2.2 Wave equation The electromagnetic field in an optical fiber can be described by Maxwell’s equations [2]:

In the absence of free charges and free currents, as in a silica fiber, we have Jf = 0 and ρf = 0 which simplify the Maxwell equations to:

0=•∇

=•∇∂∂

+=×∇

∂∂

−=×∇

B

D

DJH

BE

f

f t

t

ρ

00

=•∇=•∇

∂∂

=×∇

∂∂

−=×∇

BD

DH

BE

t

t

2.2.1

2.2.3

The flux densities D and B are the response of the medium from the applied fields E and H. They are related through the consecutive relations given by: D = εoE+P B = µoH+M where P and M are the induced polarization and magnetization of the material. In nonmagnetic materials M=0. The Maxwell’s equations can be used to find wave equation that describes light propagating in the fiber. Taking the rotation of each side of eq. (2.2.3) gives:

Now, from eq (2.2.5) we know that: B=µoH which gives:

inserting (2.2.8) in (2.2.6) gives:

from eq (2.2.4) we now get:

finally with

∂∂

×∇−=×∇×∇tBE

tt o ∂∂

=∂∂ HB µ

( ) 2

2

tt oo ∂∂

−=×∇∂∂

−=×∇×∇DHE µµ

2

2

2

2

2

2

ttt oooo ∂∂

−∂∂

−=∂∂

−PED µεµµ

2/1 coo =εµ

2.2.4

2.2.5

2.2.6

2.2.8

2.2.7

2.2.9

2.2.10

2.2.11

we get:

P in this case contains both a linear and nonlinear contributions. The linear terms in P can be directly expressed in E using the relation n2 ≡ 1+χ(1) under the assumption that the loss is very small giving a negligible imaginary part. Since in a step-index fiber the n(ω) is independent of the spatial coordinates [3] we can use the vector identity:

from (2.2.3) we have

and finally we get:

When dealing with nonlinear effects it is convenient to separate out the spatial dependence and only look at the field propagating along the fiber. If these are assumed to be monochromatic infinite plane waves and the x-axis points along the fiber, the electric field is given by:

where Eo is the envelope of the electric field. We can further simplify the expression by assuming than the envelope, Eo, only slowly varies in amplitude and phase along the fiber. This is called the slowly varying envelope approximation (SVEA) [1]. It is expressed as:

2

2

2

2

21

ttc o ∂∂

−∂∂

−=×∇×∇PEE µ

E-E)(E 2∇•∇∇=×∇×∇

0E =•∇

2

2

2

2

2

22

ttcn

o ∂∂

+∂∂

=∇ NLPEE µ

[ ])(*)( ),(),(21),( tkxi

otkxi

o etEetEtx ωω ωω −−− +=E

xEk

xE oo

∂∂

<<∂

∂2

2

2.2.12

2.2.13

2.2.14

2.2.15

2.2.16

k is the wave-vector defined as k = nω / c. Using SVEA we can reduce the second order differential equation into an ordinary one.

2.3 Second Harmonic Generation in silica Second harmonic generation (SHG) is the creation of radiation with a frequency ω2 two times that of the input radiation ω1 . ω2 = 2ω1 SHG is a second order nonlinear process. It couples the input fields through the second order nonlinear tensor χ(2). Any material that is invariant under inversion (centrosymmetric) can not have second order nonlinear processes and therefore χ(2)=0. Silica glass used in optical fibers is much like a frozen liquid. It does posses some short range order, but at long range the material is randomly oriented. If silica glass matrix is inverted around any point in space, then on a macroscopic scale, the material will still be the same and therefore it is centrosymmetric. This implies that it is impossible to generate second harmonic (SH) radiation in optical fibers. This is only true however as long as the material is symmetric. The symmetry can be broken by applying a strong electric DC field to the material. The electric field induces a polarization of the material and SH processes are no longer forbidden. This process is usually referred to as electric field induced second harmonic generation (EFISH). The arising nonlinear polarization terms are of the type [4]:

where EDC is the applied electric field, and EOPT is the propagating optical field. The 0 in the frequency arguments corresponds to the DC field. The applied external field can be permanently recorded in a so-called poling process i.e. the application of an electric field while the silica sample is hot (~280oC). This is described in more detail in chapter 4. If the external electric field is kept constant while the sample is cooled down to room temperature, a permanent field is stored and the material can be used for SHG without applying an external field. Second-order processes in silica are often described by the effective nonlinear coefficient χeff

(2), which is defined as:

NLPEnci

xo

2ωµ

=∂∂

[ ] ..)22(exp),,0(43 2)3( cctxkiEE OPTDC +− ωωωχ ω

DC)3()2( 3 Exxxxeff χχ ≡

2.2.17

2.3.1

2.3.2

2.3.3

2.4 Phase-matching In most optical systems the refractive index is different for the pump wavelengths and the generated wavelengths. In a frequency doubling experiment this means that there will be a phase mismatch between the fundamental and the SH, here expressed as a mismatch in k-vectors [1].

When the fundamental propagates along the fiber the SH generated at a certain point will interfere destructively with the SH generated earlier. The net effect is an oscillating energy flow back and forth between the fundamental and the SH. The spatial length it takes for the two waves to go from completely in phase to out of phase by π/2 is determined by the coherence length (beat length) of the process, Lc.

The fundamental of a Nd:YAG laser (λ=1064nm) launched into a nonlinear region in a fused silica glass sample will experience a refractive index nω = 1.45. The generated SH at 532 nm will experience a refractive index n2ω = 1.461188. The data for the refractive indices was interpolated from Properties of Optical and Laser-Related Materials [5]. The difference in refractive indices leads to a mismatch in k-vectors between fundamental and SH (eq. 2.4.1).The calculated coherence length using the above values for the refractive index is (eq 2.4.2) Lc = ~23,8 µm (for bulk sample). Margulis et. al made an experiment were a Ge fiber was poled optically with the high intensity pulses of a Q-switched mode-locked Nd:YAG laser. The poling process wrote a grating that phase matched the SH process. When this fiber was etched in HF a structure with a period (2Lc) of 39.2 µm was revealed [6]. The difference between the value calculated for a bulk sample and the value measured for a fiber arises since the effective refractive index is a waveguide is not solely dependent on the material properties, but also on the geometrical shape of the structure. In order for frequency conversion to be efficient the difference in refractive indices must be compensated for. There are many different ways to phase-match optical processes, which include taking advantage of existing birefringence and use different optical axis for fundamental and harmonics, using different modes and waveguide design. In this project the design of the fiber had to be compatible with standard telecom fibers in order to take full advantage of the low splicing loss. Quasi phase matching (QPM) as described in fig 2.2 was therefore the most natural choice.

0)(2)(222 222 ≠

−=−=−=∆

ooo cnn

cn

cnkkk ωωωω

ωωωωω

kLc ∆

=π

2.4.1

2.4.2

Fig 2.2. Principle of QPM. The non phase-matched (NPM) case shows how the energy oscillates between the SH and the fundamental. The QPM case shows how the SH intensity increases in a step-like manner [7]. The QPM technique is very useful in optical fibers since other means of phase-matching are less flexible and more complicated to apply. QPM is not really phase-matching in the sense that it completely removes the phase difference between the propagating waves. The efficiency of the technique comes from making the SH process less efficient (preferably completely inactive) in the areas where the SH and pump are out of phase. QPM in silica fibers gave a conversion efficiency of over 20% for nanosecond pulses at 1532nm [8]. In this case the primary coating was removed and the fiber was then polished to a D-shape and by means of lithography a periodic electrode was deposited on the flat surface. When the fiber was poled the fiber contained periodic areas with high nonlinearities with inactive areas in between. The disadvantage of this method is that the polishing makes the fiber fragile and difficult to splice. The ablation section of chapter 5 explains a method for making periodic electrodes in a fiber without having to remove the coating or polish the fiber. 2.5 Electro-Optic effect The main goal of the GLAMOROUS project is to make a wavelength converter and a switch/modulator. Phase modulation alone does not affect the intensity of the radiation. Since it is difficult to directly modulate the amplitude of the electric fields inside a fiber the common approach transfer a phase modulation and then transform the phase modulation into an amplitude modulation with a polarizer or a Mach-Zehnder interferometer. For switching purposes a Mach-Zehnder interferometer is the most natural choice since changing the phase in one arm of the Mach-Zehnder switches the radiation from one output arm to the other. A stand alone Mach-Zehnder has two input arms and two output arms and

is therefore called a 2x2 switch (fig 2.3). Any configuration of input ports and output ports can be constructed by adding interferometers. A 2x4 switch can be built by connecting one Mach-Zehnder to each of the output arms.

