Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Integrated optics asymmetric march-zehnder high-voltage sensors Huang, Lisheng 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-0009.pdf [ 2.26MB ]
JSON: 831-1.0065105.json
JSON-LD: 831-1.0065105-ld.json
RDF/XML (Pretty): 831-1.0065105-rdf.xml
RDF/JSON: 831-1.0065105-rdf.json
Turtle: 831-1.0065105-turtle.txt
N-Triples: 831-1.0065105-rdf-ntriples.txt
Original Record: 831-1.0065105-source.json
Full Text

Full Text

Integrated Optics AsymmetricMach-Zehnder High-Voltage SensorsbyLISHENG HUANGB.Sc., Zhongshan University, P.R.China 1985M.Sc., University of Washington, U.S.A. 1990A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTIlE FACULTY OF GRADUATE STUDIES(The Department of Electrical Engineering)We accept this thesis as conformingto the required standardTHE UMVERSITY OF BRITISH COLUMBIAApril 1993©Lisheng Huang, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.___________________________Department of_____________________The University of British ColumbiaVancouver, CanadaDate 4vr.\2 t993DE-6 (2/88)AbstractTwo types of integrated optics Mach-Zehnder (IMZ) interferometers, the strip-loaded IMZ and the domain-inverted IMZ, to be used for high-voltage measurements,have been investigated in this thesis. The devices are designed to have an asymmetrybetween their two branches so that the electro-optic responses in two branches aredifferent when immersed in equivalent electric fields. The intensity of the output lightreflects the strength of the electric field in which a device is immersed.In the strip-loaded IMZ the asymmetry is achieved by loading a Ti02 film ontoone of the two branches, while in the domain-inverted IMZ the asymmetry is realizedby inverting the spontaneous polarization in one of the branches. The operatingprinciples, and the designs of the devices, are presented. The mechanism for domaininversion is also studied, and a defect process is proposed for the Ti-induced domaininversio in the c side of LiNbO3.The devices have been fabricated using the Ti-indiffusion method. The lithiumout-diffusion, associated with the fabrication, is also discussed, and a pre-out-diffusionmethod is proposed to resolve the associated surface waveguiding problem. The testresults of both device types are given, in particular, the measurements of the basicdevice parameters, i.e., the half-wave voltage, the intrinsic phase, and the on/offratio. The results of stability tests and impulse tests are also presented. Also, for thefirst time in our laboratory, the annealed proton-exchange method has been used tofabricate optical waveguides.11Both device types were designed for use as high-voltage sensors. However, thedomain-inverted IMZs offer some possible advantages as high-speed modulators.111Table of ContentsAbstract 11Table of Contents.ivList of Figures viList of Tables ixAcknowledgementsChapter 1 Introduction 1Chapter 22.12.2Device DesignIntroductionBasic Concepts .2.3 Strip-Loaded Mach-Zehnder High-Voltage Sensor2.3.1 Working Principle2.3.2 Device Design2.4 Domain-Inverted Mach-Zehnder High-Voltage Sensor2.4.1 Working Principle2.4.2 Device Design556171727Chapter 33: Testing and Measured ResultsIntroductionExperiment SetupDevice Parameters3636374042455152534.4 Results for the Strip-Loaded Mach-Zehnder High-Voltage Sensor . 614.5 Results for the Domain-Inverted Mach-Zehnder High-Voltage Sensor 694.5.1 Measurement of Basic Device Parameters 704.5.2 Linearity Measurements 76Device FabricationIntroductionOptical Waveguide Fabrication in LiNbO3Surface GuidingMask DesignBasic Fabrication ProceduresFabrication of Strip-Loaded Mach-ZehndersFabrication of Domain-Inverted Mach-Zehnders3.8 Fabrication of Annealed Proton-Exchanged WaveguidesChapter Impulse Measurements.794.5.4 Stability Tests 834.5.5 Comments on the APE Method 86Chapter 5 Summary and Recommendations 895.1 Introduction 895.2 Summary 895.3 Recommendations for Future Work 91References 96VList of FiguresFigure 2.1 Schematics of a strip-loaded IMZ:(a) a plan view, (b) a cross-sectional view 9Figure 2.2 Calculation of the effective refractive index for a channel waveguide . 10Figure 2.3 Bessel functions indicating the number of modes guided inan exponential slab waveguide 14Figure 2.4 Effective index as a function of the loading-strip thickness 16Figure 2.5 The ionized defect concentration from Ti indiffusion 25Figure 2.6 The electric field calculated for the Ti concentration gradient 26Figure 2.7 Schematics of a domain-inverted Mach-Zehnder interferometer:(a) a plan view, (b) a cross-section view 29Figure 2.8 The doped Curie temperature, lateral profile, resulting from theindiffusion of a 32gm wide inversion strip and a 3.5gm widewaveguide 30Figure 2.9 The doped Curie temperature, depth profile, resulting from theindiffusion of a 32gm wide inversion strip and 3.5m widewaveguide 31Figure 2.10 The doped Curie temperature, lateral profile, resulting from theindiffusion of a 3.5gm wide inversion strip and a 3.5m widewaveguide 33Figure 2.11 The doped Curie temperature, depth profile, resulting from the‘indiffusion of a 3.5jm wide inversion strip and a 3.5m widewaveguide 34Figure 3.1 Schematics for the domain-inverted IMZ mask 44Figure 3.2 Alignment of the waveguide on an inversion strip 49Figure 3.3 A output spot of an annealed proton-exchanged waveguide 55viFigure 4.1Figure 4.2Figure 4.3Figure 4.4Figure 4.5Figure 4.6Figure 4.7Figure 4.8Figure 4.9Figure 4.10Figure 4:11Figure 4.12Figure 4.13Figure 4.14Figure 4.15Figure 4.16Figure 5.16465676871727377788081828485878894TE0 field profile without the loading stripTE0 field profile with a l2Onm thick Ti02 stripCoupling coefficient as a function of loading-strip thicknessTopograph of the coupling coefficientsOscilloscope images of responses of(a) device 82 and (b) device 91Fitted transfer function for device 82Fitted transfer function for device 90The standard error in fitting a straight line to a transfer functionas a function of the input rangeThe standard error in fitting a straight line to a transfer functionas a function of bias with a fixed input range of 20 degreeOscilloscope images of the linear responses of(a) device 89 and (b) device 91Linear fit of unfiltered data for device 91Linear fit of filtered data for device 91Oscilloscope images of impulse responses for device 90:(a) l5s rise time and (b) 1.5s rise timeIntrinsic phase drift with temperature for devices 90 and 94Intrinsic phase drift with time for device 91An oscilloscope image showing the flattenning for high voltages,taken for device 91A phase-reversal, quasi velocity-matching electrode and a waveguideviiFigure 5.2 A domain-inverted IMZ used to make quasi velocity-matchinghigh-speed modulator that may be combined with a number ofstandard microwave transmission lines 95vi”List of tablesTable 3.1 Waveguide parameters for the new mask 44Table 4.1 Measured device parameters for domain-inverted IMZs 75ixAcknowledgementsMy deepest gratitude goes to my parents and my wife for their love andsupport.I would like to express my gratitude to my supervisor, Dr. N.A.F. Jaeger, forproposing this project and providing support during the course of the work.I am most grateful to Z.Lee, B.Tsou, and M.Chen for invaluable discussionsregarding my work. I would also like to thank H. Kato for his helpful assistance inthe device fabrication. My gratitude goes to C. Backhouse for his help with theellipsometry measurements and to D. Hui for his help in device testing. My thanksextend to all those individuals in the solid state group for their help in various ways.I would like to thank the Science Council of British Columbia for the generousfinancial support for this project.xChapter 1 IntroductionSignificant research aimed at developing optical methods to measure voltagesand currents in high-voltage environments began in the 1970’s [1,2,3]. Theadvantages of optical systems, over conventional potential transformer (PT) andcurrent transformer (CT) systems, include: immunity to electro-magnetic interference,high accuracy, a non-intrusive nature, small size, large bandwidth, and, potentially,low cost. These advantages make the optical systems attractive alternatives to theconventional ones.In an optical measurement system typically an optical signal is sent to thesensing elements through either free space or optical fibres. This signal is thenusually modulated either by the electro-optic or by the magneto-optic effects,whichever is exhibited by the sensor. The modulated signal is then transmitted todetectors again through free space or optical fibres. The signal, modulated in eitherphase or in intensity, reflects the strength of the field being measured.In the development of bulk optical systems different electro-optic andmagneto-optic materials have been used as the sensing elements. For example,LiNbO3,LiTaO3,KDP, RDP, CdTe, and Bi4Ge3O12 crystals have been used forPockels effect sensors [2,4,5,6,8], Bi12SiO20 YIG, and Flint glass for Faraday effect1sensors [2,6,7,8,9], and Nitrobenzene, and chiorinate-biphenyl for Kerr effect sensors[10,11]. In most of these bulk devices the optical signals were phase modulated.Interferometers were constructed for Kerr cells and Faraday cells[1 1], however, thesetups were relatively large needing bulky lensing systems. The recent developmentsin integrated optics offer a variety of new ways to design such devices. Besidechoosing different materials with different orientations, as in the bulk sensor designs,it is now possible, with integrated optics, to actively design sensors using differentconfigurations for the waveguides and electrodes, the modal coupling betweenwaveguides, the difference in the changes of optical mode propagation constants, andeven the possible modification of the electro-optic properties of the substrate materials.This flexibility in designing integrated optics sensors has been demonstrated by theprogress made in the development of such devices here at the University of BritishColumbia. A number of optical sensors have been proposed and/or demonstratedsince the inception of this program in the early 1980’s[12,13, 14,15]. Particularly, astate-of-the-art integrated optics Pockels Cell voltage sensor, has been demonstrated byJaeger[12]. Several integrated optics Mach-Zehnder (IMZ) interferometer basedsensors have also been fabricated[13, 141. A capacitive voltage divider wasmonolithically integrated with an IMZ combining an otherwise huge voltage dividerwith an optical sensor on a small piece of wafer[14]. In the Bifurcate Optique Active(BOA) sensors proposed by Ahmed[13] optical modal couplings were utilized in thesensing. A strip-loaded IMZ sensor was also proposed[15], in which a loading strip is2deposited on one of the two branches of an IMZ, creating an asymmetric electro-opticresponse between the modes propagating in the two branches.For the integrated optics sensors mentioned above LiNbO3 was the primarycandidate for the substrate material, mainly due to its large linear electro-opticcoefficients and chemically stability. More importantly, the methods for fabricatingoptical waveguides are well developed for this material. Lithium niobate is aferroelecthc material whose Pockels effect is basically a Kerr effect biased by thespontaneous polarization[16]. Thus the electro-optic response may be reversed if thespontaneous polarization can be reversed. A novel design for an asymmetric IMZhigh-voltage sensor, utilizing such a change in the spontaneous polarization in selectedareas of the substrate, is described in this thesis.In this thesis we present studies of two asymmetric integrated optics IMZ ashigh-voltage sensors, i.e., the strip-loaded IMZ and the domain-inverted IMZ. Inboth devices the two branches of the Mach-Zehnder are designed to have anasymmetric electro-optic response. The operating principles of the strip-loaded andthe domain-inverted Mach-Zehnders, as well as the device designs are presented inChapter 2. the methods and mechanisms of the domain inversion process are alsodiscussed.In Chapter 3 the schematic of the mask used for the domain-inverted IMZs isshown, then optical waveguide fabrication techniques are summarized, lithium outdiffusion problem associated with the Ti:LiNbO3waveguide fabrication is discussed,3and the proton-exchanged method, which was used to make optical waveguides isdescribed.The results of tests performed on the devices are presented in Chapter 4. Basicdevice parameters such as the half-wave voltage, the intrinsic phase (bias), and theon/off ratios were measured and the results are presented. The electro-optic responsein the domain-inverted region is shown to be opposite to that in the uninverted region.The linearity of the device is calibrated, and the stability of the device’s performanceis investigated.A summary of the present work and suggestions for future work are given inChapter 5. The possibility of applying the domain-inversion technique in high-speedoptical modulators is discussed and new device configurations are proposed.4Chapter 2 Device Design2.1 IntroductionIn this chapter two immersion-type electric-field sensors are described. Theyare both based on the integrated optics version of the Mach-Zehnder interferometer(TMZ) in which a structural asymmetry is created between the two branches of thedevices. Due to this structural asymmetry, a different change in the phase velocity ofthe light in each of the branches occurs when both branches are subjected toequivalent electric fields. This difference in the change in phase velocity results in anoutput intensity change that can be related to a uniform field in which such a device isimmersed (hence the name: immersion-type device).In section 2.1 the basic concepts of operation, governing the behaviour ofboth device types, are presented. The operating and the design principles for thestrip-loaded UIZ, with a Ti02 loading strip on one of its branches, are described insection 2.2. The theory and the design principles for the domain-inverted IMZ, inwhich the spontaneous polarization of the substrate in a section of its branches isinverted, are presented in section Basic ConceptsWhen immersion-type, high-voltage sensors, based on IMZs, which have twostructurally different branches, are immersed in uniform electric fields, the changes inthe propagation constants of the modes will be different in the two branches and,therefore, the retardation, or the phase delay, of the modes will also be different afterpassing through the two branches of such a device. The phase difference that willresult after the modes each propagate a distance L is given