University of Southern California Dissertations and Theses

Advanced hybrid bulk/integrated optical signal processing modules

(USC Thesis Other)

Advanced hybrid bulk/integrated optical signal processing modules

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ADVANCED HYBRID BULK/INTEGRATED OPTICAL SIGNAL PROCESSING MODULES by Scott David DeMars A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Electrical Engineering) December 1995 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 9733045 UMI Microform 9733045 Copyright 1997, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by Scott David DeMars under the direction of hX5..... Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean of Graduate Studies Date .J5e.cetoJbLer..2Q.,„.L95.S.. DISSERTATION COMMITTEE Chairperson R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. A cknow ledgm ents I would first like to thank my parents for their belief in me throughout my entire education, for their love, and for their support. They have always been a reliable source of warmth and acceptance regardless of the circumstances. I am indebted to my research advisor, Professor Armand R. Tanguay, Jr., for his unwavering support and confidence in me throughout the course of this research. He has made many self- sacrifices to provide me with guidance through difficult times, for which I will always be grateful. With his vision, technical acuity, enthusiasm for research, and clear pre sentation style he has provided me with a rare education by example that has been invaluable to my own development as a scientist. My deepest thanks go to my wife, Jinie DeMars, and her family for their love and perseverance throughout my doctorate studies. Their aid with my outside responsibilities throughout the final stages of my research and the comfort that they gave to me were tremendous gifts. I would like to thank the members of the Optical Materials and Devices Labora tories at USC. Dr. Kasra Rastani developed the foundation for this work and provided many useful consultations. Dr. Zaheed Karim also contributed significantly to the results presented herein with his expertise in thin film deposition and related areas. I found the technical discussions with my colleague and friend Mr. Edward Herbulock invaluable to the advancement o f this research, and his efforts to review this disserta tion are greatly appreciated. Assistance in the device fabrication from Mr. Kartik Ananthanaryanan was critical in the final research stages. Dr. Greg Nordin, Dr. Chris Kyriakakis, Dr. Jungje Jung, Dr. John Rilum, and Dr. Praveen Asthana always volun teered their assistance when it was needed, and Ms. Karen Tierney’s exceptional administrative skills were instrumental to the flow of this research and preparation of this dissertation. I also thank Professor M. Gershenzon, Professor W. H. Steier, Professor A. A. Sawchuk, and Professor A. Willner for serving on my dissertation committee, and both General Dynamics and the Army Research Laboratory for their program support. ii R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Acknowledgm ents ii List o f Figures vii A bstract xix Chapter 1 Introduction to the Research Topics 1 1.1. Synthetic Aperture Radar (S AR) Image Formation 2 1.2. Status of Integrated Optical Technology 9 1.3. Development of Key Components 12 1.4. Extension of the Concept to Signal Correlation 18 1.5. Demonstration of Key Component Integration 23 1.6. Alternative Waveguide Materials 26 1.7. Outline of the Dissertation 27 1.8. References 29 Chapter 2 System Design Considerations 37 2.1. Acousto-Optic/CCD S AR Processor 38 2.2. Acousto-Optic Interferometric Time-Integrating Correlator 46 2.3. Preliminary Device Considerations 49 2.4. Integrated Optical Synthetic Aperture Radar (IOS AR) Processor 63 2.4.1. IOS AR Processor Geometry 64 2.4.2. IOSAR Design Parameters 67 2.5. Integrated Optical Correlator 74 2.6. Summary 76 2.7. References 81 Chapter 3 W aveguide Design and Fabrication 85 3.1. Lithium Niobate 86 3.2. Gallium Arsenide 89 3.3. Waveguide Modeling 90 3.3.1. Graded Index Waveguides 91 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.3.2. Waveguide Fabrication Sensitivities 93 3.4. Waveguide Fabrication Processes 96 3.4.1. Titanium-Indiffused Lithium Niobate Waveguides 96 3.4.2. Gallium Arsenide Waveguides 100 3.5. Waveguide Characterization 101 3.5.1. Mode Structure 101 3.5.2. Guided Mode Intensity Profile 102 3.5.3. Propagation Loss Measurements 107 3.6. Source Coupling Methods 110 3.6.1. Prism Coupling 110 3.6.2. End-Fire Coupling 113 3.7. Waveguide End-Polishing 114 3.8. Summary 116 3.9. References 117 Chapter 4 Rib W aveguide Arrays with Surface O utcoupling Gratings 123 4.1. Rib Waveguide Arrays 128 4.1.1. Incident Field Distribution 128 4.1.2. Rib Waveguide Field Distribution 133 4.1.3. Incident Field Coupling 13 8 4.1.4. Rib Waveguide Scattering Losses 140 4.1.5. Rib-to-Rib Coupling 141 4.1.6. Waveguide Component Fabrication Techniques 146 4.1.7. Rib Waveguide Array Fabrication on Lithium Niobate 152 4.1.8. Acceptance Angle Measurement 156 4.1.9. Rib Waveguide Scattering Loss Measurement 158 4.1.10. Rib-to-Rib Coupling Measurement 159 4.1.11. Rib Waveguide Array Summary 166 4.2. Surface Outcoupling Gratings 167 4.2.1. Spatial Harmonics 168 4.2.2. Grating Attenuation Constants 170 4.2.3. Grating Outcoupling Efficiency 173 4.2.4. Outcoupling Grating Fabrication on Lithium Niobate 174 4.2.5. Grating Outcoupling Uniformity 175 4.2.6. Grating Tolerance Analysis 177 4.2.7. Outcoupling Grating Summary 179 4.3. Rib Waveguide Arrays with Surface Outcoupling Gratings 180 4.3.1. Mode Beating in Rib Waveguides 181 iv R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.3.2. Surface Component Illumination Uniformity 184 4.3.3. Transverse Spread of the Outcoupled Light 187 4.3.4. Rib Waveguide Array and Surface Grating Integration 190 4.3.5. Observed Grating Outcoupling from Rib Waveguides 195 4.3.6. Observed Grating Outcoupling Uniformity 198 4.3.7. GaAs Rib Waveguide Array with Surface Gratings 201 4.3.8. Rib Waveguide Array with Surface Gratings Summary 204 4.4. References 206 Chapter 5 Embedded Lenses in Lithium Niobate W aveguides 212 5.1. Review of Waveguide Lens Structures 214 5.2. Geometric Interpretation of an Embedded Lens 218 5.2.1. Single-Element Aberration-Corrected f/1 Lens 223 5.2.2. Flat-Field Doublet 225 5.3. Guided-Wave Interpretation of an Embedded Lens 227 5.3.1. Design of the Barrier Layer Thickness 231 5.3.2. Optimization of Mode Coupling 235 5.4. Fabrication Limitations and Related Effects 238 5.4.1. Fresnel Reflections at a Tilted Interface 240 5.4.2. Mode Coupling at a Tilted Interface 241 5.4.3. Interface Roughness Scattering 244 5.4.4. Tolerance Analysis 246 5.4.5. Waveguide Anisotropy 248 5.4.6. Estimation of Embedded Lens Structure Throughput 253 5.5. Fabrication of Embedded Lenses in Lithium Niobate 253 5.5.1. Development of Key Fabrication Processes 256 5.5.2. Embedded Lens Processing Sequence 267 5.6. Experimental Characterization of Embedded Lenses 272 5.7. Summary 282 5.8. References 284 Chapter 6 Surface Acoustic W ave M odulators 289 6.1. Overview of Inter-Digital SAW Transducers 289 6.1.1. Crystal Orientation and Physical Properties 291 6.1.2. Guided Optical Wave/S AW Acoustooptic Interaction 292 6.1.3. SAW Transducer Design 295 v R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6.2. Surface Acoustic Wave Transducer Fabrication 296 6.3. Surface Acoustic Wave Transducer Electrical Testing 298 6.4. Surface Acoustic Wave Optical Modulation 301 6.5. Summary 305 6.6. References 307 Chapter 7 Integrated Optical Test Modules 7.1. Test Module Geometric Considerations 7.2. Focusing with an Embedded Lens into Rib Waveguides 7.3. Component Integration Sequence 7.4. Description of Fabricated Modules 7.5. Excitation of a Single Rib Waveguide 7.6. Selective Focusing into Several Rib Waveguides 7.7. IOSAR Processor Assessment 7.8.10 Correlator Assessment 7.9. References Chapter 8 Conclusions 342 8.1. Summary 342 8.2. Future Research Directions 347 8.3. References 353 308 310 314 319 324 325 327 328 337 340 vi R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. List of Figures Figure 1.1 Synthetic aperture radar strip-mapping geometry (after [Psaltis and Wagner, 1982]). 4 Figure 1.2 Schematic diagram of the Integrated Optical Synthetic Aperture Radar Processor (after [Bicknell, 1980; Bicknell, etal., 1985; Rastani, 1988; Tanguay, 1988]). 7 Figure 1.3 Top view of the Integrated Optical Synthetic Aperture Radar Processor. The radar return is introduced by means of a surface acoustic wave transducer and the range focused image is formed at the beginning of the rib waveguide array (from [Rastani, 1988]). 8 Figure 1.4 Integrated optical spectrum analyzer fabricated in Ti:LiNb03 (from [Mergerian, et al., 1980]). 10 Figure 1.5 Acousto-optic matrix algebra processor on a LiN b03 sub strate (from [Kar-Roy and Tsai, 1991]). 11 Figure 1.6 Space-integrating correlator architecture (from [Nishihara, etal., 1989]). 20 Figure 1.7 Schematic diagram of the integrated optical interferometric time-integrating correlator, showing the hybrid integration of the laser diode, beam expanding lens (LI) and collimat ing lens (L2), input surface acoustic wave (SAW) trans ducer, Fourier transform lens (L3) and imaging lens (L4), spatial filter (SF), and linear detector (D) on top of a rib waveguide array with grating couplers. 21 Figure 1.8 Time-integrating correlator architecture that utilizes aniso tropic Bragg diffraction (from [Liao, et a l, 1982]). 23 Figure 1.9 Schematic diagram of optical processor subsystems show ing two levels o f component integration for performance evaluation purposes. Module (a) consists of an embedded lens aligned by linear fiducial markings to a rib waveguide array and end-polished substrate. Module (b) is identical to module (a) with the addition of a SAW transducer. 26 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 2.1 Acousto-optic/CCD S AR processor. The elements of the processor include: LS-light source, AOD-acoustooptic device, F; is the focal length of the lens L{ , and the broken lines indicate the path of the rays of the reference wave (from [Psaltis and Wagner, 1982]). 39 Figure 2.2 The upper diagram depicts the range compression tech nique described in detail in the text. The lower diagram shows the azimuth compression technique (from [Rastani, 1988]). 41 Figure 2.3 A typical mask function bearing the azimuth Doppler phase history in the vertical dimension and the range dependence of the azimuth phase history in the horizontal direction (from [Rastani, 1988]). 43 Figure 2.4 One-dimensional bulk optical correlator architecture with light source S, collimating lens LO, cylindrical lens LI, acoustooptic device AO, Fourier transform lens L4, spatial filter F2, imaging lens L5, and linear detector array D (after [Cohen, 1983]). 47 Figure 2.5 Top and side views of a generalized in-line integrated opti cal architecture, in which edge and surface mounted devices are in the dark shaded areas and waveguide devices are in the light shaded areas. The devices in each region include (I) an edge-coupled source element; (II) beam expanding and collimating lenses; (HI and IV) signal input devices, imaging or Fourier transform lenses, and beam stops; (V) channelized waveguide arrays, waveguide couplers, surface mounted devices; and (VI) an edge-cou pled detector device. 50 Figure 2.6 Two schematic diagrams that show a laser diode end-cou pled to a waveguide with (a) a beam collimation lens ele ment L I, and (b) both beam expansion lens element, L2, and collimation lens element, L3. 56 Figure 2.7 Geometrical optics of the guided beam incident upon a lin ear FM index modulation and an integrated lens combina tion (After [Rastani, 1988]). 65 viii R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Figure 3.1 Schematic diagram of titanium-indiffused lithium niobate waveguide showing the orientation and polarization of the guided mode. 89 Figure 3.2 Normalized dispersion curves for a waveguide with a Gaussian refractive index profile. 95 Figure 3.3 Schematic diagram of wafer dicing patterns for waveguide fabrication. 98 Figure 3.4 Mode profile for a titanium-indiffused lithium niobate waveguide obtained by the knife-edge scanning technique. 105 Figure 3.5 Mode profile for a titanium-indiffused lithium niobate waveguide obtained by imaging with a microscope objec tive. 106 Figure 3.6 Titanium-indiffused lithium niobate propagation loss mea surements made by (a) the sliding prism method, and (b) by the CCD imaging method. 109 Figure 3.7 Prism coupling arrangement in which the gap distance between the prism and waveguide is adjusted by the applied force. 111 Figure 3.8 End-fire coupling arrangement showing the convergence angle of the incoupling beam. This angle should be less than the acceptance angle of the waveguide for efficient coupling. 114 Figure 3.9 Waveguide end-polishing mount with clamped samples and surface mounted support pieces. 116 Figure 3.10 Microphotograph of the polished ends of two lithium nio bate samples stacked together with an intermediate layer of wafer dicing tape. 117 Figure 4.1 Rib waveguide array with surface outcoupling gratings showing (a) initial intensity distribution incident on a rib waveguide array and final intensity distribution coupled from the rib waveguides by surface outcoupling gratings, and (b) cladding and substrate radiation mode propagation directions. 124 _ . ix R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.2 Schematic diagram that illustrates the formation of the field distribution at the rib waveguide array for (a) the IOS AR processor, (b) the 10 correlator, and (c) the test module. 129 Figure 4.3 Schematic diagrams of two rib waveguide geometries, in which the high index material surrounding the rib has been (a) fully removed and (b) partially removed. 132 Figure 4.4 (a) Rib waveguide array geometry and (b) region of inte gration for calculation o f the coupling coefficient. 142 Figure 4.5 Calculated coupling coefficient for 8 fim wide ribs in a Ti:LiNb03 waveguide as a function of mode number and various gap widths. The curves were calculated in (a) for a gap depth of 2500 A and in (b) for a gap depth of 5000 A. 145 Figure 4.6 One unit of a rib waveguide array pattern designed for the purpose of measuring crosstalk values between rib waveguides with different coupling lengths, gap widths, and gap depths. 147 Figure 4.7 Fabrication sequence for (a) rib waveguides and (b) sur face outcoupling gratings. These two structures may be fabricated in either order. 148 Figure 4.8 Shown are three mechanisms that lead to a tilted sidewall and inaccurate pattern transfer in the argon-ion-beam etch ing process: (a) low mask selectivity results in mask shrinkage during the etch process, (b) bevel formation in the mask eventually reaches the substrate and also results in mask shrinkage, and (c) the effects of redeposition result in a tilted sidewall even though the initial mask is ideal. 151 Figure 4.9 Schematic diagram o f grid patterns used to dice wafers for component fabrication. The wafer dicing pattern in (a) produces samples approximately 1 in x 1 in and is used for waveguide device fabrication, and in (b) produces samples approximately 1.7 cm X 1.7 cm for use in the fabrication o f the rib waveguide crosstalk samples. 153 Figure 4.10 Prism coupling arrangement for measurement of the accep tance angle of a rib waveguide array. 157 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.11 Measured dB separation between the peak intensity level of a single excited rib waveguide and the peak intensity level of the neighboring rib waveguides for a 10 mm inter action length and various rib waveguide gap widths. 162 Figure 4.12 Measured dB separation between the peak intensity of a single excited rib waveguide and the peak level of the neighboring rib waveguides for 2 /nxn gaps and various rib waveguide interaction lengths. 164 Figure 4.13 Schematic diagram of surface outcoupling gratings on a planar waveguide. The light is coupled into the slab waveguide via a rutile prism. The mode excited in the pla nar waveguide is partially outcoupled by the rectangular surface outcoupling gratings. The parameters that deter mine the performance of the gratings are also shown in this figure (after [Rastani, 1988]). 169 Figure 4.14 (a) CCD-camera-acquired image of light outcoupled from a 1.2 cm X 1 cm uniform grating with a grating period of 4 fim and grating height of 150 A, and (b) a plot of the out coupled light intensity along the 10 mm grating length. 176 Figure 4.15 Fraction of the total power radiated into the cladding for two different grating periods, plotted as a function of the grating line width variation. 179 Figure 4.16 Schematic diagram of a surface outcoupling grating on a rib waveguide. The light is coupled into the slab waveguide via a rutile prism and the guided mode is par tially coupled into a set of rib waveguides. The modes that are excited in the rib waveguides are partially outcoupled by the rectangular surface outcoupling gratings. The parameters that determine the performance of the gratings are also shown in this figure (after [Rastani, 1988]). 182 xi R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.17 Side view of a rib waveguide with surface outcoupling gratings showing (a) the propagation directions of the clad ding radiation modes for a single waveguide mode, (b) the interference between planar phase fronts of two cladding radiation modes that are directionally close to one another, and (c) the close propagation directions of two cladding modes that originate from different waveguide modes. 186 Figure 4.18 Schematic diagram of the transverse spread of light out coupled by surface outcoupling gratings from the rib waveguide array structure. 189 Figure 4.19 Two approaches to mounting azimuth mask and CCD detector array above a densely packed rib waveguide array. In (a) a transparent dielectric material is deposited onto the surface of the detector array with the mask patterned in an absorbing thin film, and (b) an intermediate substrate is used with the azimuth mask and a cylindrical lenslet array to collimate the diverging light. 190 Figure 4.20 SEM cross-section photograph of a 2 Jim gratings fabri cated on GaAs by ion-beam milling. The rough textured surface is the sample edge. The roughness is due to break age of the sample and is not a result of cleavage. 194 Figure 4.21 SEM photographs that show (a) 8 ftm wide rib waveguides and 2 (im separations etched 0.5 fim deep in a GaAs/AlGaAs waveguide, (b) identical rib waveguides that also include 2 fim period surface outcoupling gratings, and (c) a high magnification view of the rib waveguide sidewall and gratings. 196 Figure 4.22 (a) CCD-camera-acquired image of light outcoupled from surface outcoupling gratings (1.2 cm x 4 mm grating region, 4 fjm grating period, and 150 A grating height) on top of a rib waveguide array (8 fim rib waveguide widths, 2 fim gap widths, 5000 A gap depths, 1000 elements, 5 mm long) fabricated on a lithium niobate waveguide and (b) a graph of the outcoupled light intensity along the 4 mm grating length. 197 xii R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.23 The top images are magnified views of grating outcoupled light from rib waveguides in lithium niobate that have 28 [im rib waveguide widths, 2 fim gap widths, and in (a) 2 fxm period gratings. The rib waveguide structure is the same in (b) but the gratings have a 4 fim period. The lower left image (c) is a magnified view of grating outcou pled light from the 8 Jim wide rib waveguides with 2 n m gaps and 2 fjm period gratings shown in (d). 199 Figure 4.24 Experimental arrangement used for measurement of the radiation mode propagation angles and outcoupling effi ciencies from a GaAs waveguide. 203 Figure 4.25 Measured relative radiation mode outcoupling efficiencies and angles for a GaAs/AlGaAs rib waveguide array with 2 jlim period surface outcoupling gratings. 203 Figure 5.1 An embedded lens in a titanium-indiffiised lithium niobate waveguide, including schematic diagrams of (a) the cross- sectional view o f the lens thin film structure and the rela tive mode profiles for a mode-depth, d, of approximately 2 fim , and (b) the top view of a single element-aberration corrected f/1 lens design with an aperture size of W = 1 cm within the thin film overlay region. 213 Figure 5.2 Reflection loss at each interface in an integrated lens struc ture for different host waveguide materials. 2 2 0 Figure 5.3 Shown in this figure is (a) the design for a single-element aberration-corrected f/1 lens in lithium niobate and (b) the coordinate system used for the definition of the acircular interfaces. 224 Figure 5.4 Design for a field-corrected lens doublet with a 1 inch aperture showing three chirped SAW signals and the result ing focal positions of the diffracted light on the entrance plane of a rib waveguide array. Also shown is the focal position o f the undiffracted light. 227 xiii R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Figure 5.5 Schematic diagram of the embedded lens structure divided into several regions for consideration of the guided wave properties. These include: the host waveguide Regions I and VII, the thin film overlayer Regions H and VI, the lens interface Regions HI and V, and the embedded lens Region rv. 229 Figure 5.6 Comparison of (a) higher index and (b) lower index embedded lens structures. The lower index structure has small losses associated with leakage through the barrier layer(s). The higher index structure has essentially no mode leakage as long as film deposition onto the host waveguide surface is prevented. 230 Figure 5.7 Embedded S iC > 2/MgF2 waveguide leakage rate for each of the modes as a function of the barrier layer thickness. 232 Figure 5.8 Embedded S iC > 2/MgF2 waveguide lens transmission as a function of beam position in the aperture, for different waveguide attenuation constants (given in dB/cm). 234 Figure 5.9 Mode coupling efficiency for various S i0 2 layer thick nesses (ranging from 1 ftm to 2.4 /mi) as a function of waveguide structure offset. 237 Figure 5.10 Propagation losses of the lowest-order mode of an SiC>2/MgF2 waveguide for various Si0 2 layer thicknesses and as a function of the barrier layer thickness. 238 Figure 5.11 Coupling efficiency between the Ti:LiNb03 mode and each of the Si0 2 /MgF2 waveguide modes as a function of offset between the waveguide structures. These calcula tions were performed assuming a 2 fim thick Si0 2 waveguiding layer. 239 Figure 5.12 Comparison of measured mode intensity profile for the titanium-indiffused lithium niobate waveguide and the cal culated mode profile for the SiC>2/MgF2 embedded waveguide structure with a 2 fim thick waveguiding layer. 240 Figure 5.13 Schematic diagram showing (a) the slanted sidewall in the embedded lens recess, and (b) a thin prism model of the slanted sidewall. 241 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Figure 5.14 Reflectivity at the Ti:LiNb0 3/SiC> 2 interface as a function of lens recess wall tilt angle. 242 Figure 5.15 Refraction angle of the light transmitted into the lens region as a function of wall tilt angle. The refracted angle is compared with the waveguide propagation angles. 243 Figure 5.16 Effect of a tilted sidewall on coupling efficiency into the Si0 2 /MgF2 embedded waveguide with a 2 fim thick guid ing layer. 244 Figure 5.17 Mode coupling efficiency from a Ti:LiNb03 waveguide into an embedded SiC > 2/M gF2 waveguide with a 12° inter face tilt angle. 245 Figure 5.18 Circle of least confusion for the SEACF/l lens in which the (a) nominal design refractive indices are used, (b) mea sured refractive indices are used, and (c) measured refrac tive indices are used and waveguide anisotropies are included. 249 Figure 5.19 Transmittance as a function of position within the aperture of an embedded Si0 2 /MgF2 lens structure in a Ti:LiNb03 waveguide for different sidewall tilt angles not accounting for scattering, mode-coupling, and propagation losses. 254 Figure 5.20 Total transmittance of an embedded SiC>2/MgF2 lens struc ture in a Ti:LiNb03 waveguide, including scattering and reflection losses from both surfaces as well as mode mis match and embedded lens propagation losses. 255 Figure 5.21 Deposition of dielectric materials into an etched recess. The sidewall tilt and Si0 2/M gF2 structure are shown. 268 Figure 5.22 Embedded lens processing sequence. 269 Figure 5.23 Multiple exposure photomicrograph of a 1-cm aperture MgF2/SiC> 2 lens embedded in a titanium-indiffused lithium niobate waveguide (sample Lens-16). For each exposure the narrow beam of light (k = 6328 A) incident on the lens from the left was shifted in the transverse direction so that the focusing effect of the lens would be evident. The resulting focal plane is visible about 1 cm to the right of the lens center. 273 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Figure 5.24 Photomicrograph of a MgF2/Photoresist lens embedded in a titanium-indiffused lithium niobate waveguide (sample Lens-1). The dark lines on the scale in the lower region of the photograph are separated by 1 mm. The lens is illumi nated by two thin guided beams ( k = 6328 A) from the left, and the resulting focal plane is visible about 1.3 cm to the right of the lens center. 275 Figure 5.25 Focal-spot intensity profile of a 1-cm aperture f/l embed ded lens fabricated in a titanium-indiffused lithium niobate waveguide. 276 Figure 5.26 Measured transmittance of embedded lens sample Lens-8 as a function of the ray position within the aperture. 279 Figure 5.27 Photomicrograph of sample Lens-16. This is a 1-cm aper ture MgF2/SiC> 2 lens embedded in a titanium-indiffused lithium niobate waveguide. A narrow beam of light was directed through the lens center, and the intensity of the surface-scattered light was measured before and after the lens to determine the transmittance. 280 Figure 5.28 Measured transmittance of embedded lens sample Lens-55 as a function of the ray position within the aperture. 281 Figure 6.1 SAW transducer designs for operation on titanium-indif fused lithium niobate waveguides. The transducer finger width is the critical dimension, CD, and is approximately 2 jlm. 290 Figure 6.2 Surface acoustic wave transducer on a lithium niobate waveguide showing six finger pairs and a 100 ^m x 100 fim bonding pad of aluminum metal. 298 Figure 6.3 Photograph of transducer fixture for optical beam deflec tion experiments with three wire-bonded transducer pairs. 299 Figure 6.4 Frequency response of transducer IDT1 (top), and IDT2 (bottom). The transducer bandwidths are approximately 100 MHz for each design, but the central frequencies differ from one another. 300 Figure 6.5 Optical diffraction efficiency as a function of input beam angle measured from the optical axis for three different drive frequencies within the bandwidth of IDT1. 303 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.6 Internal deflection angle as a function o f frequency for the SAW transducer design described in the text. 303 Figure 6.7 Diffraction efficiency with frequency when Bragg matched to the center frequency of IDT1. 306 Figure 6.8 Diffraction efficiency as a function of transducer drive power for three different drive frequencies. 306 Figure 7.1 Schematic diagram of integrated test module on a Ti:LiNb03 waveguide with integrated transducer, lens, rib waveguide array, and alignment fiducials. 309 Figure 7.2 Geometry for focal length compensation with an external lens. The external lens (not shown) is used to focus a colli mated beam a distance D0 in front of the integrated lens. The external focus is effectively a distance D , away from the lens as a result of the refraction of the coupled beam at the prism/waveguide boundary. The position of the focal spot is given by the imaging condition. 312 Figure 7.3 Calculated coupling efficiency for a beam focused into a 6 /J.m wide rib waveguide by an embedded lens as a func tion of incidence angle. Near normal incidence, the first three symmetric modes are predominantly excited. The coupling into the anti-symmetric modes increases with larger angular offset. 317 Figure 7.4 Calculated coupling efficiency for a beam focused into an 8/j.m wide rib waveguide by a chirped SAW signal and lens doublet combination. The calculated coupling efficiency into the neighboring rib waveguides separated by 2 fjm gaps from the center rib waveguide is also shown. 319 Figure 7.5 Schematic diagram of an integrated test module showing (a) SAW transducers, (b) thin film overlayers, (c) embed ded lens, (d) rib waveguide array, (e) Ti:LiNb03 waveguide, (f) chip carrier, (g) SMA connectors, and (h) aluminum base. 323 Figure 7.6 Measured end-emission from a rib waveguide array in which light is focused by an embedded lens into a single rib waveguide. 327 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 7.7 Measured end-emission from a rib waveguide array in which individual rib elements are selectively excited by a transducer and embedded lens (lens-17) combination inte grated on a Ti:LiN b03 waveguide (second test module). Each curve corresponds to a different transducer input fre quency. 329 Figure 7.8 Comparison o f measured frequency for maximum excita tion of individual rib waveguides in an array to the calcu lated beam deflection with frequency. 330 xviii R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract The research described in this thesis focuses on the advancement of novel hybrid bulk/integrated optical and optoelectronic device technology suitable for both inherently planar and inherently non-planar optical signal processing applications. Such applications include spread spectrum communications, high frequency pulse compression, wide bandwidth correlation for automatic target recognition, and syn thetic aperture radar (SAR) image formation, among a wide variety of comparably complex multi-dimensional signal processing tasks. The primary focus was on the development of key components for an integrated optical synthetic aperture radar (IOSAR) processor. These components include wide aperture, short focal length embedded waveguide lenses for collimating, imaging, and Fourier-transform opera tions, dense rib waveguide arrays for in-plane image dissection, and surface outcou pling gratings for uniform guided-wave coupling to external components. The application of these devices to the development of an integrated optical correlator was also considered. Wide aperture, short focal length waveguide lenses are required in the IOSAR processor to provide a large processing capacity within limited substrate and waveguide dimensions. We present for the first time embedded waveguide lenses developed in titanium-indiffused lithium niobate waveguides. The lens structure was created by ion-beam milling a lens-shaped recess into the planar waveguide, followed by the deposition of an Si0 2 /MgF2 dielectric waveguide into the recess. This lens operates with f/2 performance, a 1 cm focal length, a 3 dB insertion loss, and a 0.9 /an spot size inside the waveguide. Rib waveguide arrays with densely packed elements, low crosstalk, and low propagation losses are required for the efficient and accurate dissection and propaga tion of an in-plane image. The rib-to-rib coupling characteristics of various rib waveguide array structures (8 /an wide rib waveguides, 2-1 0 /an wide separations, 2500 A and 5000 A separation depths, and 2-12 mm rib waveguide lengths) on tita- xix R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. nium-indiffused lithium niobate waveguides were investigated. Rib waveguide arrays with 2 fim separation widths and 5000 A separation depths were found to have crosstalk less than -17 dB due to rib-to-rib coupling for a 1 cm propagation distance. Uniform grating outcoupling from rib waveguides over a large distance (1 cm) is necessary in the IOSAR processor to fully illuminate surface mounted devices. Surface relief gratings with 2 fim and 4 fixn periods were fabricated on the surface of low-loss planar (0.42 dB/cm) and rib (1.6 dB/cm) titanium-indiffused lithium niobate waveguides. We observed a drop of only 15% in the intensity of light outcoupled from intentionally-designed low efficiency gratings (grating height o f 150 A) fabri cated on rib waveguides over a 5 mm outcoupling length. Densely-packed rib waveguides arrays (660 elements, 8 fim wide, 0.5 fim high, 2 ftm separations, and 1 cm long) with surface outcoupling gratings (2 fim period, 0.1 fim high) were also investigated on GaAs/AlGaAs waveguides. We found an excellent match between theoretical and experimentally observed performance. Embedded waveguide lenses, rib waveguide arrays, and surface acoustic wave transducers (100 MHz bandwidth) were integrated on a common titanium-indiffused lithium niobate waveguide to test the fabrication compatibilities and integrated device performance. The transducer was used to deflect an input beam, and the lens was used to focus the beam into a single rib waveguide within the array (6 fim rib waveguide widths, 4 ftm separations). We were able to demonstrate selective excita tion of individual rib waveguides through adjustment of the transducer drive fre quency. The systems requirements for the IOSAR processor and IO correlator were con sidered for specific case studies. The overall feasibility of these processors was eval uated with respect to the final state of advancement of the individual components and their integration into test modules. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Chapter 1 Introduction to the Research Topics Many image processing and radar signal processing tasks require intensive computation. The realization of processors capable of performing these tasks is lim ited for certain applications by practical requirements such as speed, power, cost, and processor size. Analog optical processing is an approach that is well-suited to tasks requiring intensive computation due to its inherent parallelism. In essence, optical processors have the ability to perform functions on one- or two-dimensional data fields directly, without the need for sampling, serialization, or quantization; therefore, these processors are potentially capable of high processing rates. With the selection of appropriate optical technologies, such as acoustooptics, microoptics, or integrated optics, optical processors can also be made smaller with lower power consumption than their electronic counterparts. Integrated optical processing modules, in particu lar, have the important potential benefits of low production cost and rigid construction as compared to these other optical technologies. However, the fabrication o f signal processors that deliver these benefits and simultaneously exhibit satisfactory output signal characteristics and total system bandwidth has remained a challenge. The principal objective of the research effort described in this thesis has been to address this challenge through the development of a generalized technological approach for the design and fabrication of optical signal processors that are fast, power efficient, inexpensive, compact, and rigid as well as capable of meeting appli cation-specific performance demands. The approach described herein is based on existing integrated optical technology, with the introduction of advanced waveguide 1 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. components and the use of surface-mounted bulk optical components for additional out-of-plane processing. The resulting advanced hybrid bulk/integrated optical signal processing modules benefit from relaxed design constraints, enhanced capabilities, and applicability beyond the one-dimensional signal processing limitations of con ventional integrated optical processors. We apply this concept of hybrid bulk/integrated optical technology to two prin cipal application areas. One application area is that of interferometric time-integrat ing correlation. This one-dimensional signal processing task requires the correlation of two wide-bandwidth signals, one of which acts as a reference, to recover very weak electromagnetic signals amidst a large component of background noise. The other application area is synthetic aperture radar (SAR) image formation. In this computationally intensive signal processing task, high-resolution ground imagery is formed from periodic radar probes. This inherently two-dimensional image processing task requires compression in both range and azimuth directions. For several important unmanned airborne and space-borne applications, the images must be formed in real-time with a compact and power-efficient processor. The pri mary focus of our research has been on the development of a hybrid bulk/integrated optical signal processing module for SAR image formation. The value inherent in the advancement o f an integrated optical SAR processor can be clarified through a review of the present state of development of electronic and bulk-optical SAR processors, as described in the following section. 1.1 Synthetic Aperture Radar (SAR) Image Form ation Synthetic aperture radar (SAR) consists of a moving airborne or space-borne radar system capable of producing high resolution images from radar returns that con sists of a coherent series of radar pulse echoes [Kovaly, 1976]. SAR is useful when the desired ground resolution requires a conventional radar aperture that is impracti- 2 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. cally large for airborne or space-borne platforms. The complexity associated with SAR image formation is the time-multiplexed signal processing of the radar pulse echoes. In one conventional SAR geometry called strip-mapping, shown in Fig. 1.1, the radar platform is assumed to be flown at height h along the direction 7 7 (azimuth) with a constant velocity v. Radar pulses are typically transmitted at a constant frequency (pulse repetition frequency or PRF) of about 1 kHz. The radar pulse train illuminates a patch on the ground that is to be imaged. In order to maintain peak pulse power, each radar pulse is encoded with a linear FM chirp that is later compressed in the return signal. As a single point scatterer on the ground moves through the footprint of the radar, it will be probed by many sequential radar pulses. Each scattered radar pulse will be Doppler-frequency-upshifted as the radar platform approaches the scat terer, and Doppler-frequency-downshifted as is departs. In this manner the phase of the radar return signal from this scatterer is determined by the relative azimuthal posi tions o f the scatterer and radar platform. An image of the point scatterer may be reconstructed if each radar return is first range compressed, and then the set of radar returns is time correlated against the expected Doppler phase history for the given range of the point scatterer. This correlation process completes the azimuth compres sion portion of the SAR image formation algorithm. Synthetic aperture radar systems behave linearly so that N point scatterers may each be imaged with the same resolu tion as a single scatterer. For many SAR systems, the image formation goal is to achieve that of high resolution in both range and azimuth over a large range swath. Both digital electronic and optical techniques have been employed to process the massive amount of data that is collected by SAR systems in real-time. Because the required computational throughput for real-time and near-real-time SAR process ing is extremely high (-10 Gop/sec), a digital electronics approach requires either a dedicated special purpose processor or a massively parallel computer [Subotic and Aleksoff, 1992; Mastin, et al., 1993]. An example of a real-time digital electronic 3 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 7 ] = Vt Poinl scatterer at (£0 I * » 7 0 ) Figure 1.1 Synthetic aperture radar strip-mapping geometry (after [Psaltis and Wagner, 1982]). SAR processor currently under development is the Danish airborne SAR processor as described in [Dali, et al., 1992]. This processor consists o f 20 electronic cards in two 19 inch card cages interconnected by a dedicated data path and control bus. It is capa ble of operating in various modes with range swaths from 12 to 48 km and resolutions from 2 to 8 meters, respectively. Wide-area, high-resolution digital SAR processors such as this can consume from several hundred watts to as much as 1 kW [Haney, 1994], For certain applications, such as remote satellite reconnaissance or terminally guided munitions, the requirements in processor size (< 1 ft3) and power dissipation (< 200 Watts) cannot be currently satisfied by real-time digital electronic SAR proces sors [Subotic and Aleksoff, 1992]. Consequently, several bulk acoustooptic SAR sys tems have been considered [Psaltis and Wagner, 1982; Psaltis, et al., 1983; Psaltis, 4 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1984; Pape, 1990; Subotic and Aleksoff, 1992; Haney, 1994]. Most of these proces sors are based on the time- and space-integrating (TSI) concept and perform the required signal processing with the use of one or two 1-D acoustooptic devices (AODs). These systems are currently under development and have not yet been tested onboard SAR platforms. A SAR processor variation of the TSI concept, called RAPID SAR (real-time acoustooptic programmable imaging and display for SAR), has been recently lab tested and is nearly ready for a field demonstration. It is pro jected that this system will be capable of generating 8 x 106 focused pixels in a 4 sec SAR data collection period with a power consumption of 50 W. When fully miniatur ized, the optical module is expected to occupy from 0.01 to 0.03 ft3. These perfor mance values are competitive with those expected from integrated optical implementations of SAR [Bicknell, et al., 1985; Rastani, 1988; Armenise, et a l, 1991]. However, the potential compactness (on the order o f 0.003 ft3), cost-effi ciency, and processor-rigidity of integrated optical SAR modules provide a clear advantage for certain applications. The research herein is based on the Integrated Optical Synthetic Aperture Radar (IOSAR) Processor first proposed by Bicknell, Psaltis, and Tanguay [Bicknell, et a l, 1985]. This architecture is an integrated optical adaptation of the acoustooptic/CCD SAR processor architecture proposed by Psaltis and Wagner [Psaltis and Wagner, 1982]. The initial work on the system design and component development for the IOSAR processor was conducted by Rastani [Rastani, 1988]. The realization of this processor will provide the above-mentioned reductions in size, power requirements, and costs as compared with digital electronic and bulk optical approaches. This pro cessor’s operation is also based upon the TSI concept, and hence the two-dimensional image formation task is treated as two one-dimensional signal processing tasks that are time-multiplexed. The IOSAR architecture consists of several key components in an unusual architecture that enables the implementation of this multi-dimensional sig nal processing task in a predominantly planar geometry. 5 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The IOSAR processor architecture shown in Fig. 1.2 consists o f a planar waveguide portion that performs the range compression, and out-of-plane compo nents that perform the azimuth compression. The SAR return signal is inserted into the processor through a surface acoustic wave (SAW) modulator. The laser diode source is pulsed periodically in synchronism with the PRF of the radar transmitter. Light from the butt-coupled laser diode is first collimated, then diffracted from the surface acoustic wave. The surface acoustic wave is composed of multiple chirped pulses, only one of which is shown in Fig. 1.2. The diffracted light is brought to a focus within the waveguide by imaging lenses. The range compression function for a single ground scatterer is performed in the planar portion of the processor as shown schematically in Fig. 1.3. The radar return from each such scatterer is a single chirped pulse. This return signal produces a chirped surface acoustic wave when applied to the transducer. The delay o f the pulse due to the range of the scatterer determines where the chirped pulse is in the SAW aperture when the laser is pulsed. The diffracted light is focused by the integrated lens at the front edge (plane P0in Fig. 1.2) of a rib waveguide array. The focused light couples into a single rib waveguide that corresponds to the point scatterer’s range. For a shorter range, the SAW would propagate farther from the transducer before the laser diode is pulsed and the diffracted light would focus into one of the lower rib waveguides, while just the opposite will occur for longer ranges. The light coupled into the rib waveguide will remain optically confined due to the presence o f an air gap that separates each of the rib waveguide elements. As the light propagates along the single rib element, it is outcoupled by surface gratings oriented in a direction perpen dicular to the rib waveguide as shown in Fig. 1.2. A sheet of light therefore propa gates upwards through the surface mounted components and one row of elements in the CCD array detector is illuminated. In the presence of many ground scatterers, the light distribution entering the rib waveguide array is the instantaneous range-compressed image for a single radar 6 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. CCD Array Detector (Shift and Add) Readout Buffer Mask (Encoded SAR Algorithm) Output SAR Image Plane P. Propagating Surface Acoustic Wave ------ v Gratings Rib or Channel Waveguide Array Surface Acoustic Wave IDT (Radar Return Signal Input) Laser Diode Substrate Waveguide Figure 1.2 Schematic diagram of the Integrated Optical Synthetic Aperture Radar Processor (after [Bicknell, 1980; Bicknell, et al., 1985; Rastani, 1988; Tanguay, 1988]). pulse. The rib waveguide array is used to dissect the range-compressed image and expand this image in the plane o f the waveguide, a function analogous to the anamor- phic expansion of a cylindrical lens [Rastani, 1988]. The phase of the return signal from each scatterer is a function of the azimuthal position of the scatterer within the footprint due to the previously discussed Doppler shift. To preserve this phase infor mation in the optical signal, a constant frequency tone is mixed with the radar return and input to the SAW transducer to serve as a phase reference. Grating couplers fabricated on the surface of the rib waveguides provide a means to extract this expanded range-compressed image so that a proximity-coupled two-dimensional mask and CCD array may be addressed for the completion of the azimuth compression function. The outcoupled light is first multiplied by the Dop pler phase history encoded in the transmittance of the fixed two-dimensional mask, 7 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. IOSAR PROCESSOR: RANGE COMPRESSION RIB OR CHANNEL WAVEGUIDES------ r PULSED LASER DIOOE STOP PLANAR W AVEGUIDE ON SUBSTRATE----- SURFACE ACOUSTIC W AVE (SAW) TRANSDUCER Figure 1.3 Top view of the Integrated Optical Synthetic Aperture Radar Processor. The radar return is introduced by means of a surface acoustic wave transducer and the range focused image is formed at the beginning of the rib waveguide array (from [Rastani, 1988]). and then recorded on the CCD array in the form o f a photogenerated charge distribu tion. The CCD array is synchronized with the PRF of the radar such that the photoge nerated charge distribution is continuously shifted stepwise along the direction of the rib waveguides. The relative motion between the fixed two-dimensional mask and the accumulating photogenerated charge distribution results in the required correlation in the azimuth direction. The signal at the CCD array readout buffer contains the range and azimuth compressed radar image. The physical realization of the IOSAR processor requires the development of several key waveguide components, including large aperture, short focal length lenses; densely-packed large-area rib waveguide arrays; and large-area uniform out- coupling gratings. Waveguide lenses with a large aperture are required in order to process a large range swath. The lenses should also have a short focal length for reduced processor dimensions. Rib waveguide arrays with a large number of densely- packed elements are required to dissect the range-compressed image and maintain the 8 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. range resolution. Long rib waveguides are required in order to provide a long correla tion length between the fixed mask and CCD array, and thereby increase the azimuth resolution. The channel crosstalk between the ribs should be low in order to maintain image integrity over the full propagation distance. Uniform outcoupling gratings with low efficiency are required to provide uniform illumination of the surface- mounted devices. From an analysis of the projected system capabilities for an IOSAR processor that incorporates reasonable waveguide materials and potentially feasible component capabilities, it was found that a range resolution was 3 meters over a 1 km range swath was possible [Rastani, 1988]. For a PRF o f 1 kHz, this corresponds to a system throughput of 1.3 X 106 resolution points in a 4 sec SAR data collection period. Development of the key waveguide components can potentially lead to even larger processing capacities. 