LaserPoled fibre

Bias Adjustment

FusedFibre Coupler

Detection

~Modulation

Fig.2.3. A Mach-Zehnder interferometer. The output can be shifted from one output port to the other by applying a voltage to one of the arms in the interferometer. The voltage changes the refractive index of the glass with ∆n which in turn changes the phase of the electromagnetic wave with ∆ϕ. The response of the Mach-Zehnder arises from a change in refractive index when an electric field is applied to the material. The magnitude of the voltage needed for switching is calculated below. From eq. (2.2.4), we have that.

D = εoE+P As motivated before the polarization of the material expressed as:

The linear permittivity is defined as [9]: εr = 1 + χ(1)

(2.5.2) can be rewritten as

...)( 3)3(2)2(0

)1(0 +++=+= EEEPPP NLL χχεχε

εεεχχεε

=∆+=+++= rro

EEE

D ...)( 2)3()2(

2.5.1

2.5.2

2.5.3

With

ε = n2 = (no+∆n)2 ≈ (for small ∆n) ≈ no2 +2 no ∆n

from which it follows that (ignoring terms of higher order than cubic and with χ(2)=0):

With E = Epoling + Eext and expressing in terms of χ (3)

xxxx and with eq. (2.5.5)

With

In a Mach-Zehnder the optical path length of one arm (L) must be changed by λ/2 in order to switch from one output arm to the other. The change in refractive index for switching is thus given by:

With a second-order nonlinear susceptibility χ (2)

eff = 1 pm/V and assuming that Eext << Epoling, a typical telecom wavelength, λ = 1550 nm, the unperturbed refractive index no = 1.5 and a length of the device L = 30 cm, the applied electric field needed for switching is given by:

)(21

2222)3(

2

Ennnn

nn

ooo

r

o

o χεεεε=

∆=

−=

−=∆

)323(21 2)3()2(2)3(

extxxxxexteffpolingxxxxo

EEEn

n χχχ ++=∆

polingxxxxeff E)3()2( 3χχ =

Ln

2λ

=∆

mV

LnE

eff

oext

612

9

)2( 109.3)101(3.025.1)101550(

2×=

×××

== −

−

χλ

2.5.4

2.5.5

2.5.6

2.5.7

With a distance between the electrodes d = 15 µm the voltage V required for switching is estimated to V = Eextd = 3.9×106 × 15 10-6 = 59 V. If the voltage is assumed to drop over the depleted region (where the resistance is higher) rather than across the whole distance between the electrodes the voltage for switching will be lower.

3 Silica Glass fibers Silica glasses have very small optical loss over a wide range of wavelengths; from the ultraviolet to the near infrared (Fig 3.1). This in combination with the ease of manufacturing has made it the material of choice in nearly all optical fibers used in the telecom networks [10].

Fig 3.1. Transmission loss in optical fiber. Long haul networks operate in the 1.5 µm area because of the low absorption (~0.2 dB/km). OH ions have a strong absorption peak at 2.73 µm [3] and the overtone at 1.37 µm is seen in the figure [11]. In an optical fiber the radiation is trapped in the core through total internal reflection caused by the higher refractive index in the core compared to the surrounding cladding. The refractive index difference between the core and the cladding is realized either by adding dopands to the core that increases the refractive index such as GeO2 and P2O5, or by adding dopants to the cladding that decrease the refractive index, such as Boron and Fluorine [3]. The core can be co-doped with other ions in order to increase the functionality, such as rare earth ions to make fiber amplifiers. Typically the index-step between the core and the cladding is 5x10-3 in a single-mode fiber. Fabrication of fibers is made in two stages. First a cylindrical preform (diameter ~2 cm) is manufactured with the right geometrical dimensions and the wanted refractive index profile and then the preform is drawn to a fiber. Several methods can be used for manufacturing the preform. The most common method is modified chemical vapor deposition (MCVD). In this process a fused silica tubes is used as a base. Inside the tube layers of SiO2 are deposited by mixing SiCl4 and O2 at a temperature of about 1800 oC. The fused silica tube is heated by a H2/O2 burner that sweeps along the tube to make uniform layers of deposition. When the cladding is thick enough GeCl4 or POCl3 is added to the mixture and the core is deposited as a thin layer on the inside of the tube. The tube is then collapsed before it is drawn to a fiber in a drawing tower. During fiber drawing the temperature and drawing speed needs to be precisely controlled to ensure a uniform core and cladding. The relative core/cladding dimensions are preserved and the diameter is reduced from ~2 cm to 125 µm. During fiber manufacturing the OH impurities in the fiber need to be kept at a minimum to achieve acceptable loss (fig.3.1). Typical

values of OH content is less than 10 ppb. Besides absorption from impurities there will always be small fluctuations in refractive index causing Rayleigh scattering. The loss caused by Rayleigh scattering is dominant at shorter wavelengths (R ~ 1/λ4). The fibers used in this work all have two holes placed on opposite sides of the core. Metal put into the holes is used as electrodes to apply an electric field across the core. The holes in the fibers described in this work are manufactured by drilling holes in the collapsed preform. If all the drawing parameters are carefully controlled the cross-section of the preform will simply scale down when the preform is drawn into a fiber. Typically the holes in the preform are ~ 8 mm in diameter and the diameter of the collapsed preform is ~30 mm. Since the final fiber preferably has the same diameter as a standard telecom fiber (=125 µm) the reduction factor is ~250 times. This gives a diameter of the holes in the fiber of ~ 30 µm.

4 Thermal Poling 4.1 Thermal Poling in Bulk All centrosymmetric materials, such as fused silica glass, lack a second-order nonlinear susceptibility χ(2). By applying an electric field to a silica sample the symmetry is broken and second-order nonlinearities can be present. Thermal poling consists of heating the sample to an optimum poling temperature (~250-290 oC) and simultaneously applying a high-static electric field for a sufficiently long time. In this temperature range alkali ions in the sample, and especially Na+, are very mobile and easily move through the glass matrix. The current through the glass is large when the voltage is applied and then decrease to a steady state value [12]. The steady state value is usually reached in 15-90 minutes and at this point the sample is cooled down to room temperature while the electric field is still applied. At room temperature the mobility of the Na+ is very low and the positions of the ions determined at the poling temperature are frozen. The applied field is now removed and the result is a thin optically nonlinear layer about 10 µm thick located just below the positive electrode (anode) [13]. This layer is depleted of impurity ions and is often referred to as “the depletion region”. The value of the second-order nonlinearity can be estimated by multiplying the recorded field with the intrinsic third-order nonlinearity as described in eq. (2.3.3). The intrinsic third-order nonlinearity has been found to be χ(3)

xxxx = 2×10-22m2/V2 [14]. The maximum field possible to record in the glass is limited by the dielectric strength, for fused silica this is reported to be equal to 8×108 V/m in 6 µm thick silica samples [15]. This value is quite uncertain and varies strongly with sample thickness and temperature. A measurement on twin hole fiber with 15 µm distance between the electrodes showed that it is possible to apply 6.5 kV before dielectric breakdown occurred. This gives a dielectric strength of 4.3×108 V/m The estimated maximum second order nonlinear coefficient is χeff