by[15]t4=(C2-C1)nrpEL/2 (2.1)where n6 is the refractive index of the substrate, r is the relevant electro-opticcoefficient, Ea is the applied electric field in air, and p is the reduction factor of theelectric field in the device due to the depolarization fields in the dielecthcs. Here C1and C2 are the constants of proportionality between the change in the propagationconstants in each of the branches and the change in the refractive index of thesubstrate, caused by the applied electric field. After combining the light in the twobranches in the output section of the waveguide, the intensity of the output willdepend upon the relative phase shift between the modes in the two branches. Theoptical output power from the device P is then given by[15]P0,=Pcq2{ 1 +ycos[(C2-1)nrpEJ2-i-4 11 (2.2)where is the input power, is the intrinsic phase difference (i.e., the phase6difference between the modes in the two branches before the device is immersed in thefield), and the constants a and y would be determined for a particular design. Theoptical output power intensity of the IMZ reflects the field strength in which thedevice is immersed, hence the device can serve as an electric-field sensor.2.3 Strip-Loaded Mach-Zehnder High-Voltage Sensors2.3.1 Working PrinciplesThe strip-loading method is one of the early methods for fabricating channelwaveguides in LiNbO3[17]. In this method, a channel waveguide is formed bysimply depositing a loading film on a planar waveguide[ 18,19]. Films with refractiveindices lower than the substrate’s were originally used. Films with higher refractiveindices than the substrate’s were eventually used by Uchida on LiNbO3 in order toachieve a large loading effect, i.e., large change of the effective refractive index[20].Noda et a!. demonstrated a strip-loaded channel waveguide based on a graded-indexplanar waveguide[21j. The strip-loading method was shown by Jaeger et at. to becapable of controlling the propagation constant in channel waveguides, by loading thewaveguide with a high-index strip and, thereby, controlling the electro-optic responseof a particular mode therein[15].The devices, based on the IMZ, proposed by Jaeger et al. consist of one7diffused branch being covered by a high refractive index loading strip while the otherbranch remains unloaded(see Figure 2.1). The optical mode profile is affected by thepresence of the loading strip. The presence of the high-index loading strip causes theoptical field, of a particular mode, to shift towards the loading strip, and, as thethickness of the strip increases, more light is guided in the loading strip and lessoptical field in the electro-optic substrate. Therefore, the portion of the optical fieldin the electro-optic substrate region can be controlled by the thickness of the loadingstrip. Hence, the constants of proportionality, C1 and C2 (the subscripts 1 and 2 referto the branch), will be different in the two branches.2.3.2 Device DesignThe calculation of the propagation constant of the fundamental mode of achannel aveguide under the loading strip is carried out using the effective indexmethod[22] shown in Figure 2.2. First, an exponential profile is used to approximatethe refractive index distribution in the depth direction, the z-direction in our case.The profile is treated as a slab waveguide consisting of 3 layers. The propagationconstant, and thereby the effective index n1, is calculated. Secondly, the effectiveindex method is employed to calculate the guiding properties in the lateral direction toobtain the effective index n11 of the channel.There are two major types of optical field in the waveguide, the TE-like andTM-like modes. With the approximation of TE-like and TM-like modes and the8A p[n view oF o sripIoode ñch1eIuier tnLerFeromeerA crossseclion of Lhe wo branchesLtN6O Subsi..rL.Ti. ir,diFf’._,s.d wc,’.’.Qutd.Ti02 1eedirg a1rtp(hiFigure 2.1 Schematics of a sthp-loaded IMZ: (a) a plan view,(b) a cross-sectional view.(i9The ccuIeLicn ci fhe eff’ecLve index For t chnne1wveguide cn be perFormed in 2 sleps:nsni niSIpe 1. C1cu1Le Ihe eliecLive indexn For Ihe Ldtinensiontl wveguideSlep 2. Use n1 Ic clculôIe IheefFective index n11 of the chennelwevegu ide.Figure 2.2 Caiculation of the effective refractive index for a channel waveguide.nin. +n10assumption that the refractive index changes slightly in the regions considered, thevector wave equations can be reduced to scalar wave equations[16]E+(n2_n2e00=O (2.3)0H=Owhere the n is the lateral distribution of the refractive index, neff is the effective indexfor a particular mode, and E and H are the transverse components of the field. Theeigenvalue equation for a TE or TM mode of the strip-loaded diffused waveguide withan exponential profile is given by[21]_______________2h S—tan(ht) (2.4)k0(2nAn)” 1 +Stan(ht)wherek0=2ii/A• g(O)=2dk(2nAn)1’q0(n_n)1f22 2iandJi, for TEmodesn)2/,j, for TM modes11f(n_n,)9(nj_n2, for TE modes2 2 2 21j2 2 2fliItZa)((fli ‘ia) /(nf —n,)), for 7M modeswhere n, flf and na are the substrate, loading strip, and superstrate refractive indices,Jq is Bessel function of order q, and t is the thickness of the loading strip. An and dare the maximum change and the depth of the exponential refractive index profile:n(z)=n+ A nexp(-zld). For the z-cut LiNbO3 substrates obtained from CrystalTechnology Inc., Pato Alto, CA., the refractive indices aren0=2.2869 andne=2.2019 forA0=633nm. Ti02 is chosen to be the loading film because of its highrefractive index as well as our ability to grow high quality rutile films. The refractiveindex of Ti02 obtained from our oxidation process was measured using anellipsometer from O.C.Ruldoph & Son Inc. ,Fairfield, NJ. The ellipsometer was inNull mode, and incident angle of the light, with ) 633nm, was 800. ChrisBackhouse’s ellipsometry analysis program was used to analyze the data, and therefractive index of the oxidation film was found to be 2.575. The refractive index ofthe superstrate Si02 is 1.46, also atA0=633nm[23]. In the following calculations weused these parameters.A single-mode waveguide is generally preferred for good device performance.In order to fabricate single-mode waveguides, the fabrication parameters are examinedas follows. When the loading-strip thickness is set to be zero, which corresponds tothe unloaded portion of the waveguide, the right-hand-side (RHS) of equation (2.4) is122(flifla)/(2nAn)”. For the guided optical modes, n1 is in the rangen<n1<n+An. If An=0.001 and d=3.159j.m then the RHS is 60.8 forTE modes,which is much larger than unity. However, the numerical values of the BesselfUnction Jq[g(O)1 are small as compared to the unity. Therefore, the solution of theequation should occur at Jq[g(O)]zO. If we plot Jq[g(O)] as a function of q, for agiven g(0), the number of points at which Jq[g(O)] =0 will indicate the number ofmodes that the waveguide will support. From Figure 2.3, with the conditiong(0)=5.5, one mode is guided, with a second mode about to cut-in, while withg(0)=5.0 clearly only one mode is guided. So the condition g(0) <5.5 is thecondition needed to ensure single-mode operation for a waveguide with an exponentialprofile.With the above condition, we can design a single-mode waveguide by choosingg(0) = 5, -which means (An)“2d=0.118. The resulting values are An0.001, andd=3.159um, corresponding to a diffusion of a iooA Ti strip at 1050° C for 6.5 hours.However, the mode guided under this condition, will be a weakly guided one. In theactual experiment, 400A of Ti was deposited and the diffusion was carried out at1050°C for 6.5 hours. This results in a change of the ordinary refractive indexAn0=0.0039 and diffusion depth d=3. 159am with a corresponding g(0)=8. 11.Plotting the Jq[8.l 1] shows that two modes are guided in the waveguide. The secondmode, however, is very weakly guided and, for our purposes, this waveguide can stillbe considered as being ‘near’ single moded.130.60 0.500.00 1.00 2.00 3.00 4.00 5.00 6.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00q q(a) (b)0.400.200.00—0.20—0.40 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I0.00 2.00 4.OD 6.00 8.00 1D.00q(c)Figure 2.3 Bessel functions indicating the number of modes guidedin an exponential slab waveguide14After the effective refractive index n1 of the planar structure is obtained, theeffective index method is used to calculate the propagation constant or effectiverefractive index n11 for the channel waveguide. The eigenvalue equation for a TE orTM mode of a slab waveguide istan(ht)= h(p+v) (2.5)(h2-pv)whereh=k0(n-n)’Iko(n_n)ht2, for TE modesV_’(fl/fl2)(fl2_fl2)lI2for TM modesand- J k(n1-.n), for TE modes7M modesIf a TE mode is used in calculating n1, then the TM mode equation should beused to calculate n11 and vice versa. A plot of calculated effective refractive indices ofa TE0 mode, in a strip-loaded channel waveguide, as a function of the loading-stripthickness is shown in Figure 2.4. For simulating the effect of applying an electricfield n is chosen to be 2.2879, 2.2869, or 2.2859, depending on the field strength.The operating points of the device are chosen at A and B corresponding to a loading-152.36G)-o -02.34-E0 -w -F— -2.32-:><ci) --c -Cns=2.2879——ns=2.2869. 2.30 / — - - ns=2.2859o A1:1)VLU -7-2.28 — I I I I I I I I I I I I I I I I I I I I I I I I I I0.00 0.05 0,10 0.15 0.20Loading—strip thickness in micronsFigure 2.4 Effective index as a function of the loading-strip thickness.16strip thickness of 0 and l2Onm, respectively. For a Mach-Zehnder with a loadingstrip of thickness l2Onm on one branch and no loading on the other, usingr33 =30.9x10’2m/V, the calculated value of C2-C1 for the TE0 mode is (C2C1)TE=0.455. And the corresponding value for the TM0 mode is 0.10, calculated usingr13 =9.6x10’2m/V. The resulting half-wave fields (the electric field in the air neededto create a change in phase of 7T radians between the modes in the two branches) areTE=2.6x1OSV/cm and E= 3.8x iO V/cm for the TE0 and TM0 modes,respectively; here we have assumed an immersion-type device having L= 10mm andhave used p =0.05(this value for p is obtained by assuming the depolarization field ofthe sensor substrate is similar to that of a thin disc of dielectric[141).2.4 Domain-Inverted IMZ High-Voltage Sensors2.4.1 Working PrinciplesAs was shown in the previous section, for the strip-loaded Mach-Zehnder, theportions of the optical modes, in the two branches, in the electro-optic substrate, aremade different by placing a high refractive index loading film on one of the branches,achieving a phase retardation difference in the two branches. Another, and perhapsmore elegant, method to achieve the phase retardation difference is by changing theelectro-optic properties of one of the two branches. The changes in the electro-optic17properties may include a degradation, elimination, or reversal of the electro-opticeffect. In the domain-inverted Mach-Zehnder high-voltage sensor, the spontaneouspolarization in one branch is inverted, with respected to that in the other branch, sothat the electro-optic responses in the two branches are opposite to each other whenboth are subjected to equivalent electric fields.Lithium niobate is a man made material, it is a member of the oxygenoctahedra ferroelecthcs. The structure of lithium niobate at temperatures lower thanthe Curie temperature consists of planar sheets of oxygen atoms in a distortedhexagonal close-packed configuration. As for the octahedral interstices of thisstructure, 1/3 of them are occupied by Li ions, 1/3 by Nb ions, and 1/3 areempty[241. In the para-electric phase, at temperatures higher than the Curietemperature, the Li ions are in one of the planes of oxygen ions, and the Nb ions arein the centres of the oxygen octahedra. As the crystal is cooled to the roomtemperature, the Li ions are displaced from the oxygen layers by O.715A and the Nbions from the centres of the octahedra by O.2568A [25,26]; the displacement of theions results in a dipole moment in the unit cell. If the cooling from high temperaturesis performed under the influence of an electric field (or electric current), the poling ofthe dipoles results in a spontaneous polarization and thus in the ferroelectric state ofthe crystal. Poled LiNbO3 is, therefore, a ferroelectric material, with a spontaneouspolarization P of 7OC/cm2at room temperature[27]. It is this spontaneouspolarization that is responsible for LiNbO3 exhibiting the linear electro-optic effect[16,1827]. Theoretically, the linear electro-optic effect of LiNbO3 is fundamentally aquadratic electro-optic effect biased by the spontaneous polarization. The electro-opticcoefficients are given in terms of the spontaneous polarization, for exampler33 =213(g1 +2g12g)P8,where g11, g12, and g44 are the quadratic electro-opticcoefficients of octahedra which have m3m symmetry[16]. Therefore, an inversion ofthe domains, i.e., inversion of the spontaneous polarization, will result in an inversionof the electro-optic response with respect to the uninverted region when an electricfield is applied parallel to the c-axis.When one branch of a Mach-Zehnder is fabricated in a domain-invertedregion, while the other branch is fabricated in a uninverted region, the refractiveindex in one branch will increase while the refractive index in the other will decreasewhen both branches are subjected to equivalent electric fields. The same will happento the propagation constants of the optical modes in the two branches. Hence, thedifference of constants of the proportionality C1-C2, in equation (2.1), equals 2.Thus, the device will work in a push-pull fashion when immersed in a uniform electricfield. Therefore, when used as a sensor, the device will have a relatively highsensitivity, as compared to the strip-loaded devices.Spontaneous polarization reversal, with respect to the bulk region, wasobserved when Ti was diffused into the c face of LiNbO3 samples, i.e, theferroelectric domains in the regions where the Ti was indiffused had spontaneouspolarizations opposite to those in the bulk crystal[29]. The domain structure was19revealed by chemical etching, which was performed in a solution of 2 parts HNO3 and1 part HF at 110°C; in Ti:LiNbO3,the etching rate of the c face was found to bemuch higher than that of the c face [30,3 1]. It was also found that it is possible toobtain complete domain inversion using high temperatures and short diffusion times.However, in regions that are not fully inverted, some uninverted micro-domains existand these become the sources of light scattering[321. Also, some degradation of theelectro-optic activity is found in the micro-domain-inverted regions[33J.The mechanism of the Ti-induced domain inversion, on the c face of LiNbO3,is not yet well understood. Miyazawa[291 suggested that the Curie temperature ofLiNbO3 decreases when the Ti concentration increases[34] and, therefore, thespontaneous polarization decreases with the Cuire temperature. This would mean thatthe gradient of the Ti concentration would result in a gradient of the spontaneouspolarization. It is the electric field from the spontaneous polarization gradient thatcauses the inversion. However, the spontaneous polarization is the result of theminimization of the system’s free energy. The electric field resulting from thisspontaneous polarization should not be able to change the arrangement itself.