1.2 Status of Integrated Optical Technology Realization o f the IOSAR processor requires a level of system integration that has not previously been achieved in integrated optics. Consider for comparison the integrated optical spectrum analyzer (IOSA) fabricated, for example, by Mergerian, et al. shown in Fig. 1.4 [Mergerian, etal., 1980; Mergerian, et al., 1983]. This proces sor consists of an X-cut titanium-indiffused lithium niobate waveguide (Ti:LiNb03) [Armenise, 1988], a butt-coupled laser diode, two integrated geodesic lenses [Merge rian, et al., 1979; Vahey, et al., 1980], a SAW modulator, and a butt-coupled linear photodetector array. This processor is useful for wide-band instantaneous monitoring of the electromagnetic frequency spectrum. The output power of the laser is constant, and the signal to be processed is input to the processor through the SAW modulator. The laser light is collimated by a waveguide lens and illuminates the surface acoustic wave. The angular spectrum of the diffracted light is proportional to the frequency 9 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.45 cm 2.7 cm Laser 140 E lem ent Focal P lan e Array Lo & A m plifiers RF Input H T ra n sd u c e r Figure 1.4 Integrated optical spectrum analyzer fabricated in TkLiNbOj (from [Mergerian, et al., 19801). components in the input signal. The second waveguide lens Fourier-transforms the angular distribution of the diffracted wave into a spatial intensity distribution at the back focal plane of the lens. Each position of the focused light at the photodetector array indicates a particular spectral component of the input signal. The dynamic range of this processor was measured to be in excess of 40 dB [Mergerian, et al., 1983], and the nearest neighbor crosstalk at the photodiode array was down by 15 dB [Mergerian, et al., 1980]. Another example of the degree of previously-achieved processor integration is shown in Fig. 1.5. This matrix algebra processor is capable of multiplying a 4 x 4 matrix with a 4-element vector at extremely high rates [Kar-Roy and Tsai, 1991], This processor consists of a Ti:LiNb03 waveguide with large aperture titanium-indif fused proton-exchanged (TIPE) lenses [De Micheli, et al., 1982; Zang and Tsai, 1985; Suhara, et al., 1986], input and output channel waveguide arrays, and multiple surface acoustic wave transducers. These processors are typical of the highest level of integration achieved to date 10 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. UTQ« A0« T ur« T1P6 a —m £xo*ndtng . Uutttpte T U i m ta rg t Ap*rtur« T1P£ Collimating Lana, . SAW Transoucara. ^oeuaamg Lana ST3 MCt SAW* O C i O C . 1 C - MC- Figure 1.5 Acousto-optic matrix algebra processor on a LiNb0 3 substrate (from [Kar-Roy and Tsai, 19911). for a fully functional integrated optical processor. Further examples of spectrum ana lyzers have been described in the literature [Anderson, 1978; Bamoski, et a l, 1979; Mergerian, et al., 1980; Thylen and Stensland, 1981; Suhara, et al., 1982; Mergerian, et al., 1983; Valette, et al., 1983], as well as other types of signal processors that exhibit a comparable level of integration [Verber, et al., 1981; Tsai, et al., 1985;Tsai, et a l, 1989]. An IOSAR system design, on the other hand, calls for the integration of large aperture (2.7 cm), short focal length integrated lenses (f/4 or better), rib waveguide arrays with in excess of 300 elements combined with large area surface coupling grat ings, and stacked two-dimensional surface-mounted components aligned with the rib waveguides. These specialized components are required in addition to conventional waveguide components such as the surface acoustic wave inter-digital transducer and butt-coupled laser diode shown in Fig. 1.2. Hence, an advanced state of component development and integration is required beyond that which has been currently achieved for the realization of advanced hybrid bulk/integrated optical signal process- 11 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ing modules such as the IOSAR processor. 1.3 Development of Key Components In the previous work by Rastani, several advancements were made in the devel opment of the key waveguide components for the IOSAR processor [Rastani, 1988], In addition, the results obtained by other investigators in their individual efforts to advance the state of the art in integrated optical technology have contributed to the feasibility of full IOSAR processor integration. The titanium-indiffused lithium niobate waveguides employed by Rastani in previous IOSAR component development [Rastani, 1988] are frequently selected for use in integrated optical processors [Mergerian, et al., 1980; Verber, et al., 1981; Liao, etal., 1982; Suhara, eta l., 1982; Tsai, etal., 1989] because of the typically low prop agation losses, large acoustooptic coupling coefficients, and well-established fabrica tion techniques [Carruthers, et al., 1974; Sugii, et al., 1978; Bums, et al., 1979; Holmes and Smyth, 1984; Armenise, 1988]. As a result of the selection of Ti:LiNb03 waveguides, hybrid methods of component integration were primarily employed in Rastani’s work; however, the applicability of both hybrid and monolithic integration principles to other materials systems such as silicon- and gallium arsenide-based waveguides was also discussed therein. In the research described in this thesis, we have further explored the development of several key integrated optical devices in gallium arsenide (GaAs) waveguides. However, our primary effort has been focused on further development o f key waveguide components in Ti:LiNb03 waveguides that conform to the IOSAR system design requirements identified by Rastani. Development of high-quality integrated lenses has been the subject of consider able previous investigation [Chen, et al., 1977; Yao and Anderson, 1978; Chang and Ashley, 1980; Valette, et al., 1982; Wood, et al., 1983; Vu and Tsai, 1989]. Of the dif ferent lens structures and fabrication techniques that were developed, those that were 12 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the most manufacturable were limited to small lens apertures and long focal lengths. For a number of signal processing applications, such as those discussed herein, the processing capacity may be increased through the use of wide-aperture lenses. How ever, the associated long lens focal lengths entail the need for comparably long waveguide substrates. In relation to the IOSAR processor development, Rastani addressed this problem by his investigation of two novel waveguide lens technologies that are potentially capable of both large apertures and short focal lengths (low f-num- bers), and that utilize standard planar waveguide fabrication techniques. These included titanium-indiffused proton-exchanged (TIPE) lenses in lithium niobate and thin-film-coated recessed lenses in lithium niobate. The TIPE lens, investigated initially by Zang and Tsai [Zang and Tsai, 1985], consists of a two step waveguide formation process combined with photolithographic masking techniques to provide a relative refractive index difference between the lens and the surrounding planar waveguide regions. The positive focal length lens struc ture consists of a positive meniscus titanium-indiffused proton-exchanged waveguide lens in a planar titanium-indiffused waveguide. With this technique, Zang and Tsai [Zang and Tsai, 1985] produced a 4 mm aperture, 1.5 cm focal length waveguide lens. In addition, TIPE lenses have been used as collimating and focusing elements in inte grated optical signal processors [Tsai, 1979; Zang and Tsai, 1985], Rastani developed the inversion of this process to create negative meniscus, positive focal length lenses. In this case, the lens structure consists of a titanium-indiffused lens in a titanium- indiffused, proton exchanged planar waveguide. He was able to achieved a functional I cm aperture lens with better than f/6 performance, and a 1 cm aperture lens triplet with better than f/4 performance [Rastani and Tanguay, (to be published)]. In his initial investigations into thin-film-coated recessed lenses, Rastani aimed to establish the basic lens fabrication sequences [Rastani, 1988]. This sequence con sisted of the formation of an ion-milled lens recess in a titanium-indiffused lithium niobate waveguide that was back-filled with a thin film silicon dioxide/magnesium 13 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. fluoride (SiC>2/MgF2) waveguide structure. The refractive index difference between the planar waveguide and the lens-shaped thin film waveguide leads to waveguide focusing. The photoresist mask used for the ion-milling process, however, provided an inadequate pattern transfer of the lens recess into the planar waveguide that resulted in sloped recess sidewalls. Consequently, the fabricated lenses were not functional. The potentially large refractive index difference between the lens and host waveguide regions (up to 0.7 for LiNbC>3), however, makes this a promising approach to the realization of large aperture, short focal length lenses. Others have recently reported similar structures, called embedded waveguide lenses, that were fabricated in GaAs and had a refractive index difference of 1.9 [Minot and Lee, 1990]. In their investigations, the sloped recess sidewall was also encountered. This problem was addressed by experimentation with different chemi cal etching and reactive-ion etching (RIE) processes. Vertical sidewalls were pro duced through the use of RIE and a tri-layer photoresist mask that was used to define the lens. To reduce interface reflection losses arising from the large refractive index difference, an anti-reflection coating technique was established that resulted in a lens with a throughput of 61% and f/1.8 performance for a 2.2 mm aperture [Su and Lee, 1994]. A major portion o f the research effort described in this thesis has been aimed toward the development o f embedded waveguide lens structures in Ti:LiNb03 waveguides. In the previous efforts by Minot and Lee [Minot and Lee, 1990] to pro duce embedded lens structures, great difficulties were encountered in the formation of a vertical recess sidewall; a problem that is compounded in lithium niobate by the low sputter yield and chemical inertness of the material. However, since embedded lens structures have the potential for operation with f-numbers lower than those obtained by the TIPE process, their development promises to add flexibility in processor design, reduce waveguide substrate size requirements, and increase processing capac ities. The difficulty in the fabrication of embedded lenses in LiNb03 discovered by 14 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Rastani required extensive investigation in the present work before it was resolved. Nonetheless, we have fabricated Si0 2/MgF2 embedded lens structures in Ti:LiNbC> 3 with 1 cm apertures, 1 cm focal lengths, 50% throughput for f/2 operation, a 0.9 fim focal spot size inside the waveguide, and nearly vertical etched-recess sidewalls ( 12° tilt angles). The thin film deposition techniques for the waveguide lens fabrication were developed by Karim [Karim, 1993]. An analysis o f the mode coupling at the waveguide/embedded waveguide interface was also provided by Karim [Karim, 1993] and is expanded on in this work. Compared to previously established lithium niobate based lens technologies that consist of techniques that permit the parallel manufacture of lenses (for example, pho tolithographic techniques), the performance values given above for the embedded lenses produced in this research represent an advancement in the state of the art of waveguide lenses. In particular, wide aperture (greater than a few millimeters) low f- number lenses (less than f/4) have not been previously demonstrated in this material using manufacturable technologies. The waveguide lens structure that comes the closest is the TIPE lens discussed earlier. The shortcoming of the embedded lens is the comparably high throughput loss (3 dB as compared to 1 dB for the TIPE process [Rastani and Tanguay, (to be published)]). The f-number and throughput of the embedded lens described above is also comparable to that obtained for the embedded lens structures developed by others in GaAs waveguides [Minot and Lee, 1990; Su and Lee, 1994]. However, the functional aperture size obtained for the lithium niobate embedded lens (5 mm) exceeds that pre viously obtained for the GaAs embedded lens (2 mm). Both titanium-indiffused channel waveguide array and rib waveguide arrays in LiNb0 3 were considered for incorporation into the IOSAR processor by Rastani. Because of the large refractive index difference between the rib waveguide elements and their associated air gaps, rib waveguide structures were shown to be preferable for the fabrication of high density arrays with low rib-to-rib crosstalk. Highly uni- 15 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. form rib waveguide arrays (660 guides, 8 /im wide, 1 fim high, 2 /un separations, and 1 cm long) were fabricated in Ti:LiNb0 3 waveguides and tested to determine the waveguide properties [Rastani, 1988]. The measured rib waveguide propagation losses were approximately 6 dB/cm. Light coupled into a single rib waveguide in the array led to no observable leakage to the neighboring rib waveguides after a 1 cm propagation length. Other investigators have used channel waveguide arrays in signal processors to route information in from or out to the edge of the waveguide [Ander son, 1978; Valette, etal., 1983; Tsai, et al., 1985]. For these arrays, the required chan nel densities were low enough so that sufficient isolation could be maintained by incorporating a large-enough spacing between individual elements. For the successful application of rib waveguide arrays in the IOSAR processor, the rib waveguides should generate minimal crosstalk in the range-compressed image at the CCD array detection plane. Hence, there should be minimal mode coupling and low scattering losses over a large propagation length (1 cm) since either of these effects will lead to crosstalk. Furthermore, large scattering losses can lead to signifi cant attenuation of the range image as it is expanded along the length o f the rib waveguide and thereby result in non-uniform illumination of the surface mounted devices. In the present work, we experimentally determine the rib-to-rib coupling for different rib waveguide array structures fabricated in Ti:LiNb03 waveguides. A clear dependence of the crosstalk on the waveguide structure is evident. We also show that the rib waveguide propagation losses can be significantly lowered by a reduction in the rib waveguide height. Initial investigations into the use of surface gratings on both uniform waveguides and rib waveguides were conducted by Rastani. The fabrication of grat ings on uniform waveguides is the simpler case and is discussed here first. Rastani patterned gratings holographically in photoresist on uniform Ti:LiNb03 waveguides with periods ranging from 0.6-1.6 fdm were produced by Rastani with the use of a He- Cd laser. These gratings exhibited nonuniformities that were attributed to inhomoge- 16 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. neities in either the photoresist composition or the laser illumination. To obtain better uniformity, gratings were fabricated on uniform Ti:LiNb03 by photolithographic pat terning and argon-ion beam milling. Highly uniform 2 /im and 4 /im period gratings with a 500 A grating depth were fabricated on Ti:LiNb0 3 waveguides over a 1 cm wide X 1 mm long region by this approach and were studied to determine the outcou- pling properties [Rastani, 1988]. In related research conducted by others, surface gratings have been used for waveguide input coupling [Valette, et al., 1983], for effi cient waveguide-detector coupling [Huang and Lee, 1986], for large-area focused outcoupling to an external optical disk [Suhara and Nishihara, 1986], and for the dis tributed supply of optical power to external elements [Kubota and Takeda, 1989; Takedaand Kobuta, 1991; Song, etal., 1994]. In this work, we describe the fabrication of 2 /im and 4 /im period gratings on uniform Ti:LiN b03 waveguides that cover a large area (1.2 cm x 1 cm) and provide nearly uniform outcoupling over the approximately 1 cm lengths as is required in the IOSAR processor. We also consider the theoretical aspects o f the surface outcoupling gratings that relate to the division of the outcoupled power between the surface radia tion modes and the substrate radiation modes. Since power coupled into the waveguide substrate represents a loss in the signal power as well as a possible source of background noise in the IOSAR processor, methods to increase the relative amount of light outcoupled from the surface of the waveguide are discussed. Realization of the IOSAR processor requires the fabrication of highly uniform outcoupling gratings on the surfaces of the rib waveguide array structures. This was previously accomplished by Rastani, in which the basic rib waveguide array structure described above (660 guides, 8 /im wide, 1 /im high, 2 /im separations, and 1 cm long) was combined with two different grating structures (2 /tm and 4 /im period grat ings, 1 cm x 1 mm regions, 500 A grating depths) on Ti:LiN b03 waveguides. Grat ing outcoupling from the rib waveguide arrays was also observed [Rastani, 1988]. Herein, we have combined the larger-area uniform outcoupling gratings (2 /im 17 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. and 4 flm period gratings, 1.2 cm x 5 mm regions, 250 A grating depths) with rib waveguide array structures similar to those fabricated by Rastani (1000 guides, 8 fim wide, 0.5 fim high, 2 fim separations, and 5 mm long) as well as with a new rib waveguide array structure (333 guides, 28 fim wide, 0.5 ftm high, 2 fim separations, and 5 mm long). We experimentally investigated both the macroscopic (on the scale of the rib waveguide length) and microscopic (on the scale of the rib waveguide width) outcoupling uniformity from the surface gratings and identified possible sources of non-uniformity in the illumination of surface-mounted devices in the IOSAR processor. The transverse diffraction of light outcoupled from the narrow rib waveguides (8 fim wide) implies a maximum spacing from the waveguide surface to the detection plane of the surface-mounted CCD array. The maximum spacing is cal culated and two external element mounting schemes are suggested to satisfy this requirement. The spacing requirement for a 28 fim wide rib waveguide is found to be greatly relaxed in comparison with that required by the IOSAR processor. Our investigation of this new rib waveguide array structure stems from our efforts to extend the use of the key components and basic architectural scheme of the IOSAR processor to other signal processing applications. Hence, we thereby expand upon the notion of a generalized technological approach to the implementation of sig nal processing tasks in advanced hybrid bulk/integrated optical modules, as described in the next section. 1.4 Extension of the Concept to Signal Correlation Many other signal processing applications can readily benefit from the compo nents developed in this research. We consider in particular the signal processing task of signal correlation. Essentially, correlation is used to recover extremely weak sig nals from the electromagnetic spectrum amidst a noise-filled background [Berg and Lee, 1983]. The improvement in the signal-to-noise ratio is equal to the space- or 18 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. time-bandwidth product of the processor used to perform the correlation, and is iden tified as the processing gain (PG) [Cohen, 1983]. For many important applications, such as spread-spectrum communications and radar ranging, nearly instantaneous processing is required in addition to high processing gains. Hence, extremely high computational throughputs are required. In addition, processors that can directly pro cess the detected signal (typically on the order of one to ten MHz in amplitude modu lation) without the need for sampling are desired. An analog optical correlator implementation is well suited to these requirements. In general, two principal types of correlation architectures have been identified: time-integrating and space-integrating. The requirements placed on time-integrating and space-integrating correlators are ever-increasing, and are thought to have approached the limits of analog and digital electronic signal processor abilities [Berg and Lee, 1983; Anderson, et al., 1994; Adler and Patterson, 1995]. Demands on speed, dynamic range, size, and power can potentially be satisfied with an acoustoop tic approach. A number of different correlator architectures have been successfully implemented with discrete acoustooptic components [Sprague and Koliopoulos, 1976; Casasent, etal., 1982; Berg and Lee, 1983; Ross, 1991; Riza, 1994]. Integrated optical correlators offer the potential added advantages of manufacturability, reduc tion in size and cost, and improvement in rigidity as compared with bulk optical pro cessors. Consider, for example, the integrated optical space-integrating correlator shown in Fig. 1.6. Two SAW transducers are used to insert two time-varying RF waves with carrier frequency Q c modulated by the signals g(t) and h(-t), in which h (-t) is a time- reversed reference signal. The two index gratings produced within the waveguide have modulations of g(t~x/v) and h(x/v-t) in which x is the spatial coordinate and v is the SAW velocity. The optical beam is successively diffracted by these two gratings. The final modulation of the optical beam is proportional to the product g(t-x/v)»h(x/v-t). A lens is used to spatially integrate the diffracted light by focusing 19 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Waveguide lenses Q , ( t \ P a f a r a n / * o eirm ral Laser die Diffracted wave (a o " “ 2 Qc Photodetector & ^ m Correlation signai 1 )0 n r Spatial filter / (slit) S x(t) Signai for correlation Waveguide Figure 1.6 Space-integratine correlator architecture (from [Nishihara, et al., 1989]). it onto a photodetector. The detected signal is proportional to [Nishihara, et al., 1989] in which T is the interaction time and r = x/v is the delay within the SAW aperture. The time-varying signal output of the detector consists of the desired correlation sig nal on top of a time-dependent bias. This bias term is separated from the correlation terms by post-processing. In correlator processors such as this, the gain in the signal- to-noise ratio (SNR) is equal to the time-bandwidth product (TBWP) of the SAW modulation device. Typical values for the time-bandwidth products in these proces sors are on the order IO2 [Tsai, 1979]. Comparatively larger processing gains (approximately three orders of magnitude) can be realized with the use of a time-inte grating correlator. We consider herein the implementation of a time-integrating correlator with the basic architecture shown schematically in Fig. 1.7. This architecture utilizes the same t+T/2 ( 1 . 1 ) 20 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. LD LI L2 SAW L3 SF L4 D hit) O Figure 1.7 Schematic diagram of the integrated optical interferometric time- integrating correlator, showing the hybrid integration of the laser diode, beam expanding lens (LI) and collimating lens (L2), input surface acoustic wave (SAW) transducer, Fourier transform lens (L3) and imaging lens (L4), spatial fil ter (SF), and linear detector (D) on top of a rib waveguide array with grating cou- basic components developed for the IOSAR processor. These include wide-aperture waveguide lenses, a rib waveguide array, surface outcoupling gratings, and a surface- mounted linear detector array. In this processor, one of the signal inputs, g(t), is used to intensity-modulate the laser diode light source (LD) while the other signal is used to drive the SAW transducer. The diffracted wave is modulated temporally and spa tially by the product g{t)*h{t— x/v). This diffracted wave is Fourier-transformed by lens L3 and the undiffracted light is blocked by the spatial filter (SF). The diffracted light is then imaged by lens L4 onto the front edge of a rib waveguide array. The image intensity distribution is segmented by the rib waveguides and outcoupled via the surface outcoupling gratings to a surface-mounted linear photodetector array, where it is time-integrated. The output signal is proportional to [Nishihara, et al., 1989] R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. piers. ( 1.2) 21 40 in which is the photodetector time constant. Hence, the correlation of the two input signals is spatially distributed on the detector array. In this correlator architec ture, the time window can be made very large for photodetectors with long integration times. To date, only one integrated optical implementation of the time-integrating cor relation has been reported [Liao, etal., 1982]. This processor was fabricated on a tita- nium-indiffiised lithium niobate waveguide, and utilized an elegant approach to perform the spatial filtering operation and permit the correlation to be accomplished with very few integrated optical components. This processor is shown in Fig. 1.8 and requires only a single lens. The use of surface acoustic wave anisotropic Bragg dif fraction to input one signal results in a polarization change in the diffracted signal. At the detector, the undiffracted light can be easily filtered using a properly oriented polarizer. This time-integrating correlator has been operated with a dynamic range of 27 dB and a time-bandwidth product of 4.2 X 105 [Liao, et al., 1982]. While the correlator architecture proposed herein requires more components than the correlator fabricated by Liao, it is much more flexible in terms of applicabil ity to various correlation and convolution functions. For example, space-integrating correlation, time-integrating correlation with dual SAW inputs, and matched filtering can be accomplished with few minor alterations in the architecture shown in Fig. 1.7. Another relative advantage of the correlator structure considered herein is that the devices can be made with simple and accurate planar fabrication techniques in con trast to the difficult diamond turning processes required to produce the geodesic lens in the time-integrating correlator of Liao (see Fig. 1.8). The use of planar fabrication techniques in general permits the incorporation of increased complexity in processor architectures with very few changes in the fabrication process. This is accomplished simply by modifying the photolithographic masks. There are several similarities and dissimilarities between the IOSAR processor and time-integrating correlator implementations. For example, both processors 22 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Laser diode Polarizer TM transmit) } hotodetector array ffracted wave(TM) idiffracted wave (TE) Correlation signal Signal for co i Ti: LiNb03 Waveguide 5 ^ Reference signal Figure 1.8 Time-integrating correlator architecture that utilizes anisotropic Bragg diffraction (from [Liao, etal., 1982]). require large aperture lenses with short local lengths to increase their processing capacity and reduce their substrate size. The rib waveguide array/grating structure for the correlator, on the other hand, does not require uniform low efficiency outcoupling over a long distance, but rather, high efficiency outcoupling over a short distance. Hence the issues of rib-to-rib coupling and rib waveguide scattering losses are not nearly as critical in this application. For the specific correlator design presented in Chapter 2, we find that the required density o f the rib waveguide array elements (333 guides over a I cm width, hence 28 //m rib widths with 2 ftm separations) is much less than in the IOSAR processor. This further relaxes the design constraints for the rib waveguide array structure and the surface-mounted linear detector array. 1.5 Demonstration o f Key Component Integration The full integration of the IOSAR processor and IO correlator with the compo nent technology and hybrid bulk optical components described herein will demon strate the capacity for multi-dimensional signal processing and high time-bandwidth 23 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. processing with an inherently planar technology that can be packaged in space-effi cient, power-efficient, and cost-efficient modules. In order to evaluate the feasibility o f full system integration, various levels of component integration were performed. Experiments performed previously by Rast ani included the evaluation of a module with an integrated 1.2 cm focal length TIPE lens, rib waveguide array (660 guides, 8 fim wide, 1 fim high, 2 fim separations, and 1 cm long), and outcoupling gratings (4 fim period, 0.05 fim grating height, I mm long). The rib waveguide array was aligned at the back focal plane of the lens and the gratings were defined on the surface of the rib waveguide array. The composite per formance of these combined components was demonstrated by coupling a 2 mm wide beam o f light into the waveguide and directing it through the integrated lens. The light was focused by the lens into a single rib waveguide within the array o f 660 ele ments and then outcoupled by the associated surface outcoupling gratings [Rastani, 1988]. In the present work, we have developed a new waveguide lens in Ti:LiNbC> 3 with a lower f-number than that previously achieved by the TIPE process. In consid eration of the lens requirements for the IOSAR processor and the IO correlator, two lens designs based upon the Si02/MgF2 embedded lens materials have been gener ated with Code V software. These consist of a 2.7 cm aperture field-corrected lens doublet for use in the IOSAR processor and a 1 cm aperture single element aberra tion-corrected f/1 lens for use in the IO correlator. This second lens design was used in the fabrication of all of the embedded lens structures presented herein. The combined performance of a chirped SAW waveform, flat-field lens doublet, and rib waveguide array dissection was evaluated for the IOSAR processor. This included the combined effects of diffraction of a collimated beam from a linear SAW chirp, formation o f a range-focused spot at the back focal plane of the flat-field lens doublet, and excitation of a rib waveguide element within an array of rib waveguides located at this plane. The spot intensity profile produced by the combination of the 24 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. SAW modulation and lens doublet was calculated with Code V analysis software. Mode coupling of this spot profile with the guided modes of the rib waveguides was calculated by means of overlap integrals. The excitation of the neighboring rib waveguides was also evaluated to determine the level of side-lobe-induced crosstalk. A major concern in component integration for these integrated optical proces sors is the required alignment tolerance of the various components. We have consid ered, in particular, the variability of the embedded lens focal length with fabrication process fluctuations. Both the IOSAR processor and IO correlator have several in line components that must be accurately spaced in accord with to the focal length of the lens. This is especially critical for low f-number lenses, since a small error will lead to a large focal spot blur at the intended focal plane position. The waveguide components, on the other hand, can be very accurately placed with respect to other waveguide components through the use of photolithographic techniques. Successful integration of these waveguide lenses requires careful control of the embedded lens fabrication process. We also suggest a corrective procedure to adjust the lens focal length that is capable of correcting small errors. To assess the progress described herein on the development of the embedded lens and rib waveguide arrays and the potential for full processor integration, the two processor test modules shown in Fig. 1.9 were fabricated and tested. These two test modules address two key stages of component integration. The first test module com prises the combination of the embedded lens with a rib waveguide array to demon strate the ability to focus light into a single rib waveguide element. The theoretical coupling efficiency of the light focused by the lens into a single rib waveguide was calculated for the focal spot size of the fabricated embedded lens structure. The small focal spot width (0.9 /im calculated inside the waveguide) compared to the rib spac ing (10 /im) resulted in a calculated coupling efficiency o f nearly 100%. The second test module consists of a surface acoustic wave (SAW) transducer integrated with the embedded lens and rib waveguides to demonstrate selective focusing into individual 25 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. (a) (b) Figure 1.9 Schematic diagram of optical processor subsystems showing two levels of component integration for performance evaluation purposes. Module (a) consists of an embedded lens aligned by linear fiducial markings to a rib waveguide array and end-polished substrate. Module (b) is identical to module (a) with the addition of a SAW transducer. rib waveguide elements. With this second test module we were able to show selective optical excitation across six out of ten theoretically addressable rib waveguides within the rib waveguide array. 1.6 Alternative Waveguide Materials It is of interest to extend the device fabrication techniques to other materials such as gallium arsenide (GaAs) that support active laser and detector components in order to advance the short term feasibility of a monolithically integrated IOSAR pro cessor or IO correlator. Tremendous advances in component development and inte gration have been recently made in GaAs waveguides including the embedded lens fabrication by Minot and Lee, [Minot and Lee, 1990], the integration of a laser, an ion-milled lens, and the surface gratings by Hirata et al., [Hirata, et al., 1993], and integration of ion-milled waveguide lenses with an InGaAs photodetector array [Vu, etal., 1992]. Our initial investigations into the use of GaAs-based waveguides consisted of 26 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the fabrication of rib waveguide arrays (660 guides, 8 fim wide, 0.5 /xm high, 2 /lm separations, and 1 cm long), outcoupling gratings (2 fim period, 0.1 /Jm grating height, 1 mm long), and rib waveguide arrays combined with surface outcoupling gratings in GaAs waveguides with lower index aluminum gallium arsenide (AlGaAs) barrier layers. The outcoupling efficiencies for gratings fabricated on both planar waveguides and rib waveguides were measured and compared with the corresponding theoretical outcoupling efficiencies [Rastani, etal., 1989]. 1.7 Outline of the Dissertation In the remainder of this dissertation, we consider the relationship between the system design considerations and the component requirements for these advanced integrated optical processors. These requirements are applied to the development of rib waveguide arrays, surface outcoupling grating, and embedded lenses as key inte grated optical components. The feasibility of component integration is demonstrated by the fabrication and characterization of two integrated test modules, and the exten sion of this integration to the realization of fully integrated optical signal processors is discussed. In Chapter 2, the SAR image formation and correlator processing tasks are con sidered from a systems viewpoint by reviewing examples of their bulk optical imple mentations. A generalized integrated optical processor architecture is then used to focus on key first-order processor device requirements and limitations. These requirements and limitations are subsequently refined for the specific geometries of the IOSAR processor and 10 correlator. Specific examples of IOSAR processor and 1 0 correlator designs are then given to clarify the selection of device parameters used in the following chapters on device design, development, and characterization. In Chapter 3, waveguide substrate considerations are reviewed and two particu lar materials are selected for use in this work. These are titanium-indiffused lithium 27 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. niobate waveguides (Ti:LiNb03) and gallium arsenide/aluminum gallium arsenide waveguides (GaAs/AlGaAs). The technique used to fabricate low-Ioss single mode Ti:LiNb03 waveguides is presented along with the results of waveguide loss, mode index, and mode profile measurements. The effects of process variations on the waveguide properties are determined for use in the device design discussed later in Chapter 5. In Chapter 4, the fabrication techniques used to produce uniform large-area gratings on planar Ti:LiNb03 waveguides, and gratings on rib waveguide arrays in Ti:LiNb03, are presented. The observed grating outcoupling uniformity for both of these structures is discussed, and issues regarding the illumination uniformity o f bulk optical surface-mounted devices are considered. The measured crosstalk in specially- designed rib waveguide arrays with various rib separation widths and depths are com pared with the expected crosstalk values obtained from an approximate theoretical analysis. The results of comparable rib waveguide and surface outcoupling grating integration in GaAs waveguides are also given. In Chapter 5, the theoretical aspects involved in the design of an embedded lens structure with ideally smooth and vertical lens interfaces is presented. The effects of fabrication limitations and variations on the embedded lens performance are consid ered. Two lens designs generated by Code V optical CAD software for aberration- corrected performance are introduced, of which one is then used in lens fabrication. The lens fabrication sequence is discussed, and two embedded lens structures consist ing of SiC > 2/M gF2 waveguide layers and photoresist/MgF2 waveguide layers in a Ti:LiNb03 waveguide are presented. Measurements of the S i0 2/MgF2 embedded lens focal length, focal spot size, and throughput are given. In Chapter 6 , the surface acoustic wave properties of different materials are compared with those of lithium niobate. The properties of LiN b03 are shown to be suitable for use in this research. The design and fabrication of simple 100 MHz band width transducers on lithium niobate is then discussed, followed by the results of 28 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. electrical and optical characterization of these devices. In Chapter 7, the integration of a rib waveguide array, a lens, and a SAW trans ducer on a single substrate is discussed. The results o f the integrated module are given, including the demonstration of focusing into a single rib waveguide as well as selective focusing into individual rib waveguides within a rib waveguide array using a surface acoustic wave transducer to deflect the beam. We consider the properties of each waveguide component individually, as well as when integrated onto the same substrate, and then present the expected performance values for a fully integrated IOSAR processor and 10 correlator. In Chapter 8 , we summarize the key results of the dissertation and discuss the feasibility of this approach for fully integrated optical processor implementations. Several important future research directions are also discussed in Chapter 8 . 1.8 References E. Adler and M. Patterson, “Demonstration of an Acousto-Optic Correlator for ESM Applications,” in Transition of Optical Processors into Systems, SPEE Vol. 2480, (1995). C. Anderson, A. Dorsett, R. Berinato, A. Filipov, E. Adle, and J. Pellegrino, “An Acousto-Optic ESM Receiver Upgrade,” in Transition of Optical Processors into Systems, SPIE Vol. 2236, 31-38, (1994). D. B. Anderson, “Integrated Optical Spectrum Analyzer: An Imminent ‘Chip’,” IEEE Spectrum, Dec., 22-29, (1978). M. N. Armenise, “Fabrication Techniques of Lithium Niobate Waveguides,” IEE Proceedings-J, 135(2), 85-90, (1988). 29 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. M. N. Armenise, F. Impagnatiello, V. M. N. Passaro, and E. Pansini, “Design of GaAs Acoustooptic Correlator for Real-Time Processing/' in Devices for Optical Processing, SPEE Vol. 1562, 160-171, (1991). M. K. Bamoski, B. U. Chen, T. R. Joseph, Y. M. Lee, and O. G. Ramar, “Integrated- Optic Spectrum Analyzer,” IEEE Trans. Circuits Syst., Cas-26(12), 1113, (1979). N. J. Berg and J. N. Lee, Ed., Acousto-Optic Signal Processing, Series on Optical Engineering, B. J. Thompson, Ed., Vol. 2, (Marcel Dekker, Inc., New York, 1983). T. Bicknell, D. Psaltis, and A. R. Tanguay, Jr., “Integrated Optical Synthetic Aperture Radar Processor,” J. Opt. Soc. Am. A, 2(13), 8, (1985). T. J. Bicknell, “Digital Versus Optical Techniques in Synthetic Aperture Radar (SAR) Data Processing,” Opt. Eng., 19, 157-167, (1980). W. K. Bums, P. H. Klein, E. J. West, and L. E. Plew, “Ti Diffusion in Ti:LiNb03 Planar and Channel Optical Waveguides,” J. Appl. Phys., 50(10), 6175-6182, (1979). J. R. Carruthers, I. P. Kaminow, and L. W. Stultz, “Diffusion Kinetics and Optical Waveguiding Properties of Outdiffused Layers in Lithium Niobate and Lithium Tantalate,” Appl. Opt., 13(10), 2333-2342, (1974). D. Casasent, G. Silbershatz, and B. V. Kumar, “Acoustooptic Matched Filter Correlator,” Appl. Opt., 21(13), 2356-2364, (1982). W. S. C. Chang and P. R. Ashley, “Fresnel Lenses in Optical Waveguides,” IEEE J. Quantum Electron., QE-16(7), 744, (1980). 30 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. B. U. Chen, E. Marom, and A. Lee, “Geodesic Lenses in Single-Mode LiNb03 Waveguides,” Appl. Phys. Lett., 31(4), 263-265, (1977). J. D. Cohen, “Incoherent-Light Time-Integrating Processors,” in Acousto-Optical Signal Processing , N. J. Berg and J. N. Lee, Eds., 225-288 (Marcel Dekker, Inc., New York, 1983). J. Dali, J. H. Jorgenson, E. L. Christensen, and S. N. Madsen, “Real-Time Processor for the Danish Airborne SAR,” IEE Proceedings-F, 139(2), 115-121, (1992). M. De Micheli, J. Botineau, P. Sibillot, D. B. Ostrowski, and M. Papuchon, “Fabrication and Characterization of Titanium Indiffused Proton Exchanged (TIPE) Waveguides in Lithium Niobate,” Opt. Commun., 42, 101-103, (1982). M. W. Haney, “Compact Acousto-Optic Processor for Synthetic Aperture Radar Image Formation,” Proc. IEEE, 82(11), 1735-1748, (1994). T. Hirata, M. Suehiro, M. Hihara, M. Dobashi, and H. Hosomatsu, “Demonstration of a Waveguide Lens Monolithically Integrated with a Laser Diode by Compositional Disordering of a Quantum Well,” IEEE Photonics Tech. Lett., 5(6), 698-700, (1993). R. J. Holmes and D. M. Smyth, ‘Titanium Indiffusion into LiNb03 as a Function of Stoichiometry,” J. Appl. Phys, 55(10), 3531-3535, (1984). S. Y. Huang and S. H. Lee, “Blazed Grating Couplers on LiNb03 Optical Channel Waveguides and Their Applications to Integrated Optical Circuits,” J. Lightwave Technol., LT-4, 1304-1310, (1986). A. Kar-Roy and C. S. Tsai, “New Integrated Acousto-Optic Matrix Algebra Processor Architecture,” Appl. Phys. Lett., 59(24), 3093-3095, (1991). 31 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Z. S. Karim, “Thin Film Coatings for Optical Information Processing and Computing Applications,” Ph.D. Thesis, University of Southern California, (1993). J. J. Kovaly, Synthetic Aperture Radar, (Artech House, Inc., MA, 1976). T. Kubota and M. Takeda, “Array Uluminator Using Grating Couplers,” Optics Letters, 14(12), 651-652, (1989). K. Y. Liao, C. C. Lee, and C. S. Tsai, ‘Time-Integrating Correlator Using Guided- Wave Anisotropic Acousto-Optic Bragg Diffraction and Hybrid Integration,” in the 1982 Topical Meeting on Integrated and Guided-Wave Optics, Technical Digest WA4-1 to 4, IEEE Cat. No. 82CH 1719-4, (Pacific Grove, Ca., 1982). G. A. Mastin, S. J. Plimpton, and D. C. Ghiglia, “A Massively-Parallel Digital Processor for Spotlight Synthetic Aperture Radar,” Int. J. Supercomputing Appl., 7(2), 97-112,(1993). D. Mergerian, E. C. Malarkey, and R. P. Pautienus, “High Dynamic Range Integrated Optical RF Spectrum Analyzer,” in the 4th International Conference on Integrated Optics and Optical Fiber Communication, Paper 30B3-b, (Tokyo, Japan, 1983). D. Mergerian, E. C. Malarkey, R. P. Pautienus, and J. C. Bradley, “Diamond Machined Geodesic Lenses in LiN b03,” Proc. Soc. Photo-Opt. Instrum. Eng., 176, 85, (1979). D. Mergerian, E. C. Malarkey, R. P. Pautienus, J. C. Bradley, G. E. Marx, L. D. Hutcheson, and A. L. Kellner, “Operational Integrated Optical R. F. Spectrum Analyzer,” Appl. Opt., 19(18), 3033-3034, (1980). M. M. Minot and C. C. Lee, “A New Guided-Wave Lens Structure,” J. Lightwave Technol., 8(12), 1856-1865, (1990). 32 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). D. R. Pape, “Acousto-Optic Processor for Autonomous S AR Guidance,” in Advances in Optical Information Processing IV, SPEE Vol. 1296, 206-221, (1990). D. Psaltis, ‘Tw o Dimensional Optical Processing Using One Dimensional Input Transducers,” Proc. IEEE, 72, 962-974, (1984). D. Psaltis, M. Haney, and K. Wagner, “Real Time Synthetic Aperture Radar Processing,” in the NASA Conference on Optical Information Processing for Aerospace Applications. II, (NASA, Langley, Virginia, 1983). D. Psaltis and K. Wagner, “Real-Time Optical Synthetic Aperture Radar (SAR) Processor,” Opt. Eng., 21(5), 822-828, (1982). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). K. Rastani, S. DeMars, A. R. Tanguay, Jr., N. C. Frateschi, R. Kapre, and A. Madhukar, “Grating Outcoupling from Large Area Rib Waveguide Arrays Fabricated on GaAs/AIGaAs by Selective Ion Beam Milling,” in the 1989 Annual Meeting of the Optical Society of America, OS A Technical Digest Series, Vol. 18, 116, (Optical Society of America, Orlando, Florida, 1989). K. Rastani and A. R. Tanguay, Jr., “Large Aperture Negative Meniscus Singlet and Triplet Lenses with Positive Focal Lengths Developed on LiNb03,” (to be published). N. A. Riza, “In-Line Interferometric Time-Integrating Acousto-Optic Correlator,” Appl. Opt., 33(14), 3060-3069, (1994). 33 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. W. E. Ross, “Miniature Ruggedized Optical Correlator Optimized for Space,” in LEOS Spacebome Photonics, Vol. 91068, 1-18, (LEOS, Newport Beach, Ca., 1991). S. H. Song, S. D. Jung, E. H. Lee, and S. G. Lee, “Back-Board Interconnections by Focusing Grating Coupler Arrays,” in the OSA Topical Meeting on Diffractive Optics, (Optical Society of America, 1994). R. A. Sprague and C. L. Koliopoulos, “Time Integrating Acoustooptic Correlator,” Appl. Opt., 15(1), 89-92, (1976). T. J. Su and C. C. Lee, “An Embedded Waveguide Lens with Anti-Reflection Layer,” IEEE Photonics Tech. Lett., 6(1), 89-91, (1994). N. S. Subotic and C. C. Aleksoff, “Real-Time Optical Processor for SAR Image Formation,” Opt. Comp, and Proc., June, 5-6, (1992). K. Sugii, M. Fukuma, and H. Iwasaki, “A Study on Titanium Diffusion into LiNb03 Waveguides by Electron Probe Analysis and X-Ray Diffraction Methods,” J. Mat. Sci., 13, 523-533,(1978). T. Suhara, S. Fujiwara, and H. Nishihara, “Proton-Exchanged Fresnel Lenses in Ti:LiN b03 Waveguides,” Appl. Opt., 25(19), 3379-3383, (1986). T. Suhara and H. Nishihara, “Integrated Optics Components and Devices Using Periodic Structures,” IEEE J. Quantum Electron., QE-22(6), 845-867, (1986). T. Suhara, H. Nishihara, and J. Koyama, “A Folded-Type Integrated Optic Spectrum Analyzer Using Butt-Coupled Chirped Grating Lenses,” IEEE J. Quantum. Electron., QE-18(7), 1057-1059,(1982). 34 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. M. Takeda and T. Kobuta, “Integrated Optic Array Illuminator: A Design for Efficient and Uniform Power Distribution,” Appl. Opt., 30(9), 1090-1095, (1991). A. R. Tanguay, Jr., “Integrated Optical Processing and Computing,” Optics News, 14(2), 23-26, (1988). L. Thylen and L. Stensland, “Electrooptic Approach to an Integrated Optics Spectrum Analyser,” Appl. Opt., 20(10), 1825-1832, (1981). C. S. Tsai, “Guided-Wave Acoustooptic Bragg Modulators for Wide-Band Integrated Optic Communications and Signal Processing,” IEEE Transactions on Circuits and Systems, CAS-26(12), 1072-1098, (1979). C. S. Tsai, D. Y. Zang, and P. Le, “Acousto-Optic Bragg Diffraction in a LiNb0 3 Channel-Planar Composite Waveguide with Application to Optical Computing,” Appl. Phys. Lett., 47(6), 549-551, (1985). C. S. Tsai, D. Y. Zang, and P. Le, “High-Packing-Density Multichannel Integrated- Optic Modules in LiN b03 for a Programmable Correlation of Binary Sequences,” Opt. Lett., 14(16), 889-891, (1989). D. W. Vahey, R. P. Kenan, and W. K. Bums, “Effects of Anisotropic and Curvature Losses on the Operation of Geodesic Lenses in Ti:LiNb03 Waveguides,” Appl. Opt., 19(2), 270-275, (1980). S. Valette, J. Lizet, P. Mottier, J. P. Jadot, S. Renard, A. Fournier, A. M. Grouillet, P. Gidon, and H. Denis, “Integrated Optical Spectrum Analyser Using Planar Technology on Oxidised Silicon Substrate,” Electron. Lett., 19(21), 883-885, (1983). 