(2) = (3)×(2×10-22m2/V2)×(8×108 V/m) = 0.48 pm/V (eq 2.3.3). Thermally poled samples have a reported nonlinear coefficient of up to ~1 pm/V [12]. The electric field recorded in silica samples is mainly contributed to the migration of positively charged alkali ions such as Na+, K+ and Li+, but could also arise from polarization of bonds or defects in the glass to give higher values than predicted by a model taking only charge migration into account. 4.1.2 Creation of the depletion region When a silica glass sample is heated to ~280 oC the mobility of Na+ increases to 1×10-13 m2s-1V-1 [16]. If the formation of the depletion layer was only dependent on one type of charge carriers the depletion layer would reach a steady state after a certain time and then the ion transport would stop. At this point the internal field would cancel the applied electric field and the alkali ions would be more or less static. This contradicts experimental results in which there exist an optimum poling time (i.e. 20 min for a 0.5mm thick sample at 280 oC) [17]. The existence of an optimum poling time can not be explained in a model where only one charge-carrier is taken into account. It has been proposed in literature that hydrogenated species (such as H+ or H3O+) can be injected into the glass assisted by the strong electric field at the surface of the sample [18]. This was later confirmed using Secondary Ion Mass Spectroscopy (SIMS) [19]. Explanation of experimental results require

the mobility of the hydrogenated species to be ~103-104 times lower than that of sodium. Samples used for poling typically contain a spatially even distribution of alkali ions. At the surface of the sample there is a H+ reservoir originating from the surrounding atmosphere. If the sample is poled in air this reservoir gives an almost unlimited supply of H+ ions and the main factor determining the amount of charge injection is the surface condition e.g. the type of electrodes used. If the electrodes are deposited on the glass then hydrogen already present at the surface prior to electrode deposition can be injected into the glass. If the electrodes are of contact/pressed-on type, hydrogen can pass between the electrode and the glass surface and there will be a large supply of hydrogen available, unless the sample is poled in vacuum. The electric field drives the positive ions towards the cathode to form a layer under the anode depleted of positive ions (fig 4.1). This layer forms on a time-scale of minutes. At the cathode side alkali ions are ejected out of the glass and neutralized. As the positive alkali ions are driven away from the anode side a negatively charged area is created.

The high concentration of negative charges creates a strong electric field pointing from the anode and into the glass. When the field grows strong enough the positive hydrogen ions at

Anode

Cathod

Cathod

Anode

1)(t=0) 2)(t= t1)

Cathod

Anode

NeutralNegativePositive

E-Field

Anode

Cathode

E-Field

Positiv Neutral Negative

3)(t=t2) 4)t=t_stop Fig 4.1. Creation of a nonlinear layer. In figure 1) (t=0) no voltage is applied and thesample is neutral. At 2)(t=t1) the voltage is applied. The highly mobile alkali ions drifttowards the cathode. Some ions are ejected out of the sample and neutralized at thesurface. The region that is depleted of positive ions now has a net negative charge. At3)(t= t3) the high electric field between the anode and the negatively charged region ofthe sample injects positive hydrogen ions into the glass. 4) The positive charges injectedneutralize the nonlinear region. Only a thin negatively charged region inside the sampleand a positive layer near the surface remains when the poling process is completed.

the surface of the sample are injected into the glass. Since the mobility of hydrogen in silica is much lower compared to alkali ions there will be an optimum poling time when a large depleted region has been formed and the hydrogen ions has not had the time to completely neutralize this region. Even for short poling times much of the depletion layer will be neutralized and only a thin negatively charged layer remains. This negatively charged region, together with a thin layer of positively charged ions near the surface creates a strong electric field recorded in the sample. Fused silica is very resistant to various chemicals. The only chemical that efficiently etches silica is hydrofluoric acid (HF). A solution with 40 % concentration of HF etches silica at ~0.7 µm/minute. The etching rate is reported to be depending on the electric field in the glass [20] and has been used to determine the electric field recorded in poled samples. The etching rate is easily monitored by detecting the interference fringes between a beam hitting the surface being etched and a reference surface. Etching of poled samples in 40% HF reports that the electric field is fairly constant in the entire nonlinear layer (fig 4.2) except for fluctuations near the anode surface [21]. Upon etching through the entire nonlinear layer the etching rate abruptly changes to that of an unpoled sample.

Depth in sample

Recorded E-Field

Fig 4.2. Schematic diagram of recorded E-field in a poled sample. This type of electric field distribution is usually obtained in etching experiments. Typically the width of the depletion layer is ~10 µm for a bulk sample. 4.3 Poling of fibers The mechanism when poling an optical fiber is believed to be much the same as for poling of bulk samples. The main difference is that the distance between the electrodes is ~15 µm as compared to the distance in bulk (usually 500 µm or more). The small distance between the electrodes limits the highest voltage that can be used when poling the fiber. At poling temperature (~280oC) the breakdown field is reported to be ~5×108 V/m for silica glass [21]. If this value is true for the fiber shown in fig 4.3 it should be possible to apply ~7.5 kV between two holes separated by 15 µm. We found that the maximum voltage in most poling experiments was limited to ~2-5 kV. This could be because the field intensity between two metallic cylinders is larger than between two metallic plates put apart the same distance. If the hole-to-hole spacing is increased the voltage used for poling can be increased since the electric field intensity in the glass is lower for the same applied voltage. This makes it possible to increase thickness of the nonlinear layer since the width of the depletion region scales approximately linearly with the applied voltage and not with the

electric field. The drawback of increasing the hole-to-hole distance is that once the fiber is poled, the same applied voltage will create a smaller electric field if the distance between the electrodes is increased and hence the device will be less sensitive to a modulating voltage. When poling a bulk sample the created nonlinear layer is thin in comparison with the total thickness of the sample. Since the nonlinear layer is depleted of ions and therefore has higher resistivity than the bulk glass, a voltage applied across the entire sample creates a strong electric field across the depletion region. In a poled fiber the width of the depletion layer (~10 µm) is of the same order as the distance between the electrodes (~15 µm) and thus the enhancement effect of the applied voltage will be much less profound. According to Faccio et. al. [17] the optimum poling time is shorter for thinner samples in the 100µm – 1000 µm range. These results together with simulations done indicates that the optimum poling time for fibers is < 1 minute [16]. When poling optical fiber there are several different ways of applying the electrical field. In one of the first attempts of poling a fiber a commercial polarization maintaining fiber was used. The fiber was etched to decrease the electrode-core distance and then placed between two electrodes [22]. To prevent dielectric breakdown through the surrounding air, the fiber and the electrodes were covered with polyimide. The total distance between the electrodes was ~45 µm and the poling voltage would have been limited to ~100V (Emax(air) = 3 x 106 V/m) had it not been for the polyimide. The technological advancement of preform- and fiber production has made it possible to produce micro-structured fibers with complicated cross sections. Most fiber poling experiment reported recently use fibers with electrodes embedded in holes that run along the fiber on each side of the core, as can be seen in fig 4.3.

Fig 4.3. SEM image of twin hole fiber (F030625-1E). The diameter of the fiber is 125µm. The diameter of the holes is ~35µm. The core is seen as a bright dot placed a little closer to the left hole. Index step of core ∆n = 0.023, diameter of core d = 2.3 µm, core-hole distance = 5.5µm (closest edge-to-edge).

The electrodes can be made up of any electrically conducting material and the two most common electrodes are either thin (~20 µm diameter) wires inserted from side polished openings or solid alloy electrodes pumped into the holes, see fig 4.4. Wires inserted into the holes need to be significantly thinner than the hole diameter in order to make insertion feasible. Since the wire does not fill up the entire cross section the location within the hole is random. Wire insertion is very time-consuming and the applied voltage is not fully exploited since there is usually an air gap between the wire and the glass. The air gap between the glass and the wire has the advantage of reducing the loss of the fiber as compared to a fiber with an alloy electrode filling the entire cross section of the hole. The reason for this is that the empty hole has a refractive index n = 1 as compared to the ~1.5 of the glass, so the guided radiation sees a large index step and therefore the guided field does not experience the loss causing electrode as strongly as with an alloy electrode.