• Peuzin[35} suggested that the domain inversion results from the electric field due tothe concentration gradient of the impurities. He only assumed that the electric field isantiparallel to the gradient of the impurities and thus explained why the inversionhappens in the c face but not in the c face.In order to investigate the reason that the Ti indiffusion results in domain20inversion in LiNbO3,one would have to look at the process of titanium indiffusionand the defects that are introduced by the titanium indiffusion. First we look at thedefect structure in the congruent composition of LiNbO3,which is used for waveguidefabrication. The congruent composition is 48.45 mol% Li20 and 51.55 mol% Nb205[36]. The excess of Nb ions, or the lack of Li ions, results in intrinsic defects incongruent LiNbO3. There are several defect models applied to this crystal[37,38,39].The most plausible one is the one given by Abrahams & Mash[39]. The results oftheir x-ray diffraction study show that in the Li sites 94.1 % are occupied by Li while5.9% are occupied by Nb; in the Nb sites 95.3% are occupied by Nb while 4.7% arevacant, i.e., NbL”” and V” defects exist in the congruent LiNbO3. Here we areusing the Kroger’s defect notation[40j. For example, NbLI”” is the Nb ion in a Li sitedefect with positive charges with respect to the crystal and VNb” is the Nb vacant sitewith negative charges with respect to the crystal. The results of M.E.Twigg et a!.[41] indicate an increase of NbL”” defects in Ti-doped, or Li deficient LiNbO3.Theyalso found similarities in the defect structure in Ti-doped and Li deficient lithiumniobate. Based on this, they suggested defect equations for the Ti indiffusion;however, there is a term for LiTiO3,which was undetected in Armenise’s study[421.It is well known that the first step in the diffusion process is the oxidation of thetitanium+ 20 = . (2.6)At higher temperatures, the Li ions become highly mobile and most likely diffuse into21the Ti02 layer leaving a high concentration of Li deficient states in the region close tothe surface. This effect was studied by et a!. [43], and Holmes et a!. [44].Because the octahedron is larger for the Li, the Nb ions will tend to occupy the vacantLi sites. This is the reason that, in the congruent crystal, the deficiency of Li ionresults in the Nb”” and V” defects. So, it is reasonable to suggest that thesecond step of the reaction is(2.7)The titanium ions diffuse in via the defects in the crystal, most likely via vacant Nbsites. The process can be represented by+ V= 71Nb + 4e + °2 (2.8)Combining the above three equations, the resulting defect chemistry equation for theTi diffusion is2Lili+2Nb+Ti+O=Ti+2Nb+V’+LiO 2e. (2.9)It is seen that the effect of Ti indiffusion is the introduction of some electrons into thesystem at high temperature. These electrons will likely be trapped in the NbLI sitesor trapped in Nb5 at low temperatures[45]. LiNbO3 is an insulator at roomtemperature, however, it begins to conduct at high temperatures[46,47]. In electricconductivity experiments, Jergerson[46] found that at high temperature LiNbO3 is an22n-type conducting material, especially in the reduced condition. Limb et al. did asimilar experiment and confirmed this result and suggested that the conductive agentsare ionized NbLI defects in the crystal[48]. The NbLI”” defect is a donor typedefect[40], so the indiffusion of Ti introduces donor states into the crystal.At high temperatures, the Fermi level of a high-dopant region will be greaterthan that of the low dopant one. The electrons from the high dopant region willdiffuse to the low-dopant region thus forming a space-charge region and relatedelectric field. The situation here is similar to the junction formed at room temperatureby an N and N semiconductor, where the electrons will diffuse from the N side tothe N side to form a space-charge region. The direction of this space-charge field isfrom the high-dopant region to the low dopant region. This field may be responsiblefor the domain inversion on the c face caused by Ti diffusion, because the space-charge field is in the direction opposite to that of the spontaneous polarization. Itfollows that no inversion will occur on the c face where the space-charge field will bein the same direction as the spontaneous polarization.In order to estimate the magnitude of the electric field caused by the Ti• indiffusion, tbe dopant density is assumed to be proportional to the titaniumconcentration. The Ti concentration is approximated by a Guassian profileC=C0exp(_x2/d) (2.10)where the Co=2rI(d.i7r), r is the Ti thickness before diffusion, d is the diffusion23constant. A typical donor profile is shown in Figure 2.5, where N0 is the ionizedintrinsic donor concentration of congruent LiNbO3 at that temperature. At 1050°C,the electron density, under an oxygen partial pressure of 10-12 atm, is given by Limbet a!. [47,48] as 1. 13x102°/cm3. The relation of the electron density n and oxygenpartial pressure P0 is given by [46,47,48], n P0’4, so that the electron density atatmospheric pressure can be extrapolated to be about 1. 13x10’7/cm These electronsare from the intrinsic defects, i.e., from the defects in congruent lithium niobate,hence N0= 1. 13x10’7/cm The estimated concentration of Ti for a film 400A thickdiffused into congruent LiNbO3 at 1050°C for 6.5 hours is Co=0.O1NT1using Jaegeret a!. fitted diffusion parameters (D=0.53jm2/hr,D=0.39m2/hr)[4 ]. NTI is thedensity of Ti in metal, which is 5.66x1022/cm3 The self-consistent iteration schemefrom Grummel is used to calculate the one-dimensional electric potential and electricfield of-this profile[50]. The calculated electric field, for this dopant profile, is shownin Figure 2.6. By comparison the electric field strength used to pole LiNbO3 inindustrial manufacture is about 5 V/cm[30]; one can see that the field obtained here ismore than sufficient to invert the spontaneous polarization to a depth of 6—7m.We suspect that the domain inversion on the c face of LiNbO3,caused by Tiindiffusion, is primarily attributable to the space-charge field caused by the defectsintroduced by the Ti indiffusion. This domain inversion can be utilized to constructbranches having different electro-optic responses in integrated optics Mach-Zehnderinterferometers. If a region of the substrate contains one branch that is fully inverted,24x10172.2EC)2.QC0z 1.8 -0I.C) 1.6 -C)C0I,(-)0)1.2 -N00-) 1.0 — ‘5 10 15 20Distance in depth ( m)Figure 2.5 The ionized defect concentration from Ti indiffusion.25160-140-E120-100-0>80--Dci)(_)(-)p20-w0—• 0 5 10 15 20Distance in depth ( m)Figure 2.6 The electric field calculated for the Ti concentration gradient.26while that containing the other branch is uninverted, one can make a Mach-Zehnderinterferometer working in a push-pull fashion when both branches are subjected toelectric fields having the same orientation.2.4.2 Device DesignDomain inversion in our devices is realized using a high level of Ti doping. Itis well known that the Curie temperature of Ti doped LiNbO3 (Ta) decreases as the Ticoncentration increases[29,34]. Guenais et al. gave a quantitative relation between thedoped Curie temperature T’ and the Ti concentration[34]7=1141-1O.76x (2.11)where Xm is the mole fraction of Ti in the LiNbO3 and the Curie temperature ofcongruent LiNbO3 is 1141°C [36,39]. The Ti concentration profile is modeled by aGaussian function in the depth direction and a complimentary error function in thelateral direction. The mole percent concentration can be estimated with the help ofthe calibration curves given by Minakata et al. [51]. The degree, or completeness, ofthe inversion depends on the temperature of the diffusion and Ta’. The closer that theheat treatment temperature is to the calculated T’ the easier the inversion. The aboverelation allows one to design the fabrication parameters giving the desired amount ofdomain inversion.27In our case, ideally, we wish to achieve complete inversion in one branch andno inversion in the other. In order to ensure that the entire region seen by the opticalmode in one of the branches is inverted, a thick strip of Ti is pre-diffused. Some ofthe pre-diffused strips have widths 8 times those of the waveguides. The reason forhaving some wide strips is that the change of the refractive index caused by the prediffused Ti is expected to be reasonably uniform over a large cross-sectional area.The optical waveguide to be diffused in the inverted region can then be viewed asbeing formed in a region having a somewhat higher refractive index, but otherwiserelatively uniform, see Figure 2.7, thereby making the fabrication of single-modewaveguides in both the inverted and uninverted regions possible. The calculated T’profiles, that will result from diffusing a Ti strip ioooA thick and 32gm wide at1050°C for 10 hours followed by a diffusion of a Ti strip 400A thick and 3.5gm wideat lO5O’C for 6.25 hours are shown in Figures 2.8 and 2.9. Figure 2.8 shows the T’profile in the lateral direction at the surface, it can be seen that a rise in thetemperature to 1100°C should result in an inverted region about 30gm wide. Figure2.9 shows the T’ profile in the depth direction at centre of the waveguide, it can beseen that the region with a temperature above T’ is about 4.5gm deep. While in theother branch, the one without pre-diffused strip, the minimum T’ is 1122°; thismeans that less (or no) inversion is expected in this branch.However, a waveguide formed in a large pre-diffused region may not workproperly if the change in the refractive index, in the pre-diffused region, is not as28A plan view of domininverEed IiZVoveguide regionsOomoln inverled regions(DiA crosssecLion view of’e domin’inverLed IZt Direction of the spontaneous polarizationWoveguide region ( Ti indiffusedDonain inverted region ( Ti indiffused[hiFigure 2.7 Schematic of a domain inverted Mach-Zehnder interferometer:(a) a plan view, (b) a cross-sectional view.? 1’ 1’ t t t rft ft t ft t tt tt tttttftttttftttfftttttttft tffttttttttttttttttttt:t f t t 1’ 1’ t t t t t t t f t t t t t t t tft t ft ft ft t ttt ttt t tttt ft291150-(I-)D1125-C-)0—‘1 100-DE- 1050-1)D1025-0.:J)00c::: 1000 I I I—30 —20 —10 0 10 20 30Lateral distance (i m)Figure 2.8 The doped Curie temperature, lateral profile, resulting from theindiffusion of a 32gm wide inversion stripand 3.5m widewaveguide.301150-C’):51125-C)0P1100G):5p1075-- 1050-Q)L.:50 1&25--ocI)001z1000—0 2 4 6 8 10Distance in depth (i m)Figure 2.9 The doped Curie temperature, depth profile, resulting from theindiffusion of a 32gm wide inversion strip and 3.5,m widewaveguide.31uniform as expected. Therefore, the width of some pre-diffused strips were designedto be the same as that of the waveguide’s. The calculated Curie temperature profilesfor the same fabrication conditions as those given above, i.e., a diffusion of a Ti stripof i000A thick 3.5gm wide at 1050°C for 10 hours followed by a diffusion of a Ti400A thick and 3.5gm wide at 1050°C for 6.25 hours, are shown in Figure 2.10, forlateral direction at the surface, and Figure 2.11 for the depth direction at the centre ofthe waveguide. It is shown in the figures that rise in the temperature to 1100°C willresult in an inverted region about 2m deep by 5m wide. Because this invertedregion is small, portions of the guided optical modes may not be entirely confined tothe domain-inverted regions. Therefore, a longer modulation length will likelyneeded. Here 10mm was used.To further improve the mode profiles in the waveguides, in the domain-inverted regions, the annealed proton-exchanged (APE) method may be used to formthe waveguide, replacing the second Ti indiffusion. The refractive index changeobtained from the APE method is much larger (as high as 0.12 [52]) than thatobtained from Ti indiffusion (about 0.004[53]) and, therefore, forms a betterwaveguide in the domain-inverted region.For a TM-like optical mode, in an UvIZ fabricated in z-cut LiNbO3,working ina push-pull fashion,321145-1138-o 1130-01h1081100-Q 1093-ci)-01085— I I I I I I I I I—30 —20 —10 0 10 20 30Lateral distance ( m)Figure 2.10 The doped Curie temperature, lateral profile, resulting from theindiffusion of a 3.5jm wide inversion strip and a 3.5km widewaveguide.331145-0 1130-0I1108-1100.-o 1093-a)00I I I I I0 2 4 6 8 10Distance in depth ( m)Figure 2.11 The doped Curie temperature, depth profile, resulting from theindiffusion of a 3.5gm wide inversion strip and a 3.5gm widewaveguide.34ALE=‘2nr33where L1 is modulation length, E1 is half-wave electric field (the electric field neededto obtain a phase shift of ir radians between the modes in the two branches), A is theoptical wavelength in free space, tie is the refractive index, and r33 is the relevantelectro-optic coefficient. ForA0=670m, one obtains L1E=1015.5 V assumingcomplete domain inversion and that the optical mode is entirely confined to thedomain-inverted regions. The electric field strength to be detected isE8=3000 V/mmin air, therefore, taking the dielectric depolarisation field into account using areduction factor of 0.05, the electric field sensed by the electro-optic substrate will beE0.05Ea 150 V/mm. If E1 were 150V/mm we would obtain a corresponding L1of 6.77mm. However, as a sensor, the response of the device will be, preferably, inthe linear region of the optical-intensity-out/voltage-in transfer function, hence, themodulation length should be 1/3 L1 or L, for the device, should be —2.25mm.35Chapter 3 Device Fabrication3.1 IntroductionIn an attempt to verify the theory presented in chapter two, both strip-loadedand domain-inverted Mach-Zehnder modulators have been fabricated on z-cut lithiumniobate substrates. The method for fabricating the titanium indiffused waveguides,which has been well developed in our lab, was used to fabricate both types of devices.The annealed proton-exchange (APE) method was also used, for the first time in ourlab, to make optical waveguides. The APE method may improve the deviceperforniance of domain-inverted devices, since it results in a larger refractive indexchange.A summary of the optical waveguide fabrication techniques is given in section3.2. The problem of Li out-diffusion, associated with fabricating Ti indiffusedchannel waveguides is addressed in section 3.3; a pre-out-diffusion method may offera solution to this problem. The mask designs are given in section 3.4. In section 3.5the basic fabrication processes for the Ti indiffused waveguides are described.Sections 3.6 and 3.7 cover the fabrication procedures for strip-loaded and domain-inverted Mach-Zehnders, respectively. Finally, in section 3.8 the process developedfor the fabrication of APE waveguides is described.363.2 Optical Waveguide Fabrication in LiNbO3A number of methods have been used to fabricate optical waveguides inLiNbO3[54]. Of these methods, the three most important methods are out-diffusion oflithium, indiffusion of metals, and proton-exchange. In all these methods, therefractive index of selected regions in the substrate is increased, as a result of achange in the composition of those regions.The lithium out-diffusion technique was first demonstrated by Kaminow et a!.