35 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. S. Valette, A. Morque, and P. Mottier, “High-Performance Integrated Fresnel Lenses on Oxidised Silicon Substrate,” Electron. Lett., 18(1), 13, (1982). C. M. Verber, R. P. Kenan, and J. R. Busch, “Correlator Based on an Integrated Optical Spatial Light Modulator,” Appl. Opt., 20(9), 1626-1629, (1981). T. Q. Vu and C. S. Tsai, “Ion-Milled Waveguide Lenses and Lens Arrays in GaAs,” J. Lightwave Tech., 7(10), 1559-1566, (1989). T. Q. Vu, C. S. Tsai, and Y. C. Kao, “Integration of a Curved Hybrid Waveguide Lens and Photodetector Array in GaAs Waveguide,” Appl. Opt., 31(25), 5246-5254, (1992). V. E. Wood, J. R. Busch, D. T. Moore, C. B. Woo ley, and W. H. Southwell, “Rectangular Luneburg-Type Lenses for Integrated Optics,” Opt. Lett., 8, 226, (1983). S. K. Yao and D. B. Anderson, “Shadow Sputtered Diffraction-Limited Waveguide Luneburg Lenses,” Appl. Phys. Lett., 33(4), 307, (1978). D. Y. Zang and C. S. Tsai, “Single-Mode Waveguide Microlenses and Microlens Arrays Fabrication in LiN b03 Using Titanium Indiffused Proton Exchange Technique,” Appl. Phys. Lett., 46(8), 703-705, (1985). 36 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 2 System Design Considerations In this chapter, we consider IOSAR processing and IO correlation in more detail in order to identify the system and device requirements. The basic analysis of the IOSAR processor presented by Rastani [Rastani, 1988] is contained herein for the purpose of review. This analysis presented herein includes an introduction to SAR concepts through a review of the Acousto-Optic/CCD SAR Processor [Psaltis and Wagner, 1982], a discussion of the preliminary device considerations for the IOSAR processor considered herein, a short description o f the geometrical considerations for the IOSAR processor, and an example of a synthetic aperture radar system with an appropriately designed IOSAR processor. The preliminary device considerations given herein are applied to a generic pro cessor architecture that they may suitably be applied to either the IOSAR processor or the IO correlator. We have added to the initial considerations presented by Kasra, the issue of beam expansion and collimation for a more complete analysis of the fully integrated processors. With these issues taken into consideration, the need for large- aperture, short-focal length lenses in these processors is emphasized which serves to demonstrate the underlying focus of this work. The concepts underlying interferometric time-integrating correlation are also introduced by reviewing a bulk optical system that performs the equivalent time-inte grating function [Cohen, 1983]. An integrated optical time-integrating correlator architecture similar to the one that we proposed in Chapter 1 has been previously con sidered by others [Tsai, 1979]. However, neither integrated optical correlator struc- 37 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ture has ever been fully implemented. We consider herein an example of system input signals and the corresponding processor design required to perform the correlation. The examples for the IOSAR processor and IO correlator described in this chap ter are used as the basis for the component design and system evaluation presented in the later chapters. This chapter is organized to so that the basic processor concepts are described in terms of the corresponding bulk optical processors in Sects. 2.1 and 2.2. General integrated optical device design considerations are given in Sect. 2.3, and the specific requirements for the IOSAR processor and IO correlator are given in Sects. 2.4 and 2.5, respectively. A summary of the selected system and device parameters used in the later chapters for device development and system evaluation is then given in Sect. 2.6. 2.1 Acousto-Optic/CCD SAR Processor In the Acousto-Optic/CCD SAR processor shown schematically in Fig. 2.1, the transmitted radar signal is generated by an airborne or spacebome platform with a radar antenna that operates in the side-looking geometry, as was shown in Fig. 1.1. This platform travels at height h along the direction T j (azimuth) with a constant velocity v [Kovaly, 1976; Psaltis and Wagner, 1982; Psaltis, et al., 1983]. A radar sig nal is used to illuminate a region of the ground and the reflected wave is received by the moving platform. This radar return signal contains radar reflectivity information for each of the ground coordinates within the illuminated region. The objective of the AO/CCD SAR processor is to construct a high resolution ground image by coherently processing several radar return signals over time. This image consists of point scat- terers with ground coordinates (£, Tf) within the radar beam’s path. Since the syn thetic aperture radar image formation operation is linear, the impulse response of a general point scatterer at range and azimuth coordinates (£0, Tj0) completely describes 38 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. L S ; o T SCt) •I - 2 AOO L, L , r t t ) OUTPUT f 3 + f 5 d P, T O P VIEW L S AOO S(f) r'(t) A S ID E V IE W MASK CCD L S AOO a-» OUTPUT Figure 2.1 Acousto-oplic. CCD SAR processor. The elements of the processor include: LS-ligkt source, AOD-acoustooptic device, Fj is the focal length of the lens Lj, and the broken lines indicate the path of the rays of the reference wave (from [Psal tis and Wagner, 1982]). 39 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the system. The transmitted radar signal is encoded with a linear frequency-modulated (FM) waveform in order to reduce the peak transmitted power. This technique allows for the practical use o f low-power, compact radar transmitters on both airborne and spa- cebome systems. The FM pulses have a chirp rate Bc over a duration Tand are peri odically repeated. The pulse repetition frequency (PRF) is the frequency at which the radar signal is pulsed, and the time between pulsed is given by T, which is equal to 1/PRF. The transmitted waveform is given by S(t) = rect -— — expj^jB c(t - sT )2 j exp(j27tf0t) , (2.1) 7 _ _ S in which s is the index o f each transmitted pulse and f 0 is the microwave carrier fre quency. The received radar return signal r(t) from a single point scatterer on the ground at a slant range R from the transmitter has the form r(t) = S ( t - 2 R / c ) , (2.2) in which the instantaneous range R(t) is time-dependent. The range variations can be neglected within the duration of a single pulse so that R(t) = R(sT). The distance Rq = -Jh2 + % 2 is typically much greater than the distance \h - h0\, which allows the radar return signal in Eq. 2.2 to be rewritten as KO = £ rect ■ 2 exp\jB c(t - 2 Rq/c - sTY s L T J exp [~j{27tf0/cR Q )(ysT -r]0)2] exp(y2^f0 r) • (2.3) The AO/CCD SAR system must process this radar return signal in order to con struct the image of a given point scatterer on the ground. In reference to Fig. 2.2, the radar return signal is input into the processor after it is heterodyned to the central 40 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. RANGE COMPRESSION BY SPATIAL INTEGRATION CCO ARRAY DETECTOR AO 8RAGG CELL PULSED LASER DIODE IN PU T RAOAR (SAR) SIG N A L ZERO OROER STO P AZIMUTH COM PRESSION BY TEM PORAL INTEGRATION TEMPORALLY MOOULATED, SPATIALLY UNIFORM ILLUMINATION STATIONARY MASK J t SHIFT ANO AOO CCO OETECTOR ARRAY Figure 2.2 The upper diagram depicts the range compression technique described in detail in the text. The lower diagram shows the azimuth compression technique (from (Rastani, I988J). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. operational frequency of the acoustooptic device (AOD). The input signal forms a travelling wave comprising chirped signals reflected from the ground targets. The laser source is pulsed after the reflected signal has filled the AOD aperture. The light diffracted from the AOD expands uniformly in the vertical direction and self-focuses in the horizontal direction due to the chirped FM waveforms as shown in the upper portion of Fig. 2.2. This self-focusing results in a radar focused image in the range direction only. The range-focused image is multiplied by a fixed two-dimensional mask on which the azimuth phase history has been recorded, such as that shown in Fig. 2.3. The product of this incident intensity distribution and the transmissivity of the mask is recorded on a two-dimensional CCD detector array. The photogenerated charge dis tribution is continuously shifted in a step-like fashion in the vertical direction between each laser pulse as shown schematically in the lower portion of Fig. 2.2. The relative motion of the accumulating photogenerated signal on the CCD array and the station ary two-dimensional mask results in the correlation required to perform the image compression in the azimuth direction. The signal is continuously removed from the readout buffer of the CCD array and contains the range- and azimuth-compressed ground image. In this manner, the AO/CCD SAR processor is able to form a two- dimensional image using a one dimensional AOD by performing spatial integration for the range compression function and temporal integration for the azimuth compres sion function. The IOSAR processor likewise performs spatial and temporal integra tion using a one-dimensional SAW device, as was discussed in Chapter 1. The operation of the AO/CCD SAR system is given in more detail next since the same operational considerations apply to the IOSAR processor. After each radar pulse from the airborne or spacebome platform, the radar return signal (Eq. 2.3) is heterodyned to the center frequency of the AOD, which con sists of a piezoelectric material with a metal input transducer. The resulting trans ducer modulation signal is proportional to 42 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 2.3 A typical mask function bearing the azimuth Doppler phase history in the vertical dimension and the range dependence of the azimuth phase history in the horizontal direction (from [Rastani, 1988]). in which f , is the center frequency of the AOD, and f 2 is the reference signal fre quency needed to maintain the phase information in the return signal when the light diffracted by the AOD is recorded onto the CCD array. The phase of the signal reflected from a given ground target is affected by a Doppler upshift or downshift in frequency as the radar antenna approaches or recedes from the ground target, respec tively. The light amplitude o f the source is modulated by in synchronization with the received radar signal that has a pulse repetition frequency (PRF) of 1 IT, in which T is the length of time between radar pulses. These periodic rV) = K O exp [-j2 7 c(f0 - f,)t] + A exp (j2 jtf2 t) , (2.4) (2.5) 43 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. pulses must have a duration t0 that is much shorter than the time resolution of the AOD in order to prevent smearing of the range information during the acoustooptic interaction. Another concern of importance is to maintain coherence of the light pulse over the entire pulse duration, since the azimuth compression requires interferometric processing with the reference signal. The lenses LI and L2 collimate the laser beam in the horizontal direction and focus it in the vertical direction to obtain uniform and efficient interaction with the modulated refractive index o f the AOD over the width of the device. The acoustooptic modulation signal is given by in which x is the direction of the acoustic wave propagating with velocity va. The AOD is slightly tilted with respect to this input beam so that AO interactions take place near the Bragg angle. The diffracted light is collimated by lens L4 in the verti cal direction so that the light uniformly illuminates plane P2 in this direction. Lenses L3 and L5 position the focal plane of the light diffracted by the chirped waveforms onto the plane P2 and collimate the light diffracted by the reference signal so that it uniformly illuminates plane P2. The light incident on plane P2 is multiplied by the transmissivity of a two dimensional mask in which the spatial frequency u0 is given by (f2- //) /v a and the parameter b2 [Psaltis and Wagner, 1982] is given by S,(t) = />(*)/ t + — , V va (2.6) 1 i r l a T (x ,y ) = — + —cos 2mt0x + 2 2 x (2.7) 4 f 0v2T 2va c2A y2 (2.8) 44 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Here, A y is the separation between each pair of the CCD pixels immediately behind this mask. The CCD array is operated in the shift-and-add mode so that light dif fracted by the AOD and multiplied by the mask will photogenerate charges that add to the existing charge distribution. The array is clocked in synchronism with the laser source and radar system to electronically shift this charge distribution by one pixel after each integration period. The charge generated by the 5th radar pulse will be shifted by (N - 5) pixels in the vertical direction after N exposures.The charge distri bution on the CCD array after N exposures is given by [Psaltis and Wagner, 1982] The final output of the processor is the focused image of the scatterer, impressed onto a spatial carrier with frequency u0 and shifted by a constant phase term (p. This carrier arises from the use of a reference wave in the AOD, and allows the signal to be easily separated from the bias terms. This image is continuously generated by the CCD array, producing one azimuth slice after each radar pulse. The processing for each slice is completed when the accumulated charges in one row of the CCD reach the edge of the device. A separate stage o f the CCD transfers the range image for each azimuth slice to the output pin of the CCD. The focused radar image has a sinc-function-Iike behavior in both the range and azimuth directions. The ground resolution is related to the width of these sine func tions in each direction and is given by N x s in c + bias term s (2. 10) 45 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the range direction and 5 = c*yRo x vT = cRp = X0R0 n 2N f0(vT )2 A y 2N f0vT 2 Ds (2-H ) in the azimuthal direction. The parameter Ds is the synthetic aperture of the system, and is given by NvT. This is the distance the radar platform travels during the integra tion period o f duration NT. 2.2 Acousto-Optic Interferom etric Time-Integrating Correlator The signals involved in time-integrating correlation may be processed as real quantities by non-interferometric methods, or as complex-valued quantities by inter ferometric methods that use a reference beam to retain the signal’s phase information [Cohen, 1983]. In Fig. 2.4, a bulk optical time-integrating correlator architecture is shown that employs an acoustooptic device (AOD) to input one of the correlation sig nals, and a laser source to input the other. The purpose of this processor is to form the correlation of two independent signals g{t) and h(t) that each have signal bandwidths o iB s. The laser diode signal g(t) is modulated onto a carrier with frequency f 0 and with a bias term [Cohen, 1983] so that the intensity illuminating the acoustooptic device A 02 is \E,\2 = a + g (t)c o s2 jtf0t • (2.12) The AOD is operated without a bias, so that the diffracted light intensity is pro portional to the electrical drive power. The acoustic wave consists of a reference tone at frequency f c and the signal h(t) on a carrier with frequency f 0 + fc. The light dif fracted by the AOD has an electric field distribution given by 46 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Top View S LO L I A 02 L4 F2 L5 D Side View Figure 2.4 One-dimensional bulk optical correlator architecture with light source S, collimating lens LO, cylindrical lens LI, acoustooptic device AO, Fourier transform lens L4, spatial filter F2, imaging lens L5, and linear detector array D (after [Cohen, 1983]). E2 = E ^ b + e~m o ^ ) h \ t - T ) Y m c i t ~T) • (2-13) In this equation, T= x/v, in which x is the coordinate within the acoustic aperture and v is the acoustic velocity. The diffracted light is imaged by a pair of lenses onto a lin ear detector array. The undiffracted light is removed by the spatial filter F2. The intensity of the light incident at the detector array is l3 - \E ^ . Photogenerated charges in the detector array accumulate over a period T, which is the time it takes for the entire signal hit) carried on the acoustic wave to pass through the aperture of the acoustooptic device. As a result, the charge distribution that is generated in the linear detector array after the integration period T can be written as 47 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. C(T) = \ T (,(* , t)dt = u {t) + v( t ) + b Re[ejl7*°TRgh ( r ) ] , (2.14) in which u( t ) = Tab2 , v(^) = a j T \h (t-T )\2dt , (2.16) (2.15) and R g h ( T ) = j Tj ( t ) h * ( t - t ) d t . (2.17) The signal u(t) is a signal-independent bias term and v(t) is a signal-dependent term. The right hand side of Eq. 2.14 actually has several more terms that are not shown, all of which have no frequencies in the neighborhood of zero as long as f 0 > 3By With detector integration, these terms approximately vanish. The third term in Eq. 2.14 contains the correlation of the complex-valued signals. The complex-valued correla tion can be directly processed since the correlation is on a carrier with a spatial fre quency fijv. In order to perform this integration with the appropriate grouping of terms shown in Eq. 2.14, the bandwidth of the Bragg cell B b must span the frequency range (fc, f c + f0 + Bs) and therefore must be approximately four times the bandwidth of the input signals. The desired output correlation is on a carrier. In order to recon struct the correlation signal, each cycle of the carrier must be sampled at least twice (by two detector elements). This is the minimum sampling required according to the sampling theorem and is equal to the Nyquist rate. The number of processor resolu tion elements available from this correlator architecture is R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. { s + J o (2.18) 48 in which Tg = D /va is the acoustic time aperture associated with a Bragg cell of width D and acoustic velocity v^, and ZgBg is the time-bandwidth product (TBWP) of the Bragg cell. The number o f required detector elements [Cohen, 1983] is ND= ^ LNres = 2rBBB. (2.19) B S Therefore, in order to process the complex-values correlation of the two signals g(t) and h{t) with this processor architecture, two detector elements are needed for every Bragg cell resolution element. This relationship will be used later in an example to determine the number of detector elements required to process two signals with a par ticular value selected for the signal bandwidth Bs. 2.3 Preliminary Device Considerations We consider the general in-line architecture shown in Fig. 2.5 as a first order extension into an integrated optical format of the bulk optical processors described above. The characteristics and the spatial relationships o f integrated optical compo nents with one another in such a processor determines the processing capacity and complexity achievable in the limited extent of the substrate dimensions. The con straint of substrate size is one of the primary factors that limits the flexibility and per formance of IO signal processors. In the in-line architecture, the requisite signal processing is accomplished by a single pass of an optical beam through components lined up along a common axis. Through the use of integrated mirror or beam-splitting components, larger processing capacities and more complex signal processing tasks could be realized; however, this is beyond the scope of the present work. Herein, we restrict the discussion to the simple in-line architecture in order to demonstrate the preliminary device requirements of the different components. The in-line architecture shown in Fig. 2.5 consists o f a planar waveguide with 49 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. I n in IV v VI I Figure 2.5 Top and side views of a generalized in-line integrated optical architec ture, in which edge and surface mounted devices are in the dark shaded areas and waveguide devices are in the light shaded areas. The devices in each region include (I) an edge-coupled source element; (U) beam expanding and collimating lenses; (HI and IV) signal input devices, imaging or Fourier transform lenses, and beam stops; (V) channelized waveguide arrays, waveguide couplers, surface mounted devices; and (VI) an edge-coupled detector device. waveguide components in areas indicated by light shading, and hybrid integrated components in areas indicated by dark shading. The six separate regions of this archi tecture have associated functions such as optical signal input, modulation, and output. In Region I, a laser diode is butt-coupled to the waveguide. The out-of-plane beam divergence angle of the laser diode is matched to the numerical aperture of the waveguide, and the in-plane divergence angle 0|| of the laser diode is matched to waveguide lens components in Region n. The lens components in Region II provide the required beam expansion and collimation for the rest of the components in the processor. In Region HI, signal input devices are integrated with the waveguide that operate through modulation of the guided optical wave. Common devices include those that operate based on acoustooptic or electrooptic principles. There may be one 50 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. or several such signal input devices. This region also consists of beam-blocking devices such as deep-etched grooves or surface metal films to block, divert, or attenu ate unwanted light in the waveguide and lens elements that perform imaging or Fou- rier-transform operations. Region IV is similar to Region HI, and is included to demonstrate that there can be several stages of modulation and lens devices. Region V consists o f the channelized waveguides, surface outcoupling gratings, and surface- mounted processing elements used for signal output or for further out-of-plane pro cessing. Alternatively, the signal may be output from a polished waveguide edge in Region VI and detected with a butt-coupled linear detector array. The individual devices are described in more detail below. 1. Waveguide The waveguide substrate is a fundamental part of the integrated optical proces sor, and both the substrate material and waveguide fabrication process must be care fully selected. To this point in the development of integrated optics, no single waveguide substrate has been shown to be sufficiently versatile as to concurrently sat isfy the diverse needs of optical signal processing in the way that silicon has for elec tronic signal processing. Therefore, for each integrated optical signal processing application, the crucial requirements should be identified, and a waveguide substrate should be selected that preferentially satisfies those requirements. The primary waveguide substrate requirement to consider is that of device com patibility: laser and modulator devices are needed to input signals, passive waveguide components are needed to steer and channel the information-carrying optical signal, and detector devices are needed to collect and convert the information into an electri cal output signal. The device compatibility is determined largely by the physical properties of the waveguide substrate and by fabrication process compatibilities. For many signal processing applications, waveguide substrates that possess par ticular physical properties such as high acoustooptic or electrooptic coefficients are 51 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. desirable for the integration of efficient guided-wave modulators. The advanced 10 signal processors discussed herein, for example, rely on the use of a surface acoustic wave (SAW) device as a signal input transducer. The maximum SAW interaction length with the guided beam, transducer coupling efficiency, and the highest attain able bandwidth are also determined by the physical properties of the waveguide sub strate. Surface acoustic wave propagation losses can limit both the SAW interaction length and the modulation bandwidth, while high surface acoustic wave velocities will lead to a low SAW spatial bandwidth. The refractive index of the waveguide substrate also has considerable impor tance. The wavelength of the light inside the waveguide material scales inversely with the index of refraction. This dependency has direct consequences for almost every waveguide device (e.g., the focal spot size of a lens, the deflection angle from a periodic SAW modulation, and the surface outcoupling grating properties all depend upon the waveguide refractive index). In addition, many waveguide substrates are birefringent, and the propagation constant may depend upon the direction of propaga tion of the guided wave. The selection of the optical operational wavelength is a complicated matter in which the transparency range of the material, the properties of the laser source, and the wavelength dependence of the waveguide substrate physical properties should all be considered. For example, both the waveguide scatter losses and SAW diffraction efficiencies scale inversely with wavelength. Thus, the optical wavelength should be selected such that these two properties are within tolerable limits. The waveguide depth and mode structure are important factors in the design and fabrication of the waveguide components. The laser coupling efficiency, SAW dif fraction efficiency, required embedded lens recess depth, and required rib waveguide separation depth all depend on the waveguide depth. For all of these devices, it is beneficial to have a shallow mode depth since the etch depth requirements for the embedded lens and rib waveguide components are likewise reduced. Single mode 52 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. behavior is also desirable in order to avoid intermode scattering that can occur in the separate waveguide components. For almost any application, the waveguide propa gation losses from sources such as absorption and scattering should be low enough to permit reasonable power levels to be used in the processor operation. In the fabrication of the different devices, issues o f process compatibility can dictate the device fabrication sequence. In general, thermal processing and etching should all be done first, followed by thin film depositions and then external device hybridization. A number of other waveguide substrate properties should be consid ered such as waveguide stability during high temperature processing, waveguide coefficients o f thermal expansion, waveguide susceptibility to various etchants, waveguide resistance to atmospheric corrosives, and waveguide-to-thin-film adhesion properties. 2. Laser diode Semiconductor laser diodes (LDs) are the sources normally used for integrated optical processors because of their compact size and low cost. Due to their wide applicability, laser diode production has grown into a full-fledged commercial indus try and a wide variety of source wavelengths and powers are now available. Other laser properties such as stability and reliability are well known and can be carefully controlled. With the use o f hybridization techniques, laser diodes may be coupled to waveguides of a completely different material. Therefore, the fabrication of the laser diodes can be considered independent of the waveguide fabrication. The complexity of this issue increases for the case of monolithic integration of a source on a waveguide, since all of the waveguide processing steps, waveguide structure, and operational wavelength must be compatible. The offsetting benefits of monolithic integration include the efficient coupling and the high accuracy of photolithographic alignment that can be achieved. Laser diode sources are typically butt-coupled to the waveguide substrate. The 53 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. laser diode is placed in close contact to a polished end-facet of the waveguide and the laser radiation is injected directly into the waveguide layer. The divergence angle of the laser perpendicular to the waveguide plane must be less than the acceptance angle of the waveguide for efficient coupling to occur. The waveguide acceptance angle 6a is given by in which NA is the numerical aperture of the waveguide, tif is the refractive index of the waveguide layer, and ns is the refractive index o f the substrate. For waveguides with a small refractive index difference between the guiding layer and the substrate layer, high coupling efficiencies are difficult to achieve. The acceptance angle for the Ti:LiNb03 waveguides that are discussed in Chapter 3 and used in the research described herein is 14°, which is smaller than most laser source divergence angles perpendicular to the laser junction is typically on the order of 20° to 40°). In such cases, a GRIN lens may be used to improve the coupling efficiency. The laser diode source must also supply sufficient optical power to overcome the system power losses and have enough power left to carry out the signal processing operations. Typical system losses include mode coupling losses at waveguide bound aries, waveguide propagation losses, and modulator losses. 3. Beam expansion and collimation lenses For most integrated optical processors, beam forming optics are needed to expand and collimate the input laser beam. The natural divergence of the laser beam is often used for beam expansion, and a single lens element is often used for collima tion [Anderson, 1978; Bamoski, et al., 1979; Mergerian, et al., 1983; Valette, et al., 1983]. For purposes of collimation, both geodesic [Chen, et al., 1977] and Fresnel [Suhara, etal., 1983] type lenses have been used previously. To illustrate to the spatial requirements for the beam forming optics, consider (2.20) 54 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. the case shown in Fig. 2.6(a), in which a final collimated beam width of 2H is desired. For a single element lens, the required focal length is / = ta n (« il/2 )' <2'21) For example, to form a 1-cm-wide collimated beam from the guided emission of a laser diode with divergence 0j( = 10° inside the waveguide medium, a lens with a focal length of 5.7 cm placed an equal distance from the waveguide edge is required. To prevent such a large portion of the waveguide substrate from being used for collimation purposes alone, a second beam expanding lens element may be used to increase the beam divergence as shown in Fig. 2.6(b). One possible lens combination that accomplishes this includes a beam expanding lens with a focal length f e placed a distance de from the waveguide edge. The required focal length of the collimating lens f c is given by f 1 H fc (l + d '/ f e) tan(0j,/2) (2'22) and the required spacing of this lens from the waveguide edge is d = 1 ( d e + f e d* + f e H ta n (0 i|/2 ) (2.23) For the example above, a small 1-mm-aperture f/5 defocusing lens element placed 5 mm from the waveguide edge in combination with a 1 cm aperture f/2.8 positive lens element placed 3.1 cm from the waveguide edge will collimate the beam to a 1 cm width. The use of a beam expanding element approximately halves the space required for collimation. The f-number of the beam expanding element is high and can be satisfied by a number of integrated lens technologies. The use of a beam 55 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (a) (b) Figure 2.6 Two schematic diagrams that show a laser diode end-coupled to a waveguide with (a) a beam collimation lens element LI, and (b) both beam expansion lens element, L2, and collimation lens element, L3. expanding element, on the other hand, leads to the requirement for a low f-number collimating lens element, which in practice is more difficult to achieve. The required intensity distribution of the collimated beam for illumination of the devices in Regions in and IV should also be considered. For example, the focal spot size of a lens in Region HI illuminated by the collimated beam depends on illu mination uniformity. The focal spot width and side-lobe level is an important param eter in characterizing the performance of the system. A lens with uniform illumination across the aperture will have a sidelobe level 13 dB down from the cen tral peak level. For Gaussian-weighted illumination truncated at 1.5 times the e-2 width in intensity, the highest sidelobe is -3 4 dB down. As a consequence, however, the focal spot size also increases by 40% [Tsai, 1990]. For the example given above, 0|l is specified at the full-width-at-half-maximum (FWHM) points in the angular dis tribution of the laser emission. As a result, the resulting collimated beam will have 56 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. better uniformity than for the Gaussian-weighted illumination case. If the required aperture size is the same, but even greater illumination uniformity is required, then the requirements on the collimation lenses may be more stringent. The maximum number of lens elements used in an integrated optical processor can be limited by the throughput losses. Sources of loss in a lens structure include interface reflections, mode mismatch, and scattering. Typical waveguide lens throughput losses are of the order 0.5-1.5 dB. To minimize these losses, either the number of required lens elements should be reduced or the individual lens elements should be made with a lower throughput loss. 4. SAW Modulators Next we consider the signal input devices in Regions HI and IV. Several inves tigators have used electrooptic (E-O) modulators as input devices in integrated optical processors [Verber, et a l, 1981; Tsai, et al., 1989]. In the present work, we consider the use of a surface acoustic wave (SAW) modulator [Tsai, 1979], which is another common input device. This device is used to modulate the optical beam by creating a periodic refractive index perturbation on the waveguide surface along a direction nearly perpendicular to the guided wave direction. The direction and efficiency of the diffracted beam depends on the frequency and amplitude of the phase grating. These devices are well understood and can be reliably fabricated by photolithographic tech niques with bandwidths as large as 500 MHz. A number associated with SAW modulators that characterizes the processing capacity is the space-bandwidth product (SBWP). The SBWP is equal to the interac tion width of the index perturbation with the optical beam multiplied by the range of spatial frequencies that the device can create on the waveguide surface. The attain able range of spatial frequencies is inversely proportional to the surface acoustic wave velocity. Therefore, to achieve a high SBWP, a large interaction length and a low sur face acoustic velocity are desirable. 57 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. A spatial characteristic of SAW modulators that has implications for the proces sor design is the range of deflection angles of the optical beam. The deflection angle is given by the Bragg condition for the phase grating period and can range from less than a degree to several degrees. The remaining waveguide components are oriented along the path o f the diffracted light. The deflection angles are further considered in conjunction with waveguide lenses in Regions EH and IV. The optical power loss associated with the SAW transducer is related to the dif fraction efficiency and the response linearity. In order for the deflected light power to be proportional to the SAW input power, a low diffraction efficiency is required. As a result, there will be an optical power loss of 3 dB or more in the processor that is related to the use o f SAW transducers. 5. Fourier -Transform and Imaging Lenses In Regions HI and IV, lens components can be used to Fourier transform an input spatial intensity distribution into an angular spectrum. The angular components can the be either detected or selectively filtered. This lens operation is often used in conjunction with a SAW transducer to remove the undiffracted light or unwanted dif fraction orders. In such cases, the lens aperture should match the SAW interaction length to prevent loss of information. For some applications, this may require that the lens aperture be quite large. The separation of the deflected and undeflected beams is equal to/tan(0£>), in which Op is the deflection angle. Therefore, a lens with a focal length o f /= 1 cm used in conjunction with a surface acoustic wave that gives rise to a 2° deflection angle would result in a 350 fj.m separation between the undiffracted and diffracted orders in the focal plane of the lens. This separation permits sufficient tol erance in photolithographic methods to fabricate a zero-order beam stop that removes the undiffracted light. Some processors require the use of imaging lenses in Regions HI and IV. The aperture size of the imaging lenses should also be approximately the same width as 58 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the input optical beam for the full processing capacity to be achieved. From the imag ing condition, the minimum distance between the object plane and image plane is approximately four times the focal length of the lens. Hence, imaging operations are very costly in terms o f the amount of substrate space required. In applications where imaging lenses are employed, it is especially important to use low f-number lenses in order to conserve substrate space and maintain processing capacity. 6. Channelized arrays Channelized arrays can be used in signal processing applications to dissect the in-plane intensity distributions and route this information to other waveguide or hybrid-integrated components. The arrays may consist of a variety of structures that provide two-dimensional confinement, including titanium-diffused channel waveguides in LiN b03 or rib waveguides. Channelized arrays such as these can be fabricated uniformly by standard planar processing techniques. The structure o f the channel array must be designed to match the incident field distribution at the input end as well as the device dimensions at the output end. The number of channels in an array and their separation is chosen to appropriately sample the in-plane intensity distribution so that there is little or no loss of system resolution. The processor as a whole is designed such that the available pixel-to-pixel spacing o f the surface-mounted devices is sufficient to appropriately sample the outcoupled intensity distribution from the channel waveguide array. In the dissection o f the incident field distribution, the channel waveguide array must have the proper element density to fully sample the incident field, small separa tion widths to minimize dead space and maximize the coupling efficiency, and a numerical aperture sufficient to accommodate the angular extent of the incident field. As the dissected field propagates along the length of the channel waveguides, there minimal distortion of the dissected field (that can result from coupling between rib waveguides) and low propagation losses (in rib waveguide arrays that require a large 59 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. propagation length) are desired. The minimum channel width is limited by the degree of lateral confinement that can be maintained, as well as by the photolithographic pat terning resolution (approximately 1 /an). When gratings on the surface o f the channel waveguide array are used to illuminate surface-mounted devices, the light outcoupled from each channel waveguide must accurately illuminate the pixels positioned directly above it and not those to either side. The degree of accuracy depends upon the channel waveguide width, the diffraction angle of the outcoupled light, and prox imity of the surface-mounted device. Channel waveguide arrays are designed so that the acceptance angle of each channel waveguide is large enough to accept the largest in-plane angle of propagation of the incident field distribution. The channel waveguide acceptance angle increases with greater refractive index differences between the channel waveguide and gap region. This index difference is very small for titanium-diffused channel waveguides (0.05 or less between Ti:LiNb0 3 and LiNb03) and large for rib waveguides (1.2 between Ti:LiNb03 and air). Hence the use of rib waveguides is preferred when a large acceptance angle is required. The power losses for rib waveguides include both mode coupling losses and propagation losses. The total coupling efficiency of an incident intensity distribution on the front entrance plane of the rib waveguide array is roughly proportional to the rib waveguide fill factor (rib width divided by rib period). The propagation losses of rib waveguides can be very high due to scattering from surface roughness on the rib sidewalls. 7. Surface outcoupling gratings Surface outcoupling gratings are used to couple light from the rib waveguide to surface-mounted devices. We show in Chapter 4 that highly uniform gratings with periods as small as 2 jjm can be achieved by standard photolithographic means over regions as large as 1 cm2. 60 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The properties of the grating are determined by the grating depth and profile. For smaller grating periods and greater etch depths, the outcoupling efficiency is increased. As a result, more light is outcoupled over a shorter distance. Efficient out coupling such as this is desirable for detector coupling such as that required for the IO correlator. Low grating outcoupling efficiencies are desirable for uniform outcou pling over a large distance. Gratings with low outcoupling efficiency were considered by Rastani for use in the IOSAR processor [Rastani, 1988]. A loss of optical power is associated with the use of surface outcoupling grat ings. Light is coupled into the substrate as well as into the superstrate where surface- mounted devices are situated. Light coupled into the substrate goes unused and can also reflect in part from the lower substrate surface. This reflected light may impinge on a portion of the surface-mounted devices and lead to background illumination. An increased background illumination will raise the bias level in the detected signal in the CCD array and result in a lowered dynamic range. In order to bring advanced IO signal processors to realization, these issues must be carefully considered and the por tion of power directed towards the surface-mounted devices must either be increased or measures must be taken to prevent the light directed into the substrate from reach ing the surface devices. 8. Surface-mounted devices Surface-mounted devices are used in conjunction with grating couplers to per form additional out-of-plane computations. As with the laser diode, surface-mounted devices can be fabricated independent of the waveguide and waveguide components and therefore may consist of an entirely different material specifically suited to the device’s function. We consider in particular the function of light detection from the outcoupling gratings/rib waveguide combination with surface-mounted linear or two- dimensional CCD detector arrays. Two-dimensional CCD arrays are readily available commercially with a variety o f pixel sizes, light sensitivities, and frame rates. The CCD array should be kept in close proximity to the surface of the rib waveguide array 61 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. to prevent excessive spread of the outcoupled light before it reaches the detection plane. The close placement of the CCD array can be achieved with the use of micro spacers. In addition, the CCD array should be carefully registered with the rib waveguide array to prevent run-out error. For some applications such as the IOS AR processor, a mask with variable trans mittance is used to modulate the intensity of the outcoupled light. The feasible prox imity of the CCD array is a concern in this case since the outcoupled light must travel through the mask layer before it reaches the detection plane. The CCD detector array also serves as the optical-to-electrical signal converter for the output of the signal processor. The optical power supplied to the CCD array must be sufficient to utilize the full dynamic range of the CCD, which can be as large as 107. Noise originating from waveguide scattering can contribute to a bias signal and effectively lower the accessible dynamic range. Therefore, precautions must be taken to eliminate as much o f the background scattered light as possible. Sources of noise include in-plane waveguide scattering, scattering from the rib waveguide side walls, and reflected light from the grating substrate radiation modes. The optical power supplied to the CCD detector array must be high enough to overcome these system losses if the full dynamic range o f the CCD is to be used. 9. End-coupled detector array End-coupling of a detector array to the polished edge of a waveguide is a com mon technique used to detect one-dimensional in-plane intensity distributions in IO signal processors [Bamoski, et al., 1979; Liao, et al., 1982; Mergerian, et al., 1983]. End-coupling has the limitation that it cannot easily be extended to detect a two- dimensional signal such as that produced by the IOSAR processor. An additional drawback of an end-coupled detector array is that the waveguide edge must be care fully prepared without defects, otherwise the output signal may be distorted. 62 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 2.4 Integrated Optical Synthetic Aperture Radar (IOSAR) Processor In the review of the AO/CCD SAR processor in Section 2.1, we introduced the synthetic aperture radar processing concepts and reviewed SAR processing with an optical architecture that utilizes a one-dimensional modulation device. Although the manner in which the AO/CCD SAR completes this task is somewhat more intuitive because of its bulk optical nature, in almost all respects the functionality of the IOSAR processor is analogous with the bulk optical system. The IOSAR processor shown Fig. 1.2 consists o f the laser, lens, modulator, mask, and CCD array components integrated onto a common waveguide substrate. The radar return signal input device is the surface acoustic wave modulator in this case. The light from the laser is confined in the waveguide and most of the compo nents are aligned as they are fabricated or integrated, as opposed to the bulk optical system where adjustments may be made after the processor is assembled. The acous- tooptic interaction is designed to take place near the Bragg angle of the central fre quency within the SAW bandwidth, as in the bulk optical case. The linear FM modulation of the surface acoustic wave causes self-focusing of the diffracted light within the plane of the waveguide. An integrated lens helps this focusing to occur within the substrate dimensions. Since the diffracted light is not free to expand in the vertical direction, as in the bulk optical architecture, the range-focused radar image is expanded within the plane o f the waveguide with a rib waveguide array. The addition of a constant frequency reference signal in the input SAW waveform provides a phase reference to process the Doppler information in the radar return. The relative excita tion of a rib waveguide at a given range will be dependent upon the phase difference between the range focus and the reference signal. The range-focused image is expanded within the plane o f the waveguide by the rib waveguide array and outcou pled with surface outcoupling gratings up through the azimuth mask to the CCD array. The intensity transmission functions and operation o f the CCD are similar to that described for the AO/CCD SAR processor. 