High pressure

Fiber with holes

Liquid metal

High temperature

Fig 4.4. The procedure used for filling twin hole fiber with molten alloy is illustrated in the left figure. The pressure used is ~7Bars and the temperature is kept ~25oC above the melting temperature of the alloy. The middle figure shows a fiber with alloy electrodes pumped into the holes. The electrodes fill the entire holes. The right fiber has wires inserted into the holes, which are randomly located within the cross-section. A fiber ideal for poling needs to have a number of characteristics fulfilled. The fiber should be compatible with a standard telecom fiber in terms of ease of splicing with resulting low splice loss. The overlap between the ~10 µm thick nonlinear layer and the core should be as complete as possible. For this reason the core must to be positioned close to the anode electrode. However, placing the core too close to the electrode dramatically increases the loss caused by the metal. The energy guided by the core has an approximately gaussian distribution which extents outside of the physical core. The mode field diameter is defined as the radial distance at which the electric field strength has decrease by a factor 1/e. The mode field diameter can be decreased by increasing the Ge concentration of the core i.e. by increasing the refractive index step. A comparison between the field distribution in a standard telecom fiber (SMF 28) and fiber F030625-1E, our latest fiber specially designed for poling, is shown in fig 4.5. The simulated mode field diameter is 3.2µm in the highly doped core as compared to 10.4 µm in an SMF 28. This means that for a given acceptable loss per meter, the core can be moved closer to the hole if the index step is increased.

-15 -10 -5 0 5 10 15

0,0

0,1

0,2

0,3

0,4

0,5

Core

Pow

er [a

.u.]

Radial position in fiber [um]

-15 -10 -5 0 5 10 15 20-0,10,00,10,20,30,40,50,60,70,80,9

Core

Pow

er [a

.u.]

Radial position in fiber [um]

Fig 4.5. The left picture shows a simulation of the mode field in a fiber with a core having the same index step and diameter as that in a standard telecom fiber (SMF 28). Lambda=1.55 µm, R_core= 4.15 µm, ∆n = 0.0052. The right picture shows the simulated mode field in a specially designed fiber (fig 4.3) with a highly doped core, lambda=1.55 µm, R_core = 1.15 µm, ∆n = 0.023. The index step of the right core is 4.4 times larger than that of the left one. In a commercial device an acceptable loss is roughly 3dB for the total device. Subtracting splice losses of 1dB/splice the propagation loss can be no greater than 1dB. Given a device length of ~30 cm the acceptable propagation loss level is 3 dB/m. The fiber in fig 4.6 with alloy electrodes in the holes have a loss of 2.45 dB/cm @ 1.5 µm. The fiber shown in fig 4.3 has a higher doping concentration in the core giving an index step ∆n = 0.023. The diameter of the core was decreased to 2.3 µm in order to keep approximately the same cut-off wavelength, (the wavelength at which a fiber can support higher order modes as approximately described by eq. (4.3.1). The core-to-hole distance was increased from 4 to 5.6 µm. A cutback measurement revealed that the loss was still unacceptably high (2.25 dB/cm). The cut-off of this fiber was 780 nm. Increasing the core diameter increases the cut-off and gives a smaller mode field diameter. A core size of 3.8µm gives a cutoff of 1300 nm. If the cut-off is increased beyond 1550 nm the fiber will not be single-mode at telecom wavelengths. For a step-index fiber, the cut-off wavelength λcutoff is given by the following formula:

where Rcore is the radius of the core, ncore is the refractive index of the core, and nclad is the refractive index of the cladding.

4.3.1 22

4048.22

cladcorecorecutoff nnR −=πλ

-50 0 50 100 150 200 250 300 350

1

10

100

1000

10000

Metal ends

Metal starts

Loss in the metal filled part is2.446 (+/-) 0.274 dB / cm

Inte

nsity

in fi

ber [

uW]

Position in fiber[mm]

Fig. 4.6. Loss in metal filled fiber (F030402-1F). This fiber has a loss of 2.5 dB/cm in the metal filled section. The loss is 0.4 dB/m without metal in the holes. The index step between the core and the cladding ∆n = 0.015, diameter of the core d =4.3 µm, core-hole distance = 4µm (edge-to-edge).

5 Experiments 5.1 EFISH When applying an electric field to silica glasses at room temperature the symmetry is broken and the material can frequency double. This process is called Electric Field Induced Second Harmonic generation (EFISH) and resembles that observed during poling, the main difference being that the fields during poling are permanently recorded. EFISH is an efficient way of determining how a sample responds to applied electric fields. By changing the temperature of the sample one can find important clues about the dynamics of the poling process. An experiment was done with a standard germanium core fiber. In order to test the fiber the fundamental of a Nd:YAG laser (1.064µm) was coupled into the core using a 10x microscope objective as shown in fig 5.1. The laser was Q-switched and mode-locked with a repetition rate of 3.4 kHz. Typically each Q-switched pulse contained 20 mode-locked pulses of ~120 ps duration. An estimated 1/3 of the total output power is in the pulse train, and the remaining 2/3 is in the prelase. After a few minutes the fiber was poled optically (see chapter 2.4) by the high power laser pulses used for probing which made the EFISH signal hard to resolve since the signal arising from optical poling was much stronger. To overcome this problem a new fiber was designed with a pure silica core, which supposedly would be much more resistant to optical poling. The index step of this fiber was created by doping the cladding with fluorine. The distance between the holes was ~45µm. The average power of the laser was attenuated so that 80 mW came out of the output end of the fiber. The electrodes were connected to a voltage supply that delivered square voltage pulses of variable length and amplitude. The electrode containing part of the fiber was placed in an oven with a maximum temperature of 250 oC. The fundamental at the output was blocked using broadband and interferometric filters while the SH signal (0.532 µm) was detected with a photo multiplier (PM).

1064nmPM

Voltage pulses

532nm

Fig 5.1. The experimental setup used when probing a fiber for EFISH. A voltage is applied to the alloy electrodes in a twin hole fiber. The applied field generates weak (<nW) SH signal. A picture of the cross section of the fiber taken with an optical microscope is shown on the right hand side. The core was illuminated from the other end of the fiber to make it visible.

One objective of the experiments was to determine how different electrode materials effect the response in the SH signal. Two different alloys were tested, BiSn (Bi 58%, Sn 42%) with melting temperature Tm = 138 oC and AgSn (3.5 % Ag, 96.5% Sn), Tm = 221 oC. The alloys were pumped into the holes of the fiber with ~6 bars of pressure at 175 oC for the BiSn alloy and at 250 oC for AgSn. The fibers were side polished to gain access to the electrodes and then contacted with 20 µm diameter gold-coated Tungsten wires and conducting epoxy. The total length of the electrode filled part of the fiber components was ~30 cm. Since this is much longer than the coherence length (Lc ≈ 20 µm) the SH harmonic signal will go through several (~104) maximums in the 30 cm long fiber. The output at the end of the fiber will be at maximum when the fundamental and SH have interfered constructively for a full coherence length. The EFISH response at 75 oC for the two different electrodes (BiSn and AgSn) is shown in fig 5.2. The SH signal in the fiber with BiSn electrodes makes a step increase when the voltage is switched on and then continues to grow during the whole voltage pulse. Note that there seems to be two time constants in the growth in the two rightmost pulses. The signal starts to saturate in the middle of the voltage pulse and then increases again. Note in the right graph in fig 5.2 that for AgSn, the SH signal makes a step increase when the voltage is switched on and the slowly decays. Note also that there is no clear step in the SH signal when the voltage is turned off.