[55]. When the temperature of the LiNbO3 substrate is raised to 1000-1100°C, thelow activation diffusion energy (68kcalImol) and the low activation evaporation energy(63kcal/mol) of Li ions result in the out-diffusion of lithium ions[56]. The reductionin the Li concentration near the surface is found to increase the extra-ordinaryrefractive index ne thus forming a slab waveguide at the surface. The change of theextra-ordinary refractive index and the out-diffused mole fraction of Li20 is found tobe well approximated by a linear relation, while the ordinary refractive index remainsrelatively unchanged. For slightly non-stoichiometric forms of the (Li2O)(Nb51system, with 0.48 < y < 0.50, the relation is dne/dy=-1.6 [56]. This method canproduce high quality optical waveguides with low loss ( <1dB/cm). However, thesewaveguides only guide optical modes with their E fields polarized parallel to the opticaxis of the crystal (the c-axis). Originally, the first lithium out-diffused waveguides37were slab waveguides, only recently were researchers able to fabricate channelwaveguides by Li out-diffusion[43].The indiffusion of metals such as Mg, Ni, Zn, Fe, Co, Cr, Vi, and Ti intoLiNbO3 was found to change the refractive index of the substrate[57J. In particular,Ti indiffusion has received much attention, since this process produces a relativelyhigh refractive index change in both the ordinary and the extra-ordinary refractiveindices; consequently, offering good light confinement to both ordinary and extraordinary modes. For fabricating Ti-indiffused waveguides, titanium metal is depositedonto the LiNbO3 substrate (usually by thermal evaporation) and the metal is delineatedinto channels using photolithographic processes. The thickness of the deposition isusually from iOOA to ioooA. The diffusion of the Ti into the substrate is carried outat temperatures between 950°C and 1100°C in argon, oxygen, or air. The diffusiontime ranges from 0.5 hours to 30 hours[55]. The resulting changes in the ordinaryand the extra-ordinary refractive indices are sufficiently large so that both TE-like andTM-like modes will be guided. Due to the high diffusion temperature, Ti indiffusionis accompanied by lithium out-diffusion. Whereas the Ti is patterned, the Li out-diffusion ten 4s to occur across the entire surface of the substrate. This may causecrosstalk between Ti-indiffused channel waveguides. The problems associated withthe out-diffusion of Li will be discussed in section 3.3.Proton-exchanged waveguides were first demonstrated by Jackel in 198 1[52].The proton exchange takes place when LiNbO3 is immersed in a melt of benzoic acid38(C6H5COOH), or in other acids, e.g., HNO3H2S04, and H2P04,at temperaturesbetween 160 to 249 °C[58]. The chemical reactions result in the replacement of Liin the substrate by H from the solute. This replacement results in a large increase inthe extra-ordinary refractive index, i.e., Ane0. 12[52J. However, the large refractiveindex increase is only for e’ while the change in the ordinary refractive index An0 isabout -0.05. As the result of extensive studies, the problems associated with proton-exchanged waveguides, such as the instability of the refractive index [53] and thedegradation of the electro-optic effect [60,61], have been solved by using a post-exchange annealing process[62]; a heat treatment of the proton-exchanged waveguidesat a temperature between 300 and 400°C fully restores the electro-optic coefficients[62,631. The advantages of the proton-exchanged waveguide include: quick andsimple fabrication, higher refractive index change resulting in better opticalconfinement, and better resistance to the photorefractive effect or optical damage[64].The last two advantages may be further enhanced if the proton-exchanged waveguideis fabricated on Ti-indiffused or MgO doped (5% mole fraction) LiNbO3. Theannealed proton-exchanged (APE) method is currently being used by UnitedTechnology Photonics, East Hartford, Conn., to make commercial high-speedmodulators.Other techniques such as ion-implanted waveguides[65] and voltage-inducedwaveguides[66] have also been tried. However, the Ti-indiffusion technique is welldeveloped in our lab and, therefore, it was used to fabricate the prototype devices.393.3 Surface GuidingEven though the Ti indiffusion method is frequently used, the Li out-diffusionproblem persists. Because of the high temperature at which the Ti is being indiffusedLi out-diffuses. The Li-out-diffused layer will guide light polarized parallel to the C-axis and, thus, may cause crosstalk between channel waveguides[671. Methods forsuppressing this out-diffusion have been attempted and have been reviewed by Jackelin 1988[68]. Among these methods, the most reproducible one is the ‘soakingmethod. In this method, the titanium is diffused in an oxygen environment where theoxygen is bubbled through a column of warm water before entering the diffusionfurnace. The presence of H in the surface of the LiNbO3 reduces the mobility ofthe Liby occupying the Li vacancies and, therefore, is believed to reduce the out-diffusion of the lithium[69]. In addition to the “soaking” method, a proximitytechnique can also be used in which a piece of LiNbO3 is placed close to the sampleso that the Li20 from this piece will increase the local Li20 partial pressure, therebyfurther reducing the net Li out-diffusion from the sample’s surface. However, evenwith various methods to prevent Li out-diffusion, the problem is not yet solvedbecause of problems with the reproducibility of the suppression methods.A pre-out-diffusion technique may provide a solution to the surface guidingproblem. Here the idea is to ‘pre-out-diffuse’ the wafer before the Ti is patterned and40indiffused. The pre-out-diffusion is carried out in dry oxygen at high temperatures forlong times so that a very deep Li-out-diffused layer is formed. The optical modesguided in this layer will be extremely broad; so broad in fact that the optical modesguided by the out-diffused layer will be indistinguishable from bulk modes.We attempted to use this method in the fabrication of optical waveguides.First, the wafers were pre-out-diffused at 1050°C for 24 hours in an oxygen ambient,then the Ti was patterned and in-diffused at 1050°C for 6.5 hour using the “soaking”method. After cutting and polishing the samples, the surface guiding was stillobserved. Surface guiding should have been eliminated because the depth of the out-diffusion layer should have been more than 300pm after a 24 hour pre-out-diffusion,and the beam width of the fundamental mode should be of the same magnitude[70],i.e., comparable the bulk modes. There might be some other reasons responsible forthe persistence of the surface guiding beside the Li out-diffusion. One possible reasonwas the refractive index change that might be caused by an increase in theconcentration of H, which was introduced in the wet oxygen ambient; this increaseof H has been detected by the observation of an OH absorption line at 5890A[71].As described in the previous section, the substitution of Li, by H, results in anincrease of the extra-ordinary refractive index; consequently, only the extra-ordinarymodes are guided. So, it is likely that the surface guiding that persisted was the resultof H indiffusion. If the Ti indiffusion is performed in a dry oxygen ambient,following a pre-out-diffusion of Li in a dry ambient, the surface guiding problem may41be eliminated.The effect of the stoichiometry of the LiNbO3 on the electro-optical activitywas investigated by Turner[72]. The electro-optic effect remains the same in crystalsof the (Li2O)(Nb51..system with stoichiometry variation of 0.48 < y < 0.52.The stoichiometry variation caused by out-diffusion is about y — 3 X i03, which iswell within the range in which Turner’s results apply. On the other hand,Holman[73] found that optical damage was reduced by reducing the lithiumconcentration in the crystal. Some researchers found that dry diffusion was better forsuppressing surface guiding. They did the indiffusion at higher temperatures and forlonger times in a dry oxygen ambient, therefore, their results showed no surfaceguiding[741. The reason for their results may be the Li-out-diffused layers were deep,so that they could not observe the optical mode guided in this layer. With furtherdevelopment, pre-out-diffusion may offer a practical solution to the surface-guidingproblem in the fabrication of Ti indiffused LiNbO3 waveguides.3.4 Mask DesignThe nask used for the fabrication of strip-loaded Mach-Zehnder is given byJaeger[14 (M.A.Sc. thesis)]. The masks for the waveguides consisted of sets ofMach-Zehnder patterns with different waveguide widths: 4, 6, 8, and l0m. Theangles of the Y-branches, of the Mach-Zehnders, were 2.005°, 0.8913°, 0.5730°, and0.4456°. The loading strips were 1, 5, 10, and 15mm long.42A new mask set has been designed for the domain-inverted Mach-Zehnderdevices. Based on past experience, of Ti indiffused waveguide fabrications in ourlaboratory, a pre-indiffused titanium waveguide pattern 400A thick by 3.5gm widewould more reproducibly result in single-mode waveguides. So, in the new mask, aschematic of which is shown in Figure 3.1, the widths of the Mach-Zehnderwaveguides W1 were 3, 3.5, and 4gm. The width of the inversion strips is 30gm,wider than the waveguides, and the length, L3, of the strips are 2, 3, and 5mm. Asmentioned in chapter 2, a narrow inversion-strip was also used. Here the inversionstrips were 3.5gm wide and 10mm long. The Y-branch angle was 0.5°. Such a smallangle was chosen to reduce the loss in the Y-junction. The parameters for thewaveguides and inversion-strips are summarized in table 3.1.The layout was done on a Sun SPARCstation IPC, Sun Microsystems,Mountain View, CA., using the VLSI design tool Edge, Cadence Design System Inc.,CA. Lotus 123, Lotus Development Corp., Cambridge, MA., was used to generatethe coordinates needed to form the waveguide patterns. The coordinates were thenconverted into batch command files for Edge, i.e., Skill files, to generate the layout.The use of the command files greatly increased the design efficiency. Firstly, theyreduced the tedious work of drawing with the mouse on the PC. Secondly, the layoutcould easily be regenerated, so that they were less susceptible to loss. Thirdly, it waseasy to make changes and modifications of the design using Lotus 123’s spread sheet43Dcmoininverf.ed AchZEhnder nlos____[2____________________VIL3__Domeirrinverte regim Veveguide regionFigure 3.1 Schematics for the domain-inverted IMZ mask.Table 3.1 Waveguide parameters for the new mask.Device L1 W1 W2 W3 a Num. oftype (mm) (mm) (mm) (jim) (sm) (zm) degrees devices1 7.2 3.2 3 3.5 16 15 0.5 202 6 5.6 5 3.5 16 15 0.5 203 7.2 3.2 2 3.5 16 15 0.5 204 7.2 3.2 3 3 16 15 0.5 105 7.2 3.2 3 4 16 15 0.5 106 5 11 10 3.5 16 0 0.5 2044functions. The layouts were then converted to ‘stream’ files, which were compatiblewith the electron-beam lithography machine, and were down loaded to a 1/4 inchcartridge tape. The tapes were sent to Precision Photo Mask Ltd., Quebec City,Que., who generated the masks.3.5 Basic Fabrication ProceduresFabrications of the strip-loaded and domain-inverted Mach-Zehnder modulatorswere attempted several times using Ti indiffusion. Many of the fabricationprocedures, for both types of devices, were similar and, therefore, the procedurescommon to both will be described in this section.3.5.1 Wafer StoringThe crystals used were c-cut wafers obtained from Crystal Technology Inc.,Palo Alto, CA. These optical grade wafers were optically polished, such that the in-plane scattering of the light was small, and were highly purified (the transition ionimpurity concentration was less than 2ppm), making them suitable for opticalwaveguide fabrications. The c side is usually the preferred side for the fabrication ofwaveguides, since domain inversion, which may occur on the c side, will usuallydegrade device performance[29]. Both sides of the wafers were polished, the onlydifference was that the c side was polished before the c side. The wafers were cleanas received. We evaporated 3000—4000 A aluminium onto the working surface (c side45for strip-loaded and c side for domain-inverted devices) to protect the surface. Thewafers were then cut into quadrants using a South Bay Technology Model 850 wiresaw, South Bay Technology, Temple City, CA., with 600 grit silicon carbide abrasiveslurry. The quadrants were stored with the protecting Al layer on the workingsurface.3.5.2 Thermal EvaporationMetals such as titanium (evaporation temperature of 1453°C), aluminum(evaporation temperature of 1010°C), and chromium (evaporation temperature of1157 °C) were evaporated onto the wafers using the CHA Vacuum ThermalEvaporation System, Carl Herrmann Associates Inc., Menlo Park, CA. An InficonModel 321 Thickness Monitor, Inficon Inc., East Syracuse, NY, was used to monitorthe thickness during the evaporation. However, the thickness readings from themonitorwere usually about 20—30% less than the readings from the Tencor AlphaStep-200 Profilemeter, Tenco Instrument Inc., Mountain View, CA, which was moreaccurate. This error was taken into account to when performing an evaporation.3.5.3 Thermal DiffusionThe thermal diffusions were carried out in a Mini Brute Furnace, ThermcoProduct