63 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.4.1 IOSAR Processor G eom etry In this section, the primary geometrical considerations in the design of the IOSAR processor are reviewed. Key spatial characteristics and relationships are determined for the interdigital transducer that inputs the radar signal; the large aper ture, low f-number lens that images the light diffracted from the SAW modulation; and the rib waveguide arrays that dissect and expand the range-focused image. A geometrical analysis that links these three elements was given by Rastani [Rastani, 1988] and will be summarized here as a starting point for this discussion. Consider the configuration shown in Fig. 2.7, in which the effect of diffraction from a linear FM pulse shape with geometrical length L c input into the path of a guided optical wave is schematically represented. This chirped waveform is a travel ing surface acoustic wave with the lowest frequency component (A 0c)_ 1 at the leading edge of the wave and the highest frequency component (A ,c)'* at the trailing edge. A collimated guided light beam within the waveguide is incident upon the index modu lation at an angle 6C (the Bragg angle for the SAW center frequency). The deflection angle at the front edge o f this upchirped signal is smaller than at the back edge. This variation in deflection angle results in a geometrically-determined focal length equal to the distance that the two diffracted rays travel before they intersect. This focal length was be geometrically determined by Rastani [Rastani, 1988] to be ta n 6 tc - ta n 60c ’ ( 2 '2 4) in which 90 lc are the diffraction angles from the lowest and highest spatial frequen cies, respectively, when Bragg-matched to the SAW center frequency. The associated focal spot size is o 27T f c s ' 7 t ’ a 2 5 ) 64 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Focus of two beam s diffracted at 0oc and e 1c T Collimated incident guided beam Waveguide array location r f tan Q o c f tan 6lc f tan B c Spatial frequency Chirpf_L,J_] \AlC Aqc I Focus of undiffracted beam incident at e c Integrated lens with focal length f and aperture 0 Focal plane of integrated lens Figure 2.7 Geometrical optics of the guided beam incident upon a linear FM index modulation and an integrated lens combination (After [Rastani, 1988]). in which ]3 is the longitudinal propagation constant of the waveguide. Rastani further showed that for typical waveguide and modulator parameters, the focal length f c is much larger than available waveguide substrate dimensions [Rastani, 1988]. The long focal length of the chirped surface acoustic wave necessi tated the placement of a lens with a focal length/and aperture D a distance lc away from the chirped index modulation. The new distance from the lens to the range- focused image is given by _ f c f - l c f F syS f c + f - h (2.26) The lens is designed so that Fsys is compatible with the available waveguide substrate size. The value for is approximately equal to the focal length of the le n s/fo r reasonable values off c (approximately 1 meter) and lc (a few millimeters). The focal 65 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. spot size of the chirped surface acoustic wave/waveguide lens combination is given by c I k Fsys sys~ p L (2.27) and is a measure of the minimum separation required between the focal spots of two chirped SAW signals in order for them to be resolved in accordance with the Rayleigh resolution criterion. The geometrically-determined focal field in the back focal plane of the system [Rastani, 1988] is given by ^ = { - ^ s y s -. (2.28) Jc The number of resolvable spots N res is quotient of the range of focus divided by the focal spot size of the system. This number is then equal to the space bandwidth product of the system [Rastani, 1988] given by H r e s = ( D ~ L c ) x V -A/c A0c j (2.29) in which is the spatial bandwidth of the surface acoustic wave. In order to A lc Oc preserve the full range resolution of the radar system, there should be N res rib waveguides in the rib waveguide array. In order to ensure that the light from the undiffracted beam does not impinge on any portion of the rib waveguide array, it was shown geometrically by Rastani [Rast ani, 1988] that the following inequality should be satisfied (tan.9lc + tandc)(fc- l c)> D (2.30) 66 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Notice that in the case of a downchirped SAW signal,./^ is negative, and hence this inequality cannot be satisfied. In such cases, however, the undiffracted beam focuses at the focal plane o f the lens. This focal position is situated before the rib waveguide array and is well separated from the diffracted orders. In this case, the undiffracted light is prevented from reaching the rib waveguide array by a beam stop. 2.4.2 IOSAR Design Param eters To develop a reasonable design for the IOSAR processor, the system parame ters, the physical constraints of the architecture, and the performance requirements of the application need to be identified. The important practical parameters to be deter mined include the lens aperture size and focal length, the SAW interaction length and bandwidth, and the rib waveguide width and number of elements. Some o f the physi cal constraints include the physical limitation of the substrate size, the minimum detector pixel spacing, and the maximum off-axis propagation angles required within the waveguide. The performance requirements include high channel isolation, full range resolution processing, and full utilization of the CCD detector dynamic range. The system design is based upon parameters associated with the radar system. To demonstrate the IOSAR processor design, we review the example given by Rastani [Rastani, 1988] and develop a few relationships useful in the design of the IOSAR processor. The example given by Rastani consisted of a radar system with a pulse duration T= 1 //sec and pulse repetition frequency 1 IT = 1 kHz. The radar pulse was a linear FM signal with a 50 MHz bandwidth and starting frequency f 0 of 10 GHz. The size of the range swath Rs of the airborne system was assumed to be 1 km. The chirp rate, given by the radar signal bandwidth divided by the pulse dura tion, was Bc = 5 x 1013 Hz/sec. The chirp rate and pulse duration were used in Eq. 2.10 to determine a range resolution of 8r = 3 m for this radar system. A total of 333 range resolution points 67 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. were obtained by dividing the range swath by the range resolution. The temporal record length of the radar signal Tr = IR J c+T is equal to the time required for the radar signal to make one round trip across the range swath plus the radar pulse dura tion. For this example, it was determined to be 7.6 /isec. In summary, the signal received by the radar antenna has a center frequency of 10 GHz, a 50 MHz bandwidth, and a duration of 7.6 fjsec. It consists of many over lapping linear FM pulses each of 1 fJsec duration. These overlapping signals are a result of the scattered return of the initial radar pulse from several point targets within the range swath. This radar return signal is heterodyned to the central operating frequency of the waveguide SAW transducer. The waveguide substrate has a surface acoustic velocity waveguide effective refractive index ne p and is operated at a free space optical wavelength o f Xq . The interdependencies of the IOSAR processor properties can be defined as . (2-31) D = D(va ;Rs, B ) , (2.32) ^sys = 5 (2.33) and ^ x sys= ^ xsys{^'0'<tleff'>v(X’^ ’ B,R s, l c , f } = ::SS ysNres , (2.34) in which A * ,, n e ^ , and va are the material dependent parameters; T, B , and R s are the radar dependent parameters; and lc and / are the processor design parameters. The quantities N res and D can be written in terms of their constitutive parameters as fol lows 68 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Nres = (Tr - x ) B ■ (2.35) and • V a = Trva • (2.36) The quantity D is the width of the SAW interaction region as well as the aperture size o f the integrated lens. A waveguide substrate with a low surface acoustic wave veloc ity is required to process a large range swath for practical aperture sizes. For example (from [Rastani, 1988]), an IOSAR processor composed of a y-cut Ti:LiN b03 waveguide (va = 3488 m/s for a z-propagating surface acoustic wave, n ^ = 2.2 at Aq = 6328 A) with the radar parameters given above will require a lens aperture of 2.7 cm to process the 1 km range swath. Most waveguide lens technologies produce lenses with an f-number of 4 or greater, which implies a lens focal length of 10.8 cm or longer. At best, 10.8 cm is sufficiently small to focus the range image within pres ently available lithium niobate waveguide substrate dimensions only if the input beam is externally expanded before it is coupled into the waveguide. In order to fit all of the optical elements onto a single chip, including the beam expanding and collimating elements, lenses with a much shorter focal lengths are required. This is the subject of a large portion o f the work presented in this thesis. In order to recover the full range resolution of the SAR system, the focal spot size of the chirped index modulation/lens combination must be matched to the CCD detector element size. A consequence of the use of low f-number lenses is a small focal spot size. Hence, in order to preserve the range-compressed image resolution, the rib waveguides and CCD detector array pixels must be densely packed. The high est density CCD arrays presently available have pixel spacings as low as 7 jum. For a processor with a short distance between the SAW path and the integrated lens (lc « f c,f; e.g., lc - 5 mm), the focal spot size of the system is given approximately 69 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. by (2.37) Using the values for signal bandwidth, signal duration, SAW velocity, optical wave length, and material effective index given above, and using an assumed value of 10.8 cm for the lens focal length, we find that the second term on the right hand side of Eq. 2.37 is an order o f magnitude greater than the first term. Thus, for a given radar pulse bandwidth and duration, the focal spot size is made smaller by decreasing the focal length o f the integrated imaging lens. With the present values for the SAR example, the above equation yields an 8 /tm system focal spot size. The total rib waveguide array width is likewise approximately Substituting N res = 330 (calculated using Eq. 2.35) and a system focal spot size of 8 fim into Eq. 2.38, the rib waveguide array width is 2.64 mm. For the illustrative example given by Rastani, a 10 /an rib waveguide pitch was chosen to correspond to typical CCD pixel sizes then available [Rastani, 1988]. With a 10 fjm rib waveguide pitch, 264 rib waveguides are required to cover the full range swath of the range focused image. There is some loss in the range resolution for this particular selection for the IOSAR processor design since the radar system carries information for 330 range resolution points. For the work presented herein, we have chosen to maintain the selection of a 10 fjm rib waveguide pitch so that our device fabrication results could be directly compared with Rastani’s. In Rastani’s work and in this work, the rib waveguide array is designed with 8 fim wide rib waveguides and 2 fim wide gaps. The 2 fim gap width is chosen since it is an acceptable width for photolithographic patterning purposes, and further it is wide enough to provide high channel isolation. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. res^sys • (2.38) 70 It should be noted that, despite our selection of a 10 fim rib waveguide pitch, it is technically feasible to preserve the range resolution in the IOSAR processor for the radar system example given above with the selection of an 8 fim rib waveguide pitch and CCD pixel size. In the example above, a 2.7 cm aperture lens was required to process the range swath of the radar return signal. We presumed in our calculations that a collimated beam was incident upon the SAW modulation plane. In order to achieve this, how ever, the light from a laser diode must first be expanded in the waveguide to a 2.7 cm width. The beam expansion and collimation with a single lens requires that a very large portion of the waveguide substrate be used for collimation purposes alone. In the case of a 10° laser divergence angle inside the waveguide, the light must propa gate 15.4 cm to the collimation lens (fcou = 15.4 cm) to fill the 2.7 cm aperture. This distance can be greatly reduced by using a two lens combination rather than a single lens. Following the example given in Sect. 2.3 for collimation of a beam with a 10° divergence to a I cm beam width, a lens combination was designed to collimate a beam to a 2.7 cm width. The combination consists o f a 1 mm wide f/5 defocusing lens positioned 5 mm from the waveguide edge and a second f/3 focusing lens placed 8 cm from the waveguide edge. Thus, in this 8 cm distance the beam may be expanded and collimated to a 2.7 cm width. The use of a beam expanding lens signif icantly shortens the necessary processor length. The total processor length is given by dcou + 21 c + Fsys + R[en in which the SAW plane is assumed to be situated halfway between the collimating and imaging lenses. For the example given in this section, dcou = 8 cm, 2lc = 1 cm, F ^ s = 9.5 cm, and R[en is the rib waveguide array length and is assumed to be 1 cm. Hence, the total substrate size should be 19.5 cm long (this corresponds roughly to a 7.5 inch wafer size). Substrates of this size are available for silicon-based waveguides, however, for lithium niobate this is much larger than the presently available substrate size of 4 inches and may be expensive as well as difficult to manufacture. This mismatch in 71 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. required and available substrate sizes clearly demonstrates the importance of estab lishing a wide-aperture, low f-number lens technology for IO signal processor appli cations. Later in Chapter 5, we will describe the embedded lens structure and show that it is feasible to design and fabricate an f/1.56 embedded lens combination that is appropriate for the IOSAR processor. The use of a low f-number lens combination such as this for the image-forming operation and the beam expansion and collimation operations permits the entire processing task to be accomplished on a single lithium niobate substrate. In this case, dcou - = 4.2 cm, 2lc = 1 cm, and R[en = 1 cm for a total o f 10.4 cm or approximately 4 inches. In the bulk optical SAR processor, it was necessary to add a constant frequency signal to the radar return input signal to serve as a reference and to record the Doppler phase information in the range-focused image at the CCD array. In a similar manner, a reference signal must be added to the SAW input signal in the IOSAR processor. From geometrical considerations, as long as the reference signal frequency is within the frequency bandwidth o f the input radar return signal, then the rib waveguide array will be uniformly illuminated. This reference signal may be electronically added to the radar return signal or a second transducer may be included on the waveguide sub strate to insert the reference signal separately [Rastani, 1988]. While the light diffracted from the reference wave must uniformly illuminate the rib waveguide array, the undiffracted light must not overlap any portion of the array. For the example given in this section, the inequality in Eq. 2.30 is satisfied, which indicates that the undifffacted light will not impinge upon the rib waveguide array. The azimuth compression is accomplished by the temporal correlation of the CCD recorded signal with the fixed two-dimensional mask placed directly above the rib waveguide array. The azimuth resolution was shown in Sect. 2.1 to be inversely proportional to the number of integration steps performed. Therefore, for a rib waveguide array with a 1 cm length and a CCD detector array with lO-^m-spaced 72 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. pixels, 1000 integration steps will be completed before each reconstructed azimuth slice is staged out of the CCD array. An important requirement for the laser source considered by Rastani is the pulse duration [Rastani, 1988]. This pulse duration (CCD array integration period) is kept short compared with the radar return time record length Tr divided by the number of range resolution points. This duration is the time that it takes for the range focused image to smear by one rib width due to the propagation o f the SAW modulation. This period is 20 nsec for the example given in this section. A corresponding laser modu lation frequency of greater than 50 MHz is therefore necessary. In addition, the por tion of the laser pulse duration in which the laser changes from its initial multi-mode behavior to single mode behavior must be small for advanced IO signal processing applications in which information is interferometrically processed. This duration is called the transition time. A typical Hitachi laser diode (model HLP1600) operating at 850 nm with 30 mW power, for example, has a transition time on the order of 1 nsec as determined by Haney [Haney, 1986]. These same lasers have modulation fre quencies up to 2 GHz (a switching speed of 0.5 nsec, which is comparable to the tran sition time) and a coherence length o f about 15 m [Creath, 1985] that is sufficiently large for coherent processing. Each rib waveguide should illuminate a row of the mask and CCD pixels uni formly along its 1 cm length. Therefore, low efficiency uniform gratings are required in the IOSAR processor. Over the 1000 integration steps needed to form the azimuth compressed image in the above example, the accumulated energy density within each CCD pixel should be less than the saturation energy density. The relation between the supplied laser power, the processor losses, and the CCD detection levels will be dis cussed in Chapter 7. 73 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.5 Integrated Optical Correlator The geometrical considerations for the integrated optical interferometric time- integrating correlator are nearly identical to those for the comparable bulk optical pro cessor properly scaled to the waveguide dimensions. The processor consists of a col limation lens, a modulation region, two imaging lenses with a Fourier filtering plane, and an optical outcoupling array and an external detector array. In order to discuss the critical features of an IO correlator design, we follow with an example. Consider the correlator architecture shown in Fig. 1.2 to comprise a y-cut Ti:LiNb03 waveguide with a z-propagating surface acoustic wave (va = 3488 m/s). A laser diode is used to input the signal g(t) with a bandwidth Bs of 15 MHz. Like wise, a SAW transducer with a 400 MHz centerband frequency is used to input the signal h{t), also with a bandwidth Bs of 15 MHz. The SAW modulator bandwidth required to form the correlation of these signals was discussed in Sect. 2.2, and is 60 MHz for this example. For a Ti:LiNb03 waveguide, a 1 cm surface acoustic wave/optical beam interaction length corresponds to a SAW time aperture of 2.8 fisec. These values are similar to those used in a recent bulk acoustooptic interfer ometric time-integrating correlator (7 MHz signal bandwidth, 40 MHz acoustic band width, 10 /xsec time aperture) [Riza, 1994]. In Sect. 2.2, we found that f 0 sets the spatial frequency of the correlation signal carrier, and should be at least 3Bs or 45 MHz in this example. For a reference fre quency f c of 370 MHz, the SAW modulator bandwidth will span a range of frequen cies between 370 MHz and 430 MHz. For this range o f frequencies, the deflection angle range within a Ti:LiNb03 waveguide is approximately 1.75° to 2.0°. The lens following the SAW modulation Fourier transforms both the diffracted and undif fracted light. We found from our example in Sect. 2.2 for this operation that the undiffracted light will come to a focus and be spatially separated from the diffracted orders by 350 |im if a 1 cm focal length lens is used. This provides adequate space 74 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. for the light to be spatially filtered. The integrated imaging lenses are required to have a I cm aperture in order to image the diffracted light onto the rib waveguide array. If both of these lenses are f/l, a total of 4 cm of the waveguide substrate will be occupied for this imaging operation. The processor consists of a beam expansion/collimation lens combination with parameters equal to those given in the example of Sect. 2.3 for collimation of a 1 cm wide beam. These beam-forming lenses occupy a total of 3 cm of the substrate. The SAW interaction region and rib waveguide array require very little space (less than 0.5 cm combined). The total substrate length required for this example is 7.5 cm. Hence, all of these components can to fit onto a single 4 inch substrate. Using Eq. 2.18, a total o f n = 84 resolution elements were calculated for this correlator deign. The number of detector pixels required to sample the output in order to recover the complex-valued correlation is twice the space-bandwidth product of the surface acoustic wave, or N q = 333. Hence, the processor requires a rib waveguide array and detector array with 333 elements across the 1 cm image intensity distribu tion. The corresponding center-to-center rib waveguide spacing is 30 fim. The use of 2 fim gaps between rib waveguide elements leads to a rib waveguide width of 28 fim. Highly efficient surface outcoupling gratings are then used to outcouple the light in each channel and direct it towards the detector array. The laser diode is modulated with the second correlator input signal introduced on a carrier with a 45 MHz frequency. As was mentioned earlier, these modulation rates can be easily obtained with laser diodes. The resulting correlation is recorded on a linear CCD detector array. The detected signal consists of bias terms in addition to the correlation signal on a carrier, so that post processing can be used to extract the complex correlation of the input signals. 75 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2.6 Summary In this chapter we reviewed the bulk optical implementations of a synthetic aperture radar processor and interferometric time-integrating correlator. Preliminary device considerations were given for a general integrated optical architecture. The system and device requirements of the IOSAR processor and IO correlator were each demonstrated with an example that takes into account typical input signals. For the IOSAR processor, we reviewed a processor design with a Ti:LiNb0 3 waveguide, a 50 MHz bandwidth SAW transducer, multiple integrated waveguide lens components for forming and imaging a 2.7 cm wide beam, a densely-packed rib waveguide array, low efficiency uniform outcoupling gratings, a surface-mounted mask, and a CCD detector array. This module is capable of processing 264 range res olution elements in a radar return signal that covers a 1 km ground swath and an unlimited number of azimuth resolution elements for continuous operation. Likewise, for the IO correlator we demonstrated that the complex correlation of two 15 MHz input signals could be performed using a Ti:LiNb03 waveguide with a 60 MHz bandwidth transducer, multiple integrated waveguide lenses for forming and imaging a 1 cm wide beam, a 1 cm wide rib waveguide array with 333 elements, highly efficient surface outcoupling gratings for detector coupling, and a surface- mounted high dynamic range linear detector array. These examples serve as the basis for the selection of the waveguide material and waveguide component parameters used in the following chapters. The waveguide material, component parameters, and the criteria used in their selection are summarized next. In the discussion of the design of advanced IO signal processors, we identified a number of waveguide substrate requirements, including the ease of waveguide fabri cation, low propagation losses, the support of efficient SAW transducer devices, and the fabrication compatibility with waveguide components. In consideration of these 76 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. requirements, L iN b0 3 is a suitable substrate material for use in advanced IO signal processor applications. Lithium niobate is a uniaxial crystal (n0 = 2.2869, ne = 2.2019) on which low loss (0.5 dB/cm) waveguides may be formed by the pro cess of titanium indiffusion [Schmidt and Kaminow, 1974; Sugii, etal., 1978; Bums, et al., 1979; Minakata, et al., 1979; Fukuma and Noda, 1980; Holmes and Smyth, 1984]. This waveguide formation process results in an increase in both the extraordi nary and ordinary refractive indices by as much as 0.05 and 0.005 [Armenise, 1988], respectively, and thus are capable of supporting guided modes for any propagation direction and polarization. Titanium-indiffused lithium niobate waveguides were previously fabricated by Rastani for application to the IOSAR processor. These waveguides were multi-mode and had a propagation loss on the order of 1-2 dB/cm [Rastani, 1988]. In the present work, we improved the waveguide fabrication process established by Rastani and were able to produce low-loss (0.41 dB/cm) single mode waveguides. We also obtained more accurate and thorough measurements of the waveguide properties to aid us in further device development. We used well-established techniques to deter mine the waveguide propagation losses, the waveguide mode effective index, and the waveguide mode profile. From these measurements, we were able to determine the diffusion constants related to our substrate material and indiffusion process, as well as the waveguide property variations that result from fabrication process variations. The use o f interdigital transducers for surface acoustic wave modulation of a guided optical wave in Ti:LiNb03 has been developed by a number investigators [Nguyen and Tsai, 1977; Weller, et al., 1977; Tsai, 1979; Xu and Tsai, 1991] and is especially useful for wideband modulation [Tsai, 1979; Campbell, 1989]. The surface acoustic wave velocity is va = 3488 m/s for the crystal orientation used in the research described in this thesis, and is comparable with most other waveguide mate rials [Campbell, 1989; Tsai, 1990]. The crystal orientation used in the present work is referred to as Y Z-LiN b03 (the crystal y-axis is normal to the waveguide surface and 77 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the SAW propagation direction is along the crystal z-axis). This crystal orientation provides the highest diffraction efficiency in LiN b03 [Valette, et al., 1983]. The direction o f propagation of the optical beam is along the crystallographic x-axis so that noncollinear, coplanar interaction with the surface acoustic wave may be utilized. The relatively large extraordinary refractive index change compared to the ordinary refractive index change provides tighter confinement for TE-polarized light (electric field vector along the z-axis). It has also been shown that the optical modulation bandwidth for this polarization is almost twice as large as for TM-poIarized light (polarized along the ordinary axis) in L iN b03 by Casseday, et al. [Casseday, et al., 1983]. Hence, TE-polarized light is considered in the research described herein. The transducer design used in conjunction with the Ti:LiNb03 waveguide described herein consists of a simple design with a 430 MHz center frequency and 100 MHz bandwidth. These values are sufficient to accommodate the key perfor mance characteristics given in the examples for the IOSAR processor and 10 correla tor. In the previous work by Rastani, the fabrication and integration of SAW transducers was described at the IOSAR system design level, but not pursued through the corresponding device development process. The integration of a SAW modulator with large aperture embedded lenses and a rib waveguide array was demonstrated in the present research and described herein for the first time. A critical component for advanced IO signal processors is a high performance lens with a wide aperture, low f-number, and reasonably high throughput. A lens with a 2.7 cm aperture and 10.8 cm focal length (f/4) was considered in the IOSAR proces sor example, and 1 cm lenses with 1 cm focal lengths were considered in the IO corr elator example. For collimation purposes, lenses with low f-numbers are also required. We concentrate our efforts on a single element lens that can potentially lead to such low f-number performance with a high throughput. In particular, we consider the design, fabrication, and characterization of a 1-cm-aperture embedded waveguide 78 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. lens with Si0 2 /MgF2 dielectric waveguide layers, embedded within a Ti:LiNb03 waveguide. Realization of the full potential of these lenses would satisfy the require ments for both of the signal processing tasks described above. A device that is essential for the fabrication of advanced IO signal processors to improve the processor capabilities and provide flexibility in the design is the channel ized waveguide array. In applications that require a large number of channels, these arrays need to be densely packed and yet exhibit low crosstalk. Both of these require ments can be satisfied with rib waveguides in which each o f the elements is separated from the next by an etched air-gap. The air-gap results in low evanescent field pene tration from one rib waveguide to the next, and consequently, result in low crosstalk. Rastani has previously fabricated rib waveguide arrays with up to 660 elements with a center-to-center spacing of 10 fim in a multi-mode Ti:LiNb03 waveguide. Separations between the guides were etched 1 fim deep and 2 fim wide to achieve low interchannel crosstalk. The expected crosstalk for the fabricated rib waveguide arrays was calculated and found to be low. This result agreed with observations of the end- emission o f the fabricated arrays, although the degree of crosstalk was not quantified. On the other hand, the propagation losses of the fabricated rib waveguide array struc tures were measured and found to be high (6 dB/cm). High propagation losses are undesirable in applications such as the IOSAR processor, in which outcoupling from gratings on top of the rib waveguides should be uniform over long distances (1 cm). Herein, we have advanced the state of the rib waveguide array fabrication pro cess. For the IOSAR processor, rib waveguide arrays with up to 1000 elements and 1 cm propagation lengths have been fabricated with 8 fim wide ribs and 2 fim wide gaps. The 10 fim rib pitch is commensurate with the pixel spacing within the two- dimensional CCD detector array mounted above the rib waveguide array. In the case of the IO correlator, rib widths of 28 fan were used with 2 fim separations and 333 elements so that the 1 cm width of the correlator image plane is sufficiently sampled. The minimum gap widths are limited in both of these cases by the minimum feature 79 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. size that can be uniformly patterned by photolithographic means. We also quantify the channel crosstalk for several different rib waveguide array structures in order to determine the relative trade-offs in rib waveguide array design. We have significantly reduced the rib waveguide propagation losses, and thereby improved the outcoupling uniformity that can be achieved with the integration of sur face outcoupling gratings. The surface outcoupling gratings are required to allow additional out-of-plane processing in advanced IO applications. In the case of the IOSAR processor, the grat ings should have a small modulation depth for uniform outcoupling over the 1 cm length of the rib waveguide. In addition, the period should be made small compared with the CCD pixel size to avoid local variations in the CCD illumination that may lead to processing errors. Gratings with a period o f 2 ftm are sufficient for this pur pose and may be photolithographically patterned. In the work by Rastani, surface outcoupling gratings were photolithographically generated with either 2 ftm or 4 fim periods and an interaction length of I mm on the surface of planar waveguides and rib waveguides in Ti:LiNb03 [Rastani and Tanguay, (to be published)]. As described herein, we have fabricated large area gratings (1.2 cm X 1 cm) with periods of either 2 ftm or 4 fim to study the outcoupling unifor mity over a large outcoupling distance (1 cm), as is required for the IOSAR processor. The issues related to the fabrication of the external mask and mounting of the CCD array are not addressed in depth in this thesis, but are considered in Chapter 4 in relation to the outcoupling grating function. Finally, it was mentioned previously that a suitable laser diode for such advanced IO processors is a Hitachi model HLP1600 with a 30 mW power output and 1 MHz - 2 GHz modulating range at a wavelength o f 850 nm. High power (30 mW cw) visible diode lasers such as the SDL-7311 series are now available with single mode emission, a 1 GHz modulation bandwidth, and a wavelength of 670 nm. Since 80 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. the device design, fabrication, and testing is all done at the He-Ne wavelength of 633 nm, the visible laser diodes could be directly implemented without significant changes in the processor performance. Next, in Chapter 3, we discuss in detail the selection o f the waveguide materials, the waveguide fabrication processes, and the properties o f the fabricated waveguides that pertain to device integration. 2.7 References D. B. Anderson, “Integrated Optical Spectrum Analyzer: An Imminent ‘Chip’,” DEEE Spectrum, Dec., 22-29, (1978). M. N. Armenise, “Fabrication Techniques of Lithium Niobate Waveguides,” IEE Proceedings-J, 135(2), 85-90, (1988). M. K. Bamoski, B. U. Chen, T. R. Joseph, Y. M. Lee, and O. G. Ramar, “Integrated- Optic Spectrum Analyzer,” IEEE Trans. Circuits Syst., Cas-26(12), 1113, (1979). W. K. Bums, P. H. Klein, E. J. West, and L. E. Plew, “Ti Diffusion in Ti:LiNb03 Planar and Channel Optical Waveguides,” J. Appl. Phys., 50(10), 6175-6182, (1979). C. Campbell, Surface Acoustic Wave Devices and Their Signal Processing Applications, (Academic Press, Inc., Boston, 1989). M. W. Casseday, N. J. Berg, and I. J. Abromovitz, “Space-Integrating Acousto-Optic Signal Processors Using Surface-Acoustic Wave Delay Lines,” in Acousto-Optic Signal Processing , N. J. Berg and J. N. Lee, Eds., 165-202 (Marcel Dekker, Inc., New York, 1983). 81 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. B. U. Chen, E. Marom, and A. Lee, “Geodesic Lenses in Single-Mode LiNb03 Waveguides,” Appl. Phys. Lett., 31(4), 263-265, (1977). J. D. Cohen, “Incoherent-Light Time-Integrating Processors,” in Acousto-Optical Signal Processing , N. J. Berg and J. N. Lee, Eds., 225-288 (Marcel Dekker, Inc., New York, 1983). K. Creath, “Interferometric Investigation of a Diode Laser Source,” Appl. Opt., 24(9), 1291-1293, (1985). M. Fukuma and J. Noda, “Optical Properties of Titanium-Diffused LiN b03 Strip Waveguides and Their Coupling-to-a-Fiber Characteristics,” Appl. Opt., 19, 591-597, (1980). M. W. Haney, “Acousto-Optical Time-and-Space Integrating Processors for Real- Time Synthetic Aperture Radar Imaging,” Ph.D. Thesis, California Institute of Technology, (1986). R. J. Holmes and D. M. Smyth, ‘Titanium Indiffusion into LiNb03 as a Function of Stoichiometry,” J. Appl. Phys, 55(10), 3531-3535, (1984). J. J. Kovaly, Synthetic Aperture Radar, (Artech House, Inc., MA, 1976). K. Y. Liao, C. C. Lee, and C. S. Tsai, ‘Time-Integrating Correlator Using Guided- Wave Anisotropic Acousto-Optic Bragg Diffraction and Hybrid Integration,” in the 1982 Topical Meeting on Integrated and Guided-Wave Optics, Technical Digest WA4-1 to 4, IEEE Cat. No. 82CH 1719-4, (Pacific Grove, Ca., 1982). D. Mergerian, E. C. Malarkey, and R. P. Pautienus, “High Dynamic Range Integrated Optical RF Spectrum Analyzer,” in the 4th International Conference on Integrated Optics and Optical Fiber Communication, Paper 30B3-b, (Tokyo, Japan, 1983). 82 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. M. Minakata, S. Saito, and M. Shibata, ‘Two-Dimensional Distribution o f Refractive- Index Changes in Ti-Diffused LiNb03 Strip Waveguides,” J. Appl. Phys., 50(5), 3063-3067, (1979). L. T. Nguyen and C. S. Tsai, “Efficient Wideband Guided-Wave Acoustooptic Bragg Diffraction Using Phased Surface Acoustic Wave Array in LiN b03 Waveguides,” Applied Optics, 16(5), 1297-1304, (1977). D. Psaltis, M. Haney, and K. Wagner, “Real Time Synthetic Aperture Radar Processing,” in the NASA Conference on Optical Information Processing for Aerospace Applications. II, (NASA, Langley, Virginia, 1983). D. Psaltis and K. Wagner, “Real-Time Optical Synthetic Aperture Radar (SAR) Processor,” Opt. Eng., 21(5), 822-828, (1982). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). K. Rastani and A. R. Tanguay, Jr., “Large Aperture Negative Meniscus Singlet and Triplet Lenses with Positive Focal Lengths Developed on LiNb03,” (to be published). N. A. Riza, “In-Line Interferometric Time-Integrating Acousto-Optic Correlator,” Appl. Opt., 33(14), 3060-3069, (1994). R. V. Schmidt and I. P. Kaminow, “Metal-Diffused Optical Waveguides in LiN b03,” Appl. Phys. Lett., 25(8), 458-460, (1974). K. Sugii, M. Fukuma, and H. Iwasaki, “A Study on Titanium Diffusion into L iN b03 Waveguides by Electron Probe Analysis and X-Ray Diffraction Methods,” J. Mat. Sci., 13, 523-533, (1978). 83 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. T. Suhara, T. Shiono, H. Nishihara, and J. Koyama, “An Integrated-Optic Fourier Processor Using an Acoustooptic Deflector and Fresnel Lenses in an As2S3 Waveguide,” IEEE J. Lightwave Tech., LT-1(4), 624, (1983). C. S. Tsai, “Guided-Wave Acoustooptic Bragg Modulators for Wide-Band Integrated Optic Communications and Signal Processing,” IEEE Transactions on Circuits and Systems, CAS-26(12), 1072-1098, (1979). C. S. Tsai, Ed., Guided-Wave Acousto-Optics, Springer Series in Electronics and Photonics, D. H. Auston, etal., Eds., Vol. 23, (Springer-Verlag, New York, 1990). C. S. Tsai, D. Y. Zang, and P. Le, “High-Packing-Density Multichannel Integrated- Optic Modules in LiN b03 for a Programmable Correlation of Binary Sequences,” Opt. Lett., 14(16), 889-891, (1989). S. Valette, J. Lizet, P. Mottier, J. P. Jadot, S. Renard, A. Fournier, A. M. Grouillet, P. Gidon, and H. Denis, “Integrated Optical Spectrum Analyser Using Planar Technology on Oxidised Silicon Substrate,” Electron. Lett., 19(21), 883-885, (1983). C. M. Verber, R. P. Kenan, and J. R. Busch, “Correlator Based on an Integrated Optical Spatial Light Modulator,” Appl. Opt., 20(9), 1626-1629, (1981). J. F. Weller, J. D. Crowley, and T. G. Giallorenzi, “Surface Acoustic Waveguides on LiNb03 Formed by Titanium In-Diffusion,” Appl. Phys. Lett., 31(3), 146-148, (1977). G. D. Xu and C. S. Tsai, “Novel Integrated Acousto-Optic and Electro-Optic Heterodyning Device in a LiNb03 Waveguide,” Appl. Phys. Lett., 58(1), 28-30, (1991). 84 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 3 Waveguide Design and Fabrication The selection of appropriate substrate materials and waveguide fabrication tech niques for advanced IO signal processors is crucial in that it determines which device technologies may be used, and also implies processor performance limitations. The central focus o f this chapter is on our efforts to develop a titanium-indiffused lithium niobate (Ti:LiN b03) waveguide for use in the fabrication of the waveguide compo nents described in Chapters 4, 5, and 6 , and the integration of these components described in Chapter 7. Titanium-indiffused lithium niobate waveguides were previously fabricated by Rastani for use in the IOSAR processor application. These waveguides were multi- mode and had a propagation loss on the order of 1-2 dB/cm [Rastani, 1988]. Others have found that single-mode Ti:LiNb03 waveguides can be fabricated with low prop agation losses (0.5 dB/cm) [Nishihara, etal., 1989]. In the present work, we modified the process parameters established by Rastani to produce low-loss (0.41 dB/cm) sin gle-mode waveguides. To aid us in further device development we determined the waveguide propagation losses, waveguide mode effective index, and waveguide mode profile. W ith the results of these measurements, we were able to characterize the Ti indiffusion process. We calculated the diffusion constants as well as the waveguide property variations with changes in the process parameters. These param eters are useful for further development of the waveguide process and waveguide devices. We have undergone a parallel effort to develop waveguide devices in gal lium arsenide (GaAs) waveguides, and therefore will describe the fabrication of a 85 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide structure in this material system as well. The waveguide requirements imposed by the system design considerations of the IOSAR processor and IO correlator were summarized in Chapter 2, and are con sidered in more detail in Sects. 3.1 and 3.2 to illustrate the relative advantages and disadvantages for each o f these materials. The issues that relate the waveguide design to the waveguide fabrication process, the details of the fabrication sequence, and the results of the waveguide evaluation are then discussed. This discussion is followed by a summary of the sample preparation techniques and source coupling methods that were used in the fabrication and evaluation of the devices described herein. This chapter ends with a summary o f our waveguide fabrication results and the manner in which these results are used in the device fabrication. 3.1 Lithium Niobate Lithium niobate is a synthetic ferroelectric material, which in single-crystal form exhibits large pyroelectric, piezoelectric, electrooptic, and photoelastic coeffi cients, as well as a strong bulk photovoltaic effect. As a result, it has found use in widespread device applications including acoustic wave transducers, delay lines, fil ters, optical amplitude and phase modulators, second harmonic generators, Q- switches, beam deflectors, phase conjugators, dielectric waveguides, memory ele ments, and holographic data processors [Weis and Gaylord, 1985]. For these device applications, the relative ease o f fabricating high quality optical waveguides, and the capability for wideband, efficient SAW transducers on LiNb0 3 compared with other materials, makes LiN b03 attractive for use in advanced IO signal processor applica tions. A number of hybridization techniques have been developed for LiNb03 that permit the integration o f laser sources, detectors, and various other components needed for the IO signal processor operation. Waveguides may be formed in lithium niobate by one of several diffusion pro- 86 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. cesses including titanium indiffusion, proton exchange, and lithium oxide (U 2O) out- diffusion. Titanium indiffusion is a thermally activated diffusion process that dopes the surface of the crystal with titanium, and results in an increase of refractive index that may extend from one micron to tens of microns deep from the crystal surface [Armenise, 1988]. In this region of increased index, low-loss waveguide propagation (<1 dB/cm) may take place, as well as electrooptical or acoustooptical interactions. Since LiNb0 3 is a uniaxial crystal (nG = 2.2869, ne = 2.2019 at A 0 = 6328 A, accord ing to the Crystal Technology specification sheets for the materials used for the research presented herein), the waveguide properties are polarization dependent. Titanium indiffusion leads to an increase in both the ordinary and extraordinary indi ces of refraction (Ane < 0.04 and Ana < 0.02) of the crystal; therefore, waveguide propagation may take place regardless of the relative orientation of the crystal axis and the polarization vector. This waveguide property is important for the operation of wide aperture, low f-number lenses since large off-axis propagation angles are involved. While out-diffusion o f lithium oxide is a simple process that forms low-loss waveguides, the out-diffusion depth is usually tens to hundreds of microns with a comparable lateral out-diffusion rate, which makes it impractical for the formation of densely-packed channel waveguides on the surface [Armenise, 1988]. In addition, only the extraordinary refractive index is increased, thereby permitting guided modes only for electric field polarizations with a large component along one axis of the crys tal [Carruthers, et al., 1974]. Proton exchanged waveguides can be made very low- loss (0.5 dB/cm) as well as very shallow. However, only the extraordinary refractive index is increased by this process as well [De Micheli, et al., 1982; Jackel, et al., 1982; Jackel, etal., 1983]. In the work described herein, we produced shallow (less than 3 fim deep), low- loss (0.41 dB/cm) single-mode Ti:LiNb03 waveguides by the titanium-indiffusion process. This tight confinement satisfies the requirements put forth in Chapter 2 87 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. regarding efficient guided-mode acoustooptic interaction, rib waveguide and embed ded lens fabrication depth limitations, and efficient hybrid laser diode coupling. Fur thermore, single-mode waveguide operation prevents intermode scattering or conversion in the different devices that may lead to signal error. The principal extraordinary refractive index ne applies to electric-field polariza tions along the z-axis and the principal ordinary refractive index n0 applies to electric- field polarizations in the xy plane. Waveguide substrates are available in thin wafer form with both several different crystal orientations and optical-quality polished sur faces. The crystal orientation chosen for use in the research described herein is illus trated in Fig. 3.1. In this orientation, the y-axis of the crystal is perpendicular to the wafer plane (referred to as y-cut) so that an x-propagating TE-polarized guided-wave has its polarization vector along the z-axis. The TE-polarized light is more tightly confined than TM-polarized light since the electric field vector is aligned with the crystallographic axis with the largest refractive index change. Another benefit of a y-cut crystal is that a z-propagating surface acoustic wave may be accessed. Surface acoustic waveguide propagation along this axis features a large acoustooptic coupling coefficient and a self-focus effect that greatly increases the SAW collimation length {Lc > 104Aa ) [Szabo and Slobodnik, 1973b; Szabo and Slobodnik, 1973a]. This increased collimation length leads to efficient modulation with uniformity over a large time aperture. The obtainable SAW modulation band width is approximately twice as large for extraordinary polarized light than for ordi nary polarizations [Casseday, et al., 1983]. We have found that y-cut wafers are robust when exposed to elevated temperature processes compared with z-cut wafers. Large z-cut samples are susceptible to fracturing if heated or cooled to quickly due to surface charging caused by the pyroelectric effect. 88 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. < + x> Direction of propagation <-Z> TE Figure 3.1 Schematic diagram of titanium-indiffused lithium niobate waveguide showing the orientation and polarization of the guided mode. 3.2 Gallium Arsenide Integrated optical processors fabricated on the direct band-gap semiconductor material gallium arsenide, on the other hand, may potentially consist of monolithi- cally integrated optical sources, passive waveguide components, photodetectors, and electronics [Hunsperger, 1982; Hutcheson, 1987]. This monolithic integrability is beneficial from the standpoint of manufacturability since the fabrication technology is based on one material system. In addition, power losses associated with hybrid cou pling can be avoided. Waveguides are often fabricated on GaAs substrates by the epitaxial growth of alloyed layers. A common material used for this purpose is aluminum in the forma tion of A ltGa/_xAs epitaxial layers, with the fractional content of aluminum specified by x. With increased aluminum content, the refractive index of the layer decreases from n = 3.48 for pure GaAs to n = 3.32 for an aluminum content x = 0.3 (measured at X = 1.06 /an). Thus, a waveguide is formed by the growth of a low index AiGaAs buffer layer on a GaAs substrate followed by a high index GaAs waveguide layer (GaAs/AlGaAs waveguide). The acoustooptic properties of GaAs are comparable to those of LiNb0 3 ; how ever, its small piezoelectric properties make it relatively difficult to couple power into an acoustic mode. Typically, a ZnO thin film is introduced between the transducer 89 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. and GaAs to enhance this coupling [Tsai, 1990]. The development of the key integrated optical components in GaAs waveguides may eventually lead to very powerful and efficient signal processors. In particular, the ability to very accurately define waveguide structures by epitaxial growth tech niques such as molecular beam epitaxy (MBE) will improve the design reliability o f the rib waveguide and lens components. Gallium arsenide is also isotropic, has a very high index of refraction, and lends itself well to reactive ion etching methods (deep etching with smooth vertical sidewalls) [Sonek, et a l, 1985; Vawter, et al., 1987; Behfar-Rad, et a l, 1989a; Behfar-Rad, et a l, 1989b]. All of these factors are favor able for the fabrication and operation of wide aperture, low f-number lenses. For waveguides formed on the < 100> surface, component testing is simplified due to the availability of the (110) cleavage planes for end-fire source coupling. Finally, mono lithic integration o f sources and detectors will alleviate hybridization difficulties, increase coupling efficiencies, and reduce the amount of background scattering. Although the level of component integration demonstrated by others has advanced significantly [Liou, et a l, 1989; Vawter, et a l, 1989; Hirata, et al., 1993], this development is still in its early stages. Much research remains to work out the structural details of epitaxial growth for the integration of devices with widely vary ing and sometimes conflicting requirements such as sources, waveguides, detectors, electronic components. 