0 3000 6000 9000 120000,00

0,02

0,04

0,06

0,08

0,10

Voltage

IR

SH

SH induced by voltagepulses in depressed cladding fiber @ 75 degrees BiSn Fill

SH

G [a

.u.]

Time [s]

0,0

0,5

1,0

1,5

2,0

2,5

3,0

Volta

ge [k

V]

0 2000 4000 6000 8000-0,10-0,08-0,06-0,04-0,020,000,020,040,060,08

VoltageIR

SH

SH induced by voltage pulse 1000 s @ 75 degreesAgSn fill component 3 Depressed cladding fiber

SH

G [a

.u.]

Time [s]

0,0

0,5

1,0

1,5

2,0

2,5

3,0

Vol

tage

[kV

]

Fig 5.2. The EFISH response in pure silica core fiber with different electrodes, both measurements done at 75 oC. The response from a BiSn electrode is shown on the left and the AgSn electrode is shown on the right. Note that with the BiSn electrode the SH signal slowly increases after an initial step increase when the voltage is switched on. In the fiber with AgSn electrodes the Sh signal slowly decays after an initial step increase. The three longest pulses in the left picture are 1000 s which is the length of all pulses in the right figure. In order to determine how the time constants change with temperature a new experiment designed with a fiber with AgSn electrodes. The setup was the same as described in fig 5.1 but this time the fiber was heated to 150 oC and the SH signal measured. The voltage pulses were 1000 s long and the SH signal showed the same type of behavior as at 75 oC with the

SH signal decreasing after a fast step increase when the voltage was switched on. At this temperature, as expected, the time constants were much faster (fig 5.3). At 75 oC the signal decays to half the amplitude in ~1000 s. At 150 oC the decay time is ~140 s. When the voltage is switched of the amplitude of the SH signal increases. This is an indication of a shielding field in the glass that is built up when the voltage pulse is applied and the positive ions are attracted to the cathode and the negative ions to the anode. After 1000 s the signal is almost gone and the shielding field is completely canceling the external field. When the external field is removed, only the shielding field is left. The amplitude of the signal when the voltage is remove is much lower than when the voltage is turned on. This is most likely because the decay of the shielding field is faster than the time response of the measurement system. In order to filter out noise a lock-in amplifier with build in averaging was used. Note also that the signal when the voltage is switched on gets smaller for each voltage pulse. This could be caused by a remaining shielding field that does not completely decay away from pulse to pulse.

0 2000 4000 6000 8000-0,006-0,004-0,0020,0000,0020,0040,0060,0080,0100,012

Voltage

IR

SH

SH induced by voltage pulse 1000 s @ 150 degreesAgSn fill component 3 Depressed cladding fiber

SH

G [a

.u.]

Time [s]

0,00,51,01,52,02,53,03,54,04,55,0

Vol

tage

[kV

]

Fig 5.3. EFISH response with AgSn electrodes in pure silica fiber at 150 oC. The decay of the SH signal is much faster than at 75 oC. Notice the response in SH when the voltage pulses are switched off, indicating a shielding field. 5.2 Indiffusion of silver in fiber with thin film silver electrodes The SH response with different electrode materials was further investigated in a pure silica fiber with silver film electrodes. The electrodes were created by flowing a mixture of AgNO3 mixed with NH4, and a reducing agent (dextrose) inside the holes for about 1 hour [23]. The procedure creates a ~1 µm thick film on the inside of the holes. The electrodes

were contacted and the experiment was done in much the same way as the EFISH experiments described in fig 5.1, i.e. square voltage pulses 1000 s long were applied to the electrodes at different temperatures and the SH signal generated from a Nd:YAG laser was monitored. The SH response was much like that of the AgSn electrode (fig 5.3) with a step response when the voltage was switched on and then a slow decay. At 50 oC the SH response looks about the same from pulse to pulse (fig 5.4 left). When the voltage pulse is applied there is a step increase in the signal which then decays with a time constant of ~500 s. In pulses number 1, 3 and 4 it looks like more than one time constant. When the same 1000 second voltage pulses were applied to a fiber at 100 oC the behavior was completely different. When the second pulse was applied there was a huge increase in second harmonic combined with a fast decrease in IR pump power (fig 5.4 right). After the second voltage pulse both SH and IR powers were below the noise floor of our detection system. Since the decrease in signal power was not abrupt the fiber could not have broken and the only possible explanation is that the attenuation of the fiber had increased dramatically.

0 2000 4000 6000-0,15-0,10-0,050,000,050,100,150,200,250,30

SH in pure silica silver electrode, 50oC , 8 mW IR @ output150 mW input, 4.4 kHz Q-switched 1000s pulses

SH

gen

erat

ed @

532

nm [a

.u.]

Time [s]

0

500

1000

1500

2000

IR

Voltage SH

Volta

ge [V

], I

R @

out

put [

100

is 1

0 m

W]

0 1000 2000 3000 4000 5000

0

1

2

Ag electrode, 100 oC, 8 mW IR output 150 mW input4.4kHz Q-switch 1000s pulses

SH

gen

erat

ed @

532

nm [a

.u.]

Time [s]

-1,0-0,8-0,6-0,4-0,20,00,20,40,60,81,0

IRVoltage

SH

Vol

tage

[kV

], IR

@ o

utpu

t [1

is 1

0 m

W]

Fig 5.4. EFISH experiment with thin film silver electrodes in a pure silica core fiber. The picture on the left shows the SH response at 50 oC. There is no increase in the SH signal as more pulses are applied. On the right hand side the same measurement was done at 100 oC. Note the big increase in SH combined with a sharp decrease in IR pump power as the second voltage pulse is applied to the fiber. Thermal poling of optical fibers usually does not significantly alter the guiding properties. Since Na+ is very immobile at 100 oC it is highly unlikely that the fiber was poled. The most likely explanation of the surprising results is that the elevated temperature combined with the high electric field drove silver into the core, or to the proximity of the core. The silver atoms then formed clusters that grew to nano-scale particles. Silver nanoparticles are known to have a resonant enhancement of SH processes with a peak around 440nm [24]. This enhancement extends to 532 nm and could explain the large increase in the SH signal. Silver nanoparticles could also explain why the IR drops so dramatically since most metals are very efficient absorbers of light. In visualize where in the fiber the IR drops a 10 mW HeNe (632 nm) laser was coupled into the fiber using a 40x microscope objective. In fig 5.5 only the right part of the fiber was in

the oven at 100 oC. There is a distinct point at which the scattered intensity increases strongly. This point coincides with the beginning of the heated region of the fiber and visual inspection showed no localized scattering indicating that the fiber was not broken. At the output end of the fiber light could be seen, but it was hardly detectable with an unaided eye, indicating an output intensity of <10-11 W. The estimated input was ~5mW giving a very rough estimate of the total loss to be ~83 dB over 20 cm (4 dB /cm).

Fig 5.5. The light from a HeNe laser (632 nm, 10mW) was coupled into the fiber with assumed silver in the core region. The light was launched from the left and as the light propagates in the fiber intensity drops dramatically at the part of the fiber that was heated to 100 oC during the EFISH experiment. The substrate glass on which the fiber was mounted is seen on the right hand side of the picture. This strong absorption was an indication of silver, but it was not a definite confirmation. A piece of the fiber was send to Jacob Fage-Pedersen (DTU, Denmark) who probed it with a white lightsource. The light was coupled in and out by means of standard telecom fibers that were aligned on xyz-stages to small pieces (2 cm) of poled and unpoled fibre, respectively. The absorption spectrum showed no significant peak just above 400 nm as would have been expected for silver nanocluster absorption [24] fig 5.6. Secondary Ion Mass Spectroscopy (SIMS) measurements were also performed to see if there was any detectable amount of silver in the glass. The silver electrodes gave a huge signal, but the concentration of silver in the core was less than the estimated detection limit of 0.02 atomic %.