Corporation, Orange, CA. In order to suppress Li out-diffusion from thewafer, oxygen was bubbled at a rate of 1 f/mm through an 8-inch, deionized (DI)water column at 80°C.A dust blower was used, to remove any dirt accumulated on the samples, prior46to placing them into the oven tube. The temperature was then ramped up to 1050°Cin about 2 hours. The temperature controller setting was set to 1050°C to maintainthe temperature at 1050 ± 0.5°C. The titanium was indiffused for 6.5 hours whenfabricating the waveguides and for 10 hours when creating the domain-invertedregions, after which the temperature was lowered to room temperature at a rate ofabout 20°C/mm. Sometimes, following the indiffusion, some dirt particles were seenon the surface of the samples; these were attracted by the large electric fieldsresulting from the pyroelectric effect of the LiNbO3 during the heating and coolingcycles. An ultrasonic bath was used to remove these dirt particles.3.5.4 Waveguide PatterningBefore the fabrications, the Al layers protecting the crystal surfaces wereremoved using an Al etchant; which was a solution consisting of 57 ml DI water, 21ml nitric acid (HNO3), and 222 ml phosphoric acid (H2P04).A titanium layer, having a thickness of about 400A, was thermally evaporatedonto the wafers using the CHA thermal evaporator. Positive photoresist, Shipley PR1400-27 from Shipley Company Inc., San Jose, CA, was spun onto the wafer in aspinner at 4000 RPM for 35 seconds, resulting in a photoresist film between 1 and1 .2gm thick. The samples were then loaded into a oven to pre-bake the photoresist at95°C for 30 minutes.After being cooled, following the pre-bake, the samples were placed in theKarl-Suss MJB 3 Contact Aligner, Karl Suss America Inc., Waterburgh, VT. The47photomask was carefully aligned with the samples. When the wafers came intocontact with the mask, they tended to shift relative to each other, hence minoradjustments were needed during the contact process. Figure 3.2, shows the alignmentof a waveguide and a domain-inverted region. After proper contact was made, thephotoresist was exposed for about 50 seconds to UV light, at ). = 320nm with a powerintensity of 25mW/cm (the exposure time should be adjusted according to the age ofthe photoresist and the UV light intensity). After the exposure, the samples werepost-baked at 95°C for 20 minutes; however, when using a lift-off method, no postbake was needed but an additional treatment of the photoresist, in which the samplewas soaked in chlorobenzene for 8 minutes, was needed to produce an over-hanginglip[75].The samples were then developed in a cooled (— 10°C) solution of 50%Shipley.MF-312 and 50% DI water at 15°C or, alternately, Shipley MF-319 at roomtemperature; constant stirring was needed. After the pattern was clearly seen, andfree of dissolving photoresist, an additional 15 seconds were needed for patterns to beused in lift-off procedures, in order to ensure the undercut, before the samples wererinsed in a DJ water bath. Experience was essential here to obtain a gooddevelopment; some of the variables that needed to be considered included whether thepattern needed a dark field or light field mask, the condition of the developer, and thetemperature. The developed patterns were examined under the microscope; here afilter was used since the normal illumination from the microscope will tend to expose48Figure 3.2 ‘1aLgnment of the waveguide with the inversion stripthe photoresist. If the pattern was under developed, the sample could be placed backinto the developer for an additional 15 seconds. Once the desired pattern wasobtained, the samples were rinsed in the DI water bath for 5 minutes and then wereblown dry with nitrogen.The portion of the Ti film that was not covered by the photoresist was etchedusing a Plasmatherm System AMNS-500E, Plasma Therm Inc., Kresson, NJ. Theauto loading factor of about 90 and tuning factor of about 86 were used to optimizethe operation of the system so that the forward RF power was 150 watts with thereflected power was less than 5 watts. With these conditions, a CF4 plasma was used49to etch off 400A of Ti, at 75°C, within 20 minutes. Acetone or Microstrip 2001were used to dissolve the photoresist revealing the Ti waveguide pattern, at whichtime the Ti pattern was ready for thermal indiffusion.After the waveguide was indiffused, a Si02 optical buffer layer, about 2000Athick, was sputtered onto the wafer. The sputterer used to deposit the Si02 was aPerkin Elmer Sputtering System Model 3140, Perkin Elmer, Norwalk, CT. The Si02target was first sputter-cleaned for one hour before the deposition. Using 18 mtorr ofargon, 5 mtorr of oxygen, and a forward RF power of 150 watts with reflected powerof less than 5 watts, the Si02 was deposited at a rate of about 2000AIhour.3.5.5 Cutting, Grinding, and PolishingThe wafers were cut using a South Bay Technology Mode 850 Wire Saw. Theabrasive slurry used was a mixture of DI water (20 grams), glycerine (100 g), and600 grit silicon carbide (60 g). Another piece of c-cut LiNbO3, the same size as thesample, was used as a cover piece for protection in the grinding and polishingprocesses. After cutting, the samples and their respective cover pieces were carefullycleaned. A thin layer of wax was then applied between the sample and the coverpiece, with the cover piece covering the working surface of the sample. The sampleand the cover piece were clipped together and put in an oven at 100°C for 2 hours tosqueeze the wax out from between them. The thinner the wax layer the better for thelater grinding and polishing processes; since the sample’s edges will easily chip offwith a thick wax layer. The thickness of the wax layer between the sample and the50cover piece was in the range of 2 to 6pm.The grinding and polishing were done on a Buehler Polisher with AutometPower Head, Buehler Ltd. Lakebuff, IL., The sample and the cover piece wereclamped into the polishing jig so as to expose a 1 to 2mm edge for grinding andpolishing. Grinding papers of 600, 800, and 1200 grit were used for the grinding (inthat order). Large scratches were removed before entering the polishing steps.Polishing slurries, with 1 and 0.5 micron sizes, were used to polish the samples. Forproperly polished samples, no scratches should be seen, especially in the regions closeto the working surfaces of the samples. The samples were then soaked in warmtrichioroethylene to remove the wax, and an ultrasonic bath in methanol to remove thedirt particles.3.6 Fabrication of Strip-Loaded Mach-ZehndersThe main difficulties encountered in the fabrication of the strip-loaded MachZehnder interferometers were the alignment of the loading-strip with one of thebranches of the device and the control of the thickness of the Ti02 film. With thehigh magnification of the microscope on the mask aligner, the misalignment could becontrolled to within about 0.5m; while the control of the thickness of the Ti02 wasless sure, due to the 20—30% error in the thickness monitor and the variation in thedensity of the deposited Ti film. The procedures for fabricating the strip-loadedMach-Zehnders were as follows:51a. Etch off the protecting Al layer.b. Evaporate 400A of Ti on the wafer’s c side.c. Pattern the Mach-Zehnder waveguide photoresist pattern (dark field mask).d. Evaporate Al onto the pattern and lift-off the Al with acetone.e. Plasma etch the Ti not covered by Al.f. Etch off the covering Al, revealing the Ti pattern.g. Indiffuse the Ti pattern at 1050°C for 6.5 hour.h. Evaporate 800A Ti on the wafer.i. Pattern the loading-strips (light field mask).j. Plasma etch the Ti not covered by the photoresist.k. Clean off the photoresist.1. Oxidize the Ti strips at 600°C for 2 hours.m. Sputter deposit a 2000A Si02 optical buffer layer.n. Evaporate 3000A Al electrodes on both sides of the wafer.o. Cut, grind, and polish the wafer edges.3.7 Fabrication of Domain-Inverted Mach-ZehndersDomain-inverted Mach-Zehnders were fabricated on the c side of the LiNbO3wafers. The fabrication involved the indiffusion of a titanium strip to create thedomain-inverted regions followed by a waveguide fabrication. The waveguides werealigned in such way that one branch of the Mach-Zehnder resided in the middle of the52domain-inverted region while the other branch resided outside the domain-invertedregion. The fabrication steps were as follows:a. Etch off the protecting Al layer.b. Evaporate ioooA of Ti on the c side of the sample.c. Pattern the inversion strips.d. Plasma etch the Ti not covered by the photoresist,e. Clean off the photoresist.f. Indiffuse the Ti pattern at 1050°C for 10 hours.g. Evaporate 400A Ti onto the sample.h. Pattern the waveguides.i. Plasma etch the Ti not covered by photo resist.j. Clean off the photoresist revealing the Ti waveguide patterns.k. Indiffuse the waveguide titanium at 1050°C for 6 hours.1. Raise the temperature to 1100°C for 15 minutes to enlarge the desired domain-inverted regions.m. Sputter deposit a 2000A thick Si02 optical buffer layer.n. Evaporate 3000A Al electrodes on both sides of the wafer.o. Cut, grind arid polish the wafer edges.533.8 Fabrication of Annealed Proton-ExchangedWaveguidesThe procedures for fabricating APE waveguides were adapted fromreferences[52,62,76]. Straight waveguides were first used to develop the process.The waveguide patterns were created by photolithography. A layer of 500 Athick Cr was evaporated onto the wafer. The waveguide regions were revealed bylifting off the photoresist. The Cr masked wafers, with the waveguide regions notcovered by the Cr, were then ready for proton exchange.The proton exchange was performed in molten benzoic acid. The powderedbenzoic acid was heated to 200°C on a stove. The samples were pre-heated to 100°Cin a beaker of DI water. The wafers were then dipped in the molten benzoic acid for15 minutes. The resulting proton layer was a step profile with a depth of 0. 1pm,and a refractive index change, e’ of 0.12. The samples were withdraw from theacid and left to cool in air. To avoid thermal shock, care was taken not to dip thewafers into the solvent immediately. The Cr mask was then etched off using asolution of 50% HC1 and 50% glycerine. After the proton exchange, the sampleswere annealed for 2.5 hours in the Mini Brute Furnace at a temperature of 350°C for2.5 hours. The annealing process helps to reduced the stressed induced by the protonexchange and to smooth out the proton profile, from a step profile to a complimentary54error or a Guassian profile, providing improved guiding properties. An Si02 bufferlayer was sputtered onto the wafers. The samples were then cut and polished by thesame process as for the Ti waveguides. These APE waveguides were the first to befabricated in our lab. The output spot of such a waveguide is shown in Figure 3.3.By using the APE method, we have added to our options for fabricating integratedoptical devices.Figure 3.3 The output spot of an annealed proton-exchanged waveguide55Chapter 4 Device Testing andMeasured Results4.1 Introduction.In order to verify the theory and the designs of our devices, samples werefabricated and tested. All of the asymmetric Mach-Zehnder senors were eitherfabricated or proposed and fabricated for the first time. Therefore, the primaryemphasis of the testing was on the device functionality as compared to the designexpectations. For example, by measuring the half-wave voltage of the domain-inverted Mach-Zehnder, we can verify the domain inversion caused by the Tiindiffusjon and, thereby, determine the electro-optic response of the domain-invertedregions. As regards the devices’ potentials as sensors, basic device parameters, suchas the half-wave voltage, the intrinsic phase, and the extinction ratio, were measuredusing the optical-intensity-out/voltage-in transfer function. Since such a device maybe used in metering, the linearity of the devices, as well as the stability of the deviceparameters with respect to time and temperature, were investigated. Impulsemeasurements were also done on some devices.The experimental setup is described in section 4.2 and the basic deviceparameters to be measured are described in section 4.3. Results for the strip-loaded56IMZs are given in section 4.4, while the measurements and results for the domain-inverted IMZs are presented in section Experimental Setup.The testing apparatus was setup on an optical breadboard. Devices under testwere placed on a sample holder that was isolated from the breadboard. The tests onthe strip-loaded Mach-Zehnders and the first domain-inverted Mach-Zehnder weredone with a laser, a LAS-0670-10 GaA1As laser from LaserMax Inc., Rochester, NJ,wavelength A =0.67gm and a polarization ratio of 100:1. The light from the laserwas coupled into a polarization maintaining fibre and was, in turn, butt coupled to thedevices. The output light from the devices was focused on a detector using a 25xobjective lens.As the testing progressed, the light source for the domain-inverted devices,fabricated with the new mask, was a LAS-1300-6 laser, also from Laser Max, havingA =1. 