3.3 Waveguide M odeling Many planar waveguide structures need to be analyzed in the design of the pro cessor waveguide and the waveguide components. In particular, the Ti:LiNb0 3 waveguides have a gradient refractive index profile, the GaAs/AlGaAs waveguides can be treated as 3-layer structures, and the embedded lens is a 4-layer structure that may be considered a leaky waveguide. While exact and approximate analytical solu- 90 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tions to the wave equation exist for some specific cases, a generalized numerical tech nique has been adopted for the analysis of waveguide structures with arbitrary index profiles [Ghatak, et al., 1987; Ghatak, etal., 1988; Ramadas, eta l., 1989; Ghatak, et al., 1990; Tomer, et al., 1991]. This technique is based on the definition of a field transfer matrix for each layer of different refractive index in the structure, with graded index waveguides modeled with a stepwise approximation to its continuously varying refractive index. This numerical approach is used later in Sect. 3.5 to model the elec tric field mode profile of graded index Ti:LiNb0 3 waveguides, for use in comparison with a measured mode profile. In Chapter 5, an analysis of the 4-layer embedded lens structure utilizes this technique to determine the required lens barrier layer thickness. In general, an arbitrary two-dimensional waveguide structure may be analyzed by this method for a particular polarization and wavelength in order to determine the normal ized complex electric field profile of each waveguide mode and its corresponding complex propagation constant. 3.3.1 Graded Index W aveguides Due to the nature of the titanium-indiffusion process, Ti:LiN b03 waveguides have a refractive index profile that gradually transitions from a high surface value to slightly lower substrate value without an abrupt discontinuity. Since there is no clear boundary between these regions, the depth of the guided modes cannot be determined at first glance, but rather must be derived from an analysis of the waveguide mode structure that incorporates the graded index. Next, we review the relationship between the waveguide fabrication process and the formation of the graded index pro file, and clarify the relationship between the process control parameters and the resulting waveguide mode properties in Sect. 3.3.2. Diffusion of titanium into lithium niobate is accomplished by first vacuum evap orating or sputtering a thin layer ( 10-100 nm) of titanium onto the crystal surface and then heating the crystal to temperatures ranging between 900-1150° C in an argon, 91 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. nitrogen, oxygen, or air ambient for 0.5 to 30 hours. As the diffusion temperature is increased, the surface titanium layer undergoes several changes. It first begins to oxi dize at 500° C and forms polycrystalline TiC> 2 [Hunsperger, 1982]. As the tempera ture increases between 700° C and 950° C, it is believed that large crystallites of the (Tio 65Nbo.35)02 phase form on the surface. At 1000° C, this surface layer acts as the source of the indiffusion [Armenise, etal., 1983]. Alternatively, it has been proposed that Li within the surface layer participates in the formation of a stable (Lio.25Nty).75C>2)o.42(Ti02)o.58> which then acts as the diffusion source [Rice and Holmes, 1986]. According to Sugii, et al., Ti diffuses substitutionally into lithium niobate, occupies Nb sites, and is +4 valence [Sugii, et al., 1978]. For long diffusion times compared with the time it takes for all of the metal to enter the crystal, the titanium concentration profile approaches a Gaussian function distribution [Schmidt and Kaminow, 1974] in which y is the depth below the surface, D is the diffusion depth, and O L a is the num- film of thickness tji- This functional dependence has been confirmed by microprobe and X-ray microanalysis studies [Sugii, etal., 1978; Minakata, etal., 1979]. The dif fusion depth is given by in which t is the diffusion time and the temperature-dependent diffusion constant D c is given by (3.1) ber of atoms per unit volume (of order 5.65 X 1022 atoms/cm3) within the deposited D 2 = 4D ct , (3.2) 92 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (3.3) The parameters D 0 and T0 are the material-dependent diffusion constant and activa tion temperature, respectively, and T is the diffusion temperature. Much work has been done by others to relate the titanium concentration profile to the index change profile for both ne and n0 refractive indices. It was found that ne in y-cut samples is related to the titanium-to-niobium ratio, and is sensitive to the oxi dation state of the crystal [Griffiths and Esdaile, 1984]. For y-cut samples, a linear variation of the extraordinary refractive index change with titanium concentration is observed [Minakata, et al., 1978]. Accordingly, the refractive index profile is assumed to vary in depth with the same functionality as the titanium concentration and is normally expressed as [Griffiths and Esdaile, 1984] in which ns is the substrate refractive index and An is the refractive index change at the crystal surface. This relation is assumed to hold for the waveguides fabricated in the present work and described herein. 3.3.2 W aveguide Fabrication Sensitivities A property inherent to device fabrication is that of process variation. Process variations lead to corresponding variations in the device characteristics, which in turn must be within some acceptable tolerance level defined for the application. In Chapter 5, we estimate the sensitivity of the embedded lens focal length to waveguide process parameters for both the host waveguide and the embedded waveguide. The sensitivity of the single-mode Ti:LiNb0 3 waveguide to variations in the fabrication R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. n(y) = ns + A n exp (3.4) 93 process was calculated and is presented next. The primary characteristic of the waveguide considered in these calculations was the effective refractive index ne j^ since the properties of the embedded lens depends directly on this waveguide param eter. The parameters involved in the formation of a waveguide by the titanium-indif- fusion process are given in Eqs. 3.1 through 3.4. The process parameters with the greatest variability are the initial titanium layer thickness and the diffusion tempera ture. Herein, we assumed that variations in the thin film density, diffusion time, and diffusion constant are small in comparison. The diffusion constants may vary from one crystal wafer to another; however, we assume that in production the diffusion constants for each wafer are characterized and appropriate adjustments are made in the waveguide processing sequence to compensate for the variations. The result of the diffusion process is the modification of the refractive index of the material near the waveguide surface as described by Eq. 3.4. In this equation, both the refractive index change An and the diffusion depth D vary as a result of the process variations. This two-parameter variation was taken into consideration in the calculation of the waveguide mode properties. In order to determine the correspond ing changes in the effective refractive index of the waveguide, the normalized eigen value equation for a waveguide with a Gaussian index distribution was considered. The generalized solution to this equation is graphically expressed in the form of dis persion curves that relate the normalized waveguide index b to the normalized waveguide depth V as shown in Fig. 3.2 [Nishihara, et a l , 1989]. To a good approxi mation, the relationship between these normalized parameters in a single-mode waveguide (2 < V < 4) is considered linear. The relationship between the waveguide effective refractive index n e f f and b is given in the expression for the normalized index 94 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 0.6 m = 0 0.4 m = 1 m = 2 0.2 m = 3 o .o 2 4 6 8 10 12 14 Figure 3.2 Normalized dispersion curves for a waveguide with a Gaussian refrac tive index profile. b K/-V), ( « / - n s2) (3.5) in which «y= ns + An is the refractive index at the crystal surface. The refractive index nycan be easily related to the physical parameters tty and d through the expres sion for the normalized depth V- 2 n d ■ f. (3.6) With the assumption that the index change is small compared with the substrate refractive index (i.e., «y= ns), we derived an expression for the variation in effective refractive index as it is related to the diffusion temperature variation, 95 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. and the effective refractive index as it is related to the titanium film thickness varia tions, dn 7 nf An dtTi — = ™ - ^ T 3 V - 4 . (3.8) n 80 neffA tTi Substituting the values for these parameters used for the present work C ns = 2.2019, neff= 2.2069, An = 0.015, V = 3 ,T 0 = 3 x 104 K, and T = 1273 K) into Eqs. 3.7 and 3.8, we found that for a ±10% variation in the titanium film thickness typically observed in our process, there is a ±0.05% variation in the waveguide effec tive refractive index {ne f f ~ 2.2069 ± 0.001). For the ±10 K variation in the indiffu sion furnace temperature, which is approximately the specification for our furnace, there is a ±0.015% variation in the effective refractive index 2.2069 ± 0.0003). As can be seen, the variation due to the titanium thickness fluctuations is the larger of the two effects in our process. 3.4 Waveguide Fabrication Processes The titanium-indiffusion process that was used to produce low-loss, single mode waveguides in lithium niobate is discussed next. The waveguide substrate han dling procedures described here are used for sample preparation in the fabrication of the waveguide devices discussed in the later chapters. A summary of the GaAs/AlGaAs waveguide fabrication process and material structure is also given. 3.4.1 Titanium -Indiffused Lithium Niobate W aveguides Optical grade LiNbC> 3 wafers were supplied by Crystal Technology, Inc. in Palo 96 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Alto, California. These wafers are congruent to ±0.005 mole% and contain less than 2 ppm transition metals, including iron. Lithium niobate is uniaxial with refractive indices of na = 2.2869 ± 0.002 and n0-n e - 0.085 ± 0.001 (at A = 6328 A and 20 °C) as noted by the supplier. The supplier uses the notation <+Y> and <-Y> to specify the plus and minus crystal planes whose surface normal is parallel to the crystallo- graphic y-axis, respectively. Similar notation for the other crystal planes as well. The wafers used in this research are cut such that the <-Y > crystal surface is the top sur face of the wafer. The flatness of the <-Y> surface is a maximum of 7 fim T.I.R. and has a 10/5 scratch dig finish while the back surface has a 60/30 scratch/dig finish. Wafers were supplied with a 3” diameter and a 0.040” thickness. Prior to waveguide formation on these substrates, each wafer was usually sliced into three 1” wide strips with the cuts parallel to the jr-crystal axis as shown in Fig. 3.3. The cuts were made with a diamond-embedded titanium wire on an auto matic wire saw. The wafer surfaces were protected from damage during this process by application of a low-tack wafer dicing tape to the front and back surfaces. In addi tion to scratch protection, the tape seals the wafer from the wire-saw residue and sim plifies the wafer cleaning procedure before the waveguide formation process. After the wafer was diced, the tape was removed and the <+Y> surface of each substrate slice was lightly scribed with a short line parallel to the z-axis to identify the crystal orientation. Later on during the device fabrication, each time the samples were diced into smaller pieces, the unmarked portions o f a substrate were likewise scribed in this manner to indicate the direction of the crystallographic axis. The substrates were then carefully cleaned in preparation for the titanium depo sition. The cleaning procedure followed is similar to that given previously by Rastani [Rastani, 1988]. The samples were rinsed in 18 M Q-cm de-ionized (DI) water and diluted Micro-cleaning solution (purchased from Cole Palmer, Chicago, 111.), hence forth referred to as dilute soapy water, to remove the wire-saw residue. A dampened cotton swab was used to gently remove residue from the surface of the substrates that 97 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. +x Figure 3.3 Schematic diagram of wafer dicing patterns for waveguide fabrication. wouldn’t readily rinse free, and residue was removed from the cut edges by rubbing the edge with a damp, soapy Texwipe cloth. The samples were then blown dry with nitrogen gas and visually inspected to locate any large particles or residue of tape adhesive. These contaminants were removed by gently cleaning the area with a cot ton swab soaked in acetone. When all of the wafers appeared clear of gross contami nation, they were immersed in 1,1,1 trichloroethane (40° C) for 10 minutes followed by immersion in DI water, acetone, and methanol for 5 minutes each in the presence of ultrasonic agitation. This solvent cleaning is done in order to remove the cleaning solution residue, grease, and other organics. An additional cleaning process was used for the preparation of samples for the waveguide indiffusion process. The samples were rinsed in DI water and then immersed in a hydrochloric acid-hydrogen peroxide-DI water (1:1:5) solution (at 75° C) for 20 minutes in order to remove metallic ion contaminants from the surface of the substrates that can lead to waveguide scattering if left on the wafer. A teflon wafer carrier was used to protect the substrates from damage due to vibration in the boiling liquid. After this cleaning treatment, the substrates were rinsed again in DI water, blown dry with nitrogen gas, and baked in a convection oven at 150° C for 30 minutes. 98 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The titanium depositions were performed in a Balzers BAK-640 boxcoater with a multi-pocket hearth for electron beam evaporation. The distance from the hearth to the sample planetary plates in this system is approximately 30 cm. In order to keep the titanium film contamination level as low as possible, the system was specially cleaned prior to titanium deposition. All of the interior liners were exchanged with a set of liners that had been blasted clean with glass beads and degreased. All traces of deposits were removed from the unshielded surfaces of the vacuum chamber, and the chamber heaters, hearth, planetary, crystal monitor holder, and viewport windows were cleaned. The system was then brought to high vacuum and degassed with the heaters set at 300° C for 2 hours. After the chamber returned to room temperature, 99.99% pure titanium pellets were loaded into a molybdenum liner that was inserted in one of the hearth pockets. The pellets were pre-melted with the source shutter closed. Titanium was then deposited onto photolithographically-pattemed witness samples for deposition rate calibration of the crystal thickness monitor. With the chamber cleaned, the titanium source material pre-melted, and the dep osition rate calibrated, we proceeded with the titanium deposition onto the waveguide substrates. The fully-cleaned substrates were mounted on the Balzers planetary plates with the <-Y> surface facing the source material. After the chamber pressure was in o the low 10 Torr range, titanium was deposited onto the mounted substrates to a thickness of 250 A using planetary rotation to achieve thickness uniformity. During evaporation, the chamber pressure was typically in the low 10' 7 Torr range, the elec- tron-gun voltage was 6 kV, and the deposition rate was approximately 2 A/sec. No chamber heating or background gasses were used during the titanium deposition. The titanium-coated substrates were then loaded onto a quartz boat and inserted into a furnace with an inner quartz tube. The furnace is a three-foot-long Lindberg Model 5744-A-ET-PLS three-zone furnace with programmable Eurotherm Model 820 controllers. Only one wafer (consisting of two or three sliced sections) was indiffused at a time, with the wafer sections placed in the region of greatest temperature unifor- 99 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. mity at the center of the furnace. During the indiffusion, wet oxygen was flowed through the quartz furnace tube. The use of wet O2 as the Ti-indiffusion ambient atmosphere has been shown by others to reduce the outdiffusion o f Li20 that can complicate the waveguide structure [Jackel, etal., 1984; Canali, etal., 1988], and additionally leads to an observed reduc tion in photoreffactive sensitivity [Esdaile, 1978]. The oxygen was supplied from a high purity gas cylinder and bubbled through a column of de-ionized water that was heated to 45° C. The resulting oxygen flow of 1.5 liters/minute was moisture-satu rated (nearly 100% relative humidity). The furnace was ramped from room temperature to 1000° C in 2 hour. After the furnace reached 400° C, the wet oxygen flow was turned on. The titanium was indif fused for 2 hours at 1000° C, and then the furnace was turned off and allowed to cool naturally back down to room temperature. After the temperature dropped below 600° C, the wet oxygen flow was turned off. Once the furnace reached 100° C or less, the wafer slices were drawn from the oven. They appeared transparent with no visible surface defects. 3.4.2 Gallium Arsenide W aveguides Semiconductor waveguides were prepared using molecular beam epitaxy (MBE) of GaAs/AlGaAs layers on semi-insulating GaAs substrates by K. C. Rajku- mar. The waveguide structure consists of a <100> cut semi-insulating substrate with a 2 pan. epitaxial layer of Al0^G aojA s (n = 3.32) to act as a barrier layer for a 1 /an thick GaAs (n = 3.48) guiding layer. This structure supports two modes for each polarization. Compositional variations of ± 6 % are inherent to the epitaxial growth process, which leads to a refractive index variation in the AlGaAs layer of ±0.45%. Consequently, this leads to an effective refractive index variation of ±0.02% for the lowest order TE mode of the waveguide structure {ne ff= 3.4527 ± 0.0004). These 100 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguides were used in the fabrication and characterization o f rib waveguides with surface outcoupling gratings as described in Chapter 4. 3.5 W aveguide Characterization Our primary concern in the fabrication of Ti:LiNb0 3 waveguides was to estab lish a process that would produce low waveguide losses and single-mode behavior. The characterization methods described in this section are designed to measure the waveguide propagation constants, the intensity profile of the guided modes, and the waveguide attenuation constants. These parameters were used to aid in the design and evaluation of the waveguide components and integrated modules discussed in the later chapters. 3.5.1 Mode Structure The effective refractive index of a waveguide mode is an important parameter in the design of many waveguide devices such as refractive elements (e.g., lenses, prisms) and diffractive elements (e.g., surface acoustic wave modulators, outcoupling gratings). A very accurate method to determine the number of modes supported by a waveguide and the effective refractive indices of these modes is the prism coupling method [Nishihara, et al., 1989]. By this method, individual waveguide modes are selectively excited and their propagation constants calculated in direct relation to the prism incoupling angle. Under well-controlled conditions, it is possible to determine these angles with an accuracy o f 20 arc sec = IO-4 rad. Subsequently, the effective refractive indices can be determined to an accuracy of 10" 4 [Nishihara, et al., 1989], The prism coupling technique is described further in Sect. 3.6.1. After fabrication of the Ti:LiNbC>3 waveguides, the prism coupling method was used to determine the waveguide mode structure and effective refractive indices. A rutile prism with refractive indices nQ = 2.584 and ne = 2.871 and a prism angle of 101 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48.87° was used to prism incouple light from a 15 mW He-Ne laser (^=6328 A). A second prism was used to outcouple the waveguide light at the opposite end of the sample. The presence of a guided mode was observed by the appearance o f a dim streak along the waveguide surface (due to surface scattering) and by the presence of a beam outcoupled from the second prism. All of the Ti:LiNb03 waveguides that were fabricated support one TE-polarized mode and no TM-polarized modes. This mode structure is advantageous in that waveguide scattering or device scattering can not lead to mode conversion in the waveguide. From sample to sample, less than 4% variation in the coupling angle was observed and the calculated effective refractive index is 2.2069 ±0.001 on average. This effective refractive index variation is approximately what we expect as a result of the titanium layer thickness variations. In order to determine the waveguide anisotropy present in these samples, TE- polarized light was prism coupled into the waveguide so that it would propagate along the z-axis. The measured effective refractive index along this direction for a single sample was 2.2877 ± 0.0001. The index ellipsoid for the TE-polarized mode can be expressed by i sin2 e n 2(0) (2.2877)2 (2.2069)2 ’ in which 6 is the propagation angle with respect to the x-axis. The effects of this anisotropy on the performance of the embedded lens structure are considered further in Chapter 5. 3.5.2 G uided M ode Intensity Profile Aside from the effective refractive indices of a waveguide, the design o f many waveguide devices depends on the electric field distribution of the guided modes. The electric field profile is used in overlap integrals to determine the laser diode butt- coupling efficiency, the surface acoustic wave diffraction efficiency, the waveguide 102 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. lens throughput, and the rib or channel waveguide incoupling efficiency. A direct measurement of the guided-mode intensity distribution is achieved by placing a scanning knife-edge detector in close proximity to a polished waveguide edge to profile the end-emission. A SpotScan profiler supplied by Photon, Inc. is capable o f a resolution on the order of 20 nm for intensity distributions from 0.2 fim to 30 //m wide. The SpotScan head consists of a 1 inch wide chrome-coated glass plate with several small (-75 fim wide) windows opened through the chrome. The glass plate is rapidly translated back and forth in front of a detector. The unit is then placed in the path of a beam of light such that during half of the cycle the radiation is incident upon one of the windows and reaches the detector, and during the other half the radiation is blocked by the chrome coating from reaching the detector. Several windows are provided as a matter of convenience for alignment and to provide addi tional openings in case there is surface contamination or a scratch on one of the win dows. Any one of the windows can be used to perform the measurement although care should be taken not to illuminate more than one window. The time derivative of the detector response in the transition between these two states is proportional to the spatial intensity distribution of the electromagnetic field. The difficulty associated with measurement of the near-field profile of the waveguide by this approach is that the wide glass plate head must be placed within close proximity to the waveguide edge. Placing the head in close proximity becomes problematic if the width of the waveguide is more than a few millimeters. For example, if the intensity profile of a single-mode waveguide with a 3 fim width is to be measured by this method, the knife edge must be place within 25 fim of the edge to be within the near field of the end-emission [Hunsperger, 1982]. If the waveguide is 1 inch wide, the angular mis alignment of the head and the waveguide should be less than 3 arc minutes. This rota tional accuracy is difficult to achieve with the SpotScan device and translation mount. A mode profile measurement was performed on a waveguide fabricated as detailed in Sect. 3.4.1 that was cut to a width of 3 mm. In this experiment, light from 103 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. a He-Ne laser (Aq = 6328 A) was prism coupled into the waveguide. In this case, the angular deviation of the waveguide and head need only be less than 30 arc minutes (0.5°) to place the scan head within the near-field of the waveguide end-emission. The rotational resolution of the SpotScan stage was better than this (of order 0.1°). We can conservatively estimate that the head was actually positioned within 10 fim of the waveguide edge (1.5 mm X tan(0.1°) = 3 fim). The result of the SpotScan mea surement is shown in Fig. 3.4. We are uncertain of the position of the air/Ti:LiNb03 boundary. The error bar on the graph corresponds to our guess of the region where the waveguide surface is likely located. The depth of the waveguide mode estimated from this measurement is 3 fim between the 1/e2 points in intensity. Since the scan ning head can’t be placed in direct contact with the waveguide edge, this measured width is slightly larger than the actual near-field profile. The measurement is also subject to multiple reflections between the polished waveguide edge and the reflective surface o f the chrome coating that can lead to error in the measured profile. A second approach that employed a microscope objective to produce a magni fied image of the waveguide end-emission on a CCD detector array permitted a more accurate measurement of the waveguide mode intensity profile [Helms, et al., 1990]. In this approach, the resolution is limited by the numerical aperture of the objective. As a result, very shallow waveguide structures cannot be accurately imaged. The rel ative advantage of this approach is that the objective can be easily positioned at the objective’s working distance away from the waveguide edge, in contrast to the previ ous experiment. This measurement was performed with a 40x microscope objective (NA = 0.65) and a CCD detector array placed 12 cm from the objective. The estimated resolution of this approach (0.5 ^m) was found sufficient to obtain a clear image of the mode intensity profile. The CCD detector response to light from a He-Ne laser (Aq = 6328 A) was calibrated. The measured mode profile for the 3-mm-wide edge- polished waveguide used in the direct measurement approach is shown in Fig. 3.5. 104 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 - Air LiNbO. c 0.6 ® T 3 O 2 HH ■ O ® N 15 0.4 1 o 2 0.2 0.0 0 1 2 3 4 Depth (microns) Figure 3.4 Mode profile for a titanium-indiffused lithium niobate waveguide obtained by the knife-edge scanning technique. Note that the measured mode depth is approximately 2.2 fim between the 1/e2 points in intensity, and that the long evanescent tail on the substrate side of the profile closely resembles the expected profile for a Gaussian index distribution waveguide. The discrepancy between the two measured mode profiles is explained by the two drawbacks of the direct measurement method mentioned above, both of which tend to broaden the measured mode profile. The matrix method for waveguide analysis discussed in Sect. 3.3 was used to determine the index change An and diffusion depth D that leads to a single-mode Ti:LiNbC> 3 waveguide with a mode profile that closely matches the measured profile. A Gaussian refractive-index-distribution waveguide was evaluated with different combinations of An and D. Only waveguide structures that produced an effective refractive index that matched the value measured by the prism coupling method {ne ff= 2.2069) were considered. The change in refractive index that satisfied these 105 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 .« 0.8 0 ) c ( D Measured Calculated Neff = 2.2069 Ane = 0.0137 D = 1.531 c 0.6 < D T 3 O 2 " g 0.4 N (0 E O 0.2 Z 0.0 0 2 1 3 4 Depth (microns) Figure 3.5 Mode profile for a titanium-indiffused lithium niobate waveguide obtained by imaging with a microscope objective. conditions is An = 0.0137 ±0.0009, and likewise the effective waveguide depth is D = 1.53 ±0.13 fim. The calculated mode profile of a waveguide characterized by the mid-values of these two parameter ranges is shown in comparison with the mea sured profile in Fig. 3.5. These waveguide parameters were used to estimate the diffusion constant for our substrates and diffusion process. The temperature-dependent diffusion constant was calculated to be D c = 8.13 x 10' 13 cm2/sec using Eq. 3.2. Since the theoretical description of D c has two unknown material constants (i.e., D 0 and T0 as given in Eq. 3.3), absolute determination of D0 and T0 is not possible with our present data. However, reasonable combinations of values for these constants that yield our mea sured value for D c can be compared to those measured by others. For the diffusion temperature of T= 1273 K used in our process, combinations of the material-depen dent diffusion constant and activation temperature found independently by Holmes, 106 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Minakata, and Sugii yield values for D c that most closely match our measured value for this coefficient [Minakata, et al., 1978; Sugii, et al., 1978; Holmes and Smyth, 1984], as given in Table 3-1. Note that in this table, the measured material constants Table 3-1: Measured Diffusion Constants Do (cm /sec) Ambient Gas Tq (K) Reference 0.01830 N 2 and 0 2 3.248 X 104 [Griffiths and Esdaile, 1984] 0.00750 Ar 2.9 X 104 [Bums, etaL, 1979] 0.00022 Air 2.529 x 104 [Sugii, etal., 1978] 0.00042 Air 2.622 x 104 [Minakata, etal., 1978] 0.00030 W e t0 2 2.575 x 104 [Holmes and Smyth, 1984] 0.03470 Air 2.853 x 104 [Fukuma and Noda, 1980] 0.02685 Wet Ar and 0 2 3.19 x 104 [Dahan, etal., 1991] measured by separate investigators varies by as much as two orders of magnitude. This variation is a result of variation in the quality of the lithium niobate substrate material that was available over the period that the experiments were performed and the significant differences in the diffusion processes used by individual investigators (such as process apparatus and ambient gasses). 3.5.3 Propagation Loss M easurements Loss of optical power in the guided mode occurs due either to absorption or scattering. Both types o f losses will serve to lower the dynamic range of an optical processor. Therefore, fabrication of waveguides that reduce these losses is essential before one can reduce the proposed applications to practice. The primary mechanisms that contribute to scatter are surface scattering due to 107 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide roughness at the air interface and in-plane scattering due to index inhomo geneities [Hutcheson, 1987]. Well-confined modes scatter more as a result of surface roughness since mode is closer to the scattering surface with a large refractive index difference; therefore, it is not desirable to fabricate Ti:LiNbC> 3 waveguides in which the guided mode is too shallow (1 jjm deep or less). Reportedly, optical damage at high power densities for visible light in excess of 1 x IC T 4 mW/|im2 can lead to scattering in lithium niobate waveguides and limit pro cessor performance [Hutcheson, 1987]. However, power densities more than three times larger than this value at Aq = 0.6328 A have been produced within the Ti:LiNb03 waveguides used in the present work before any optical damage effects were observed. A simple method for measuring propagation losses is a non-destructive tech nique called the sliding prism method [Nishihara, et al., 1989]. The loss measure ment is accomplished by selectively exciting a waveguide mode by prism incoupling as described in Sect. 3.6.1, and then prism out-coupling the guided-mode light at sev eral different positions along its path. The loss coefficient is determined by the expo nential drop of the out-coupled power as a function distance. Care must be taken to maintain consistent prism incoupling and outcoupling efficiencies to obtain an accu rate measurement. The propagation loss in a Ti:LiNb03 waveguide fabricated by the processes outlined in this chapter was measured using a 15 mW He-Ne laser (Ao = 6328 A). The relative power outcoupled from the second prism as a function of distance between the prisms is shown in Fig. 3.6(a). The measured drop in optical power with propagation distance was determined to be 0.45 ± 0.1 dB/cm. This value for waveguide attenuation is comparable with those achieved by a number of other investigators (typically as low as 0.5 dB/cm) [Armenise, 1988; Nishihara, et al., 1989]. Waveguide propagation loss measurements can be obtained by imaging the guided streak onto a CCD detector array [Okamura, et al., 1983; Okamura, et al., • 108 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. > is ® G C -4 1 Loss = 0.45 ±0.1 dB/cm • Measured data Linear fit - 5 L l -1_1—L I- I L_1 L I i I J _ l -I I I—1 —1_ I_ l I I I I I I I I 1 I I I I I I I I I 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Distance (cm) co ■ o C O c a) C D > © C L 1 0 -1 -2 -3 Loss = 0.41 ± 0.019 dB/cm Measured data Linear fit -4 5 0.0 0.5 1.0 1.5 2.0 3.0 2.5 3.5 4.0 (b) Distance (cm) Figure 3.6 Titanium-indiffused lithium niobate propagation loss measurements made by (a) the sliding prism method, and (b) by the CCD imaging method. 109 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1985]. The difficulties associated with repeatable prism outcoupling can be avoided by this technique, leading to a more accurate measurement. In this method, it is assumed that the intensity o f the scattered light is roughly proportional to the guided mode power at a given position along the path of the guided mode. The attenuation constant is obtained in this case from the decay in the guided streak intensity. Light from a 15 mW He-Ne laser (Xq = 6328 A) was coupled into the same Ti:LiNb03 sam ple that was used in the sliding prism measurement. The relative scattered light inten sity is shown as a function o f the guided-streak propagation distance in Fig. 3.6(b). The propagation loss obtained by this measurement was 0.41 ±0.019 dB/cm. The measured propagation loss values obtained by linear least-squares fits of the data taken by the two methods are in close agreement. The imaging method is easier to perform, however, and produces more accurate results. 3.6 Source Coupling M ethods A fully integrated processor will operate with a laser diode source butt-coupled to the waveguide edge. For component testing purposes, it is more convenient to use either the prism coupling method or the end-fire coupling method to excite a guided- wave. In these techniques, the specific properties of the input beam may be conve niently formed by an external optical system. In the case of prism coupling, testing is simplified since a high quality edge-polish is not required. 3.6.1 Prism Coupling Prism coupling is a well-known method for waveguide excitation [Nishihara, et al., 1989]. It consists of the placement of a carefully cut and polished prism into inti mate contact with the waveguide surface as shown in Fig. 3.7. The refractive index of the prism is higher than that o f the waveguide to be tested. A beam of coherent light is then directed on the base o f the prism where it is internally reflected within the 110 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. F Laser Illumination Waveguide v Substrate Figure 3.7 Prism coupling arrangement in which the gap distance between the prism and waveguide is adjusted by the applied force. prism. Although the light is reflected at the base of the prism, an evanescent field pen etrates the gap region and extends into the waveguide layer. Light will couple into the waveguide when a phase-matching condition between the propagation vector of the light in the prism and the propagation vector of one of the waveguide modes is satis fied. The mode that is excited can be selected by the appropriate orientation of the incoupling beam at an angle 0t - with respect to the normal of the prism face. In this way, individual waveguide modes are selectively excited and the effective refractive index for the excited mode is determined by a direct relation with the incoupling angle as given below in which y/p is the prism angle, 0L is the incoupling angle relative to the normal of the prism entrance face, and np is the prism index of refraction. The utility of prism cou pling is that it provides a very convenient means of selectively exciting a waveguide mode in planar or channel waveguides since the prism can be repeatedly attached and (3.10) 111 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. removed. In a similar fashion, a second prism can be placed in the path of a guided wave in order to outcouple the light. For prism incoupling with a tapered gap between the prism and waveguide, cou pling efficiencies can approach 80%. The tapered contact is necessary to limit the region of intimate contact; otherwise, light coupled into the waveguide at the front of this region may couple back into the prism before it passes beyond the back-edge rather than remain confined in the waveguide. A typical method to obtain a tapered gap is to deform the waveguide by applying pressure to a point on the back-side of the waveguide substrate. The incoupling efficiency is low for our Ti:LiNb03 waveguides since the substrate is 1 mm thick and not easily deformed by applying pressure to the back side. A tapered gap is difficult to achieve in this case. Incoupling efficiencies were typically between 5-15% for the Ti:LiNb03 waveguides described in this chap ter. However, we found that we could typically achieve prism outcoupling efficien cies ranging between 50 and 100%. High efficiencies in prism outcoupling can be achieved because a taper is not required and the region of intimate contact may be made large. Light coupled into the prism is not coupled back into the waveguide because it is coupled into an unbound mode that only propagates away from the waveguide. The primary drawback of the prism coupling method is that the coupling unifor mity and repeatability is difficult to control, since it relies greatly on the cleanliness of the prism and waveguide surfaces and the flatness of the polished surfaces. Also, the interaction width of the incoupling region is very narrow so that defects in the prism or the waveguide surface in this interaction region can lead to phase distortion across the beam aperture. Prism couplers are not generally used as the source coupling tech nique in fully integrated processors since the precision ground-and-polished prisms are expensive and require a special coupling mount to hold the prisms firmly to the sample. 112 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3.6.2 End-Fire Coupling End-fire coupling is the most basic method of waveguide excitation, and is accomplished simply by focusing an external laser source onto a polished waveguide edge as shown in Fig. 3.8. In order for the coupling to be efficient, the numerical aperture of the focusing element should match that of the waveguide [Nishihara, et al., 1989]. Individual modes cannot be selectively excited by this method, as they can by the prism-coupling method. Rather, the coupling efficiency into each mode is determined by the overlap integral of the input beam electric field distribution with the characteristic electric field distributions of the various guided modes. In the case of Ti:LiNb03 waveguides, end-fire coupling requires special prepa ration of the waveguide edge by edge-polishing. The edge-polishing technique used for the work described herein is presented in Sect. 3.7. Gallium arsenide waveguides, on the other hand, can be easily prepared for end-fire coupling by cleaving along one of the < 100> directions. The losses that are typically incurred by this method include mode coupling losses and Fresnel reflection losses. The Ti:LiNb03 waveguides described earlier have numerical apertures approximately equal to NA = 0.148. To couple into this waveguide, either a fully illuminated 5x microscope objective or a high power 60x objective with a 1 mm wide beam input is required to match the numerical aperture. The focal spot size obtained with these focusing elements is approximately 2 fim, and therefore will closely match the waveguide mode profile. For the GaAs/AlGaAs waveguides, a microscope objective with NA = 1.04 or less is required for efficient coupling. Since most microscope objectives satisfy this condition, the microscope objective that produces a focal spot most closely matched to the mode profile o f the waveguide should be used. For the 1 fim thick waveguide layers characteristic o f the GaAs/AlGaAs waveguides described earlier and an illumination wavelength of Ao = 1.06 fim, a microscope objective with NA = 0.5 is suitable (such as a fully illu- 113 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Input Beam W aveguide Substrate h— H Figure 3.8 End-fire coupling arrangement showing the convergence angle of the incoupling beam. This angle should be less than the acceptance angle of the waveguide for efficient coupling. minated 25x objective). The refractive index difference between air and the waveguide material leads to a Fresnel reflection loss. For lithium niobate and gallium arsenide waveguides, the end-facet Fresnel reflections are 14% and 30%, respectively. In our later experi ments, these reflection losses were taken into account whenever the optical power coupled into or out of a waveguide by this method was calculated. 3.7 W aveguide End-Polishing The optical properties of the waveguide devices discussed herein were evalu ated for the most part using the prism coupling technique. However, in several cases we found it necessary to polish the waveguide ends to end-fire coupled light into the waveguide or to observe the waveguide end-emission. The end-polishing that was performed in the present work is similar to the techniques used by others [Furch, et al., 1983]. Several samples were end-polished for the research described in this the sis, including the planar Ti:LiNb03 waveguides that were used for mode-profile mea surement as described in this chapter; the rib waveguide array samples used for crosstalk measurements as described in Chapter 4; the embedded lens samples used for measurement of the lens focal spot size as described in Chapter 5; and the test- 114 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. modules used for demonstration of component integration as described in Chapter 7. For waveguide evaluation purposes, the edge of the sample where the waveguide meets the end to be polished should be flat and free of chips or scratches, since this is where the waveguide coupling takes place. In our end-polishing proce dure, several samples are clamped face-to-face and placed into a brass polishing mount as shown in Fig. 3.9 to prevent the edge from chipping. The mount is prepared with several lithium niobate support pieces 1 mm thick attached with wax to the sur face to provide a large polishing area and thereby prevent edge rounding. Before the samples are inserted into the mount, the waveguide surfaces are covered with vinyl wafer dicing tape to prevent the waveguide surface from becoming scratched. The stack of samples is clamped with the ends to be polished flush to one another so that they protrude from the face of the mount by about 0.5 mm. All of the polishing steps were performed on an R.H. Strasbaugh 6DA-1 auto matic polishing machine. The wafer edges and support pieces were first lapped on 600 grit SiC paper to obtain planarity. After the samples were lapped, the lapping platen is removed and replaced by another platen with a Texmet pad attached to the surface. The Texmet pad, supplied by Buehler, Inc., is prepared by impregnating the surface with 15 fim diamond paste. The samples are then automatically polished on this pad for approximately 3 hours with a diamond extender solution (also supplied by Buehler, Inc.) used to keep the pad wet. The platen is sequentially exchanged with other platens prepared with smaller diamond grits (9 ^m , 6 fj.m, and 1 fim) until the final desired polish quality is achieved. The polished ends were observed by micro scope to determine when the platens should be exchanged. In general, the platens should be exchanged when the surface quality no longer improves with the same grit 115 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Lithium niobate waveguides Lithium niobate support piece Clamping screws \ Figure 3.9 Waveguide end-polishing mount with clamped samples and surface mounted support pieces. size. Typically, it takes a minimum of 3 hours of polishing per step. The mount and samples are carefully cleaned with soapy water and then in an ultrasonic DI water bath for about 5 minutes between each step to prevent the polishing pads from becom ing contaminated with the larger grit sizes from the previous step. In Fig. 3.10, a pho tomicrograph of final polished waveguide edge is shown. We estimate that at least 80-90% of the entire waveguide edge of the sample is free of chipping or scratches as a result of this process. The end-polish shown in Fig. 3.10 is typical of the quality that we achieved for our samples with this process. 3.8 Summary In this chapter, we discussed the merits of two waveguide materials for imple menting advanced 10 signal processors. In particular, the formation and characteriza tion of low-loss, single-mode Ti:LiNb03 waveguides on y-cut substrates was described. The waveguide loss (0.41 dB/cm) was accurately measured and found to be much lower than that previously achieved by Rastani (1-2 dB/cm). The effective refractive index (ne f f = 2.2069) o f this structure was measured and the effects of pro cess variations were considered in preparation for discussion of the embedded lens 116 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 3.10 Microphotograph of the polished ends of two lithium niobate samples stacked together with an intermediate layer of wafer dicing tape. design and fabrication in Chapter 5. The greatest variations were attributed to fluctu ations in the titanium layer thicknesses and resulted in a refractive index change of ±0.001. In addition, the mode profile for this waveguide structure was measured and compared with theory. The mode depth of 2.2 fim is sufficiently shallow for the waveguide devices to allow single-mode behavior, but not so shallow as to cause excessive waveguide scattering losses. A gallium arsenide waveguide fabrication technique was discussed, and a GaAs/AlGaAs waveguide structure with a 1 fim thick waveguide layer and a 2 fim thick barrier layer was described. This waveguide was prepared for the development of rib waveguide arrays with surface outcoupling grat ings as described in Chapter 4. 3.9 References M. N. Armenise, “Fabrication Techniques of Lithium Niobate Waveguides,” IEE Proceedings-J, 135(2), 85-90, (1988). M. N. Armenise, C. Canali, M. D. Sario, A. Camera, P. Mazzoldi, and G. Celotti, “Characterization of (Ti0 65^ 0.35)0 , Compound as a Source for Ti Diffusion During Ti:LiNb03 Optical Waveguide Fabrication,” J. Appl. Phys., 54( 1), 62-70, (1983). 117 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. A. Behfar-Rad, S. S. Wong, J. M. Ballantyne, B. A. Soltz, and C. M. Harding, “Rectangular and L-Shaped GaAs/AlGaAs Lasers with Very High Quality Etched Facets,” Appl. Phys. Lett., 54,493-495, (1989a). A. Behfar-Rad, S. S. Wong, R. J. Davis, and E. D. Wolf, “Masking Considerations in Chemically Assisted Ion Beam Etching of GaAs/AlGaAs Laser Structures,” J. Electrochem. Soc., 136, 779-782, (1989b). W. K. Bums, P. H. Klein, E. J. West, and L. E. Plew, “Ti Diffusion in Ti:LiNbC> 3 Planar and Channel Optical Waveguides,” J. Appl. Phys., 50(10), 6175-6182, (1979). C. Canali, C. D. Bemardi, M. D. Sario, A. D'Orazio, and S. Morasca, “Steplike Refractive-Index Increase Induced in Planar Ti:LiNb03 Waveguides Diffused in 0 2:H20 Atmosphere,” Appl. Opt., 27(19), 3957-3958, (1988). J. R. Carruthers, I. P. Kaminow, and L. W. Stultz, “Diffusion Kinetics and Optical Waveguiding Properties of Outdiffused Layers in Lithium Niobate and Lithium Tantalate,” Appl. Opt., 13(10), 2333-2342, (1974). M. W. Casseday, N. J. Berg, and I. J. Abromovitz, “Space-Integrating Acousto-Optic Signal Processors Using Surface-Acoustic Wave Delay Lines,” in Acousto-Optic Signal Processing , N. J. Berg and J. N. Lee, Eds., 165-202 (Marcel Dekker, Inc., New York, 1983). R. Dahan, N. Croitoru, and S. Ruschin, “Studies on the Relation Between the Diffusion Process and Optical Properties in Ti-Diffused Planar Optical Waveguides,” Appl. Opt., 30(30), 4396-4401, (1991). M. De Micheli, J. Botineau, P. Sibillot, D. B. Ostrowski, and M. Papuchon, “Fabrication and Characterization of Titanium Indiffused Proton Exchanged (TIPE) Waveguides in Lithium Niobate,” Opt. Commun., 42, 101-103, (1982). 1 1 8 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. R. J. Esdaile, “Closed-Tube Control of Out-Diffusion During Fabrication o f Optical Waveguides in LiN b03,” Appl. Phys. Lett., 33(8), 733-734, (1978). M. Fukuma and J. Noda, “Optical Properties o f Titanium-Diffused LiN b03 Strip Waveguides and Their CoupIing-to-a-Fiber Characteristics,” Appl. Opt., 19, 591-597, (1980). B. Furch, E. Bratengeyer, and H. Rauch, “Fast High-Quality Edge Polishing of LiNb03,” J. Opt. Commun., 4(2), 47-50, (1983). A. K. Ghatak, I. C. Goyal, and R. L. Gallawa, “Mean Lifetime Calculations of Quantum Well Sturctures: A Rigorous Analysis,” IEEE J. Quantum Electron., 26(2), 305-310,(1990). A. K. Ghatak, K. Thyagarajan, and M. R. Shenoy, “Numerical Analysis o f Planar Optical Waveguides Using Matrix Approach,” J. Lightwave Tech., LT-5(5), 660-667, (1987). A. K. Ghatak, K. Thyagarajan, and M. R. Shenoy, “A Novel Numerical Technique for Solving the One-Dimensional Schroedinger Equation Using Matrix Approach- Application to Quantum Well Structures,” IEEE J. Quantum Electron., QE-24(8), 1524-1531,(1988). G. J. Griffiths and R. J. Esdaile, “Analysis of Titanium Diffused Planar Optical Waveguides in Lithium Niobate,” IEEE J. Quantum. Electron., QE-20(2), 149-159, (1984). J. Helms, J. Schmidtchen, B. Schiippert, and K. Petermann, “Error Analysis for Refractive-Index Profile Determination from Near-Field Measurements,” J. Lightwave Tech., LT-8(5), 625-633, (1990). 