100 200 300 400 500 600 700 800 900 100010-12

10-10

10-8

10-6

10-4

10-2

TRA

NSM

ISSI

ON

(A.U

.)

12mm unpoled fibre12 mm poled fibre

White-light source #2

100 200 300 400 500 600 700 800 900 10005

10

15

20

WAVELENGTH (nm)

POLI

NG

-IND

UC

ED L

OSS

(dB

/cm

)

Fig 5.6. The top graph shows the spectrum of a fiber subjected to an EFISH experiment at 100 oC that possibly indiffused silver, as compared to an untreated fiber. The bottom graph shows the absorption introduced by the EFISH experiment (computed as the difference in absorption in the two curves in the top figure). In order to confirm that it was silver that caused the strong SH signal the experiment was repeated several times. The results were quite confusing. Many times the SH signal increased and one time it grew as much as 1000 times larger than when the voltage was first switched on. The time for the strong signal to arise varied between the experiments. The same experiment was done using the same type of silver electrodes in a fiber with a Ge doped core. In this case there was no enhancement in the SH signal. More experiments will be done in the future in order to gain full understanding of this interesting process. The diffusion technique looks like a very promising way of increasing the nonlinearity of the fiber. 5.3 Measuring the temperature in the fiber When doing poling experiments the temperature needs to be carefully monitored. This can be done in many different ways of which the easiest is to use an external temperature sensor that measures the temperature of the atmosphere surrounding the fiber. The temperature inside the fiber can be estimated by measuring the resistance change of the electrodes inside the fiber, or by monitoring the wavelength shift of a fiber Bragg grating written in the core of the fiber. These methods require some type of preparation of the fiber

and to overcome this difficulty a new method was invented that can be used to measure the temperature inside the fiber without making any preparations of it. A small fraction of the fundamental from a Nd:YAG laser (1064 nm) was converted to 532 nm radiation with a frequency doubling crystal (fig 5.7). Both the fundamental and the SH signal were coupled into the fiber using a 10x microscope objective. The output from the fiber was sent through another frequency doubling crystal where the two beams at 532 nm interfered. Depending the phase relationship between the beams they either interfered destructively or constructively.

ϖ

2 ϖ

1064nm ϖ + 2ϖ

hotplate

Fiberϖ

2 ϖ

ϖ + 2ϖ 2ϖ

2ϖ Fig 5.7. Experimental setup used for measuring the temperature of a fiber. 532 nm radiation generated with a frequency doubling crystal was launched into a pure silica fiber together with the fundamental of a Nd:YAG laser (1064 nm). At the output of the fiber some of the 1064 nm radiation was converted to 532nm using a second crystal. The interference between green generated before and after the fiber was measured. Since the change in refractive indices with temperature are different at 1.06 µm and 0.532 µm, a change in temperature modulates the output intensity in the setup (eq.5.3.1). The phase modulation then transforms into an amplitude modulation at the output of the second crystal. Since both beams travel the same physical path the interferometer is very insensitive to mechanical and thermal fluctuations. The introduced refractive index difference is almost linear if the temperature interval is small and can be expressed as:

where nω is the refractive index at the fundamental wavelength, n2ω is the refractive index at the SH wavelength and k is a constant. The temperature was measured with an external thermometer. As can be seen in fig. 5.8 the amplitude modulation goes through a total of 20 periods as the temperature drops from 147 -> 63 oC, that is 4.2 oC / period. If the amplitude difference in one separate period can be determined to within 10 discrete steps it is possible to determine the temperature in the fiber with an accuracy of 0.42 oC. The length of the heated fiber was ~15 cm. A fiber twice the length would go through one interference period in half the temperature change. Once the fiber is calibrated the temperature in the fiber can be found by measuring the starting temperature and counting the interference fringes.

5.3.1 knndTd

≈− )( 2ωω

0 100 200 300 400 5000,480,500,520,540,560,580,600,620,640,660,680,70

Interfecence fringes

Interference between green generated from IR before and after a dep cladding fiber

SH

G [a

.u.]

Time [s]

60

80

100

120

140

160

Temperature

Tem

pera

ture

[o C]

Fig 5.8.The resulting interference pattern from SH generated before and after the fiber. Heating the fiber introduces a phase shift, which moves the fringes of the interference pattern. 5.4 Poling experiments. Most reports on fiber modulators found in the literature describe setups were HeNe lasers (633 nm) are used to probe the nonlinearity. The main reason for using a HeNe laser is to reduce the loss in the setup. At a shorter wavelength (compared to the standard telecom wavelength), the optical mode is more confined to the core and the loss is greatly reduced. A fiber suitable for telecom applications usually has a cutoff wavelength of 1200 nm, so at 633 nm these fibers can support more than one propagating mode. When modulating a fiber supporting many modes it is difficult to distinguish between phase modulation and mode modulation. Since the GLAMOROUS project aims at telecom applications we tried to characterize our components at longer wavelengths, preferably 1550 nm. However, some fibers had too high loss at 1550 nm and 980 nm had to be used to get a detectable signal. One of the aims of the GLAMOROUS project is to make a switch with a response time faster than 100 µs. A setup was built for testing the rise time of a fiber component. This fiber had 0.1 ppm of alkali impurities and a cross section which is shown in fig 4.6. The fiber was poled at 2.6 kV and 260 oC. The voltage was ramped to 2.6 kV during 5 minutes and kept at 2.6 kV for 10 minutes for a total poling time of 15 minutes. The electrodes used were gold-coated Tungsten wires manually inserted into the fiber though polished holes on the side of the fiber. The total length of the fiber where the electrodes overlapped, which is the active part of the device, was 11 cm long. The light source was a 1550 nm linearly polarized laser diode which was coupled through an erbium doped fiber amplifier (EDFA) and then into the poled fiber component using a bare fiber holder. The output of the fiber component was coupled to a standard telecom fiber using another bare fiber holder and then fed into a fiber coupled polarization splitter with one of the output arms connected to a 2.5 Ghz detector. 20 ns square voltage pulses with an amplitude of 400 V were applied to the electrodes in the fiber and the optical response was measured. The refractive index change in the two optical axis was different for the same applied voltage and hence the

optical response to a voltage pulse could be chosen to give a positive or negative response at the detector depending on the polarization state fed into the fiber component. When the voltage pulse was switched our measurement system picked up electrical noise that completely stopped us from detecting the real optical response. To overcome this problem a 5 km delay fiber was spliced between the polarization splitter and the detector. This gave a delay of ~26 µs between the switching of the voltage pulse and the arrival of the modulated optical signal. This time was long enough for all electrical noise to decay. The optical response of the component is shown in fig. 5.9. The risetime of the component is ~6 ns which is much faster than the 100 µs required in the GLAMOROUS project. The oscillations in the signal mainly arise from the electrical reflections caused by the impedance mismatch between the coax cable transporting the voltage pulse from the voltage generator to the fiber component and the electrodes in the fiber.

980 990 1000 1010 1020 1030-1,0-0,50,00,51,01,52,02,53,03,5 26 us fiber delayline

Intensity response from 400V 20ns pulse. 2.5Ghz detectorF030402-1G 11 cm long wolfram wires 2.6kV 15 min

Diff

eren

tial i

nten

sity

[a.u

.]