3(Lm and a polarization ratio of 319:1. The photorefractive effect in the LiNbO3was greatly reduced at this wavelength[77J. In these tests, a 25x, long-workingdistance, objective lens was used to end-fire couple light into the devices. The laserand the lens were mounted on a 3 dimensional translational positioner so that the lightcould be focused onto the input of the waveguides. The light that emerged from thedevices was collected with a lOx objective lens. The image was projected onto aninfra-red sensor card which converted the JR light to visible light so that one could57observe the output signal by eye during alignments. The JR card needed to berecharged by visible light every 15 minutes or so.Once the optics were aligned, we detected the optical power using a Newport818 infra-red optical detector, Newport Corp., Fountain Valley, CA, together with aNewport 835 digital power meter. However, the low bandwidth precluded it’s use athigh frequencies, e.g., the fastest rise time of the amplifier was 1 .Oms. For testing athigh frequencies, a New Focus 1811 JR high speed detector, from New Focus Inc.,Mountain View, CA., capable of measuring up to 125 MHz was used. Theresponsivity of this detector was 0.82 A/W with a voltage gain of 4OmV/A. Thesignal from the detector was sent to an HP 54600A digital oscilloscope, from Hewlett-Packard, Colorado Springs, CO, for display. An RS232 link was setup between theHewlett Packard oscilloscope and an IBM compatible PC so that the data could betransferred and stored in the PC using a software package called ‘scopelink’. Thestored data could be further processed for analysis. The applied modulating signalwas connected to channel 1 of the scope and the modulated optical signal wasconnected to channel 2. Therefore, in the data files the first column was time, secondcolumn was the applied voltage, and the third column was the output response.Both the strip-loaded and domain-inverted Mach-Zehnders are immersion typesensors. Therefore, in order to simulate the immersion of the sensors in the fields,planar electrodes were fabricated on the devices. The electrical modulating signal wasapplied directly to the device by probing the electrodes. The high voltage was58obtained using a high-voltage transformer.4.3 Device ParametersThe device characteristics can be well described by the optical-intensity-out/voltage-in transfer function[15]I=(Ij2)[l+cos(JtV/VE+4I)] (4.1)where‘in and ‘out are the input and output optical signals, V is the applied voltage,V, is the half-wave voltage, and is the intrinsic phase. The half-wave voltage isthe applied voltage required to obtain a difference of ir radians between the phases ofthe propagating fundamental modes in two branches of the Mach-Zehnder. Inequation (4.1), the extinction ratio is assumed to be 100%. However, for an actualdevice the extinction ratio will be less. In order to use this device for metering onhigh-voltage lines, the sensor should, preferably, have a linear response. Our devices,having sinusoidal transfer functions, do not have linear responses. However, forsmall x the sin(x) function may be approximated by x, i.e., sin(x)—x. If we definethe linear region by requiring that (sin(x)-x)/x < 0.3%, we find that x should be inthe range -0.135 <x < 0.135. In other words, the linear region of the transferfunction is ±0.135 radians or ±7.7 degrees for a bias point of ±9O°. The deviceshould, preferably, be operated in this linear region of the transfer function. Taking59into account that the applied voltage is also a sinusoidal function of time,V=V0Sfl(mt), when the device is biased at -90° the transfer function becomes‘Out= 1 +sin(-1tsin(() mt)) = 1 +sin(sin( mt)) (4.2)where x =irV0/V1,corresponds to the x above, and m is the angular frequency of themodulating signal. If we expand the transfer function in a series of Bessel functionsthensin(sin( mt)) 1(x)( mt)+2J3()sjn(3 mt)+2J5()sjn(5co mt) +... (4.3)where J(x) is a Bessel function of order n. We can see that higher order harmonicsare generated in the signal. The coefficients for the harmonics will indicate thestrengths of the harmonic compositions. Take the case of the above analysis,x=O.l35 where 2J1()- is 0.23% of x. This means that if the device is operatingin the linear region defined above, the higher harmonic components are less than0.3% of the fundamental. If the device is biased at an arbitrary , equation (4.3)becomessin(sin( ,,,t) +dt, ) = sine [.J0(x)-‘-2J2()sin(2w mt)+2J4()sin(4 (4.4)+cos41[21(X) in( net)+2J3()sin(3c net) +...].For a pure sine-wave modulating signal applied to the device, we can calculate the60intrinsic phase and the on/off ratio of the device by Fourier transforming the deviceresponse signal, and relating each harmonic term in the Fourier series to thecoefficients of equation (4.4). The analysis becomes complicated if higher harmonicsare contained in the applied modulating signal, which may be generated in a high-voltage transformer. So, in practice, we fit the transfer function to a straight lineusing the least square method and analyze the standard error.Measurements were also made on the intrinsic-phase drift with respect totemperature and time.4.4 Results for the Strip-Loaded IMZ DevicesAs expected, the sensitivity of the strip-loaded Mach-Zehnder was low, whichwas suiible for high-voltage sensors. However, the response of the devices wasfound to be so weak that we could not be sure if the detected modulated signal wasthat of the device or was from other sources which might have been coupled to oursystem. The small amount of modulation led us to believe that the loss in the strip• loaded branch might be so large that the output power measured was mostly from theunloaded branch.There were several factors which could account for the loss in the strip-loadedbranch: the coupling loss at the junction between the loaded section and the unloadedsection; the scattering at the interface of the Ti02 and the LiNbO3;and the absorption61in the Ti02 layer. The scattering at the interface could be reduced by carefultreatment of the sample and a high-temperature oxidation of the Ti film would resultin only one oxidation state of Ti4, giving the rutile phase of Ti02, and thus leadingus to expect low losses in the loading films[78,791. No detectable absorptioncoefficient of the Ti02 film was measured using ellipsometry. So the coupling loss atthe junction between the unloaded region and the loaded region was thought to be themain source of loss in the devices.To calculate the coupling between the fundamental modes in the two regions,the field profiles were obtained using the approximate method given by Martcatili[80j.Again the exponential distribution of the refractive index was used as in chapter 2.The TE fields in the simplified waveguide structure in Figure 2.3 in chapter 2 aregiven asE1(x,z)=E1cos(px)exp[—y1(z—t)], in region( 1)E2(x,z)=2cos(p,x)cos(y,z+ ce), in region(2)E3(x,z)=E3COS(PX)Jq[g(Z)], in region(3) (4.5)4(x,z)exp[—p( —wI)1Jq(S(z)1 in region(4)E5(x,z)=Eexp[p(x+w/2)JJq g(z)], in region(5)where k0=2ic/2 21/2y1=k0(na)2 21/2y2=k0(nf—n1)2 2.1I2P1=p3k(n1-ip4=p5=k0(n_n)1I2g(z)=2dk0( n3n)’exp(z/2d),62andtan(cz)= (y12)—tan(yt)1+(y/y)tan(y)Here w and t are the width and thickness of the loading strip, d is the depth of theexponential profile, and a’ n, flf, n1, n11, and An are as defined in chapter 2. TheTE0 field mode profiles for t=O and t=120 nm are shown in Figures 4.1 and 4.2,respectively. The coupling coefficient between the TE0 mode in the loaded region andthe TE0 mode in the unloaded region was determined by the overlap integral of thetwo fields at the junction of the two regions[81]2tfl(°)l)ht2fE°kx,z)E°(x,z)dxdz (4.6)p(p+p)where P is the normalized optical power defined as.i= fE(0)(x,z)E(0)*(x,z)dxdzand E° and E1 are the electric field mode profiles and and are thepropagation constants of the unloaded and loaded regions, respectively. Here thepropagation is assumed to be in the y-direction. The calculated coupling coefficientbetween the TE0 modes, in the loaded region and unloaded region, as a function ofthe loading-strip thickness, is shown in Figure 4.3.63Figure 4.1 TE3 profile without the loading film.64Figure 4.2 TE field profile with a l2Onm thick loading film.65A loading-strip thickness of l2Onm for one of the branches was chosen inorder to achieve a reasonable difference in the propagation constants in the twobranches. However, when 120nm of Ti02 was used, theoretically only about 10% ofthe power coupled from the unloaded section into the loaded section. This wouldmean that most of the power was lost to radiation modes and to higher-order modes.Since there were two such junctions, we expected at most an on-off ratio of 1 % forthis particular arrangement.From Figure 4.3, it is obvious that a Ti02 thickness of more than 80 nm willresult in a large insertion loss. Therefore, a taper is needed at the junction betweenthe loaded and the unloaded sections to minimize such loss.Coupling coefficients between TE0 modes under loading films of differentthicknesses were also calculated. In the calculations, a iooA step increase in the Ti02thickness was used because our experimentally controlled thickness of Ti02 can beexpected to be accurate to about ± 100 A. A topographical figure of the couplingcoefficients is shown in Figure 4.4. The best result for the total coupling of opticalpower, coupled from the unloaded section into the 120 nm loaded section bysuccessive steps of 100 A Ti02 films, was found to be —74%. However, this involves12 fabrication steps to load the TiC2 onto one branch. Still, from Figure 4.4, onefinds that the coupling remains above 90% for a Ti02 thickness of up to 700A.Therefore, the first few steps can be replaced by larger steps without seriouslyaffecting the results. So a 5 step combination of 600A, 200A 100A iooA and 200A661.000.80CC-)0.600U0.40-oC)0.200.00 — i I I I I I I I I I I0.00 0.04 0.08 0.12 0.16Loading—strip thickness in micronsFigure 4.3 Coupling coefficient as a function of loading-strip thickness.670,01 0.02 0.03 0.04 0.05 0.06 0.07 0,08 0.09 0.10 0.11 0.12 0.13 0.14 0.150.14 0.130. ,,)JL///’ ‘ ‘ ‘ :0.15,0.12 0.12CEo.ii - 0.1100.20.10•°‘°0.7 —Cl) 0.8 0.08o.oa •0.9- 0.07 •5.90.031(1oojI( 0.030.01 1IIII-o0020,02-I___________________-J0.010.01 0.02 0.03 0.04 0.05 0.06 0.07 0,08 0.09 0.10 0.11 0.12 0.13 0,14 0.15Loading film thickness in micronsFigure 4.4 Topograph of the coupling coefficients68will result in 69% coupling, while a 4 step combination of 800A iooA iooA and 200Awill result in 65% coupling. Since another tapering is needed to couple light from theloaded section back into the unloaded section. Taking all these variables into account,the total insertion losses were estimated to be about 2.63dB, 3.20dB and 3.72dB for a12-step, a 5-step, and a 4-step tapering, respectively. The respective on-off ratios ofthe modulation signals are expected to be 16.46dB, 14.58dB and 13.51dB. As thecost of masks for such devices would be quite high, combined with the fact that wewere getting useful results from our domain-inverted devices, further attempts atmaking strip-loaded IMZs was not pursued.4.5 Results for the Domain-Inverted IMZ Devices.The testing of the first domain-inverted device was done with a laser having)0=0.67m. It was found that the half-wave voltage was 116 volt as compared to theexpected half-wave voltage of 101 volt. Once the concept was thus verified a newmask was designed. For devices fabricated using the new masks, the modulationlengths of the’devices were 2, 3, and 5mm with wide inversion strips and 10mm withnarrow inversion strips. The waveguiding properties of the devices with wideinversion strips were found to be poor as compared to the devices with narrow sthps,i.e., the 10mm devices with narrow inversion strips had the better guiding andmodulation characteristics. Therefore, only the results for the devices with 10mm69long inversion strips are presented.4.5.1 Basic Device Parameter MeasurementsThe optical system was aligned to obtain maximum optical output from thedevice under test. The best output obtained from the detector was about 700mV,corresponding to 2OW of optical power. The applied voltage was supplied by avariac through a high-voltage transformer. The voltage was applied to the planar Alelectrodes on both sides of the sample. The voltage was monitored on anoscilloscope. It was necessary to use a divider to step down the applied signal. Theratio of the applied and the monitored voltage was 1.7. To obtain the modulationfrom the devices, the applied voltage was increased so that it was obviously largerthan the half-wave voltage. Oscilloscope images of the applied and response signalswere then collected. The oscilloscope images of two typical devices are shown inFigures 4.5 a and b. Besides the scope images, the data for each channel were alsocollected in ASCII form. To analyze the collected data, a Matlab program was usedto least square fit the measured transfer function toI=dc+Acos(itV/V+41). (4.7)The fitting parameter dc is the dc offset of the signal, A is the amplitude of the signal,VT is the half-wave voltage, and is the intrinsic phase. The on/off ratio can becalculated from (dc+A)/(dc-A). The results of the fits for two of the devices are70(a)--I-‘p-pC 1)B2.50 V Vp-pC2)74.4mV VvgC2)=39S.6rriV(b)Figure 4.5 Oscilloscope images of responses of (a) device 82 and (b) device 91.71Cl)4)C:5..._D.0..0•CCl):50—I I I—100.0 —60.0 —20.0 20.0 60.0 100.0Applied voltage (volt)Figure 4.6 Fitted transfer function of device 82.72(1) •a_e.—I——) ...C ..:3.-D0CCCl):30:3.0I I I I—120.0 —80.0 —40.0 0.0 40.0 80.0 120.0Applied voltage (volt)Figure 4.7 Fitted transfer function of device 90.73shown in Figure 4.6 and 4.7, where the dots are the collected data and the solid linesare the fitted curves. In Table 4.1 the half-wave voltages, the intrinsic phases, andthe extinction ratios of several devices are listed.The equation for calculating V, for the push-pull situation isdXiç= ° (4.8)2nr33Lwhere d is the thickness of the wafer, e is the refractive index of the LiNbO3,r33 isthe relevant electro-optic coefficient, and L is the modulation length. For A =1 .3jLmfle=214M[83], and r33= 34x10’2m/V[84]. d was measured to be 0.5mm, and Lwas 10mm. This gave a calculated half-wave voltage of 96.6 volt. As shown in thetable, the measured half-wave voltages were consistently higher than the expectedpush-pull values but less than twice the push-pull ones. This confirms that thespontaneous polarizations in the Ti-pre-diffused regions were, in fact, inverted andthat the electro-optic coefficient r33 in the inverted regions was opposite to that in thenon-inverted regions; otherwise it would not have been possible to obtain half-wavevoltages less than twice of the calculated push-pull values, not even if one haddestroyed the electro-optic effect in the Ti-pre-diffused branch. There were severalpossible reasons for the degradation of the half-wave voltages from the calculatedpush-pull ones. Firstly, the domain inversion in the Ti-pre-diffused region was notcomplete, with some of the micro-domains still in the original direction, reducing the74cumulative electro-optic effect. Secondly, partial domain inversion also occurred thebranch with no Ti pre-indiffusion. Presummably, partial inversion occur when in-diffusing the Ti used to form the waveguides. Thirdly, long, high-temperature, heattreatments might also induce inversion in the wafers[84]. It might well be the casethat, to some extent, all of the above phenomena occur during the fabrication of thedevices. It was obvious that the intrinsic phases of the devices were random. Thereasons this was observed might be that the completeness of the inversion varied fordifferent devices, the Ti concentration in the inverted regions varied for differentdevices, or the drift in the intrinsic phases with respect to time, as we will see later inthe stability tests. The extinction ratios of most of the devices were reasonably good.Table 4.1 Measured device parameters for domain-inverted IMZs.Device No. V1(volt) %err in V 1(degree) on/off ratio82 106.7 10.4 116.6 