119 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. T. Hirata, M. Suehiro, M. Hihara, M. Dobashi, and H. Hosomatsu, “Demonstration of a Waveguide Lens Monolithically Integrated with a Laser Diode by Compositional Disordering of a Quantum Well,” IEEE Photonics Tech. Lett., 5(6), 698-700, (1993). R. J. Holmes and D. M. Smyth, ‘Titanium Indiffusion into LiNb03 as a Function of Stoichiometry,” J. Appl. Phys, 55(10), 3531-3535, (1984). R. G. Hunsperger, Integrated Optics: Theory and Technology, (Springer-Verlag, New York, 1982). L. D. Hutcheson, Ed., Integrated Optical Circuits and Components, Series on Optical Engineering, B. J. Thompson, Ed., (Marcel Dekker, Inc., New York, 1987). J. L. Jackel, A. M. Glass, G. E. Peterson, C. E. Rice, D. H. Olson, and J. J. Veselka, “Damage Resistant LiN b03 Waveguides,” J. Appl. Phys., 55, 269-270, (1984). J. L. Jackel, C. E. Rice, and J. J. Veselka, “Proton Exchange for High-Index Waveguides in L iN b03,” Appl. Phys. Lett., 41(7), 607-608, (1982). J. L. Jackel, C. E. Rice, and J. J. Veselka, “Composition Control in Proton Exchanged LiNb03,” Electron. Lett., 19, 387-388, (1983). K.-Y. Liou, U. Koren, S. Chandrasekhar, T. L. Koch, A. Shahar, and C. A. Burrus, “Monolithic Integrated InGaAsP/InP Distributed Feedback Laser with Y-B ranching Waveguide and a Monitoring Photodetector Grown by Metalorganic Chemical Vapor Deposition,” Appl. Phys. Lett., 54(2), 114-116, (1989). M. Minakata, S. Saito, and M. Shibata, ‘Two-Dimensional Distribution of Refractive- Index Changes in Ti-Diffused LiNb03 Strip Waveguides,” J. Appl. Phys., 50(5), 3063-3067, (1979). 120 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. M. Minakata, S. Saito, M. Shibata, and S. Miyazawa, “Precise Determination of Refractive-Index Changes in Ti-Diffused LiNb03 Optical Waveguides,” J. Appl. Phys., 49(9), 4677-4682, (1978). H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). Y. Okamura, S. Sato, and S. Yomamoto, “Simple Method of Measuring Propagation Properties of Integrated Optical Waveguides: An Improvement,” Appl. Opt., 24(1), 57-60, (1985). Y. Okamura, S. Yoshinaka, and S. Yomamoto, “Measuring Mode Propagation Losses of Integrated Optical Waveguides: A Simple Method,” Appl. Opt., 22(23), 3892-3894, (1983). M. R. Ramadas, E. Garmire, A. K. Ghatak, K. Thyagarajan, and M. R. Shenoy, “Analysis of Absorbing and Leaky Planar Waveguides: A Novel Method,” Opt. Lett., 14(7), 376-378,(1989). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). C. E. Rice and R. J. Holmes, “A New Rutile Structure Solid-Solution Phase in the LiN b030 8 -Ti02 System, and its Role in Ti Diffusion into LiN b03 ,” J. Appl. Phys., 60(11), 3836-3839,(1986). R. V. Schmidt and I. P. Kaminow, “Metal-Diffused Optical Waveguides in LiN b03,” Appl. Phys. Lett., 25(8), 458-460, (1974). G. J. Sonek, L. Jian-Zhong, E. D. Wolf, and J. M. Ballantyne, “SiCl4 Reactive Ion Etching for GaAs Optical Waveguides,” J. Lightwave Tech., LT-3, 1147-1150, (1985). 121 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. K. Sugii, M. Fukuma, and H. Iwasaki, “A Study on Titanium Diffusion into LiNb03 Waveguides by Electron Probe Analysis and X-Ray Diffraction Methods,” J. Mat. Sci., 13,523-533,(1978). T. L. Szabo and A. J. Slobodnik, Jr., “Acoustic Wave Diffraction and Beam Steering,” Report Number AFCL-TR-73-0302, (Air Force Cambridge Research Laboratories, Bedford, Mass., 1973a). T. L. Szabo and A. J. Slobodnik, Jr., ‘T he Effect of Diffraction on the Design of Acoustic Surface Wave Devices,” IEEE Trans. Sonics Ultrason., SV-20(3), 240-251, (1973b). L. Tomer, J. Ferrer, and F. Canal, “Applicability of the Lorentzian Peak Method to Analyze Leaky and Lossy Optical Waveguides,” Appl. Opt., 30(18), 2418-2420, (1991). C. S. Tsai, Ed., Guided-Wave Acousto-Optics, Springer Series in Electronics and Photonics, D. H. Auston, etal., Eds., Vol. 23, (Springer-Verlag, New York, 1990). G. A. Vawter, L. A. Coldren, J. L. Merz, and E. L. Hu, “Nonselective Etching of GaAs/AlGaAs Double Heterostructure Laser Facets by Cl2 Reactive Ion Etching in a Load-Locked System,” Appl. Phys. Lett., 51, 719-721, (1987). G. A. Vawter, J. L. Merz, and L. A. Coldren, “Monolithically Integrated Transverse- Junction-Stripe Laser with an External Waveguide in GaAs/AlGaAs,” IEEE J. Quantum Electron., QE-25(2), 154-162, (1989). R. S. Weis and T. K. Gaylord, “Lithium Niobate: Summary of Physical Properties and Crystal Structure,” Appl. Phys. A, 37, 191-203, (1985). 122 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Chapter 4 Rib Waveguide Arrays with Surface Outcoupling Gratings Rib waveguides with surface outcoupling gratings have the simple yet crucial function in advanced IO signal processors to dissect an in-plane image and to dissem inate the image to surface-mounted components with efficiency and accuracy. In this chapter, we evaluate the capacity of fabricated rib waveguide and grating structures to accomplish this function sufficiently to fulfill the system requirements for the IOS AR processor and IO correlator. The image dissemination function is illustrated in Fig. 4 .1(a), in which a portion of the rib waveguide array with surface outcoupling gratings is shown. For the IO correlator, the surface outcoupling gratings should provide efficient outcoupling over a distance as small as 100 fim, whereas in the IOSAR processor the outcoupling should be uniform over a 1 cm distance. The surface outcoupling gratings couple light into several radiation modes, which propagate at characteristic angles dependent upon the size of the grating period relative to the wavelength of the light inside the waveguide. The propagation directions of the radiation modes are indicated in Fig. 4.1(b) for an arbitrary grating. A detector array placed in close proximity to the grating surface will be illuminated by the upward-propagating modes. Four important issues must be considered in our evaluation of rib waveguides and grating structures that implement the desired image dissemination function. The first issue is the dissection of the incident field distribution. The rib waveguide array 123 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (a) (b) Figure 4.1 Rib waveguide array with surface outcoupling gratings showing (a) ini tial intensity distribution incident on a rib waveguide array and final intensity distribu tion coupled from the rib waveguides by surface outcoupling gratings, and (b) cladding and substrate radiation mode propagation directions. must have the proper element density to fully sample the incident field, small separa tion widths to minimize dead space and maximize the coupling efficiency, and a numerical aperture sufficient to accommodate the angular extent of the incident field. The second issue is the propagation of the dissected field along the length of the rib waveguides. There must be minimal distortion of the dissected field due to coupling between rib waveguides, and low propagation losses in rib waveguide arrays that require a large propagation length. The third issue is the grating outcoupling effi ciency of the dissected image. A significant portion of the total outcouped light must impinge upon the surface-mounted devices. The fourth issue is the illumination of the surface-mounted devices. The light outcoupled from each rib waveguide must 124 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. accurately illuminate the pixels positioned directly above it and not those to either side. For the IOSAR processor, the illumination must be uniform along the length of the rib waveguide. The rib waveguide array and grating structure in the IOSAR processor requires a much higher element packing density, a longer array length, and a larger outcou pling region than the rib waveguide array and grating structure in the IO correlator. In this chapter, we focus primarily upon the fabrication and evaluation of rib waveguide arrays and grating structures as required for the IOSAR processor. The results, in general, encompass the less severe requirements for the IO correlator as well. In our evaluation o f densely-packed rib waveguide arrays with large-area, uni form surface outcoupling gratings, we expand upon the theoretical analysis first pre sented by Rastani [Rastani, 1988]. As a part of our analysis, we identify sources of crosstalk that lead to distortion of the initial in-plane image. Sources of distortion that arise from the initial image dissection and rib-to-rib coupling previously considered by Rastani are addressed in greater detail herein. In addition, we consider for the first time in the context of the IOSAR processor and IO correlator the contributions to crosstalk from rib waveguide scattering and transverse diffraction (across the width of the rib waveguide) of the light outcoupled from the rib waveguides. In another part of our analysis, the illumination uniformity of surface-mounted components provided by surface outcoupling gratings on rib waveguides is considered in depth. An initial analysis and characterization of surface outcoupling gratings with a short interaction length (1 mm) integrated on rib waveguides in Ti:LiNb0 3 waveguides was presented by Rastani. The guidelines for uniform grating outcoupling over a large area devel oped therein are revisited in the present work and applied to the fabrication of uni form surface outcoupling gratings with a larger interaction length (1 cm). We consider additional sources of illumination non-uniformity such as rib waveguide propagation losses, coherent interference of the outcoupled modes at the external detector plane, interference between multiple modes supported by the rib waveguides, 125 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. and inhomogeneities associated with the component fabrication. In the third part of the our analysis, the rib waveguide and grating power transfer efficiencies are consid ered. In the previous work by Rastani, reference was made to the sources of power loss. These include inefficient input field coupling to the rib waveguide array, direc tion of light into the waveguide substrate by the surface outcoupling gratings, and rib waveguide propagation losses. In the present work, we cover these issues and pay careful attention to the efficiency of power distributed to the surface-mounted devices by the outcoupling gratings, since the loss of light into the substrate can be quite high. Others investigators have used channel waveguide arrays in signal processors to route information in-from or out-to the edge of the waveguide [Anderson, 1978; Val- ette, et al., 1983; Tsai, et al., 1985]. For these arrays, the required channel density is low so that sufficient isolation can be maintained with an appropriate spacing between elements. Channel waveguides in LiNb0 3 have been shown to be unsuitable for densely packed arrays since the coupling between channels is high [Rastani, 1988]. Since crosstalk due to rib-to-rib coupling is a possible limiting factor in the per formance of high density rib waveguide array structures, we have experimentally investigated rib-to-rib coupling for various rib waveguide element spacings, separa tion depths, and interaction lengths for structures fabricated on Ti:LiNb03 waveguides. The coupling between rib waveguides is analyzed theoretically through calculation of the rib waveguide mode properties with an approximate analytical method [Marcatili, 1969], estimation of shallow-etched rib waveguide separations by the effective index method [Bums and Milton, 1975], and calculation of the coupling coefficient between rib waveguides [Kuznetsov, 1983]. The results o f extensive theoretical investigations have been used to describe the coupling between parallel waveguides [Marcatili, 1969; Marcuse, 1971; Yariv, 1973; Kuznetsov, 1983; Hardy and Streifer. 1985; Kawakami and Haus, 1986; Burke, et al., 1992]. Coupling between parallel waveguides has application to areas such as modulators [Ramaswamy, et al., 1978; Minakata, 1979], directional couplers [Haus 126 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. and Molter-Orr, 1983], wavelength filters [Burke, et al., 1992], and semiconductor laser arrays [Scifres, et al., 1979; Butler, et al., 1984; Kapon, et al., 1984]. Coupling between parallel rib waveguides has also been analyzed [Burke, et al., 1992; Mao and Huang, 1992; Ratowski, et al., 1992]. In the majority of these works, coupling between parallel waveguides is desired and hence the coupling constants under con sideration are large. In the present work, however, the coupling constant should be small so that the rib waveguides are effectively isolated from one another. The performance of the IOSAR processor azimuth-compression function is crit ically dependent upon the rib waveguide and surface outcoupling grating lengths. In the present work, we have experimentally investigated large-area gratings on both planar waveguides and rib waveguide arrays. Surface outcoupling gratings have been previously used for waveguide input coupling [Valette, et al., 1983], for efficient waveguide-detector coupling [Huang and Lee, 1986], for large-area focused outcoupling to an external optical disk [Suhara and Nishihara, 1986], and for distributed supply of optical power to external elements [Kubota and Takeda, 1989; Takeda and Kobuta, 1991; Song, et al., 1994]. In work performed earlier by Rastani [Rastani, 1988], two coupling methods were considered for use in the IOSAR processor including surface grating coupling and thin film eva nescent coupling. The use of surface outcoupling gratings was shown to have some advantages over thin film couplers, including relaxed requirements for the surface- mounted component proximity and increased dynamic range for shorter grating peri ods. Most of the component development conducted in the present work is based on Ti;LiNbC> 3 waveguides. However, we have also experimentally investigated the fab rication of rib waveguide array and grating structures on GaAs/AlGaAs waveguides because of the long-term potential for a monolithically integrated processor. In this investigation, we measured the basic surface outcoupling grating properties such as the grating outcoupling efficiency for the separate radiation modes. These measure- 127 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ments were performed in a manner similar to the initial characterization performed by Rastani for the same devices fabricated in Ti:LiNb03 waveguides. The process used in the fabrication o f rib waveguide and grating structures on GaAs/AlGaAs waveguides is also similar to that used by Rastani for the same structures fabricated on Ti:LiNb03 waveguides [Rastani, 1988]. Rib waveguides have also been fabri cated by others in GaAs-based waveguides [Somekh, et al., 1974; Deri, et al., 1987]. In this chapter, we discuss the theory, fabrication, and characterization of rib waveguide arrays on planar Ti:LiNb03 waveguides followed by a similar discussion for surface outcoupling gratings on planar Ti:LiNb03 waveguides. The issues related to the integration of these two waveguide components are discussed in the final sec tion. The GaAs/AlGaAs rib waveguides and surface grating structures are also dis cussed in the last section. 4.1 Rib Waveguide Arrays In this section, we discuss the issues of incident field dissection by a rib waveguide array and o f propagation of the dissected field along the length of the rib waveguides. We identify a connection between the rib waveguide dissection capabil ities and the propagation properties that is inherent in the design of the rib waveguide gap depth and width. 4.1.1 Incident Field Distribution The design of the rib waveguide array for a particular processor depends upon the characteristics of the incident field distribution that is to be dissected. The inci dent field distribution varies for different processor geometries. In Fig. 4.2, the geom etries are shown for the IOSAR processor, 10 correlator, and integrated test module. The differences in angular extent of the incident field distribution and image size for these three examples are schematically represented. To efficiently dissect these input 128 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. SAW Rib Waveguide Array (a) IOSAR processor SAW Rib Waveguide Array (b) 1 0 correlator SAW Rib Waveguide Array (c) Test module Figure 4.2 Schematic diagram that illustrates the formation of the field distribution at the rib waveguide array for (a) the IOSAR processor, (b) the 10 correlator, and (c) the test module. 129 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 749 fields, the acceptance angle of the rib waveguides should be commensurate with the angular extent o f the incident field distribution, and the rib waveguide array gap width should be small compared with the rib waveguide width (large rib waveguide fill fac tor). The numerical apertures (NA) of the rib waveguides determine the permissible range of angles o f the incident beam for efficient incoupling. This consideration is similar to that encountered when light is coupled into an optical fiber. The primary difference is that fiber coupling usually takes place at a air/fiber interface, whereas the coupling for these processors takes place at a planar waveguide/rib waveguide inter face. The refractive index of the planar waveguide must therefore be taken into account when calculating the rib waveguide numerical aperture: In this equation, 9a is the acceptance angle of the rib waveguide, ne j j is the effective refractive index o f the planar waveguide portion of the processor, n, is the refractive index of the rib waveguide, and ng is the refractive index of the gap region between the rib waveguides. From the ray-optics perspective, light coupled into the rib waveguide at an angle less than the acceptance angle will undergo total internal reflection (TER.) at the rib waveguide sidewalls. Total internal reflection occurs when n ,> n g and the angle of incidence of the ray on the rib waveguide sidewall is greater than the critical angle 6C defined by sin 6C = ngln,. Light coupled into the rib waveguide at an angle greater than the acceptance angle will undergo a reflection loss with each reflection from the rib waveguide sidewalls. This coupled light is highly attenuated as it propagates. From the wave-optics perspective, light coupled into the rib waveguide at an angle greater than the acceptance angle will excite both waveguide modes and free space modes. The excitation of the guided modes, how ever, will be very weak. (4.1) 130 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. The minimum required value of Q a can be calculated for each processor through analysis o f each case as discussed in the examples of Chapter 2. The IOSAR proces sor forms a 2 .6 mm wide range-focused image with an angular spectrum bounded by the maximum off-axis propagation angles of the light through the lens. For the f/4 lens used in the IOSAR example, this leads to maximum off-axis angles of ±7°. The 10 correlator, on the other hand, forms a 1-cm-wide image of the SAW modulated light on the rib waveguide array. The maximum off-axis angles in this case are about ±0.25°, which is the SAW deflection range. The integrated test module consists of an f/1 lens; hence, the maximum off-axis propagation angle is given by tan'*(0.5) = 27°. To determine the acceptance angle of the rib waveguide array, the rib waveguide structure must be considered. The basic rib waveguide structure is shown in Fig. 4.3(a) and consists of a rectangular waveguide region with a homogeneous refractive index n, on top of a substrate with a refractive index n2 This waveguide is surrounded by a third medium of refractive index n0 such that n, > n2 > n0. Regions 1 1 and HI on either side of the rib waveguide constitute the “gap” or separation between each two rib waveguide elements within an array. In order to obtain a large acceptance angle, the value of the gap refractive index, ng in Eq. 4.1, should be made as small as possible. In the structure shown in Fig. 4.3(a), the gap depth is equal to the waveguide depth b, and hence the minimum value for ng = 1 (air) is obtained. The other parameters, ne f f and n h are determined by the selection of the substrate material and the waveguide formation process. The maximum acceptance angle that can be achieved for rib waveguides fabricated on the single-mode Ti:LiNb03 waveguides (described in Chapter 3) is calculated by substitution of the values for the planar waveguide effective refractive index (ne j j = 2.2069), the rib waveguide refrac tive index (n, - 2.2156), and the minimum gap refractive index (ng = 1) into Eq. 4.1. The calculated acceptance angle is 63°, which is more than sufficient for any of the processors discussed above. The second factor that determines the coupling efficiency of the incident field 131 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. n0 (a) n I 111 ~b n0 n, n0 b __ .1 n 2 i ► y Substrate i i -------- a ---------► n n0 i I ni b 1 h HI n0 h n, _ _ _ n, a 1 n 2 i y Substrate Figure 4.3 Schematic diagrams of two rib waveguide geometries, in which the high index material surrounding the rib has been (a) fully removed and (b) partially removed. distribution into the rib waveguide array is the relative width of the rib waveguides and the gaps. The rib waveguide fill factor is a measure of the fraction of the total array width that can support and propagate the incident field and is given by T )r -d !W , in which d is the rib width as shown in Fig. 4.3(a) and W is the rib waveguide spacing within the array (hence, the gap width is W - d ) . For an incident field distribution that is fairly uniform, the total incoupling efficiency can be esti mated by assuming that all o f the light incident on a rib waveguide is incoupled and all of the light incident on a gap is either reflected or scattered out of the waveguide. In this case, the incoupling efficiency would be proportional to the rib waveguide fill factor. 132 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. To obtain a high coupling efficiency, the rib waveguide fill factor should be high. Correspondingly, the gap width in high density rib waveguide arrays should be small. To illustrate this, consider the high-density rib waveguide array required for the IOSAR processor and the comparatively lower-density rib waveguide array required for the 10 correlator. The rib waveguide fill factor for the IOSAR processor is T\r = 0.8 (8 /xm/10 /an) and for the 10 correlator is rfr = 0.93 (28 /an/30 /an). An increase in the gap width will lead to a much more severe drop in the coupling effi ciency for the IOSAR processor than for the 10 correlator. It is not always practical or beneficial to produce a rib waveguide array with a large gap depth or narrow gap width. There are processing difficulties related to the formation o f gaps with smooth rib waveguide sidewalls and controlled gap widths in Ti:LiNb0 3 when the required gap depth is comparable with the photolithographic res olution. This is typically the case for high density arrays such as those required for the IOSAR processor. A large gap depth can also lead to a high guided-wave attenua tion in the rib waveguides due to scattering from surface roughness on the rib waveguide sidewalls. High levels of scattering from the rib waveguides can lead to both crosstalk between parallel channels of the surface-mounted devices and nonuni form illumination of these devices along the length of the rib waveguide array. A nar row gap width can lead to increased coupling between rib waveguides within the array. In order to provide a more detailed description of how the rib waveguide array properties vary with the selection of gap width and gap depth, we consider next the field distributions of the rib waveguide structure. 4.1.2 Rib Waveguide Field Distribution We first consider a more general rib waveguide structure than that described ear lier, in which the gap depth is less than the waveguide depth b, as is shown in Fig. 4.3(b). In this case, a layer of material with refractive index n, remains on either side of the rib. An accurate mode analysis of this rib waveguide structure is possible 133 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. with a number of different numerical methods [Stem, 1988; Feit and Fleck, 1990; Vezzetti and Munowitz, 1990]. However, these methods are complex to implement. In the present work, we chose to use a simple method to describe the rib waveguide mode structure for arbitrary gap depths. This method is based on an approximate ana lytical method [Marcatili, 1969] and is valid for structures with abrupt refractive index discontinuities. By this method we obtain approximate analytical expressions for the electric field distributions in the rib waveguide and surrounding regions as well as the associated propagation constants for each of the guided modes. We have extended this analysis to include the first order effects of an arbitrary etch depth h in the gap regions of the rib waveguide structures as shown in Fig. 4.3(b). The structure shown in Fig. 4.3(b) is simplified by assuming that Regions II and HI are filled uniformly with a material of refractive index ng, which is a func tion of the gap depth h. The highest value that ng can take on is the refractive index of the rib waveguide n h which corresponds to no gap depth, and hence no lateral con finement. The lowest value that ng can take on is the refractive index o f air, which corresponds to a gap depth equal to or greater than the waveguide layer thickness as discussed previously. The effective refractive index of the gap region will decrease monotonically from the upper to the lower refractive index value for increasing gap depth. In Chapter 3, we found that the Ti:LiN b0 3 waveguide has a gradient refractive index in depth with a Gaussian profile. For this analysis, the waveguide layer in Region I is assumed to have a refractive index equal to the surface refractive index of the Ti:LiN b0 3 waveguide (2.2156) and a thickness b equal to the effective waveguide thickness for a gradient refractive index waveguide. This thickness is defined by n(y = b) = ne f f in which n(y) is given by Eq. 3.4 [Nishihara, et al., 1989]. For single mode Ti:LiN b0 3 waveguides, this waveguide thickness is approximately 1.53 jU m . The refractive index for n2 is 2.2019 for the LiN b0 3 substrate and n0 is 1 for the case in which air surrounds the rib waveguide. 134 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. With these changes, we consider the simplified rib waveguide geometry (Fig. 4.3(a)) with the refractive index in Regions II and III replaced by ng (the refrac tive index of the gap region). The eigenfunctions and analytical expressions for the field distributions in dielectric optical waveguides derived by Marcatili [Marcatili, 1969] were similarly modified to incorporate the refractive index ng of the gap region in place of n0. The solutions that we considered in particular are for modes with an Ef-polarized electric field vector (i.e., the electric fields are predominantly parallel to the x-axis of the coordinate system). The transverse electric field distributions in the waveguide and gap regions are given by I: Ex = F /H " '2*2 ~ ^ 2)sinkx(X+Sx)cosky{y+ 5y) ’ (4.2) II: f - ;p lc-c r Y s2 + ng2k 2 ) Y eP c o skx [x + 8X) c o s ky(y + ) ex p y g(x + d ) > (4.3) and HI: Ex = - iE 0 2 2 i 2 \ Yg +ns k I Y eP cos kxSx cos k j y + i5v j . (4.4) In Regions II and HI the field decays exponentially (evanescent field) with dis tance from the rib waveguide while the field in Region I is sinusoidal in both x and y. The constant terms in each equation are required so that the boundary conditions are satisfied for the Ex polarization. The electric field amplitude E0 is chosen arbitrarily. The evanescent field decay constants are given by y g = 4 n '2 ~ ng2)k 2 ' kx2 ’ (4.5) 135 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 o = J ( n l2 - no2)k 2 - ky2 ’ (4.6) and n I2 _ n 22) ^ 2 ~ ky2 (4.7) for the gap, cladding, and substrate regions, respectively. In the equations above, k is the wave number in free space, kx is the standing wave number in the jc-direction in the waveguide, and ky is the standing wave number in the y-direction in the waveguide. The reciprocal of each decay constant is a measure of the penetration depth of the evanescent field into the surrounding medium. For smaller differences in dielectric permittivities between the waveguide and gap, A e = n ,2-ng2, the field pene trates farther into the gap region as seen from Eq. 4.5. The deeper field penetration leads to stronger coupling between neighboring rib waveguides. The longitudinal propagation constant is given as in which ne g- is the effective refractive index of the waveguide. In general, several discrete values of the longitudinal propagation constant exist that may be used in Eqs. 4.2 through 4.4, each of which is a solution to the eigenvalue equations arrived at by the application of the boundary conditions (4.8) 2 n 2n 2kxy, I ,lg x f g (4.9) tan kxSx (4.10) 136 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4.11) and tan kySy - - ~ A .v 7 (4.12) The phase shift parameters 8X and 5y are required so that the boundary condi tions are satisfied. In addition, Eqs. 4.9 and 4.10 correspond to solutions in the x- direction and may be solved independently o f Eqs. 4.11 and 4.12, which correspond to solutions in the y-direction. Once the characteristic values for the longitudinal propagation constant /? and the electric field distributions are determined, the electric field for a given mode in the rib waveguide may be fully expressed by in which the angular frequency (0 - iTtclk, and c is the velocity of light in free space. In order to simplify the above equations, we assume the known physical param eters of the Ti:LiNb0 3 waveguides and determine the vertical field distribution of the rib waveguide structures. The refractive index at the surface of the waveguide was found in Chapter 3 to be rif = 2.2019 + 0.0137 = 2.2156, and the effective waveguide thickness is 1.53 fim, which also happens by chance in this case to be the 1/e depth of the Gaussian refractive index profile. With the values and assumptions stated above regarding the effective waveguide thickness, Eqs. 4.11 and 4.12 were solved for X = 6328 A. As expected, only a single vertical mode is permitted with the solution k y - 1.551 yum -1 and 8y = -0.962 jjm. These values were used in Eqs. 4.6 and 4.7 to find the values for y0 and y2. The final field profile in the vertical extent was then cal culated by substituting these values into Eq. 4.2. The electric field profile as a func Ex (x, y ,z ,t) = Ex {x,y) • exp[/(filf - /fe)] , (4.13) 137 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tion of depth in Region I was compared with the measured field profile for a planar single-mode Ti:LiNb03 waveguide, as obtained in Chapter 3, and was found to be nearly identical. The rib waveguides will support multiple transverse modes. The denominator of Eq. 4.9 was examined to determine the range o f values for kx that permit possible solutions for the eigenfunctions. From this range and the known rib waveguide width, the maximum number of modes can be estimated by This relationship indicates that rib waveguides with greater widths or a lower gap refractive index will support a greater number of transverse modes. With these simple expressions for the rib waveguide mode profiles, the coupling efficiency of the incident field distribution into the rib waveguide array can be described more accurately than the previous estimations made with the acceptance angle and rib waveguide fill factor, as described next. Actual calculations of the inci- dent-field coupling for two specific cases are presented later in Chapter 7 where the issues of system integration are discussed. 4.1.3 Incident Field Coupling The mode coupling efficiency from the planar portion of the waveguide to the rib waveguide array is determined by mode overlap integrals. In our theoretical treat ment of the rib waveguide field distributions, we found that several transverse modes may be supported by the rib waveguides. We also found that the field profile in the vertical direction is nearly identical for the rib waveguide and planar waveguide; hence, the overlap integral in the vertical extent is nearly unity. As a result, only the transverse field distribution needs to be considered to obtain an approximate measure R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 'max (4.14) 138 of the mode overlap. These discrete modes, in addition to the continuum of radiation modes, form an orthonormal and complete set o f wave functions. The field distribu tion incident upon the rib waveguide array can be represented by a modal expansion among this set with the associated mode amplitudes a mn (mth mode of the nth rib waveguide) determined by the overlap integral f E™ {x-nW)E*dist{x)dx J— o o a mn -oo 2 r — r E f { x ) d x (4.15) In this expression, Exm is the TE-poIarized electric field distribution for the nth rib waveguide as given by Eqs. 4.2 through 4.4 and E^ist is the TE-polarized transverse electric field distribution o f the light incident on the front of the rib waveguide array. The coupling efficiency into mode m of the nth rib waveguide is equal to \amnI 2. This expression will be used in Chapter 7 to study the incident field distribution coupling efficiency for the cases of a test module arrangement and of an IOSAR processor arrangement. In general, the field distribution incident upon the rib waveguide array will excite more than one waveguide mode. This excitation will occur unless the incident field exactly matches the field profile of one of the rib waveguide modes. In this case, the numerator in Eq. 4.15 will integrate to zero for all mode numbers m other than the matching mode profile. The multiple excited modes will propagate independently of one another along the length of the rib waveguide array, subject only to propagation losses due to scattering and perturbations due to surface outcoupling gratings. The guided mode attenuation due to scattering from the rib waveguide sidewalls is dis cussed next. 139 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.1.4 Rib Waveguide Scattering Losses The roughness in the rib waveguide sidewalls is a result of the rib waveguide fabrication process. The majority of the light scattered by the sidewall roughness will propagate into the substrate [Hewson-Browne, e ta i, 1989]. The light that propagates upward toward a surface-mounted device is likely to illuminate the pixels directly above the rib waveguide as well as the pixels above the adjacent rib waveguides, and hence lead to crosstalk. In the IOSAR processor, high rib waveguide scattering losses can also lead to non-uniformity in the illumination of the surface-mounted mask and CCD array along the length o f the rib waveguide. In the presence of a high scattering loss, the power in the guided mode is significantly depleted before it reaches the end of the rib waveguide. The grating outcoupled power, in this case, will drop in proportion to the power in the guided mode. In order to achieve uniform outcoupling, a low outcou pling efficiency is required so that 80-90% of the light remains in the rib after a 1 cm propagation length. Two possible ways to reduce the scattering losses in the rib waveguides are to improve the fabrication process to produce structures with minimal surface rough ness, or to reduce the refractive index difference between the rib waveguide and the gap region. In the work by Rastani, the fabrication o f rib waveguides (8 fjm wide ribs, 2 jum wide gaps, 1 fim gap depth) with highly uniform sidewalls was demon strated, yet the propagation losses were on the order o f 6 dB/cm. With this guided wave attenuation, 75% of the guided optical power is scattered from the waveguide over a 1 cm propagation length. Waveguide surface scattering is proportional to the difference in dielectric per mittivity A e between the waveguide material and the bounding medium [Hutcheson, 1987]. The attenuation coefficient due to surface scattering can be expressed in terms of the refractive indices of the materials as follows 140 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. a s o= A e = n,2 - n 2^ (4.16) in which 0 C S is the attenuation coefficient due to scattering and all of the other param eters were defined above. An increase in the refractive index of the gap region, there fore, will lead to lower waveguide scattering. This increase in the gap refractive index can be achieved by decreasing the gap depth between the rib waveguides. A direct consequence of doing this, however, is that there will be increased penetration of the guided mode through the gap region and hence an increase in the coupling between the adjacent rib waveguides, as is shown next. 4.1.5 Rib-to-Rib Coupling The transfer of optical power between two adjacent rib waveguides through the barrier gap region is a source o f crosstalk. A schematic representation of coupling between rib waveguides is shown in Fig. 4.4, in which rib waveguides of width d are spaced by a distance W. The degree of coupling is determined by the overlap of the electric-field evanescent tail from the mode in one rib waveguide with the electric - field distribution of the mode inside the second rib waveguide. The overlap integral is generally expressed by the coupling coefficient in which E ,x and E2x are the transverse field distributions of like modes of the first and second rib waveguides, respectively. For the pair o f coupled rib waveguides which z is the propagation direction along the rib waveguide. For rib waveguide (4.17) shown in Fig. 4.4, light that is initially within the first rib waveguide I is periodically coupled into rib waveguide 2 and back into rib waveguide 1 according to cos2 Kz, in 141 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.4 (a) Rib waveguide array geometry and (b) region of integration for cal culation of the coupling coefficient. arrays with low crosstalk, the coupling length I tfK is large compared with the rib waveguide length. The coupling coefficients for two-dimensional structures with abrupt interfaces have been determined previously through the use of mode-coupling theory [Kuz netsov, 1983; Hardy and Streifer, 1985; Kawakami and Haus, 1986; Haus, et al., 1987]. The expression derived by Kuznetsov [Kuznetsov, 1983] and used by Rastani [Rastani, 1988] in his analysis is also used here: In this equation, all of the parameters have been defined previously, and it is evident that the coupling coefficient decreases exponentially with gap width (W - d ). The functional dependence of the coupling coefficient on the other parameters such as the gap region effective index and the mode number, however, is not evident. To illustrate the dependence of this coupling coefficient on the refractive index ^ ^ n ^ fk /y g e x p l-iW -d ) ^ ^ ^ Pd k x2 +{n,/ng) j4yg 2 (4.18) 142 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. of the gap region ng, we assume that the gap region has a mean dielectric permittivity of £g = ng2 = ano 2 + C 1 “ aW . (4-19) in which a = h/b is the fraction of the surrounding waveguide film that has been removed. This assumption is similar to that made for the dielectric permittivity of relief gratings in grating outcoupling analysis [Tamir and Peng, 1977]. In the relief grating structure, the grating layer consists partially of the waveguide material and partially o f the cladding material. The use of this approximation in the grating analy sis is valid as long as the waveguide mode is only slightly perturbed by the gratings. Similarly, we can expect Eq. 4.19 to be accurate in our analysis o f rib waveguides for weakly bound transverse modes (a very small gap depth) in a rib waveguide with transverse symmetry. The transverse symmetry is essential in this assumption to ensure a smooth transition from an unbound to a bound state as the gap depth is per turbed from zero. Likewise, we can expect Eq. 4.19 to be accurate for a nearly fully etched gap depth. In this case, the lowest order rib waveguide mode is well confined in the rib waveguide by air gaps on either side. This rib waveguide mode is then per turbed if the gap depth is slightly less than the rib waveguide depth. The relation in Eq. 4.19 was used to determine ng for two gap depths (2500 A and 5000 A) for rib waveguides arrays fabricated in the planar Ti:LiNb0 3 waveguides described in Chapter 3. The effective waveguide depth b of 1.53 fim leads to values for ng of 2.066 and 1.906 for gap depths of 2500 A and 5000 A, respectively. These values were used to calculate the field distributions and propagation constants for rib waveguide array structures with 8 jim rib waveguide widths and various gap widths (W - d ) ranging from 2 to 10 /an. The required parameters for the coupling coeffi cient (Eq. 4.18) were then inserted for each guided mode. The calculated coupling coefficients are shown as a function of the waveguide transverse mode number for 143 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. several different gap widths in Fig. 4.5(a) (2500 A gap depth) and in Fig. 4.5(b) (5000 A gap depth). For the higher assumed value of ng (2500 A gap depth), approximately 15 trans verse modes are supported by the rib waveguide structure, compared with 2 0 modes approximated with Eq. 4.14. For the lower assumed value of ng (5000 A gap depth), approximately 24 transverse modes are supported by the rib waveguide structure compared with 28 approximated with Eq. 4.14. We can surmise that Eq. 4.14 gives only an approximate prediction of the number of supported modes, and that the num ber of modes increases with larger gap depths. In Fig. 4.5, the coupling coefficients increase with mode number for a given gap width due to increased penetration of the evanescent field through the gap region. For incident field distributions that excite these higher order modes, the crosstalk due to coupling between rib waveguides is increased. The rib waveguide array should be designed such that the highest order mode that is excited by the incident field distribution will still produce low crosstalk. By comparison of both graphs in Fig. 4.5, it is evident that the waveguide cou pling coefficient decreases strongly with increasing gap depths. From our previous considerations, we found that the associated decrease in the refractive index of the gap region will lead to higher rib waveguide scattering and a larger acceptance angle for the rib waveguide array. Furthermore, for a given gap depth, the coupling coeffi cient decreases for larger gap widths. Rib waveguide arrays designed with larger gap widths to achieve low crosstalk due to coupling will have a lower rib waveguide fill factor, and hence the incident field coupling efficiency will be reduced. The graphs in Fig. 4.5 illustrate the general behavior of the coupling coefficient formula described in Eq. 4.18. However, we do not expect the values for the coupling coefficients obtained through our assumption of the gap refractive index dependence on gap depth to have absolute accuracy. To more closely examine the coupling present in fabricated rib waveguide array structures, a special rib waveguide array was designed and fabricated on single-mode Ti:LiNbC> 3 waveguides. This rib 144 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. M °de N um ber (m icro n s) 10 15 20 F ig u re 4 .5 c„ . . M o < * e N um ber agaP * P ‘h o f S 0 0 0 A 145 Pe™ ssto" 0f» .ec0py^ h, waveguide array consists of several rib waveguide units that follow the pattern shown in Fig. 4.6. Each unit consists of a central rib waveguide that extends from one end of the substrate to the other, and a pair of adjacent rib waveguides on either side with length L c and separation distance that extend to only one end of the sample. The coupling length Lc is varied from one unit to the next and ranges from 0 mm to 12 mm in increments of 2 mm. Likewise, the gap width Wg is varied from one unit to the next and ranges from 2 fjm to 10 fim in increments of 1 jjrn. On a given substrate, these patterns are repeated twice for a total of 126 rib waveguide units. The rib waveguides were arranged in the manner shown so that the waveguide coupling can be observed for several interaction lengths, gap widths, and gap depths. Since the neighboring guides in each unit do not extend all of the way to the left hand side of the sample, the central rib waveguide can be selectively excited by end-fire coupling. The leakage of light from the central guide to the neighboring guides is observed at the right-hand side of the sample where all three rib waveguides end. The end-polished facet permits a relative measurement of the end-emission peak intensity from each guide. The fabrication issues and processes for these structures are described in the next two sections. Two samples were fabricated with 8 fjm wide rib waveguides and gap depths of 2500 A and 5000 A. In addition, the waveguide ends were polished so that light could be coupled in by the end-fire technique. The results of the experimen tal characterization of these structures are given in Sect. 4.3.6. 