Time [ns] Fig 5.9. The optical response from a 20 ns square voltage pulse applied to the electrodes in a fused silica fiber. The oscillations in the optical signal are due to multiple reflections of the voltage pulse since the electrical connections were not properly impedance matched. The fiber was poled in air for 15 minutes at 2.6 kV. Not only the speed of the fiber component, but also the switching voltage Vπ is important. The GLAMOROUS project aims at delivering a switch with Vπ = 50 V. The same fiber as used for the rise time measurement described above was used for measuring Vπ. Instead of wire electrodes the fiber was filled with an AuSn alloy (80% Au, 20% Sn) with a melting temperature of 282 oC. The fiber was polished from the side to expose the electrodes and then thin wires were glued to the electrodes with conducting epoxy. The fiber had a fused silica cladding with 0.1 ppm alkali impurities and a poled region that was ~12 cm long. Since the primary coating of the fiber becomes black when the fiber is heated to 300 oC it was impossible to tell exactly where the electrodes started and ended and the accuracy of the 12cm is to within ± 1 cm. The fiber was poled at 260 oC with the voltage ramped to 2.3 kV during the first 5 minutes (in order to minimize the risk of dielectric breakdown) and then kept at 2.3 kV for another 10 minutes for a total poling time of 15 minutes. The poled

fiber was placed in one arm of a Mach-Zehnder interferometer built with two 50/50 fiber couplers. A tunable 1.5 µm laser from was connected to the input coupler of the Mach-Zehnder setup. The two arms of the coupler at the output side of the interferometer were coupled into a differential detector/amplifier (New Focus 2017 Nirvana). One arm in the Mach-Zehnder interferometer was ~15cm longer than the other so the bandwidth of the Mach-Zehnder was very limited, meaning that a small change in wavelength led to a modulation of the output intensity of the interferometer. The wavelength of the input laser was tuned as to give a maximum modulation for a given voltage. The optical response is shown in fig 5.10. The maximum optical power just to the right of the voltage peak and the minimum optical power close to the minimum voltage defines a phase modulation of the light with π radians and the voltage between the maximum in the intensity just to the right of the voltage max and the minimum, right at the bottom of the voltage is curve is Vπ=1.59 kV. To the best of our knowledge this is the lowest (perhaps only) value obtained at 1.5 µm.

-5 -4 -3 -2 -1 0 1 2 3 4 5 6

0.0

0.5

1.0

1.5

2.0

Vol

tage

[kV]

Time [ms]

-5 -4 -3 -2 -1 0 1 2 3 4 5 6-10

0

Mach-Zehnder respons of f031023-1C, ~25 cm long

Por

t #1

Inte

nsity

[a.u

.]

Fig 5.10. The optical response of a Mach-Zehnder interferometer with a poled fiber in one of the arms. The applied triangular voltage modulates the phase of the light in the fiber to switch the output from one arm to the other. The switching voltage Vπ is determined as the voltage needed to go from a maximum to a minimum in intensity. In this case 1.59 kV. 5.5 Ablation of silver electrodes The maximum nonlinear coefficient obtained in thermally poled optical fibers is ~1 pm/V. Because of the possible long length of components made from optical fibers this value is more than enough for making an all-fiber optical switch. When making a wavelength converter it is not only the nonlinear coefficient that determines the efficiency of the device, but also the phase matching properties. Since different wavelengths experience

different refractive indices the spatial length over which the process is constructive is often limited to tenths of micrometers. To make an efficient device the process must be phase matched. This has been achieved in different ways, e.g. by quasi phase matching (QPM) (see chapter 2.4) in D-fibers with periodic electrodes created with photolithography [8]. We have invented a new way of doing QPM in fibers. A thin film electrode is created on the inside of the holes in a twin hole fiber. The film is then periodically ablated to generate a periodic electric field in the fiber when a voltage is applied to the electrodes. The electrodes are deposited by mixing two solutions, one containing dextrose solved in H2O and the other containing AgNO3 and NH3 solved in H2O. When the two solutions are mixed the silver ions are reduced to atomic silver and deposit on any surface in contact with the solution. The two solutions are mixed and pumped into the holes of the fiber. As soon as the solutions are mixed different waste products are formed. The particles quickly grow larger than the diameter of the holes in the fiber (~30 µm) and therefore clog them. To overcome the clogging problem a Y-junction was built where the two solutions are mixed just before entering the fiber. This limited the size of the particles and the solution could flow for several hours before the holes were clogged. This time was sufficiently long to form a silver film (1µm thick) that was conductive for fiberlengths > 2 m. Setting up the Y-junction cell was rather time consuming so the two solutions were cooled down in a fridge (~8 OC), mixed and kept at this temperature during the whole filling process. The lower temperature of the solution slowed down the reaction speed so that the size of the waste particles was kept small. One end of the fiber was kept in the fridge, but the main part of it was kept at room temperature where the reaction speed was much faster. It takes only ~30 seconds for the solution to travel through the fiber and this time is much shorter than the time it takes to form particles larger than the hole diameter. There was a sharp gradient in the thickness of the silver film in the region where the fiber left the fridge caused by the change in deposition speed with temperature. With this filling technique we are able to make conducting electrodes up to 3 m long. Typically the deposited film have a resistance of 1000 Ω/m and. it is possible to splice the fiber to a standard telecom fiber without removing the metal. Once the electrodes are created they are structured by periodically ablating them from the side using a frequency doubled mode-locked Q-switched Nd:YAG laser giving 0.532 µm radiation. The average power needed for ablation varies with the thickness of the silver film. Typically 17 mW focused with a 30x microscope objective was sufficient. Every pulse train carried 1/3 of the total power and 2/3 was carried in the prelase. Typically the repetition rate of the Q-swithced pulses was set to 1 kHz and gave ~20 pulses, ~150 ps long in each Q-switched pulse. An attenuator was used to lower the power so that the system could be aligned without ablating the electrode. The laser beam was then scanned along the fiber and the periods written by periodically exposing the electrode to the beam during less than 1 s. After ablation the electrodes were still continuous and conducting since only the part of the electrode closest to the core was ablated as shown in fig 5.11. A picture of the ablated electrode is shown in fig 5.12.

Ablating beam

Coating Ag electrodes

After ablation

Fig 5.11. The left pictures shows how the ablating laser beam removes a part of the electrode. After ablation the electrode is still conducting since the part of the electrode the farthest from the core is untouched. The right picture shows the silver film inside one of the holes prior to ablation.

Core and innercladding

Ag electrode

Ablated spot

40 µm

Fig 5.12. A fiber with periodically ablated silver film electrodes. The period here is 40 µm, but it can easily be changed simply by translating the beam farther between each ablated spot. In order for the spatial resolution of the electric field to be good the resistance in the ablated spots must be very high. To measure the increase in resistance introduced when ablating a silver film a bulk experiment was done. A silver stripe, 20 mm long and 1.5 mm wide was deposited on a fused silica glass disk with a mask, 1.5 mm thick and 25 mm in diameter (fig 5.13). The silver film was contacted with thin gold plated Tungsten wires (20 µm diameter) attached to the film with conducting silver epoxy. The silver film was connected in series with a current meter and a 19.7 MΩ resistor. A voltage was applied to the circuit and the current was monitored. The focused beam of a frequency doubled Nd:YAG (the same used when ablating the fiber electrodes) was scanned across the film to cut it off. Two

experiments were made. In the first one the 19.7 MΩ resistor was shortcut and beam scanned in steps of 100 µm across the film while the resistance through the film was monitored. During the first 16 steps the resistance was 47 Ω, ablating another 100 µm increased the resistance to 48 Ω. During the last 100 µm the resistance increased to more than 2 MΩ. This indicates that only a very small fraction of the cross section is needed to support the small currents during poling. In the second experiment 4 ablations were made across the entire film while the current was monitored. The resistance of the silver film was found using Ohm’s law. Each new ablated stripe on average introduced an extra 4.55 MΩ of resistance to the circuit. Using an optical microscope the ablated stripes were estimated to be 45 µm wide. The small current measured after the film was ablated mainly flows through the glass sample on which the film is deposited. Since the resistance increases dramatically when the electrode is ablated this technique can be used to define an electric field in the fiber with good spatial resolution.

U =230V R =19.7M Ω

Fused silica d iskW ith electrode15x1,5m m .

A

A bla ted lines 1 .5 m m apart

45 µm w ide

Fig 5.13. Setup for measuring the resistance introduced when ablating stripes across a thin silver film deposited on a fused silica substrate. The current is monitored while the silver strip is ablated. For each ablation the resistance of the circuit increases by 4.55 MΩ..