22:183 106.3 10.0 151.1 21:184 103.7 7.3 132.8 96:186 98.6 2.0 25.2 47:189 115.7 19.7 106.7 40:190 97.1 0.5 -12.3 29:191 109.8 13.6 65.8 219:192 133.1 37.7 -52.9 7:193 109.6 13.5 -105.5 13:194 106.3 10.0 -27.7 7:195 131.3 35.9 178.7 20:196 107.5 11.2 166.5 12:1754.5.2 Linearity Measurements.In order to obtain the linearity of the devices, a small signal was applied tothose devices having the proper intrinsic phase, i.e., biased at around ±900. Toanalyze the collected data, we used a least-square fit of the output signal to a straightline. In doing so, we found that the actual dynamic range that satisfied the standarderror, for the fitting of a straight line to the transfer function, of less than 0.3% waslarger than we expected from the argument given above in section 4.2.To find the relation between the standard error and the signal range, a Matlabprogram was written to fit a sinusoidal function to a straight line and to calculate thestandard error of that fit. Figure 4.8 shows that the standard error of the fit increasesas the input range increases. However, it is obvious that the standard error is lessthan 0.1 % for a range less than 19 degrees, and less than 0.3% when the range is lessthan 28 degrees. This means that, if the device is properly biased then, the standarderror will less than 0.3% in the range from -28 to 28 degrees. So the restriction ofthe intrinsic phase of our device seems not to be as strict as we had previouslyexpected, i.e., -7° <x < 70 On the other hand, if we fix the input range to be ± 10degrees, or 20 degrees, and plot the standard error as a function of the intrinsic phase,we obtain the result shown in Figure 4.9. The error is less than 0.3% if the intrinsicphase is within 37 degrees. In other words, if we keep the input range within 20degrees (or ± 10°), then, the intrinsic phase of the device can be off by as much as 37760130130Cr)O.OE+000Figure 4.8 The standard error in fitting a transfer function to a straight lineas a function of the input range.4.OE—0033.OE—0032.OE—0031 .OE—0035 10 15 20Input range (degrees)774.OE—003 -3.OE—0030ci)2.OE—003 -t7C0I.Ct) -1.OE—003 -0.OE+000-5 10 15 20 25 30 35 40 45Bias (degrees)Figure 4.9 The standard error in fitting a transfer function to a straight lineas a function of bias for fixed input range of 20 degree78degrees while still maintaining the standard error to within 0.3%. This greatlyreduces the resthction on the intrinsic phase of our devices.Oscilloscope images of the small signal responses of two devices are shown inFigure 4.10. To filter out the high frequency noise, a Butterworth filter of order 2,with a cutoff frequency of 600Hz was used. Figure 4.11 shows the fitting of thelinear response of device 91 without filtering, and the solid line is the least square fitto the data. The x-axis is the applied voltage, normalized with respect to the half-wave voltage, presented in degrees, i.e., 180° corresponds to an applied voltage equalto V1. The intrinsic phase of the device was 70° when this data was collected. Therange of the applied signal of ± 5° corresponds to an applied voltage of 10.4 voltspeak to peak. Figure 4.12 shows the fitting of the filtered data. The standard errorwas reduced from 0.8% to 0.3%, which was still slightly larger than that is suggestedfor metering[85,86]. While for device 89, whose intrinsic phase is 106°, the filteringreduced the standard error from 1.8% to 0.3%.4.4.3 Impulse MeasurementsTransient measurements, including high-voltage, switching impulses, andlightning impulses, are typical applications for high-voltage sensors. The standard testimpulse for switching impulse has a virtual rise time of 250jts and tail time of 300Os,while the standard test impulse for lightning impulses has a virtual rise of 1 .5s and atail time of 50s. Here, the defmition of the virtual rise time is 1.67 times the time it79(a)p-pC 1)= . 063 V Vp—pC2) 142 .2mV VavgC2)= 163 .6rnV1 4,5.OOV 2 20.0?j +—O.OOs 2.O0/Figure 4.10 Oscilloscope images of the linear responses of(a) device 89 and (b) device 91.801.0 -0.8 -Unfilterred data- Fitted curve(I) ñR_4-1 .J.s.)CD0.3--D00C-(1)-4-’1Do—0.8-I— I .L..)— I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I—6.0 —4.0—2.0 0.0 2.0 4.0 6.0Applied voltage normalized to the halfwave voltage (in degrees)Figure 4.11 Linear fit of unfiltered data for device 91.811.0 -0.8 -Filterred dataFitted curve(JD.4_JCD0.3 -0.00-.9—0.2 -(1):3——0.8 -—6.0 —4.0 —2.0 0.0 2.0 4.0 6.0Applied voltage normalized to the halfwave voltage (in degrees)Figure 4.12 Linear fit of filtered data for device 91.82takes the signal to rise from 30% to 90% of the peak value[85]. The impulsemeasurements were performed by applying a sharp, raised-triangle wave to thedevices. The magnitude of the applied signals was 1/3 of the half-wave voltage. Thesignal was supplied from a waveform generator. From the oscilloscope images shownin Figure 4.13, one can see that the responses could follow the input signals for the15s raise times but failed to do so for the 1 .5s raise times. This was probably dueto the large capacitance between the planar electrodes on the faces of the devices,which was measured to be 250nf. These devices should be capable of sensingswitching impulses and might be able to perform the lightning impulse measurements,with better results when used as individual devices having smaller capacitances.4.4.4 Stability TestsThe stability of the device parameters is an important aspect of a sensor if it isto serve as a monitoring device. Tests on the stability of the intrinsic phases, withrespect to temperature changes and time, were carried out. The thermal tests werecarried out by placing the sample on a heated stage in which a temperature sensingelement was assembled to monitor the temperature of the devices. The intrinsicphases of the devices changed quite drastically with increasing temperature. Figure4.14 shows the results of tests done on devices 90 and 94. The rate of change of theintrinsic phases of these two devices were 3.9°/°C and 4.9°/°C, respectively, whichwere much larger than those of the integrated optics Pockels cell[85], but were83Figure 4.13 Oscilloscope images of impulse response for device 90:(a) 15s raise time and (b) l.5Ms raise time.(a)(b)8420.O(1) 0.0-ci-)—20.0 - A-o . 0 ——40.0 0 Z—60.0 A/o — A-c-8O.ODD°°° Device 94—100.0:Device 90—120.0 - 111111 11111111111 I I 11111111 I 11111 I I I I I I I I150 20.0 25.0 30.0 35.0 40.0Temperature (in degree C)Figure 4.14 Inthnsic phase drift with temperature for devices 90 and 94.85comparable with those of other IMZ devices tested in our lab.The intrinsic phases of the devices were also found to drift with respect totime. This drift with respect to time might be responsible for the random intrinsicphases we observed in table 4.1. Figure 4.15 shows the intrinsic phase drift of device91. One reason for the drift could be the photorefractive effect, which, presummably,is increased in the regions with the high Ti concentrations. This problem of theintrinsic phase drift has to be solved before the device can be used in a monitoringapplication. However, if these devices were to be used as high-speed modulators, thelow frequency drifts in the intrinsic phase could be controlled using a counter actingelectric field.The outputs of some devices was flattened when large positive voltages wereapplied. A case of this is shown in Figure 4.16. The reason for this is not currentlyunderstood.4.5 Comments on the APE MethodTo improve the device characteristics, an annealed proton-exchanged method issuggested to form the waveguides after the Ti-induced inversion; annealed proton• exchanged waveguides have larger refractive index changes, which also, will improvethe guiding properties of the waveguides in the Ti-inverted regions. The annealedproton-exchange process is a low temperature process and therefore, will not inducedomain inversion. Thus, it should improve the cumulative electro-optic effect in bothof the branches.86110-100-90- *(I)ci)80-*ci)- 70-*60-(I)C *_c J040*(1)C30- *--JC20-10— I0 1 2 3 4 5 6 7 8 9 10 11Time ( hours )Figure 4.15 Intrinsic phase drift with time for device 91.87Figure 4.16 An Oscilloscope image shows the flattenning for high voltages,taken for device 91.j 1b51).L)y 2 100V —tj.OQs [j88Chapter 5 Summary andRecommendations5.1 IntroductionA summary of results for two immersion-type integrated optics asymmetricMach-Zehnder high-voltage sensors, namely the strip-loaded IMZ and domain-invertedIMZ, will be given in section 5.2. In section 5.3, along with some otherrecommendations, several devices based on domain-inverted IMZs are proposed tomake velocity-matched high-speed modulators.5.2 SummaryTwo immersion type integrated optics asymmetric Mach-Zehnder high-voltagesensors have been designed, fabricated, and tested. One is the strip-loaded IMZ witha loading strip on one of its two branches, and the other is the domain-inverted IMZwith domain inversion in one of its two branches.For the strip-loaded devices, the properties of the loading TiO2 films werestudied using ellipsometry and X-ray photoelectron spectroscopy (XPS). Therefractive index of the film was found to be 2.575 using ellipsometry and it was89shown in the XPS study that only one chemical state of Ti, Ti4, was present and,therefore only the rutile phase of Ti02 was present, which resulted from the hightemperature oxidations at 600°C. Such devices were fabricated and tested. The testresults indicated that there was a large loss in the strip-loaded branch. Theoreticalcalculations were performed to fmd possible reasons for the loss. The conclusion ofthis work was that the loss was probably due to mode coupling at the junctionsbetween the loaded section and unloaded section of the waveguide. It was shown thata tapering of loading strips could significantly reduce this loss. It was found that 5steps of iooA thick Ti02 films were needed to achieve 70% coupling. While such afabrication should be possible, it was felt that our facilities were not sufficient toguarantee the successful fabrication of these devices.Ti-induced domain inversion in the c face of LiNbO3 was studied. A defectprocess for the Ti indiffusion was proposed to explain the Ti-induced domaininversion. According to the theory of ferroelectrics, the electro-optic coefficient r33 inthe domain-inverted regions will be opposite to that of the bulk. The domain-invertedIMZ was first proposed, and then designed, fabricated, and tested. The measuredhalf-wave voltages of the devices ranged from 97 to 133 volts; the theoretical pushpull value was 96.6 volts. These results verified that the domain-inverted regionshave electro-optic responses opposite to those of the non-inverted regions whensubjected to equivalent electric fields. The differences between the measured valuesand the calculated values may be due to incomplete inversion in the inverted branches90or possibly some unintentional inversion in the uninverted branches. The linearities ofthe devices were tested; with proper filtering, the standard errors were found to beclose to the suggested 0.3% for an input range of 50 Impulse measurements on thedevices showed that they can follow impulses having up to 15s rise times but failedto do so for fast impulses with 1.5is rise times, this was probably due to the largecapacitance formed by the metal electrodes fabricated on the two faces. Thermalstability and time drift measurements on the devices were also conducted. Thedevices’ biases were found to be sensitive to the rise in temperature, havingtemperature coefficients about 5°I°C for 10mm long inverted regions. The biaseswere also found to drift with time. The performances of the devices generally agreedwith the design requirements; however, some problems, such as bias drift, need to beaddressed before these devices can be used as sensors.A. pre-out-diffusion method was proposed to resolve the surface guidingproblem associated with the Ti:LiNbO3waveguide fabrications. The method consistedof a pre-out-diffusion of Li at 1050°C for 24 hours and fabricating waveguides usingthe ‘soaking’ method. However, the surface guiding persisted, this was probably dueto H indiffusion in the wet ambient instead of lithium out-diffusion.5.3 Recommendations for Future WorkMore experiments should be done on the pre-out-diffusion method tocharacterize the effect of the wet ambient on the surface guiding. Perhaps one can91distinguish the reasons that may be responsible for the surface guiding in the wetambient. We suspect that for long diffusion time, the surface guiding will likely becaused by the indiffusion of H. By improving the pre-out-diffusion method, we maybe able to provide a practical solution to the surface guiding problem associated withTi-indiffused waveguide fabrication.Because proton-exchanged waveguides have a high refractive index change anda strong resistance to the photorefractive effect, further development of the proton-exchanged waveguide fabrication method should allow us to improve the deviceperformances of the domain-inverted devices.As the Ti-induced domain inversion method is improved, we should be able toutilize the push-pull effect between the inverted regions and uninverted regions tomake new modulators. One of the possibilities is to use this effect to make velocitymatching and quasi velocity-matching integrated optics high-speed modulators. Oneof the problems with high-speed optical modulators is the mismatch of the microwavevelocity and the optical wave velocity[87]. There are two general methods to resolvethis problem. One is to match the phase velocity of the optical waves and themicrowaves; for example, the velocity matching for devices based on LiNbO3 wasshown by increasing the microwave velocity to match the optical velocity [88,89].The other method is to use periodic electrodes to reverse the direction of the appliedelectric field within the optical waveguide at the ‘walk-off’ distance,( i.e., the distancein which the optical wave and the microwave to become 180° out of phase, due to the92velocity mismatch) to compensate for the phase walk-off between the microwave andoptical wave[90,91 ,92]. One such electrode, with a waveguide, is shown inFigure 5.1.Using domain-inverted regions, it would be straight forward implement thequasi velocity-matching method. Here, the domain-inverted regions can be patternedperiodically along the waveguides with a periodic length equal to that of the walk-offdistance between the microwave and optical wave, as shown in Figure 5.2. A quasivelocity-matched high-speed modulator can be made combining this IMZ with either amicrostrip line, a slot line, a coplanar strips, a coplanar waveguide, or a number ofother standard microwave transmission lines, where the periodic domain inversionallows us to realize the phase reversal effect. Here, there is an advantage to begained, for the straight electrode, the microwave loss is substantially reduced, ascompared to the periodic electrodes of Figure 5.1.93A plan view of phsereversl qusiveIocinching elecEnode wiEh wveguideN\\\\____________Elecirodes FS VveguideFigure 5.1 A phase-reversal, quasi velocity-matching electrode anda wavegude.94An 1iZ wilh periodic dornininverLed regionsdoveguide regionsOornininverLed regionswlkoFF disLnceFigure 52 A domain-inverted IMZ can be used to make a quasi velocity-matchedhigh-speed modulator that may be combined with a number of standardmicrowave transmission lines.95References:[1] A.J.Rogers,”Optical technique for measurement of current at high voltage’,Proc. lEE, vol.120, no.2. pp.261,1973.