4.1.6 Waveguide Component Fabrication Techniques In the research described in this thesis, a number of waveguide components were fabricated including rib waveguide arrays, surface outcoupling gratings, and embedded lenses. In this section we consider some of the requirements and limita tions of the fabrication processes. We consider in particular the fabrication of the rib 146 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ti:LiNb0 3 Substrate T C * w 9 g \ : > - " . . v * : * - : ' . . ' t \ - " ■ \ Figure 4.6 One unit of a rib waveguide array pattern designed for the purpose of measuring crosstalk values between rib waveguides with different coupling lengths, gap widths, and gap depths. waveguide and grating structures described in this chapter. The general concerns cov ered here, however, also apply to the case of embedded lens fabrication. More details regarding embedded lens fabrication are given in Chapter 5. The rib waveguide and grating fabrication sequences generally consist of a pho tolithographic step used to create a photoresist mask on the waveguide substrate, and an argon-ion-beam dry etch step used to anisotropically etch the exposed regions of the waveguide substrate as shown in Fig. 4.7. The remaining differences between the various structures and materials are the choice of photoresist thickness to give the maximum protection and most accurate pattern transfer during the etch sequence, and the choice of etch time to achieve the desired device depth in the given waveguide substrate material. Both rib waveguide and grating structures require high uniformity, high resolu tion, and smooth patterns over a large area ( 1 cm X 1 cm). The selection of the photo lithographic process and the etching process is guided by these requirements. The region of required uniformity is small compared with the typical wafer han dling capabilities of photolithographic processing equipment (e.g., the Carl Suss 147 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Patterning of nb array with photoresist I AT 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 Patterning of gratings wth photoresist 1 AT I I l I I I I I I I i I I I I I I I I I I I t i I I I I I I I I I I Argon-tan miSng 1 Argocvton mU Bng I Removal of photoresist Removal of photoresist (a) (b) Figure 4.7 Fabrication sequence for (a) rib waveguides and (b) surface outcoupling gratings. These two structures may be fabricated in either order. MJB3 contact mask aligners used in the research described herein handle up to 3” wafers). As a result, standard photolithographic processes can be used to pattern a photoresist layer with excellent definition and uniformity. The standard photolitho graphic process consists of the application o f photoresist by a spin-on method, photo resist curing to remove excess solvents, patterned exposure of the photoresist with ultraviolet light, and chemical development to remove the UV exposed photoresist. Uniformity can be achieved with this process as long as the device is patterned suffi ciently far from the sample edge. Some space at the sample edge is preferred since the photoresist thickness is non-uniform within 1-2 mm of the substrate edge due to a characteristic of the photoresist spin-on process called edge-beading. The smallest feature size that may be patterned in photoresist is on the order of the thickness of the photoresist. The exposure wavelength of the mask aligning sys- 148 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tem will limit the smallest feature size due to diffraction effects. For the MJB3 aligner, a minimum feature size o f approximately 1 jmv is possible with a sub-micron thick photoresist. The edge definition and uniformity in the photoresist is determined by the smoothness of the photomask. The photomasks for the waveguide devices dis cussed herein consist of electron-beam-written chrome-on-glass substrates with a res olution of 0.1 /un. The roughness transferred to the photoresist is typically smaller than 0.1 fim in amplitude, since the photoresist chemical developing process tends to smooth out this roughness in the patterned photoresist on the sample surface. For planar device fabrication, either wet chemical etching, dry etching, or reac tive etching may be suitable depending on the materials and device requirements. Lithium niobate is resistant to most acids and does not lend itself well to chemical etching techniques. Dry etching techniques such as argon-ion-beam etching (IBE) and reactive ion beam etching (RIBE) are more suitable for this material [Matsui, et al., 1980; Zhang, eta l., 1984; Huang and Lee, 1986; Belanger and Yip, 1987]. These dry etching methods are anisotropic, insensitive to the crystal structure, and can result in a very accurate pattern transfer from the overlaying mask. Gallium arsenide is reactive to a number of chemical etches, most of which etch preferentially along certain crystallographic directions. Dry etching techniques are also useful to produce accurate anisotropic etching in GaAs. The physical sputter rate is about three times that of lithium niobate, which leads to simplified device fabrica tion. In the research described herein, we used argon-ion-beam etching to define the rib waveguide and grating structures in the waveguide substrate material. The ion- beam etching was done with an Ion-Tech argon-ion gun capable of ±5% etch depth uniformity over a 3 cm diameter region. The etch depth uniformity o f the rib and grating structures is typically better than ±3% over a 1 cm x 1 cm region. Even greater etch depth uniformity can be achieved if sources with larger beam diameters are used. 149 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. The anisotropic etching provided by the argon-ion-beam etching process leads to very accurate pattern transfer compared with isotropic chemical and dry etch pro cesses [Nishihara, et al., 1989]. In isotropic processes, the substrate material tends to be etched under the photoresist mask edge. This undercut leads to a shrinkage in width of the device features. Therefore, the resolution that can be realized by these techniques is limited. There are also limiting factors in the achievable resolution with the argon-ion- beam etching process that are related to the etch depth. The depth to which pattern transfer accuracy is maintained in the argon-ion-beam etching process depends upon the substrate and mask materials, the mask sidewall profile, and the mask thickness. To describe these dependencies, several terms related to the argon-ion-etch pro cess are defined next. During the etch process, both the mask material and the exposed substrate are etched. The selectivity of a mask is defined as the ratio of etch rates of the substrate material to the mask material. For a mask with low selectivity, the etch process sometimes results in the lateral erosion of the mask. This mask ero sion is referred to as mask shrinkage and is illustrated in Fig. 4.8(a). For high mask selectivities, the mask shrinkage is slowed and accurate pattern transfer can be attained. In such cases, a bevel can still form in the mask (see Fig. 4.8(b)). A bevel is a facet that develops in the edge of the mask with an angle that corresponds to the angle o f incidence of the ion beam that produces the highest etch rate in the mask material. The effective mask thickness is a measure of the thickness of the mask mate rial that can be etched away before the bevel reaches the substrate surface. At that point in the process, mask shrinkage will again take place [Cantagrel, 1975; Lee, 1980]. If the mask is made sufficiently thick and has a high selectivity, an inaccurate pattern transfer can still result from redeposition of the substrate material onto the recess sidewall (see Fig. 4.8(c)). Redeposition is an effect whereby material sputtered from the substrate surface impinges on the recess sidewall and adheres to it. The physical properties of the substrate material will determine the rate of redeposition 150 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 4.8 Shown are three mechanisms that lead to a tilted sidewall and inaccurate pattern transfer in the argon-ion-beam etching process: (a) low mask selectivity results in mask shrinkage during the etch process, (b) bevel formation in the mask eventually reaches the substrate and also results in mask shrinkage, and (c) the effects of redepo sition result in a tilted sidewall even though the initial mask is ideal. that occurs during the etching process. The redeposition leads to a net decreased etch rate for the sidewalls, and consequently the sidewalls become tilted. For both lithium niobate and gallium arsenide, this effect is noticeable for etch depths greater than about 0.5 /Jin. These artifacts of the argon-ion-beam etching process have implications for the fabrication of deep embedded lens recesses (-3 fjm ) as discussed in Chapter 5, and must also be considered in the fabrication of rib waveguides with narrow and deep gap regions. It was shown by Rastani [Rastani, 1988] that photoresist could be used effectively as an etch mask for rib waveguides with two micron wide gaps etched 1 ftm deep. The surface roughness of those rib waveguides was very minor and could not be identified even when observed at high magnifications with a scanning electron microscope. To obtain this degree of sidewall smoothness, it is essential that the 151 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. argon-ion-beam milling be performed at a reduced rate to prevent heating o f the sam ple and reflow of the photoresist. Photoresist reflow occurs when the photoresist is heated above its softening temperature of about 120° C. If this occurs, the photoresist profile becomes rounded and the edge definition becomes distorted. 4.1.7 Rib Waveguide Array Fabrication on Lithium Niobate A general process for the fabrication of rib waveguides on a multi-mode, planar Ti:LiNb03 waveguide and a measurement of the guided-wave attenuation of these structures were given in an earlier work by Rastani [Rastani, 1988]. That investiga tion has been extended herein to include fabrication of highly uniform rib waveguide arrays with various rib waveguide widths, gap widths, and propagation lengths in sin- gle-mode Ti:LiNb03 planar waveguides and GaAs/AlGaAs waveguides. The fabri cation of rib waveguides on GaAs/AlGaAs planar waveguides will be described later in Sect. 4.3 along with the grating fabrication process in GaAs. Rib waveguide arrays with 8 fim rib waveguide widths and 2 fim gap widths etched to a depth of 5000 A were fabricated in single-mode Ti:LiNb0 3 waveguides. These arrays consisted of 1000 elements of 5 mm length. Similarly, rib waveguide arrays with 28 /an rib widths and 2 fjm gaps etched to a depth of 5000 A were fabri cated in single-mode Ti:LiNb03 waveguides. These arrays consisted of 333 elements of 5 mm length. For all of the device fabrication in lithium niobate described here and in the later chapters, the single-mode planar Ti:LiN b03 waveguides described in Chapter 3 were used. The waveguide strips (1” wide X 3” long) were sliced, scribed, and cleaned according to the process outlined in Chapter 3 before the rib waveguide array process ing sequence was started. The samples prepared for the rib waveguide arrays were sliced as shown in Fig. 4.9(a), while those prepared for the rib waveguide crosstalk measurements were sliced as shown in Fig. 4.9(b). After the final DI water rinse in 152 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. I inch 1 7 cm (a) (b) Figure 4.9 Schematic diagram of grid patterns used to dice wafers for component fabrication. The wafer dicing pattern in (a) produces samples approximately 1 in X 1 in and is used for waveguide device fabrication, and in (b) produces samples approximately 1.7 cm X 1.7 cm for use in the fabrication of the rib waveguide crosstalk samples. the cleaning process, the samples were dried with nitrogen and dehydration baked on a hot plate at 120° C for 2 min to improve the photoresist adhesion to the waveguide surface during the photolithography and etch steps. A layer of AZ5214-E positive photoresist was first applied to the sample surface by the spin-coating technique to a thickness of 1.4 fim. This was obtained by spin ning the resist on at 4 krpm for 30 seconds. With this thickness, the rib waveguide array may be accurately and uniformly patterned into the photoresist, and the photore sist will act as an adequate mask for the accurate transfer of the pattern into the planar waveguide by ion-beam etching. The samples were then soft baked on a hot plate at 120°C for 2 minutes to remove excess solvents from the photoresist. The hot plate temperature is stable to ±0.5° to provide repeatable results. This method is preferable to baking the samples in a convection oven since it reduces the processing time signif icantly (2 min. as opposed to 30 min. for the convection oven). Both the rib waveguide array photomask and the rib waveguide crosstalk photo- 153 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. mask were obtained through the Army Research Laboratory in Adelphi, MD. For exposure o f the photoresist through either mask, the samples were always oriented such that the rib waveguides would run along the crystallographic jc-axis of the sub strate. The exposure time was typically 7 sec on the Carl-Suss MJB3 mask aligner with an exposure intensity setting o f 6.8 mW/cm2 at the 365 nm exposure wave length. Development was performed in 400K developer diluted 1 part to 4 parts DI water. The samples were immersed in this solution for approximately 40 sec, after which they were immersed in DI water and then blown dry with nitrogen. After development, the patterned photoresist was observed under an optical microscope and was found to have very smooth and sharp edges. A final UV blanket exposure for 2 min on the mask aligner was performed after the photoresist development step to facilitate the removal of the photoresist with solvents after the ion-beam-milling pro cess. Once the samples were patterned, they were prepared for argon-ion-beam mill ing. This process was performed in a vacuum system that sustains a vacuum of 2 X 10' 7 Torr. The sample platform used in this chamber is a copper water-cooled stage that can be tilted along either axis and was normally leveled with respect to the ground. The ion beam source was an Ion-Tech 3 cm ion-gun with standard pyrolytic- graphite grids. The ion-gun was inverted above the mount and directed toward the mount center (normal-incidence ion beam milling). A calibrated fiducial was used to position a patterned waveguide sample on the mount in the region o f greatest milling uniformity. To perform the milling process, a sample was attached to the mount with Mung II, a thermally conductive adhesive made for vacuum processing, and exposed portions of the copper mount were shielded with graphite sheets to protect them from the ion beam. The argon-ion-beam milling was accomplished using a 5.0 seem flow rate of high-purity argon (99.999% pure) into the ion source at a chamber pressure of approximately 5 x 10' 5 Torr, a beam voltage of 750 V, a beam current of 30 mA, and 154 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. an accelerator voltage of 500 V. The distance between the gun orifice and the sample was about 25 cm and the ion current density was 1 mA/cm2. Samples milled under these conditions for 30 minutes resulted in 5000 A etch depths. Further experimenta tion at these milling parameters for different time durations indicated that the milling rate was approximately 1 ^m/hour. The water-cooled stage was found to provide ade quate cooling to prevent photoresist reflow or burning even with continuous operation of the ion gun. After each sample was ion beam milled, it was removed from the vacuum and the photoresist layers were dissolved with acetone. Dilute soapy water was used along with slight agitation with a cotton swab to remove any residual particles that were adhered to the surface of the samples. The samples were then rinsed in DI water and blown dry with nitrogen. The rib waveguide arrays were normally patterned at one end of the sample so that a guided mode could be prism incoupled into the planar waveguide region and directed into the rib waveguide array. The rib waveguide arrays were also patterned in this manner to leave room on the planar portion of the sample to integrate other devices for demonstration of the test module operation. The rib waveguides etched for crosstalk measurements, on the other hand, extended from one end of the sample to the other as described earlier. A piece of each rib waveguide crosstalk sample was sliced off with a wire-saw approximately 1 mm from the end where the rib waveguides met the sample edge. The samples were sliced in this manner in order to remove the portion of the etched rib waveguides that is non-uniform due to the photoresist edge-beading effect men tioned earlier. Both ends of these samples were then end-polished with the process described in Sect. 3.7 to facilitate coupling into and out of the rib waveguides. The rib waveguide array fabrication process described above is different than that described by Rastani. In the work by Rastani, a 3 cm ion beam gun was used in 155 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. conjunction with an aluminum substrate holder that was not water cooled and was approximately 15 cm from the gun orifice. The photoresist patterned samples etched therein were subject to heating. To circumvent reflow or burning of the photoresist during the etch process, the samples were first hard-baked in a convection oven at 140° for 2 hours to temperature stabilize the photoresist. The argon-ion-beam etching was then conducted at low settings (150 V beam voltage, 7 mA beam current, 0.1 mA/cm2 ion current density) to maintain a low sample temperature [Rastani, 1988]. The use of a water-cooled substrate holder in the process described herein has alleviated these requirements. In addition, the larger spacing of the sample from the gun orifice (25 cm in the present work compared with 15 cm in the work by Rastani) has yielded improved etch depth uniformity. 4.1.8 Acceptance Angle M easurem ent The value of the effective refractive index of the gap region between the fabri cated rib waveguides can be determined to first order through measurement of the rib waveguide acceptance angle. We attempted to measure the acceptance angle in a sample with 8 fjxa wide rib waveguides and a 5000 A gap depth. The sample was attached to a prism coupling mount at an angle with respect to the incoupling prism as shown in Fig. 4.10. A narrow collimated beam from a 15 mW He-Ne laser (A = 6328 A) was prism coupled into the waveguide and directed toward the rib waveguide array at various angles of incidence. An outcoupling prism was mounted on the rib waveguide array to observe the rib waveguide excitation. It was found that the rib waveguides could be excited by the input beam for angles of incidence as large as 24°. For angles larger than this, it became difficult to mount the prisms and sample due to the physical constraints of the prism coupling mount. This approach to mea surement of the rib waveguide acceptance angle has the limitation that the rib waveguide excitation is not abruptly cut-off when the acceptance angle is exceeded by the input beam. However, it was noted that the observed excitation efficiency 156 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Incoupling Prism Outcoupling Prism He-Ne Beam Sample Waveguide Array Figure 4.10 Prism coupling arrangement for measurement of the acceptance angle of a rib waveguide array. started to drop for angles larger than about 10°. From our approximation of the gap refractive index made with Eq. 4.19, the acceptance angle should be about 30°. A fairly rough estimate of the gap refractive index made by this measurement (assuming 9a = 10°, n e ff= 2.2069, n , = 2.2156 and with the aid of Eq. 4.1) is ng = 2.182. Later in Sect. 4.1.10, we will show that this is fairly close to the value obtained by measure ment of rib-to-rib coupling. The reason for the inaccurate value for the acceptance angle obtained through the use of Eq. 4.19 will also be discussed. With this experimental configuration and an incident beam angle of 0°, the guided streak of the light coupled into the rib waveguides could be easily discerned. The surface scattering of the rib waveguides provides a convenient approach to the measurement of propagation losses as described in Chapter 3. In the next section, we describe the measurement of scattering losses in rib waveguides fabricated in single mode Ti:LiNbC> 3 waveguides. 157 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 4.1.9 Rib Waveguide Scattering Loss Measurement The propagation losses of a rib waveguide array (8 fjm rib waveguide widths, 2 fim gap widths, 5000 A gap depths) fabricated on a single-mode Ti:LiNb0 3 waveguide were measured with the CCD imaging technique described in Chapter 3 and the experimental arrangement described above. Over a streak length of approxi mately 2.5 mm, the surface scattering dropped off by 0.4 dB, which indicates a propa gation loss rate of 1.6 dB/cm. This loss rate is lower than that found by Rastani (6 dB/cm) partially due to the lower losses of the initial planar Ti:LiN b03 waveguide used in the research described herein (0.41 dB/cm compared with approximately 2 dB/cm measured by Rastani). The remaining difference in the measured loss rate can be attributed to the differences in the rib waveguide array structure, as we describe next. The attenuation constant due to rib waveguide sidewall scattering is propor tional to the difference in the squares of the indices on either side of a waveguide interface (Eq. 4.16). The approximate refractive index of the gap region ng obtained with Eq. 4.19 substituted into Eq. 4.16, leads to the relative expression a{n,2-n 02) for the scattering that occurs for different gap depths. This expression can be interpreted as the scattering that arises from the portion of the waveguide thickness a that con sists of an air/waveguide boundary. This portion a is equal to unity {i.e., 100% of the waveguide height is bounded by air in the rib waveguide gap region) for the rib array fabricated by Rastani [Rastani, 1988] and is approximately 0.32 (0.5 /zm/1.53 ^m) for the rib waveguide array described above. The attenuation constant due to scattering of the latter rib waveguide structure can be calculated from the scattering loss for rib waveguides measured by Rastani (4 dB/cm) to give 0.32 x 4 dB/cm = 1.28 dB/cm. The total attenuation constant is given by 1.28 dB/cm + 0.41 dB/cm (planar Ti:LiNb03 waveguide) = 1.69 dB/cm. This estimated value agrees very well with our measured value of 1.6 dB/cm. As a result of this lower measured propagation loss, the illumination nonuniformity of a surface-mounted device over a 1 cm rib 158 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide array length is improved to a 30% drop in outcoupled light (under the assumption that a weak but uniform outcoupling grating is used). Consequently, the fabrication of a variable height grating structure for improved illumination uniformity becomes much more practical since not nearly as much light is lost to scattering. A reduction in rib waveguide scattering will also reduce the amount of crosstalk present at the surface-mounted device. 4.1.10 Rib-to-Rib Coupling M easurement In our theoretical analysis of rib waveguides, we found that there may be a trade-off between rib waveguide scattering and rib waveguide crosstalk due to tunnel ing that is inherent in the selection of the gap depth. The lowered rib waveguide attenuation due to scattering for shallow-etched rib waveguide gaps was confirmed in the last section compared with the loss value obtained by Rastani for deeply-etched gaps. In this section we present the results of the rib waveguide coupling measure ments. From these measurements, we show that there is an increase in the coupling constant for smaller gap depths and gap widths. In the final design of an advanced 10 signal processor, the trade-off in these two factors as well as the outcoupling unifor mity must be carefully considered. Two samples were prepared for these measurements as described in Sect. 4.1.7. Each sample consisted of 8 fim wide rib waveguides with various gap widths and coupling lengths. One sample was etched to a depth of 2500 A and the other to 5000 A. The rib waveguides are arranged in units of three as previously described. Each unit is spaced from the next by 50 fim so that they are sufficiently isolated from one another. A single rib waveguide from each unit extends to both ends of the sam ple. Light from a 5 mW He-Ne laser (A = 6328 A) was end-fire coupled into one of these rib waveguides and the end-emission from the three rib waveguides at the oppo site end of the sample was observed in order to measure the waveguide coupling. The peak measured intensity in the central rib waveguide was compared with the peak 159 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. intensity in the neighboring rib waveguides. The log of the ratio of these peak inten sities was then calculated to determine the degree of isolation of the guides. This measurement was repeated for nearly all of the units on each sample. For some units, the rib waveguides were slightly chipped at the polished waveguide sample edge and hence an accurate measurement of the coupling could not be made. Since each unit v/as duplicated twice on each sample, the occurrence of damage in both o f the identi cal units was small. In all of these measurements, the lowest-order mode was excited in the central rib waveguide. Hence, the measurements of the rib waveguide coupling that are presented here apply only to the lowest order mode coupling. Since the adjacent rib waveguides in each unit do not extend all of the way to the end of the sample where the incoupling is performed, they are less likely to be excited by the incoupling beam. This assumption is especially true for shorter inter action lengths since the distance from the incoupling objective is greatly increased. The longest coupling length Lc is 12 mm and the sample length is 14 mm. Therefore, some of the adjacent rib waveguides come within 2 mm o f the incoupling end of the waveguide and are subject to slight excitation from the focused incoupling beam. To excite the central rib waveguide in a particular unit, the numerical aperture of the Ti:LiNb03 rib waveguide was matched as described in Sect. 3.6.2 with a I mm input beam focused by a 60x objective. Variable neutral-density optical attenuators were used to control the amount of input light. The end-emission from the rib waveguides was imaged onto a CCD array with a 40x objective. A beam-splitter cube was placed between the objective and CCD array so that the relative power level of the end-emission could be monitored with a detector. The power measured with this detector was typically between 1 nW and 1 fj.W. The signal from the CCD cam era was then synchronized to an internal trigger on a Hewlett Packard oscilloscope so that a line scan of the end-emission intensity profile could be displayed. The auto matic gain and bias controls of the CCD array were turned off. The rib waveguide coupling was then measured. First, the input optical power was adjusted to provide a 160 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. clear signal on the CCD array from the central rib waveguide end-emission and to bring the CCD response voltage up to a reference level (chosen to be about 3/4 o f its full range voltage). The input power was then increased with the variable attenuators to bring the signal voltage of the highest sidelobe up to the same reference voltage level. The relative separation of the peak and sidelobe intensity in dB was then calcu lated through comparison of the optical power from the end-emission incident on the detector for the two different input powers. The results of the measurements on these two samples for a 10 mm interaction length and various gap widths are shown in Fig. 4.11. The presence of significant end emission from the adjacent rib waveguides for small gap widths in both of the sam ples indicates the presence of coupling between the rib waveguides. The optical power in the central rib waveguide is proportional to cos\ kL c) and in the neighbor ing rib waveguides is sin2(KLc) after a coupling length Lc [Saleh and Teich, 1991]. The coupling coefficient is approximately related to m, the measured value in Fig. 4.11, by 20 log(tan(fcLc)) = m, in which m is measured in dB, Lc = 1 cm, and K has units o f cm-1. For gap widths of 2 jum on these two samples, the calculated cou pling coefficient is 0.35 cm-1 and 0.14 cm-1 for 2500 A and 5000 A etch depths, respectively. From our experiments, we found that the coupling decreases for larger gap widths as expected. For gap widths that lead to a rib waveguide crosstalk of less than approximately -22 dB, the intensity level of the light in the neighboring rib waveguides falls below the background illumination level caused by light coupled into the substrate. We found it difficult to obtain an accurate measure of the crosstalk below this level with this experimental arrangement. Another concern in this experiment is that of inadvertent excitation of the neigh boring rib waveguides from the end-fire coupling to the center rib waveguide. How ever, the drop in the power measured in the end-emission of the neighboring rib waveguides for larger gap widths indicates that the rib-to-rib coupling is responsible 161 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 0 -5 33, -10 q T t -1S T O O - J -20 o -25 -30 0 1 2 3 4 5 6 7 8 9 10 11 12 Gap Width (microns) Figure 4.11 Measured dB separation between the peak intensity level of a single excited rib waveguide and the peak intensity level of the neighboring rib waveguides for a 10 mm interaction length and various rib waveguide gap widths. for the neighboring rib waveguide excitation and not the end-fire coupling. For the measurements shown in Fig. 4.11, the interaction length is 10 mm and the sample length is 14 mm, hence the neighboring rib waveguides start 4 mm from the waveguide edge where the end-fire coupling occurs. For such a large distance, the excitation of the neighboring rib waveguides should be insensitive to the small change in gap width and the measured power in the neighboring guides would be nearly constant in the absence of coupling between the rib waveguides. The nearly linear drop-off on the logarithmic scale in the coupling with an increase in gap width for both samples is indicative of the exponential dependence of the coupling coefficient on the gap width indicated in Eq. 4.18. We have shown here that the rib waveguide crosstalk can be reduced through fabrication o f a rib waveguide array with either a greater gap depth or a greater gap width. The degree of crosstalk due to coupling also depends upon the rib waveguide 162 1— I r Interaction length = 10 mm • Gap depth = 2500 A □ Gap depth = 5000 A a S a • • • a □ J I I L R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. interaction length, as previously noted. The coupling between the central and neigh boring rib waveguides of units that have 2 fim gaps and various interaction lengths is shown in Fig. 4.12 for both the 2500 A gap depth and 5000 A gap depth. For both samples, the coupling increases with increasing interaction length. The theoretical channel crosstalk due to coupling is approximately equal to 20 log(tan( kLc)). This function is plotted in Fig. 4.12 for the samples with both 2500 A and 5000 A gap depths. For both samples, the measured crosstalk drops for shorter interaction lengths. However, this drop in measured crosstalk is not as fast as that predicted by theory for the shortest interaction lengths (less than 4 mm). This departure from the ory may be due either to the background noise from substrate illumination or contri bution to the coupling from some other mechanism such as sidewall scattering. The measured coupling coefficients are much larger than the values obtained in our example of the rib waveguide coupling trends (see Fig. 4.5), in which the gap refractive index was estimated from the average dielectric permittivity in the region. However, since we have a model for the rib waveguide field distributions, the gap refractive indices required to produce the measured coupling constants can be calcu lated. The rib waveguide mode parameters were calculated with several different assumed values for the refractive index of the gap region to determine the values required to produce coupling coefficients with Eq. 4.18 that are commensurate with the measured coefficients o f 0.35 cm-1 for the sample with a 2500 A gap depth and 0.14 cm-1 for the sample with a 5000 A gap depth. Gap refractive indices of 2.2052 and 2.199 were found to correspond to these two cases, respectively. Earlier, we assumed that the refractive index in the gap region was given by the mean dielectric permittivity, which was determined by the gap depth. In the limit of large gap depths (comparable with the waveguide thickness), this estimation may be fairly accurate since there will be very little penetration of the mode evanescent field into the gap region and the refractive index is approximately that of air. In the oppo site extreme of small gap depths, however, the penetration of the rib waveguide mode 163 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 0 -5 □ o ' ~ -10 C L a g 5 -15 - i o i— -20 -25 0 2 4 6 8 10 12 14 Interaction Length (mm) Figure 4.12 Measured dB separation between the peak intensity of a single excited rib waveguide and the peak level of the neighboring rib waveguides for 2 fim gaps and various rib waveguide interaction lengths. evanescent field into the gap region can be rather large and it is not as clear what value for the refractive index in the gap region should be used in our approximate analysis. In this case, a different method is required to make this estimation. For small gap depths, the gap region consists o f a thinned Ti:LiNb03 waveguide. If the thinned waveguide still supports a guided mode, the gap refractive index can be estimated by the effective refractive index of the guided mode. This method for estimation of the refractive index of the bounding material in a 2-D waveguide is called the effective index method [Bums and Milton, 1975]. The effec tive refractive index was calculated for the single-mode Ti:LiNb03 waveguide under the assumption that the fabricated waveguide was subsequently thinned by various amounts. We performed these calculations with the matrix method for waveguide evaluation described in Chapter 3. The refractive index profile was assumed to be Gaussian with a diffusion depth of 1.53 fj.m and a refractive index change of 0.0137 T Calculated (k = 0.35 cm'1 ) Gap width = 2 pm • Gap depth = 2500 A □ Gap depth = 5000 A Calculated (ic = 0.14 cm'1 ) • i □ j o - ' " a / Jcl L 164 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. as described earlier. The effective refractive indices for the gap region obtained by this method were 2.2053 (for a 2500 A gap depth) and 2.2038 (for a 5000 A gap depth). Note that the calculated effective refractive index of the gap region for the 2500 A gap depth is very close to the value derived from the experimental data obtained from the crosstalk measurement, whereas the calculated effective refractive index of the gap region for the 5000 A gap depth is somewhat higher than that obtained by the crosstalk measurement. The effective index method will give a less accurate estimate of the gap refractive index as the thinned Ti:LiNbC> 3 waveguide approaches cut-off. For our Ti:LiNbC> 3 waveguides, this will occur at a gap depth of approximately 8000 A. As the gap depth approaches this value, the refractive index of the gap region will tend towards the lower values obtained through calculation of the mean dielectric permittivity. If the gap refractive indices obtained by the measured coupling constants are representative of the fabricated rib waveguide array structure for the two gap depths that were considered, other properties of the rib waveguides may be calculated from these values such as the transverse mode profiles and acceptance angles. For a gap refractive index of 2.199, which corresponds to the 5000 A etch depth, the 8 fim wide rib waveguides will support approximately 7 transverse modes (from Eq. 4.14) and will have an acceptance angle of approximately 7° (from Eq. 4.1). This rib waveguide acceptance angle should be sufficient for both the IOSAR processor and IO correlator applications. However, we found in Sect. 4.1.5 that the rib waveguide coupling increases for higher waveguide mode numbers. Therefore, rib waveguide arrays with 8 fim rib waveguide widths, 5000 A gap depths, 2 fim gap widths, and I cm lengths may experience crosstalk greater than -17 dB (measured for the lowest order mode) if the higher order modes are excited. In order to prevent inordinately high coupling between rib waveguides in this case, the gap depth should be made deeper or the gap width larger. W ithout an accu rate model or experimental data to describe the rib waveguide coupling constant for 165 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. gap depths greater than 5000 A, the gap depth required to provide low rib waveguide crosstalk values for all of the supported modes is difficult to accurately predict. We expect, however, that the rib waveguide coupling should rapidly decrease for gap depths greater than 8000 A, which corresponds to the mode cut-off depth of the thinned waveguide in the gap region. 4.1.11 Rib Waveguide A rray Summary In this section, we considered the different incident field distributions and the associated rib waveguide array requirements for the IOSAR processor, IO correlator, and test module. To efficiently dissect the incident field, we found that the rib waveguides should have a high rib waveguide fill factor and a sufficiently large acceptance angle. For a high density array, such as that required for the IOSAR pro cessor, this implies that the gap separating the rib waveguides should be narrow and deeply etched. However, a deeply etched gap region can lead to excessive rib waveguide mode attenuation due to scattering from the rib waveguide sidewalls. This scattering loss can lead to nonuniformity in the illumination of surface-mounted devices for long rib waveguide lengths (1 cm) as well as crosstalk due to the genera tion of stray light that may illuminate neighboring pixels in the surface-mounted devices. The scattering losses are reduced for rib waveguides with shallow gap depths. We theoretically and experimentally determined that the amount of light cou pled between rib waveguides increases for shallow gap depths and narrow gap widths. A trade-off between external element illumination uniformity and crosstalk due to rib waveguide coupling has been identified that is inherent in the selection of the rib waveguide gap width and depth. We have experimentally measured the crosstalk due to rib waveguide coupling. The measured rib waveguide crosstalk of the lowest order mode for an array of 8 /im wide rib waveguides with 2 /am gap widths, 5000 A gap depths, and 1 cm interaction lengths was -17 dB. In advanced IO processors, the desired level of crosstalk should 166 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. be much lower than this (of order -4 0 dB). It has not yet been fully established that crosstalk due to rib waveguide coupling is the limiting crosstalk mechanism in these structures. We do know, however, that this crosstalk can be reduced as was noted above. It was previously shown by Rastani that gap depths as large as I fim can be achieved with photolithographic patterning and dry etching techniques. Although such large gap depths can lead to vastly improved rib waveguide isolation, the increased rib waveguide attenuation is problematic for the IOSAR processor perfor mance. This situation is tenable in the case that the rib waveguide sidewall roughness can be reduced by improvements in the fabrication process. The experimentally measured crosstalk values due to rib waveguide coupling were used in a simple theoretical model of the rib waveguide mode properties to esti mated the rib waveguide gap region refractive indices for two gap depths. These esti mated values were in good agreement with those obtained by calculation o f the gap region refractive index by the effective index method. The mode properties of rib waveguide arrays with 8 fim wide rib waveguides, 2 fim wide gaps, and a 5000 A . gap depth were estimated based upon these measurements. We determined that these rib waveguides would support approximately 7 transverse modes and have an acceptance angle of 7°. This acceptance angle is sufficient for both the IOSAR processor and IO correlator, but is too small for the test module. In our characterization of the test modules, discussed later in Chapter 7, we take this mismatch in the input field and rib waveguide properties into account in our evaluation of the test module performance. 4.2 Surface Outcoupling Gratings Surface outcoupling gratings are required in the advanced IO signal processors described in this thesis to outcouple light from the rib waveguide arrays to external surface-mounted devices. Coupling between confined waveguide modes and radia- 167 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tion modes takes place in the presence of a periodic perturbation in the dielectric per mittivity along the direction of wave propagation AeHy, z). Both rigorous [Marcuse, 1976; Chang, etal., 1980] and approximate [Ogawa, e ta i, 1973; Handa, etal., 1975] theories have been established to describe this coupling. Several investigators have used these theories to analyze gratings for both input and output coupling [Nishihara, et al., 1989]. Similarly, the issues of grating coupling to external detectors have been previously considered [Huang and Lee, 1986]. Large-area, nonuniform gratings have been studied for use in externally focusing an outcoupled beam for a optical disc pick up head [Suhara and Nishihara, 1986], and to outcouple and collimate the light pro duced by a laser diode source integrated on a GaAs substrate with a waveguide and ion-etched waveguide lens [Hirata, etal., 1993]. Herein, we consider grating outcoupling from both Ti:LiNb03 waveguides and GaAs/AlGaAs waveguides. The theoretical description of grating outcoupling pre sented next is applicable to either waveguide material. 4.2.1 Spatial Harmonics W hen light confined in a planar waveguide propagates into a region with sur face outcoupling gratings, a portion of the optical power is coupled out of the waveguide and into several free-space radiation modes. The radiation modes from surface outcoupling gratings on a planar waveguide propagate in both the cladding (air) and substrate (LiNb03) medium and are referred to as spatial harmonics. These radiation modes have characteristic propagation constants fiq [Nishihara, et al., 1989] given by nck sin 0 ^ = nsk sin 9 ^ = f3q = ne ^ k + qK , q = 0,±1, ±2, ±3,K (4.20) In this expression, 6q c^ and 6q^ are the propagation angles of the outcoupled light in the cladding and substrate, respectively, measured in the plane formed by the normal 168 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Rutile Coupling Prism Radiation mode e j c> out Slab ^ Waveguide Slab J Coupled Mode <3 c-axis out LiNb0 3 Substrate Figure 4.13 Schematic diagram of surface outcoupling gratings on a planar waveguide. The light is coupled into the slab waveguide via a rutile prism. The mode excited in the planar waveguide is partially outcoupled by the rectangular surface out coupling gratings. The parameters that determine the performance of the gratings are also shown in this figure (after [Rastani, 1988]). to the waveguide surface and the propagating guided wave. The angles are measured with respect to the normal to the waveguide surface and to the direction of the propa gating guided wave (see Fig. 4.13), is the propagation constant for the q\h spatial harmonic, K = lit!A is the grating wave number, and A is the grating period. Since nc < ns < ne jy< rif, the spatial harmonics are limited to order q < ~ 1. If the grating region is large compared with the wavelength of the outcoupled light, the radiation modes can be approximated as plane waves. In Eq. 4.20, n e f f is the effective refrac tive index of a planar waveguide in the presence of the gratings. The value of ne^ is approximately the same as the effective refractive index for a guided mode in the same waveguide without gratings [Nishihara, et al., 1989]. The relation given above in Eq. 4.20 is used later in this chapter to calculate the number of radiation modes and their propagation directions for grating on both lithium niobate and gallium arsenide waveguides. The important issue raised here is that light can be coupled into both the 169 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. cladding region and substrate region. From Eq. 4.20, the acceptable values of fiq for the cladding modes fall in the range - n jc to - ‘ rnjc, while the acceptable values of fiq for the substrate modes fall in the range -n jc to +njc. Since nc < ns, there will always be as many or more substrate radiation modes than cladding radiation modes. Later we will show that it is unfavorable to have light coupled into the substrate for the advanced IO processor configurations discussed in Chapter 2, and will discuss approaches to minimize this effect. 4.2.2 G rating Attenuation Constants For a uniform grating depth, the outcoupled light intensity decreases along the length of the ribs since the power in the guided mode is continuously coupled to the radiation modes. From a coupled mode analysis [Nishihara, et al., 1989], the power in the guided mode after a propagation distance L under a uniform grating is P(L) = P0 ex p (-2 a rL ) , (4.21) in which P0 is the input power and O tr is the sum of the attenuation constants for all of the radiation modes. The fraction of power coupled into any given radiation mode after the propagation distance L (the power transfer efficiency or coupling efficiency) is equal to / f ( L ) = ( a f / a ^ l l - e x p (-2 a rL )}, (4.22) in which the attenuation constant for the qth mode in a q^ with (i) = (c) for the clad ding modes and (s) for the substrate modes. The values for a q^ may be calculated from the results of a perturbation analysis by Tamir and Peng [Tamir and Peng, 1977]. In their analysis, a transmission-line approach was used to analyze the case o f surface relief gratings. 170 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The mean value of the relative dielectric permittivity in the grating region for the coupler shown in Fig. 4.13 is (4.23) in which ng in this case is the grating refractive index (equal to rij for the case of etched relief gratings) and ag = d ,/A is the aspect ratio of the grating. The amplitude of the q\h Fourier component of the grating is With these definitions, the attenuation constants for the radiation modes from a TE- polarized waveguide mode are expressed as in which the parameter A qa is defined by and Te f f is the effective waveguide thickness determined by the Goos-Haenchen shift [Nishihara, et al., 1989]. The parameter A qa arises from a weak periodic fluctuation in C X . These fluctuations are due to interference effects that stem from the discontinu ity at the upper and lower grating interface. The decay constants given by Eq. 4.25 increase with grating height as a result of the increased penetration of the guided mode into the grating region. The decay constant for the qth mode is proportional to the q\h Fourier component of the grating. Thus, for a rectangular relief grating with an aspect ratio of ag = 0.5, the even Fourier components disappear (see Eq. 4.24) so (4.24) (4.26) 171 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. that there is no outcoupling into these modes. The outcoupling efficiency is the same for cladding and substrate modes with the same mode index q. However, since there are generally more substrate modes, a greater portion of the light is coupled down wards than upwards. This can be interpreted physically to correspond to the weaker confinement at the waveguide/substrate boundary than at the waveguide/air boundary. For the case that ngc < ne g, the guided wave penetrates as an evanescent wave into the grating layer to a depth given by larger grating depths, the attenuation constants do not increase significantly. Instead, the gratings tend to scatter light from the waveguide. The attenuation constants in this saturation region are given by Herein, we refer to the grating height hc as the saturation height of the gratings. For the purpose of grating outcoupling to external elements, grating heights less than the saturation height are preferred to prevent scattering. The waveguide parameters that determine the attenuation constants for the radi ation modes are the effective refractive indices of the guided modes, the waveguide thickness, and the refractive indices of the waveguide structure in the vertical dimen sion. The outcoupling efficiencies for specific waveguide and grating structures are discussed next. R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. k (4.27) The attenuation constants in Eq. 4.25 increase as tg2 until tg reaches the value hc. For 172 4.2.3 Grating Outcoupling Efficiency In two case studies performed by Rastani to identify potential grating structures for use in the IOSAR processor, surface outcoupling gratings on Ti:LiN b03 waveguides with periods of either 2 fim or 4 /tm were considered. The angles and efficiencies were calculated for the cladding radiation modes with Eqs. 4.20 and 4.23 through 4.28. Rastani found that there are 7 cladding modes for the 2 fim period grat ings and 13 cladding modes for the 4 /an period gratings. Rastani also found that the grating outcoupling efficiency could be increased through an increase in the grating height or a decrease in the grating period. For a 500 A grating height and a I mm long grating, 33% of the light would theoretically be outcoupled from the 2 fixn period grating and 15% of the light would be outcoupled from the 4 ftm period grat ing [Rastani, 1988]. There are two issues that we considered in the present work with regard to these previous results. The first issue is that the IOSAR processor requires uniform outcou pling over a length as large as 1 cm. As presented herein, we have reconsidered the grating depth required to provide uniform outcoupling over such a large distance. The second issue that we have considered herein is the power loss due to light outcou pled into the substrate. This issue was not addressed in the earlier work by Rastani. The values o f hc and O C r for different grating heights tg were calculated for 2 fim and 4 /im period gratings on our single-mode Ti:LiNb03 waveguides. In these calcu lations, the aspect ratio of the grating (as shown in Fig. 4.13) was assumed to be 0.5, ne ff= 2.2069, nc = 1, /y = 2.2156, ns = 2.2019, and A = 6328 A. For a grating period of 2 /im, the calculated saturation height is hc = 700 A. constant for a grating height of 100 A is a r = 0.05 cm-1, constant, the cladding and substrate radiation modes Eq. 4.20. In addition to the 7 cladding radiation modes, were identified. The attenuation constants for all of the 173 The total grating attenuation To calculate this attenuation were first determined with 13 substrate radiation modes radiation modes were calcu- R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. lated with Eqs. 4.23 through 4.26. The total grating attenuation constant a r is the sum of all of these attenuation constants. This low value of 0.05 cm-1 corresponds to outcoupling of approximately 10% of the total waveguide power after a 1 cm propa gation distance (according to Eq. 4.21). For a grating height of 250 A, a r = 0.35 cm-1 and 50% of the guided light is coupled into the radiation modes after a 1 cm propagation distance. These calculations were repeated for a 4 fim period grating. The saturation height of a 4 /im period grating remains the same but the associated values of d r for the 100 A and 250 A grating heights decrease to 0.03 cm-1 and 0.2 cm-1, respec tively. These values correspond to 6 % and 33% outcoupling of the guided power after a 1 cm propagation distance. There are 13 cladding radiation modes and 27 sub strate radiation modes over which this outcoupled power is spread. For both the 2 /im and the 4 /im period gratings, only one tenth of the total out coupled light is coupled into the cladding radiation modes. The rest of the outcoupled light is coupled into the substrate. For the cases just considered, only 0.6% (one tenth o f the 6% for the 100 A deep 4 /im period grating) to 5% (one tenth of the 50% for the 250 A deep 2 /im period gratings) of the guided power incident upon the gratings will illuminate the surface-mounted devices. Although the illumination is uniform (between 6 % and 50% variation along 1 cm), this represents a significant power loss in the processor. Methods to reduce the power loss into the substrate and at the same time maintain outcoupling uniformity are considered later in Sect. 4.2.6. 