6 Conclusions Optical fibers have a great potential to take over some of the functions that are performed by nonlinear crystals in today’s optical networks. The main advantage of using optical fibers is the possibly low cost and the ease of splicing components to standard transport fibers. The two functions that are dealt with specifically in this thesis are wavelength conversion and optical switching. Wavelength conversion was demonstrated in a fiber with periodic electrodes. The efficiency of the device is mainly determined by the number of periods and the precision of the periods, both with respect to the precision in the definition of the electrodes, and with respect to smearing of the electric field in the glass due to surface current and fringing fields. The work on optical switching has had major problems because of the extremely high attenuation in the available fibers (~2.5 dB/cm) at 1.5 µm wavelength. Some results have been obtained and at present the lowest switching voltage obtained is ~1600V. This value is far from the goal of 50 V. The measured switching voltage in the components produced so far has been almost independent of the poling procedure. There are many possible reasons why this happens, the first one being that the fiber is not poled e.g. since there are too many impurities injected into the glass on the anode so that no depleted region is formed. The second possible reason is that the formed nonlinear layer does not overlap with the core. Since the conductivity of the glass is lower in the region depleted of alkali ions this could give a high resistivity region outside of the core region were most of the voltage drop occurs. If this happens poling can actually increase the voltage needed for switching. The main effort in the future will be on decreasing the switching voltage. This will be done by developing new fibers with lower attenuation and the core placed at optimum distance from the hole in order to increase the overlap with the poled region. A new stabilized Mach-Zehnder interferometer will be constructed so that the effect induced by poling the fiber can be studied in real time. This will give a much better understanding of the poling process. Up until now the fibers have only been characterized before and after poling and no conclusion about the poling dynamics have been made.

References

Reference List

1. P.N.Butcher, D.Cotter: The Elements of Nonlinear Optics. Camebridge University Press, 1990.

2. M.V.Klein and T.E.Furtak,. "Optics". (1986), John Wiley & Sons.

3. G.P.Agrawal: Nonlinear Fiber Optics. Academic Press, 1995.

4. I.C.S.Carvalho, W.Margulis, and B.Lesche,. "Second-Order Nonlinear Effects in Optical Fibers". (2003)

5. David N.Nikogosyan: Principles of Optical and Laser-Related Materials - A Handbook. John Wiley & Sons Ltd, 1997.

6. W.Margulis, F.Laurell, and B.Lesche,. Nature , Vol: 378, pp 699. (1995)

7. P.G.Kazansky and V.Pruneri,. "Electric-field poling of quasi-phase-matched optical fibers". Optical Society of America , Vol: 14(11), pp 3170-3179. (1997)

8. V.Pruneri, G.rate, P.G.Kazansky, D.J.Richardson, N.G.Broderick, J.P.de Sandro, C.Simonneau, P.Vidakovic, and J.A.Levenson,. "Greater than 20%-efficient frequency doubling of 1532-nm nanosecond pulses in quasi-phase-matched germanosilicate optical fibers". Optics Letters , Vol: 24, pp 208-210. (1999)

9. R.Kashyap,. "Fiber Bragg Gratings". (1999), Academic Press.

10. A.J.Ikushima, T.Fujiwara, and K.Saito; Silica glass: A material for photonics. Applied Physics (2000), Vol:88(3), 1201-1213.

11. C.Chang-Hasnain,. "Optical fiber ". (2002), EECS 233.

12. R.A.Myers, N.Mukherjee, and S.R.J.Brueck; Large second-order nonlinearity in poled fused silica. Optics Letters (1991), Vol:16, 1732

13. P.G.Kazansky, A.R.Smith, and P.S.J.Russell; OSA Technical Digest Series (1995), Vol:22, 175

14. L.Canioni, M.O.Martin, B.Bousquet, and L.Sarger; Optics Communication (1998), Vol:151, 241

15. F.Haberl, J.Hochreiter, J.Zehetner, and A.J.Smith; Electrical breakdown in ge-doped silica glass fibers. International Journal of Optoelectronics (1990), Vol:4(5),

16. A.Kudlinski, G.Martinelli, Y.Quiquimpois, and H.Zeghlache,. "Microscopic model for the second-order nonlinearity creation in thermally poled bulk silica glasses". pp TuC3. (2003)

17. D.Faccio, V.Pruneri, and P.G.Kazansky,. "Dynamics of the second-order nonlinearity in thermally poled silica glass". Applied Physics Letters , Vol: 79(17), pp 2687. (2001)

18. T.G.Alley, S.R.J.Brueck, and R.A.Myers,. " Space charge dynamics in thermally polied fused silica". Journal of Non-Crystalin Solids , Vol: 242, pp 165. (1998)

19. T.G.Alley, S.R.J.Brueck, and M.Wiederbeck,. "Secondary ion mass spectroscopy study of space-charge formation in thermally poled fused silica". Journal of Applied Physics , Vol: 86(12), pp 6634 . (1999)

20. B.Lesche, F.C.Garcia, E.N.Hering, W.Margulis, I.C.S.Carvalho, and F.Laurell; Etching of Silica Glass under Electric fields. Physical Review Letters (1997), Vol:78(11), 2172

21. A.L.C.Triques et.al.; Depletion region in thermally poled fused silica. Applied Physics Letters (2000), Vol:76(18), 2496

22. X.-C.Long, R.A.Myers, and S.R.J.Brueck; A Poled Electrooptic Fiber. IEEE Photonics Technology Letters (1996), Vol:8(2), 227

23. N.Myrén, M.Fokine, O.Tarasenko, L.Nilsson, and W.Margulis,. "Periodic internal fiber electrode fabricated by laser ablation". Vol: BGPP 2003 Conference Preceedings, pp TuC1. (2003)

24. A.Podlipensky, J.Lange, G.Seigfert, H.Graener, and I.Cravetchi; Second-harmonic generation from ellipsodial silver nanoparticles embedded in silica glass. Optics Letters (2003), Vol:28, 716

List of publications I. ”Widely tunable fiber-coupled single-frequency Er-Yb:glass laser” G. Karlsson, N. Myrén, W. Margulis, S. Tacheo, F. Laurell. Applied Optics, Vol. 42 Issue 21 Page 4327 (July 2003) II. “Comparison of characterization techniques and the effect of surface condition in poled silica” C. S. Franco, G. A. Quintero, N. Myrén, A. Kudlinski, H. Zeghlache,H. R. Carvalho, A. L. C. Triques, D. M. González, P. M. P. Gouvêa, G. Martinelli, Y. Quiquempois, B. Lesche, I. C. S. Carvalho, W. Margulis. Submitted to Journal of Applied Physics, November 2003 III. “In-fiber electrode lithography” N. Myrén, W. Margulis, M. Fokine, L. Kellberg, O. Tarasenko. Submitted to Optics Letters, November 2003 Additional publications by the author: IV. “Periodic electrodes fabricated by laser ablation” N. Myrén, W. Margulis, M. Fokine, L. Kellberg, O. Tarasenko. Conference on Bragg grating, poling and photosensitivity (BGPP), Monterey, California, September 2003. V. “Comparison of characterization techniques and the effect of surface condition in poled silica” C. S. Franco , G. A. Quintero, H. R. Carvalho, I. C. S. Carvalho, D. M. González, P. M. P. Gouvêa, A. L. C. Triques, B. Lesche, N. Myrén, W. Margulis, G. Martinelli, Y. Quiquempois, A. Kudlinski, H. Zeghlache. Conference on Bragg grating, poling and photosensitivity (BGPP), Monterey, California, September 2003. VI. “Tuning of an Er-Yb:glass mini-laser using external feedback from fiber Bragg grating” N. Myrén, W. Margulis, G. Karlsson and F. Laurell, S. Taccheo. Conference on Lasers and Electro-Optics (CLEO) Los Angeles, California, May 2002