[2] G . A. Massey ,D. C. Erickson,and R. A. Kaldec, ‘Electromagnetic fieldcomponents: Their measurement using linear electro-optic and magneto-opticeffects’ Appl.Opt., vol.14, no.11, pp.2712, 1975.[3] R.Hebner,R.A.Malewski, and E.C.Cassidy,’Optical methods of electricalmeasurement at high voltage levels’, Proc. IEEE vol.65, no.11, pp. 1524,1977.[4] T. Mitsui, K.Hosoe, H.Usami, and S.Miyamoto,’ Development of fibre-opticvoltage sensors and magnetic-field sensor’, iEEE Trans. Power Deliv., vol. 2,no.1, pp.87, 1987.[5] Y. Kuroda, Y.Abe, H.Kuwahara, K.Yoshinaga,’Field test of fibre-opticsvoltage and current sensors’, SPJE vol.586, Fibre Optic Sensors, p.30,1985.[6] M. Kanoi, G. Talcanhashi ,T. Sato, Migaki, E. Mori, K. Okumura, ‘Optical voltageand current measuring system for electric power systems’ ,IEEE Trans. PowerDeliv. vol. 1,no. l,pp.9l, 1986.[7] T. Yoshino,’Optical fibre sensors for electric industry’ ,SPIE, vol.798,pp.258,1987.[8] S.J.Huang, D.C.Erickson,’The potential use of optical sensors for the96measurement of electric field distribution’ ,IEEE Trans. Power Deliv. vol.4no.3, pp.1579, 1989.[91 T. Sawa, K.Kuruosawa, T.Kaminishi,and T.Yokota,’Development of opticalinstrument transformers’ ,IEEE/PES 1989 Transmission Distributionconference, 89TD 380-7 PWRD.[10] R.E.Hebner and S.R.Booker,’A portable Kerr system for the measurement ofhigh voltage pulses’ ,Proc. IEEE SOUTHEASTCON., vol.1, pp.1A-i-i, 1975.[11] I.E. Thompson, M. Kristiansen,and M.O.Hagler, ‘Optical measurement of highelectric and magnetic fields’, IEEE Trans. Instrwn. Meas., vol. 25, pp.1, 1976[12] N.A.F.Jaeger,’Integrated Optics Pockels Cell Voltage Sensor’ ,U.S.Patent #5,029,273, July 2, 1991.[13] M.J.Ahmed,’Integrated optical devices in lithium niobate’,Ph.D. dissertation,Univ. British Columbia, Vancouver,B.C., Canada, 1981.[14] N. A. F .Jaeger, ‘Integrated optical devices in lithium niobate’ ,M. A. Sc. thesis,Univ. British Columbia, Vancouver, B.C. ,Canada, 1985.[15] N.A.F.Jaeger and L. Young, ‘Asymmetric Slab and Strip-Loaded IntegratedOptic Devices for the Measurement of Large Electric Fields’, J.Light. Tech.,vol.LT-5, pp. 745, 1987.[16] M. Di Domenico and R. Wemple,’Oxygen-Octahedra Ferroelectric I: Theoryof Electro-optical and Nonlinear Optical Effects’, J.Appl. Phys., vol. 40, pp.720,1969.97[17] I.P. Kaminow, ‘Optical waveguide modulators’, IEEE Trans. Micro. TheoryTech., vol.23, pp.57, 1975.[18] H.Furuta, H. Noda,and A.Ihaya, ‘Novel optical waveguides for integratedoptics’, App!. Optics, vol. 13, PP. 322, 1974.[19] V.Ramaswamy, ‘Strip-loaded film waveguides’, Bell Syst. Tech. J., pp.698,1974.[20] N.Uchida, ‘Optical waveguide loaded with high refractive-index strip film’,Appl. Opt. vol.15, no.1, pp.179, 1976.[21] J.Noda,M.Fukuma,S.Saito, ‘Strip-loaded waveguide formed in a graded-indexLiNbO3 planar waveguide’, Appi. Opt., vol. 17, pp. 1953, 1978.[22] T. Tamir, ‘Guided wave optoelectronics’, Sringer-Verlag,Berlin, Heidleberg,pp.63,1990.[23] G.V.Sansonov,’The oxide handbook’,IFI/Plennm. Data Company, 1982.[24] R.S. Weis and T.K. Gaylord, ‘Lithium Niobate: Summary of PhysicalProperties and Crystal Structure’, Appl.Phys.A, vol.37, pp. 191, 1985.[25] S..Abrahams,J . M. Reddy,and 3. L. Bernstein, ‘Ferroelectric lithium niobate 3.Single,.crystal x-ray diffraction study at 24°C’, J.Phys. Chem.Solids, vol.27,pp.997, 1966.[26] S.C. Abrahams,J.M. Reddy ,and J . L. Bernstein, ‘Ferroelectric lithium niobate 4.Single crystal neutron diffraction study at 24°C’, J.Phys. Chem.Soiids,vol.27,pp.997, 1966.98[27] S.H.Wemple, M.DiDomenico,Jr. ,and I.Camlibel, ‘Relationship between linearand quadratic electro-optic coefficients in LiNbO3,LiTaO3 and other oxygen-octahedra ferroelectrics based on direct measurement of spontaneouspolarizations’, Appi. Phys. Lett., vol.12, no.6, pp.209, 1968.[29] S.Miyazawa, ‘Ferroelectric domain inversion in Ti-indiffused LiNbO3 opticalwaveguide’, J.Appl. Phys. vol. 50, pp. 4599, 1979.[30] K. Nassau,H.J. Levinstein,and G.M. Loiacono, ‘Ferroelectric lithium niobate 1.growth,domain structure, dislocations and etching’, J.Phys. Chem.Solids,vol.27, pp.9831966[31] N.Ohnishi and Lizuka,’Etching study of microdomains in LiNbO3 singlecrystal’, J.Appl.Phys., vol.46, no.3, pp.1063, 1974.[32] S. M. Shapiro,R.W. Gammon ,and H. Z. Cummins, ‘Visual observation offerroelectric domains in TGS’, Appl.Phys.Lett., vol.10, pp.113967.[33] S.Thaniyavarn,T.Findkly,D.Booher,and J.Moen, ‘Domain inversion effects inTi:_LiNbO3integrated optical devices’, Appl.Phys.Lett. vol. 46, pp. 933,1985.[34] B.Guçnais,M.Baudet,M.Minier,and M.Le Cun,’Phase equilibria and Curietemperature in the LiNbO3-xTi2system, Investigated by DTA and x-rayDiffraction’, Mat. Res.Bull.vol .16, pp.643, 1981.[35] J.R.Peuzin, ‘Comment on “domain inversion effects in Ti_LiNbO3integratedoptical devices”, Appl.Phys.Lett. vol.48, pp.1104, 1986.99[36] O’Bryan ,H. M. Gallagher,P.K. Brandle,C . D. ‘X-ray studies of congruentLiNbO3’, J.Am.Ceram.Soc. vol.68, pp.4.93, 1985.[37] P.Lerner, C.Legras, and J.P.Dumas,’Stacking faults in congruent LiNbO3’, J.Ciyst. Growth, vol.3/4, pp.231, 1968.[38] H.Fay, W.J.Alford,and H.M.Dess,’Dependence of second-harmonic phasematching temperature in LiNbO3 crystals on melt composition’,Appl.Phys.Lett.,vol. 12, no.3, pp.89,(l968).[39] S.C. Abrahams and P.Marsh,’Defect structure dependence on composition inlithium niobate’, Acta Cryst. B43, 61,(1988).[40] F.A.Kroger,’The chemistry of imperfect crystals’, Amsterdam, NorthHolland, 1964.[41] M.E.Twigg, D.J.Eaglesham, D.M.Maher, S.Nakahara, and R.J.Holmes,’Thecrystallography and defect chemistry of structure faults in lithium niobate’,J.Appl.Phys., vol.63, pp.52951988.[42] N.M. Armenise,C. Canali,M.De Sario,A. Carnera,P.Mazzoldi, and G. Celotti,‘Characterization of Ti02 LiNbO3 and (Ti065Nb035)02 compound growthobserved during Ti LiNbO3 optical waveguide fabrication’,J.Appl.Phys.,vol.54, pp.62,1983.[43] C.S.Lau,P.K. Wei, C.W.Su,and W.S. Wang,’Fabrication of magnesium-oxide-induced lithium outdiffusion waveguides’, IEEE Photon. Tech. ,Vol.4, no.8, pp.872, 1992.100[44] R.J.Holmes and D.M.Smyth,’Titanium diffusion into LiNbO3 as a function ofstoichiometry’, J. Appl.Phys. vol.55, no.10, pp.3531, 1984.[45] H.Jhans,J.M.Honig,and C.N.R. Rao,’Optical properties of reduced LiNbO3’,J.Phys.C: Solid State Phys.,voLl9, pp.3649, 1986.[46] R.J. Jergerson and R.W. Bartlett,’High temperature transport process inlithium niobate’, J. Phys. Chem. Solids, vol.30, pp.2639,1969.[47] Y.Limb, N.Cheng,and D.H.Smyth, ‘Composition and electrical properties inLiNbO3’, Ferroelectric, pp.813,1981.[48] D.M.Smyth, ‘Defects and transport in LiNbO3’, Ferroelectric, vol.50,pp.93,1983.[49] N.A.F.Jaeger and B. Tsou, ‘Calculation of the fundamental mode-sizes inoptical channel waveguides using Guassian quadrature’ ,IEEE Trans.MicrowaveTheoiy Tech. to appear in June 1993.[50] H.K. Gummel,’A self-consistent interactive scheme for one-dimensional steadystate transistor calculation’ ,IEEE Trans.EIec.Dev. ,pp.445,l964.[51] M.Minakata,S.Saito, and S.Miyazawa, ‘Precise determination of refractive-index changes in Ti-diffused LiNbO3 optical waveguides’, Appi. Phys. Lett.vol.49, no. 9 pp.4.6’T7, 1978.[52] J.L. Jackel, ‘Proton exchanged for high-index waveguide in LiNbO3’,Appl.Phys.Lett., vol.41, 607,1982.[53] M.Fukuma,J.Noda,and H.Iwasaki,’Optical properties in titanium-diffused101LiNbO3 strip waveguides’,J.Appl. Phys. vol.49, no.7,pp.3693, 1978.[54] M.N.Armenise, ‘Fabrication techniques of lithium niobate waveguides’, lEEproceedings vol.135, pp.85,1988.[55] I.P.Kaminow and J.R.Carruthers,’Optical waveguiding layers in LiNbO3 andLiTaO3’, App!. Phys.Lett. ,vol.22, pp.326,1973.[56] J.R. Carruthers, I.P.Kaminow,and L.W.Stulz,’Diffusion kinetics and opticalwaveguide properties of outdiffused layers in Lithium Niobate and LithiumTantalate’, App!. Optics, vol.13,no.10, pp.2313, 1974.[57] R.V.Schmidt,et al,’ Metal-diffused optical waveguides in LiNbO3’,App!.Phys.Lett., vol.25, no.8, pp.458,1974.[58] K.K Wong,’ Integrated optical waveguides and devices fabricated by protonexchange: a review’, SPIE vol. 933 Integrated Optical Circuit Engineering VI,pp.13, 1988.53[59] A. Yi-Yan, ‘Index instability in proton-exchanged LiNbO3 waveguides’,App!.Phys.Lett., vol42, pp.633,1983.[60] R.A.Becker,’Comparison of guided-wave interferometric modulators fabricatedon LiNbO3 via Ti indiffusion and proton exchange’, Appl.Phys.Lett. ,vol.43,131,1983.[61] M.Minakata,K.Kumagai,and S.Kawakami, ‘Lattice constant changes andelectro-optic effect in proton-exchanged LiNbO3 optical waveguide’,Appi. Phys.Lett. vol.49,pp.922, 1986.102[62] P.G.Suchoski, et al, ‘Stable low-loss proton-exchanged LiNbO waveguidedevices with no electro-optic degradation’, Opt. Soci.Am. ,vol. 13 ,pp. 1050, 1988.[63] A.Loni,et al,’Proton-exchanged LiNbO3 waveguides: The effects of post-annealing and buffered melts as determined by infrared spectroscopy, opticalwaveguide measurements and hydrogen isotopic exchange reactions’, J. Light.Tech. vol. 7, pp.911, 1989.[64] J.Jackel, A.M.Glass, G.E.Peterson, C.E.Rice, D.H.Olson, and J.J.Veselka,‘Damage-resistant LiNbO3 waveguides’, J. Appi. Phys., vol.55, no.1, pp.269,1984.[65] B.L.Weiss,’The characteristics of optical waveguides fabrication in y and zcut LiNbO3 by He implantation’, J.Appl.Phys., vol.80, pp.4.64, 1986.[66] N.A.F.Jaeger and L.Young,’Voltage-induced optical waveguide modulator inLithium Niobate’, IEEE J. Quantum. Elec., vol.25, no.4,pp. 720,1989.[67] J.Noda, et al, ‘Electro—optic amplitude modulation using three-dimensionalLiNbO3 waveguide fabricated by Ti02 diffusion’, Appi. Phys.Lett. vol. 27,191975.[68] J.L.Jackel,’Suppression of outdiffusion in titanium diffused LiNbO3: AReview’ J. Opt. Commun, vol.3,82, 1982.[69] M.De Sario, M.N. Armenise, C.Canali, A.Carnera, P. Mazzoldi, andG.Celotti, ‘Ti02 LiNbO3,and (TiNbO1.JO2compound kinetics during theTi, LiNbO3 waveguide fabrication in the presence of water vapours’, J. Appi.103Phys., vol.57, pp. 1482, 1985.[70] E.M. Conwell, ‘Modes in anisotropic optical waveguides formed by diffusion’,IEEE J.Quanrum.Elec., vol.10, no.86, pp.608, 1974.[71] C. Canali, C.De Bernardi, M. De Sario, A. Loffredo, G. Mazzi, andK.Morasca, ‘Effects of water vapour on refractive index profiles in Ti:LiNbO3planar waveguides’, J. Light. Tech., vol.4, no.7, pp.951, 1986.[72] E.H.Tumer,F.R.Nash,and Bridenbough, ‘Dependence of linear electro-opticeffect and dielectric constant on melt composition in lithium niobate’, J.Appl.Phys., vol.41, no.3, pp 5278, 1970.[73] Holman,P.J.Cressman, and J.F. Revelli, ‘Chemical control of optical damage inlithium niobate’,AppLPhys.Lett., vol.32, no.5, pp.280, 1978.[74] O.Eknoyan, A.S. Greenblatt, W.K.Bums, and C.H. Bulmer, ‘Characterizationof Ti:LiNbO3deep waveguides diffused in dry and wet oxygen ambient’, Appi.Optics., vol.25, no.5, pp.737, 1986.[75] R.Willians,’Modern GaAs processing methods’,Artech House,pp.135, 1989.[76] J.J.Veselka and G.A.Bogert, ‘Low-insertion-loss channel waveguides in LiNbO3fabricated by proton exchange’, Elec.Lett., vol.23, no.6, pp.265, 1987.[77] G . T. Harvey, G. Astfalk, A. Y. Feldblum ,and B. Kasun,’The photorefractiveeffect in titanium indiffused lithium niobate optical directional couplers atl.3m’, IEEE J. Quantum.Elec., vol.22, no.6, pp.939, 1986.[78] G. Hass, ‘Preparation, properties and optical application for thin film of104titanium dioxide’, Vaccum., vol.11, pp.331, 1952.[79] J.Stringer,’The oxidation of titanium in oxygen at high temperature’, ActaMetal., vol.8, pp.’759, 1960.[80] E.A.J. Marcatili, “Dielectric Rectangular Waveguide and Directional Couplerfor Integrated Optics”, Bell Sys. Tech. J. pp 2071, 1969.[81] D. Marcuse,”Radiation Losses of Tapered Dielectric Slab Waveguides”, BellSys. Tech.J. ‘pp 273, 1970.[82] C.J.G.Kirkby,’Refractive index of lithium niobate, wavelength dependence:tables’, RN= 16002 in ‘Properties of Lithium Niobate’, EMIS DataviewsSeries no 5. 1988.[83] C .J. G . Kirkby, ‘Electro-optic coefficients of lithium niobate: tables’,RN= 16013 in ‘Properties of Lithium Niobate’, EMIS Dataviews Series no 5.1989.[84] K.Nakamura, H.Ando,and H.Shimizu,’Ferroelectric domain inversion cause inLiNbO3 plates by heat treatment’, Appl.Phys.Lett. vol.50, no.20, pp. 1413,1987.[85] F.Rahmatian,’Integrated optics Pockels cell high voltage sensor’, M.A.Sc.thesis, Univ. of British Columbia, Vancouver, B.C. Canada, Jan.,1993.[86] D.C.Erichson, ‘The use of fibre optics for communications, measurement andcontrol within high voltage substations’,IEEE Trans. P.A.S., vol.99, no.3,pp. 1057, 1980.105[87] R. C. Alfemess,S .K.Korotky,and E. A.J. Marcatili, ‘Velocity-matching techniquesfor integrated optic travelling wave switch/modulators’ ,IEEE J. Quan. Elec.,vol.20, no.3, pp.301, 1984.[88] K. Kawano,T. Kitoh,H.Jumonji,T. Nozawa,and M. Yanagibashi, ‘Velocitymatched LiNbO3 waveguide modulator with 20 GHz bandwidth and 4.7Vdriving voltage at 1. 52gm wavelength’, Electr. Lett. ,vol.25 ,no. 20,pp.1382,1989.[89] T. Yoneyama,K. Niinuma,and S. Kanno, ‘Velocity-matched LiNbO3 waveguideoptical modulator using inverted slot line’ ,IEEE Microwave Guided Wave Lett.,vol.1, no.8,pp.192,1991.[90] S.K.Korotky,and J.J.Veselka, ‘Efficient switching in a 72-Gbitls Ti:LiNbO3binary multiplexer/demultiplexer’ , OFC’90,TUH2 ,pp.32, 1990.[91] M. Nazarathy,D.W.Dolfi,and R.J.Jungerman, ‘Spread spectrum frequencyresponse of coded phase reversal travelling wave modulators’, J.Light. Tech.,vol.5, no. 10,1987.[92] R.L.Jungerman and D.W.Dofi, ‘Coded phase-reversal LiNbO3 modulator withbandwidth greater than 20 GHz at 1.3gm wavelength’,Electr.Lett., vol.23,no.4, pp.172,1987.3. 23106


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items