4.2.4 Outcoupling Grating Fabrication on Lithium Niobate In the fabrication of surface outcoupling gratings in the present work, the emphasis was on producing large-area, uniform gratings with a low outcoupling effi ciency. Next, we describe the fabrication process used to produce uniform gratings over regions as large as 1.2 cm x lc m on waveguide substrates. The process 174 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. sequence is illustrated in Fig. 4.7 and is nearly identical to that for the fabrication of rib waveguides. Large-area, uniform surface outcoupling gratings o f 2 fim and 4 fim periods were fabricated on single-mode Ti:LiNb0 3 waveguides sliced into 1 inch X 1 inch samples as described above for the rib waveguides. After a dehydration bake, these samples were spin-coated with 1805 positive type resist at 4 krpm for 60 sec to pro duce a photoresist thickness of 0.5 fim. This thin photoresist coating facilitates the patterning of uniform gratings over the 1.2 cm x 1 cm region. The coated samples were then baked on a hot plate at 110° C for 1 min before the UV exposure. A photo mask with the large-area grating patterns supplied by Photoscience, Inc. in Torrance, CA. was used for these experiments. The photoresist-coated samples were exposed on the MJB3 mask aligner such that the grating lines were parallel to the crystallo- graphic z-axis and thus perpendicular to the optical propagation direction. The expo sure intensity was 5.6 mW/cm2 at 365 nm, and the exposure duration was 1 sec. The samples were developed with MIF-353 developer for 2 min 30 sec, rinsed in DI water, and blown dry with nitrogen. They were then blanket-exposed for 2 min as was done for the rib waveguide samples. The argon-ion beam-milling was performed with the same parameters described for the rib waveguide arrays except that the run duration was usually between 36 sec and 6 minutes in order to obtain grating etch depths of 100 A to 1000 A. After etch ing the gratings, the samples were removed from the chamber and the photoresist was removed with acetone. Dilute soapy water and DI water were used to remove small traces of contamination as was done in the fabrication of the rib waveguide arrays. 4.2.5 Grating O utcoupling Uniformity Highly uniform grating outcoupling over a 1 cm propagation distance was observed from surface outcoupling gratings on a Ti:LiNb03 waveguide fabricated 175 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. (a) 1.0 0.8 c o c ^ 0.6 C O 0.4 C O c C D c 0.2 0.0 -2 0 2 8 10 4 6 Position (mm) (b) Figure 4.14 (a) CCD-camera-acquired image of light outcoupled from a 1.2 cm X 1 cm uniform grating with a grating period of 4 fim and grating height of 150 A, and (b) a plot of the outcoupled light intensity along the 10 mm grating length. 176 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. with a small grating height. The grating had a period o f 4 /w i and a height o f 150 A. Light from a 15 mW He-Ne laser was prism coupled into the planar waveguide por tion of the sample and directed into the grating region. The light outcoupled from a large-area grating into a radiation mode was imaged onto a CCD array as shown in Fig. 4.14(a). The input beam propagated from left to right and was approximately 3 mm wide. A uniform streak is visible where the guided light reaches the grating region. The outcoupling intensity along the direction o f propagation is plotted in Fig. 4.14(b). This data was obtained from a calibrated line scan of the CCD camera response. The outcoupled intensity over this length is relatively uniform due to very weak coupling between the guided modes and radiation modes. The intensity fluctua tions of ±10% may be due to slight nonuniformities in the grating height. The non uniformities are not likely due to coherent interference effects since we imaged only one of the radiation modes. The issues of nonuniformity in the grating structure are discussed next. 4.2.6 Grating Tolerance Analysis In this section, we consider the variations in illumination of the surface- mounted components that result from fluctuations in the grating structure. Limita tions in the planar fabrication techniques can lead to these variations. The photolitho graphic and ion-etching techniques that were discussed earlier can lead to variations in the fabricated components. For example, in the ion-beam etching process we have found an etch depth uniformity of ±5% over a 3 cm x 3 cm region. To determine this, a large silicon wafer was patterned with photoresist, centered in front of the ion-gun, and etched. The photoresist was then stripped from the wafer and the etch depth was measured with a surface profilometer at several different positions on the wafer. From these results, we found that for the fabrication of a 1 cm2 grating centered in front of the gun, the grating height variation over this area should be less than about ±3% (or ± 8 A for the 250 A high gratings discussed in the previous section). This 177 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. small of a variation is difficult to detect. Since the outcoupled power is proportional to the grating attenuation constant, which is in turn proportional to tg2, this variation will lead to a ±6 % fluctuation in the illumination intensity over the 1 cm2 region. In the photolithographic definition of the gratings there is an absolute linewidth error. For a well controlled process, the photolithographic mask pattern can be pre cisely transferred into the photoresist. The grating photomasks used in this work were fabricated by electron-beam photolithography with a linewidth tolerance specifi cation of approximately ±0.1 //m. The total amount of light coupled into the cladding region is a function of the grating linewidth, since the linewidth determines the grat ing aspect ratio ag. Thus, variation in this parameter will lead to variation in the illu mination of any surface-mounted devices. The fraction o f power outcouple into the cladding region per unit length of the grating compared with the total outcoupled power per unit length is defined as the power division. The power division indicates how efficiently the outcoupled light is used to illuminate the surface mounted devices. The power division was calculated as a function of grating line width variation for a 2 fjm period grating with a 250 A grating height on a single-mode Ti:LiNb0 3 waveguide. The results are shown in units of dB in Fig. 4.15. The power division is relatively insensitive to line width variations for this grating period. As previously noted, only about 10% of the outcoupled light per unit length is directed toward the surface-mounted device and the rest of the outcoupled light is lost into the substrate. In applications where the external element must be placed in close proximity to the rib waveguide array (as in the IOSAR processor), the selection of a different grating period can improve the power division between the surface and substrate modes. In Fig. 4.15 the power division for a 1.54 fim period grating at the same grating depth is shown to be between 4 and 6 dB higher than that for a 2 fim grating. However, the sensitivity to linewidth fluctuations is much greater in this case. If the line width can be accurately controlled, this may be an alternative approach to improve the power division into the cladding radiation modes. Note that it is not necessary to specify the 178 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1.54 urn period 2 pm period o o -10 -12 -14 - 0.10 -0.05 0.00 0.05 0.10 Line Width Variation (jxm) Figure 4.15 Fraction of the total power radiated into the cladding for two different grating periods, plotted as a function of the grating line width variation. grating heights of the 2 /tin and 1.54 /tin gratings in the above comparison of power division. The fraction of power coupled into a given mode, given in Eq. 4.22, is pro portional to the attenuation constant of that mode divided by the total gradng attenua tion constant. Likewise, the power division is equal to the sum of the cladding-mode attenuation constants divided by the total grating attenuation constant (i.e., Z?a 9(c)/ a r) with no dependency on the grating length. From Eq. 4.25, the attenua tion constants for all of the radiation modes are proportional to the grating height squared (tg ). Hence, in the ratio of attenuation constants there is no dependency on grating height. 4.2.7 Outcoupling G rating Summary In this section, we have experimentally demonstrated that uniform grating out- coupling can be achieved over an outcoupling length of 1 cm for very small grating 179 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. heights. We have established that small fluctuations in grating height can lead to sig nificant fluctuations in the outcoupling efficiency. Small localized fluctuations in the outcoupled light intensity along the grating have been observed but have not been correlated with grating height fluctuations since these are experimentally difficult to measure (on the order of 10 A). We have theoretically calculated the division of power between cladding and substrate radiation modes for the case of grating outcoupling. For 2 fim and 4 fim period gratings in L iN b0 3 waveguides, we have determined that the fraction of power that is directed into the cladding region is only about 10% of the total outcoupled light. This low power division represents a potential significant power loss for imple mentations of the IO signal processor discussed herein. In addition, light coupled into the substrate can reflect or scatter from the substrate lower surface and potentially lead to background illumination of the surface-mounted devices. This background illumination would effectively decrease the signal-to-noise ratio in surface-mounted detector arrays as well as lower the dynamic range. A grating structure was consid ered that would improve the power division into the cladding region. It was found, however, that this design would be more susceptible to process variations than the gratings with a low power division into the cladding region. Other methods to improve the power division into the cladding region include the use of sub-micron grating periods as well as grating blazing techniques [Chang and Tamir, 1980]. Further investigation should be conducted to address the 10 dB power loss suffered in grating outcoupling devices due solely to power division of the radiation modes. 4.3 Rib W aveguide Arrays with Surface Outcoupling Gratings In this section, we apply the grating outcoupling analysis to gratings fabricated on rib waveguides. Although there is additional transverse confinement in rib 180 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. waveguides as opposed to the planar waveguide case, the general grating outcoupling behavior is assumed to be approximately the same for the two cases. Our experimen tal measurements o f grating outcoupling from planar waveguides and grating outcou pling from rib waveguide regions on the same substrate show good agreement in the outcoupling properties. Such agreement indicates that the stated (equivalence) assumption is valid to first order. The situation for grating outcoupling from rib waveguides is shown in Fig. 4.16. In this case, a beam coupled into the planar waveguide portion o f the sample couples into a set of rib waveguides. This coupling occurs as previously described, and results in the excitation of one or more rib waveguide modes. These modes then propagate to the grating region and are outcoupled. In general, these modes will have effective refractive indices slightly different from those of the planar waveguide. We assume herein that it is valid to insert these values in Eqs. 4.20 through 4.28 to calcu late the propagation angles of the grating spatial harmonics and the associated cou pling efficiencies. Since each guided mode in the rib waveguides can lead to a different set of radiation modes with slightly different propagation angles, we con sider next the possibility of illumination nonuniformities at the surface-mounted detector plane due to codirectional or contradirectional coupling between the modes within a single rib waveguide as a result of the grating perturbations. 4.3.1 M ode Beating in Rib Waveguides In the IOSAR processor, the two-dimensional external mask/CCD array combi nation mounted above a rib waveguide array with surface outcoupling gratings should be uniformly illuminated along the length of a given rib waveguide. Since the photo generated charge in a detector element is proportional to the incident power, the effect of power exchange between a pair of guided mode in the rib waveguide that are per turbed by the surface gratings could possibly lead to a periodic variation in the detected intensity along the rib waveguide. Earlier we stated that the power outcou- 181 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Radiation mode Rutile Coupling Prism out Slab/Rib Boundary In cid e n t^ He-Ne Beam LTLTLTLn Slab ^ Waveguide Slab y Coupled Mode Riby x Coupled Mode c-axis out LiNb03 Substrate Figure 4.16 Schematic diagram of a surface outcoupling grating on a rib waveguide. The light is coupled into the slab waveguide via a mtile prism and the guided mode is partially coupled into a set of rib waveguides. The modes that are excited in the rib waveguides are partially outcoupled by the rectangular surface out coupling gratings. The parameters that determine the performance of the gratings are also shown in this figure (after [Rastani, 1988]). pled by a surface grating from a waveguide at any position z along the grating interac tion region is proportional to the guided mode power passing through the rib waveguide at that position (see Eqs. 4.21 and 4.22). Hence, if power is periodically exchanged between two modes, one of which is preferentially outcoupled, this beat pattern may be present in the grating outcoupled illumination of the surface-mounted detector array. Since the grating lines are perpendicular to the direction of the propagating modes, any power that is coupled from one mode to another will be coupled between collinear propagating modes. The angle between the collinear modes is either 0° for the case of codirectional coupling, or 180° for the case of contradirectional coupling. In addition, coupling can occur between modes of alike or different polarization. For the case of the Ti:LiNb0 3 waveguides considered herein, the polarization state o f the 182 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. incident light is TE. The waveguide perturbation under consideration is a periodic change in the dielectric permittivity of the waveguide. This incremental change, Ae, is a scalar quantity in the case under consideration and hence there is no coupling between modes o f dissimilar polarization [Nishihara, et al., 1989]. This may occur for a periodic modulation of the refractive index induced through the electrooptic or acoustooptic effect, in which case A e is a tensor quantity with off-axis components. Such coupling could not occur in the Ti:LiNb03 waveguides considered herein since these waveguides do not support light with TM polarization. Efficient exchange of power between two modes generally occurs when the Bragg condition for phase matching is satisfied. However, for the surface-relief grat ing periods considered herein, there is a large deviation from exact phase matching given by 2 A B = p b - ( p a + qK ), in which # = ±1, ± 2 , K (4.29) The quantities Pa and Pb are the propagation constants of two arbitrary modes under consideration, q denotes the order of coupling, K = 2 /zM is the grating wave number, and A is the grating period [Nishihara, et al., 1989]. Consider first the case of codirectional coupling. For 2 fim period gratings, K = 3.1 and the codirectional mode spacing, A ft = Pb~Pa, for 8 fim wide rib waveguides in Ti:LiN b0 3 is on the order of 5 x 10-2 fim~l. Although there may be 20 to 30 transverse modes supported by the rib waveguides, the minimum magnitude deviation from exact phase matching for codirectional coupling is approximately AB = 1.5 f a n 1 . In the case o f contradirectional coupling with 2 fim period gratings, the contra- directional mode spacing, A/3 = Pb-(-P a), for 8 fim wide rib waveguides in Ti:LiNb03 is on the order of 44 /un-1. The order of coupling that gives approxi mately zero deviation from phase matching is q = 15. The q\h Fourier component of 183 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the surface relief grating perturbation given by A eq - (n f - n%){sin(^a^) / qn], (0 < a < 1, q * 0) , (4.30) in which rif is the waveguide-layer refractive index, nc is the cladding-layer refractive index, and a is the grating aspect ratio [Nishihara, et a l, 1989]. The 15th Fourier component is approximately 2% of the value of the first Fourier component. The grating period required to obtain contradirectional phase matching with the first Fou rier component is A = 0.14 fim. Such a grating will provide much stronger coupling by comparison than the 2 ftm period grating considered here. In fact, strong contradi rectional coupling is the primary precaution in the use of submicron gratings for out coupling elements in the advanced 1 0 signal processors. The large deviation from exact phase matching present in the codirectional cou pling case and the high order of coupling involved in the phase matched contradirec tional coupling case leads to negligible coupling between modes for an 8 fim wide rib waveguide in Ti:LiNb0 3 with 2 ftm period gratings. A similar argument can be made 4 f im period gratings. Hence, we do not expect that power exchange between modes due to surface grating perturbations will lead to significant nonuniformities in the illu mination of the surface mounted elements for the cases considered herein. Next, we consider the possibility of illumination nonuniformities due to coher ent interference of cladding radiation modes at the surface mounted detector plane. 4.3.2 Surface Com ponent Illum ination Uniform ity The multiple-outcoupled spatial harmonics described by Eq. 4.20 and the effect that they have on the illumination uniformity of the external elements along the length of the rib waveguide is now considered. The propagation directions of the cladding spatial harmonics for a Ti:LiN b0 3 waveguide with 2 fim gratings are represented in Fig. 4 .17(a). The light outcoupled from the grating region below one pixel of the sur- 184 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. face-mounted element diverges in seven discrete angles ranging from +70° in the direction of copropagation with the waveguide mode (measured from the vertical) to 70° in the direction of contrapropagation with the waveguide mode. Conversely, the light incident on a given pixel in the external device array converges from several dis crete directions. Since the light is coherent, it will constructively interfere and the measured intensity will be proportional to the sum of the electric field vectors squared. However, since the angular range of the light is so large, the spatial varia tions in the intensity will be much smaller than the pixel dimensions. For example, under the assumption that the radiation modes are plane waves, the intensity variations along the detector plane due to the interference of two plane waves, such as is shown schematically in Fig. 4.17(b), is given by means of Eq. 4.20. For 2 fan period surface outcoupling gratings, the smallest angu lar separation for two radiation modes is 18° (this case occurs for one mode that prop agates normal to the waveguide surface and a second mode that propagates in the +18° direction). These two modes will interfere with one another and create an inten sity variation along the detector with a period of about 2 fim. To calculate this period, the argument of the cosine term in Eq. 4.31 was set equal to 2k, and z solved for under the assumption that A = 6328 A. Thus, for a 10 fim square pixel, the intensity will be averaged over 5 cycles. Since this is the largest interference period created by the 2 fim gratings, there should not be significant periodic nonuniformities from pixel to pixel. For 4 fim period gratings, on the other hand, the smallest angular separation between radiation modes is approximately 9° (this case occurs for one mode that propagates normal to the waveguide surface and a second mode that propagates in the +9° direction). Hence, the largest interference period along the detector array in this (4.31) in which 6, and d2 are the propagation angles of the radiation modes determined by 185 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. o f cladding radiation modes Surface grating 1 1 — 'A H Waveguide mode I I agA lib waveguide cross-section___* (a) Planar wave phase fronts Waveguide mode Propagation directions of two cladding radiation modes separated by a small angle (b) i One cladding radiation mode from waveguide mode 1 — | r--------1 r f 1 One cladding radiation mode K — from waveguide mode 1 1 -------- 1 I--------1 ► P — Propagation directions of two cladding radiation modes Two waveguide modes * Pb (c) Figure 4.17 Side view of a rib waveguide with surface outcoupling gratings show ing (a) the propagation directions of the cladding radiation modes for a single waveguide mode, (b) the interference between planar phase fronts of two cladding radiation modes that are directionally close to one another, and (c) the close propaga tion directions of two cladding modes that originate from different waveguide modes. 186 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. case is 4 /an, and each pixel will average over 2.5 periods. For grating periods much larger than this, it is possible that there will be a resulting variation in the detected illumination from one pixel to the next due to this interference pattern. Previously, we noted that there may be multiple guided modes in the rib waveguides. When outcoupled, each waveguide mode can lead to a different set of radiation modes, all o f which interfere at the detector plane. Two cladding radiation modes of the same order that originate from different waveguide modes will propa gate in directions separated by a very small angle as shown in Fig. 4.17(c). The beat length of this interference pattern is determined by the substitution of Eq. 4.20 into Eq. 4.31 for equivalent radiation orders of two rib waveguide modes. The interfer ence pattern that is formed has a period equal to XtAnep in which Ane f f is the differ ence in effective refractive indices of the two modes. For 8 fim wide rib waveguides formed in single-mode Ti:LiNb0 3 waveguides, Ane j f ranges from 0.001 to 0.005. Hence, the interference period is between 100 fim and 600 fim . Since this period is so much larger than the pixel spacing in the detector plane, such interference is not observable in the detection configuration under consideration. It is important to note at this point that the radiation modes formed by the rib waveguide/grating structure are not plane waves due to the rib waveguide transverse confinement. The transverse confinement leads to a transverse spread of the outcou pled light due to diffraction from a finite aperture width. In the next section, we dis cuss the detector coupling requirements imposed by the transverse spread of the outcoupled light. 4.3.3 Transverse Spread o f the Outcoupled Light For the specific case of gratings on rib waveguides, the light outcoupled from the surface outcoupling gratings is tightly confined in the transverse direction by the rib waveguide width. This situation is in contrast to the ideal case of a grating on an 187 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. unconfined planar waveguide. Due to the finite rib waveguide width, the outcoupled light will diffract in the transverse direction with an angle between minima in the far field. This situation is shown schematically in Fig. 4.18 with the angle < j > indicated and for a near field beam width d equal to the rib width. The collimation length is given by and is an approximation of the maximum distance that an external element should be placed above the rib waveguide array to prevent overlap of the light from one rib waveguide with the neighboring pixel in the external element [Huang and Lee, 1986]. For rib waveguides with 8 /Jm widths, 0=9° and the collimation length x c = 50 fiin (for A = 6328 A). Since this distance is rather small, it would be very dif ficult in the case of the IOS AR processor to include two external elements on separate substrates (azimuth phase mask and CCD detector array) within this distance since substrates with thicknesses less than a two hundred microns are usually fragile and are not very flat. Two approaches that could be used in this situation are shown in Fig. 4.19. The first approach makes use of a thin transparent dielectric material deposited on the surface of the CCD detector array to act as a planarizing support for a variable thickness metal layer, which in turn acts as the azimuth mask. Alterna tively, the spatially-varying transmittance of this mask may be produced by area- encoding the surface of the device with an optically opaque film. This modified CCD array may then be easily mounted within the collimation length above the rib waveguide array. In the second approach, a cylindrical lenslet array on a separate (4.32) 188 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Detector Array U J T A T iAr i i Transverse spread of outcoupled radiation modes Rib -Waveguide Array — d Figure 4.18 Schematic diagram of the transverse spread of light outcoupled by sur face outcoupling gratings from the rib waveguide array structure. substrate is registered above the rib waveguide array to narrow the far-field width of the diverging outcoupled light. The variable transmittance mask is included on the front or back side o f this substrate. Therefore, the CCD detector array could be used as supplied commercially without further modifications. If the lenslet array is placed 50 fim above the rib waveguide array and the substrate thickness is 300 fim , the required lenslet f-number is approximately f/5. Cylindrical lenslet arrays supplied by Nippon Sheet Glass (NSG) have an 85 ftm width, a 100 finl pitch, and are f/2.5. These values are approaching the sizes required for this example. In the case of the IO correlator, the rib waveguide width of 28 ftm leads to a smaller divergence angle of 2 .6° and to a much larger collimation length of 600 fim. A single device or two stacked devices could be surface-mounted and illuminated without significant spread of the outcoupled beam. In applications such as the IO cor relator where the rib waveguide widths are large and the collimation distance x c is several hundred microns, it may be possible to mount the detector on the backside of 189 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Detector Array Cylindrical microlens | array Rib Waveguide Array (a) (b) Figure 4.19 Two approaches to mounting azimuth mask and CCD detector array above a densely packed rib waveguide array. In (a) a transparent dielectric material is deposited onto the surface of the detector array with the mask patterned in an absorb ing thin film, and (b) an intermediate substrate is used with the azimuth mask and a cylindrical lenslet array to collimate the diverging light. a thinned substrate in order to capture the higher-power substrate radiation modes. 4.3.4 Rib W aveguide Array and Surface G rating Integration The surface outcoupling gratings with improved outcoupling uniformity over a large grating region, discussed in Sect. 4.2, were integrated with the rib waveguide arrays discussed in Sect. 4.1. The fabrication procedures for the individual structures were unchanged in the integration process. Either the rib waveguide array or the sur face outcoupling gratings can be fabricated first. It was shown by Rastani that highly uniform gratings could be fabricated on the surface of rib waveguide arrays with deeply etched gaps. For the research presented herein, we fabricated integrated rib 190 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide/grating structures in the same order used by Rastani and in the reverse order. We have found that both approaches lead highly uniform structures. Rib waveguide arrays with 8 ftm rib waveguide widths, 2 ftm gap widths, and a 5000 A gap depth were fabricated in single-mode Ti:LiNb0 3 waveguides. These arrays consisted of 1000 elements of 5 mm length. Similarly, rib waveguide arrays with 28 /mi rib waveguide widths, 2 ftm gap widths, and a 5000 A gap depth were fabricated in Ti:LiNb03 single-mode waveguides. These arrays consisted of 333 ele ments of 5 mm length. We typically followed the rib waveguide array fabrication with the fabrication of large-area ( 1.2 cm x 1 cm) outcoupling gratings with periods of either 2 ftm or 4 fim. The gratings were integrated on the surface of these samples in both planar waveguide and rib waveguide regions. Rib waveguide and grating fabrication techniques were extended to the GaAs material system by fabrication of these structures onto a GaAs/AlGaAs waveguide. A surface outcoupling grating with a 2 ftm period and 1000 A grating height was fabri cated onto the waveguide surface first. A rib waveguide array with 8 ftm rib waveguide widths, 2 ftm gap widths, and a 5000 A gap depths was then fabricated with 660 elements of 1 cm in length onto the GaAs/AlGaAs waveguide in regions both with and without gratings. The full fabrication procedure for rib waveguides and gratings on lithium nio- bate is similar to the individual procedures for each of the structures done separately. In this case, the individual procedures are performed one after the other with the only added detail of structure alignment. Once one structure is ion-beam milled into the surface of the waveguide, the second structure can be easily aligned (such that the grating lined are perpendicular to the rib waveguides) during photolithography. Since the individual fabrication procedures for the rib waveguides and gratings in GaAs have not been discussed yet, these are given next. The full integration procedure is described in this example. 191 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The fabrication procedure used for the GaAs rib waveguides is a direct adapta tion of the process described by Rastani [Rastani, 1988] for fabrication of rib waveguides on multimode Ti:LiNb03 waveguides. The initial planar GaAs/AlGaAs structure was described in Section 3.4.2. The structure consists of a (100) cut semi- insulating substrate with a 2 /tm epitaxial layer of Al0 3Gao 7As to act as a barrier layer for a 1 /a n thick GaAs guiding layer. A waveguide substrate slightly larger than 1 cm x 1 cm in size was cleaned and dehydration-baked by the method described ear lier in Section 4.1.7 for the lithium niobate samples. This cleaning was done to pre pare the sample for photolithography. The integrated rib waveguide and grating structures were fabricated according to the sequence shown in Fig. 4.7 with the fabrication o f the gratings preceding that of the rib waveguides. The sample was coated at 4 krpm with positive photoresist of type Aspect 812 to a thickness of 0.8 fim. The sample was baked in a convection oven at 90° C for 30 min. prior to the pattern exposure. The masks that were used to pattern the gratings and rib waveguides were supplied by ASET Electromask Co. in Woodland Hills, CA. The grating mask consisted of a 1 cm wide x 1 mm long grating with a grating period of 2 /an. The lines of the grating were parallel to the long dimension of the grating region. For the grating exposure, the sample was oriented with respect to the mask such that the gratings extended in from the sample edge about 6 mm, rather than all of the way across the sample. The gratings were patterned in this manner so that after the subsequent rib waveguide fabrication, one region of the sample would have rib waveguides with no gratings, a second region would have combined rib waveguides and gratings, and a third region would have gratings alone on the planar waveguide. The mask aligner lamp power was 380 W and the sample was exposed for 14 sec. The pattern-exposed sample was then developed and blan ket-exposed for 5 minutes on the mask aligner. Finally, the sample was hard-baked at 140° C for 30 minutes. The GaAs/AlGaAs sample was then mounted in a vacuum chamber with a 192 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Commonwealth 3-cm ion gun. The beam diameter was 3 cm and the distance to the sample mount was about 15 cm. The sample holder was an aluminum mount without water-cooling. After a vacuum base pressure of about 4 x IO " 6 Torr was obtained, ultra-high purity (99.999% pure) argon was bled into the chamber to bring the pres sure up to 2 X IO" 4 Torr. The ion gun was then operated with a beam voltage of 70 V, an accelerator voltage of 70 V, and a beam current of 7 mA for 15 min. A low beam voltage and beam current were used to prevent burning o f the thin photoresist during the etching process and to give a low etch rate that could be easily controlled. After the milling was complete, the sample was removed from the chamber. The photore sist was stripped with acetone, rinsed in DI water, and blown dry with nitrogen. From DekTak measurements of the surface feature height, we found that an ion beam mill ing duration of 15 minutes resulted in 0.1 ftm deep gratings. The result of the fabrica tion o f high-quality surface outcoupling gratings with a 2 ftm period on a bare GaAs substrate with this same process is shown in Fig. 4.20. The GaAs/AlGaAs waveguide sample was then coated at 4 krpm for 30 sec. with positive photoresist of type Aspect 10X to a thickness of 1.5 fim for the rib waveguide array fabrication. After application of the photoresist, the sample was baked for 30 min. at 90° C in a convection oven. The rib waveguide array mask pat tern consisted of 660 ribs with 8 ftm rib widths, 2 ftm gaps widths, and a 2 cm rib length. The rib waveguides that were patterned extended all of the way across the sample and perpendicular to the lines of the grating structure. Before exposure of the photoresist through the photomask, the sample was oriented such that the rib waveguide lines were parallel to the edges of the sample. The exposure time was also 14 sec. at a 380 W lamp setting on the mask aligner. The exposed sample was devel oped, rinsed in DI water, and blown dry with nitrogen. The sample was then blanket- exposed for 5 min. on the mask aligner prior to a 30 min., 140° C hard-bake in a con vection oven. To etch the rib waveguide pattern, the sample was mounted in the same system 193 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Figure 4.20 SEM cross-section photograph of a 2 ftm gratings fabricated on GaAs by ion-beam milling. The rough textured surface is the sample edge. The roughness is due to breakage of the sample and is not a result of cleavage. with the Commonwealth ion gun used for the grating fabrication. In the same manner as previously described, argon was admitted into the chamber. The ion beam milling parameters in this case were an 80 V beam voltage, an 80 V accelerator voltage, a 7 mA beam current, and a 30 min. milling duration. The sample was then removed and cleaned in the same manner as previously described. The milling duration of 30 minutes at the higher gun settings resulted in a 0.5 fim rib waveguide depth. The end pieces of about 1 mm width were cleaved from the original sample at the beginning and the end o f the rib waveguides to remove those portions of the sam ple where the pattern transfer was nonuniform due to the photoresist edge-bead, and to prepare the waveguide end-facets for end-fire coupling. High magnification SEM photographs that show the results of the fabrication process are given in Fig. 4.21. The rib waveguide definition was excellent, with very smooth sidewalls. The grating aspect ratio is approximately 0.6, hence it is apparent in Fig. 4.21 that there is some inaccuracy in the pattern transfer of these features. Slight fluctuations in the grating features are also evident. These fluctuations are to be expected, however, since the feature size is near the limit of the photolithographic patterning process. The grating pattern is also evident in the rib waveguide gap region of Fig. 4.21(c). This artifact of the fabrication process may result in excess 194 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. scattering of the guided mode. Next, we describe the observed uniformity of light outcoupled from 4 fim period gratings on a Ti:LiNb03 rib waveguide array that was fabricated in this work. The observed cladding-radiation-mode outcoupling angles and efficiencies o f 2 fim and 4 ftm period grating on lithium niobate waveguides were previously character ized by Rastani [Rastani, 1988]. Herein, we assume that the cladding radiation mode angles of these gratings are similar to those measured by Rastani for gratings with a 4 ftm period. Later, we will return to the subject of gratings and rib waveguides on GaAs/AlGaAs waveguides in order to discuss the results of our characterization of the grating outcoupling angles and efficiencies from these structures. 4.3.5 Observed Grating Outcoupling from Rib Waveguides Grating outcoupling over a 4 mm propagation distance was observed from sur face outcoupling gratings (1.2 cm x 4 mm grating region, 4 fim grating period, and grating height o f 150 A) on a rib waveguide array (8 fim rib waveguide widths, 2 fim gap widths, 5000 A gap depths, 1000 elements, 5 mm long) fabricated on a Ti:LiNb03 waveguide. Light from a 15 mW He-Ne laser was prism-coupled into the planar waveguide portion of the sample and directed into the rib waveguide array in an arrangement such as that shown in Fig. 4.16. The light outcoupled from a iarge- area grating was imaged onto a CCD array as shown in Fig. 4.22(a). The input beam propagates from left to right and is approximately 1 mm wide. A uniform streak is visible in the planar waveguide portion of the sample. The streak is slightly brighter after it is coupled into the rib waveguides. The guided streak is very bright in the grating region as the radiation modes are imaged onto the CCD camera. The outcou pling intensity along the direction of propagation is plotted in Fig. 4.22(b). This intensity plot was obtained from a calibrated line scan of the CCD camera response. The outcoupled intensity over this length drops by about 15%. This drop in intensity can be accounted for by the rib waveguide propagation losses alone, which indicates 195 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (C) Figure 4.21 SEM photographs that show (a) 8 ftm wide rib waveguides and 2 ftm separations etched 0.5 ftm deep in a GaAs/AlGaAs waveguide, (b) identical rib waveguides that also include 2 ftm period surface outcoupling gratings, and (c) a high magnification view of the rib waveguide sidewall and gratings. 196 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Rib W a v e g u id e B o u n d a ry jG intm g Boundary Coupling Pi ism t Grating Width O u tc o u p le d Light (a) 1.0 0.8 « c 0.6 -e Start of grating region c o 0.4 V) c C D Start of rib waveguide array c 0.2 0.0 -2 0 2 4 Position (mm) Figure 4.22 (a) CCD-camera-acquired image of light outcoupled from surface out coupling gratings ( 1.2 cm X 4 mm grating region, 4 fan. grating period, and 150 A grating height) on top of a rib waveguide array (8 fJia. rib waveguide widths, 2 /tm gap widths, 5000 A gap depths, 1000 elements, 5 mm long) fabricated on a lithium niobate waveguide and (b) a graph of the outcoupled light intensity along the 4 mm grating length. 197 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. that the gratings outcoupling efficiency is very low, as expected for a grating height of 150 A. The outcoupled intensity also fluctuates along the length of the rib waveguides. This fluctuation is similar to that seen for the gratings fabricated alone (without rib waveguides) on a planar waveguide. These large-scale fluctuations are likely due to variations in the grating structure and possibly the rib waveguide struc ture. The most likely cause is a variation in the grating height since, as we found ear lier, small changes in height can lead to significant changes in grating coupling efficiency. These large-scale fluctuations in the outcoupled intensity as a function of posi tion can potentially be measured and corrected for in the fabrication of the Doppler phase history mask in the IOSAR processor. Small scale fluctuations, on the other hand, may be much more difficult to predict and compensate for in the design of the rib waveguide structure. In the next section, we describe the observed microscopic fluctuations (on the order of the width of the rib waveguides) in the intensity of the outcoupled light from several rib waveguide and grating structures. 4.3.6 O bserved G rating Outcoupling U niform ity In order to observe the uniformity o f light outcoupled from gratings on rib waveguides, light was prism coupled into several lithium niobate rib waveguide/grat ing samples in a manner similar to described shown in the last section. However, in this case a 10X microscope objective was used to magnify and image the light outcou pled from the rib waveguides onto a CCD array. Captured images are shown for 28 ftm wide rib waveguides and 8 fim wide rib waveguides in Fig. 4.23. The images of the rib waveguides can be identified running from left to right, in which the rib waveguides with the gratings on the surface appear the brightest and the gaps appear the dimmest. The top two images are of two separate rib waveguide samples with 28 fim wide ribs, 2 fim wide gaps, and 5000 A etch depth. The rib waveguides in Fig. 4.23(a) have 2 /an period gratings etched 150 A deep into the surface. The out- 198 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (C) ' (d) Figure 4.23 The top images are magnified views of grating outcoupled light from rib waveguides in lithium niobate that have 28 pan rib waveguide widths, 2 pan gap widths, and in (a) 2 pan period gratings. The rib waveguide structure is the same in (b) but the gratings have a 4 /an period. The lower left image (c) is a magnified view of grating outcoupled light from the 8 pan wide rib waveguides with 2 /an gaps and 2 /an period gratings shown in (d). coupled light from each rib waveguide appears to consist of several narrow streaks that run along the length of the rib waveguide. The origin of these streaks is not clear; however, it is possible that they arise from resolution limitations in the imaging sys tem used. The rib waveguide outcoupling was observed using a 10X objective (A/A = 0.25) to image the light onto a CCD detector array placed about 30 cm from the objective. The effective resolution of this imaging system is about 1.2 /an. Hence, it is likely that these streaks result from coherent interference effects where some of the object features (such as rib waveguide edges) are under resolved. If this 199 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. is the origin o f these streaks, then the observed nonuniforraity will not be an issue in the proximity-coupled detector arrangement. The rib waveguides in Fig. 4.23(b) have 4 //m period gratings etched 250 A deep. A slight periodicity in the intensity along the rib waveguide is discernible. This periodicity corresponds roughly to the grating period o f 4 fim. The presence of these nonuniformities suggests that small grating periods compared with the surface- mounted detector array pixels are preferable in cases where high uniformity is required. The lower images in Fig. 4.23 are of a rib waveguide array with 8 /m i rib waveguide widths, 2 /a n gap widths, 5000 A gap depths, and 2-/an-period surface outcoupling gratings. The image in Fig. 4.23(c) is a magnified view of one region of the rib waveguide array. In this magnified view, there are no apparent periodic non uniformities present along the rib waveguides. However, it appears as if the illumina tion is brighter near the edges of the ribs. This again may be due to the imaging arrangement described above. In all of the images in Fig. 4.23 there are large-scale inhomogeneities in the detected illumination, which may be due in part to the imaging method and nonuni formities in the CCD array response. The observed large-scale fluctuation in the rib waveguide and grating structures may also be due in part to nonuniform mask contact in the photolithographic process, uneven development of the photoresist, or etch depth fluctuations. All of these nonuniformities can lead to illumination variations. The only observed small-scale periodic illumination fluctuations that occur along the rib waveguide seem to be associated with the grating period. As long as the grating period is sufficiently small compared with the detector pixel size, these nonuniformi ties will not be resolved. In practice, the CCD array will be placed in close proximity to the rib waveguide array surface. To unambiguously determine whether the coherent illumi- 200 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. nation from several outcoupled modes leads to periodic illumination of the CCD array, experiments should be conducted with surface-mounted CCD arrays. There is no evidence from the imaging arrangement discussed in this section that such periodic nonuniformities will be present in the fully integrated rib waveguide array/grat ing/surface-mounted detector structure. In addition to the measurements and observations performed on the rib waveguide array/grating structures in Ti:LiNbC> 3 waveguides, the propagation angles and outcoupling efficiencies of radiation modes outcoupled from gratings fabricated on GaAs/AlGaAs rib waveguides were also studied. These results are presented next. 4.3.7 G aAs Rib W aveguide Array w ith Surface G ratings In Sect. 4.3.4, we described the integration of a rib waveguide array (8 fim rib waveguide widths, 2 ftm gap widths, 5000 A gap depths, 660 elements, 1 cm length) and surface outcoupling gratings (2 fim period, 1000 A grating height, 1 mm interac tion length) on a GaAs/AlGaAs planar waveguide. In this section, we present the results of characterization of the 2 fim grating structure on both planar and rib waveguide portions of the sample. As mentioned earlier, the ends of the sample were cleaved so that the rib waveguides could be individually excited. Light from a 0.5 W NdrYAG laser (k = 1.06 fim ) was end-fire coupled into a single rib waveguide within the array, and the end-emission from the opposite facet was observed. Light was visible only in the excited rib waveguide and not the neighboring rib waveguides. The absence of light in the neighboring rib waveguides after a propagation distance of about 1 cm indi cates good waveguide confinement (on the order of 20 dB isolation or better). The relative grating outcoupling angles and efficiencies were then measured with the experimental arrangement drawn schematically in Fig. 4.24. In this arrangement, a cylindrical lens was used to simultaneously excite several 201 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. rib waveguides on which surface outcoupling gratings had been fabricated. A portion of the light in each rib waveguide was grating outcoupled into several radiative modes. The propagation angles of these radiative modes were determined by means of Eq. 4.20. The material parameters used in these calculations were given in Chapter 3 in the description of the GaAs/AlGaAs waveguide structure. A calibrated CCD camera was placed at several locations far from the waveguide sample in order to detect the individual radiation modes. The relative power in each cladding radia tion mode (relative coupling efficiency) was indicated by the magnitude of the camera response to the incident light. These measurements were repeated for light outcou pled from the surface outcoupling gratings in the planar portion of the waveguide. The results of these relative measurements for four cladding modes are shown in Fig. 4.25 as compared with the theoretical values obtained from the relations pre sented in Sect. 4.2.3. Since the fabricated grating height is greater than the saturation grating height (600 A) calculated from Eq. 4.27, the grating attenuation constants were calculated by using Eq. 4.28. The relative measured coupling efficiencies given in this figure are normalized by the highest theoretical coupling efficiency so that a direct comparison can be made. The measured propagation angles and relative coupling efficiency from one cladding radiation mode to the next correspond well with theory. This correspon dence indicates that the photolithographic replication of the gratings in this material produced an accurate pattern transfer of the grating period and shape. Such control can eventually be used to modify the shape of the gratings and therefore enhance the coupling efficiency into a single order (e.g., by blazing the gratings). Another inter esting point to notice is the similarity between the measured coupling efficiency for the planar waveguide region and the rib waveguide region. The fact that these values are nearly identical indicates that the process of integration of the rib waveguides with the gratings has not significantly affected the grating performance. In addition, this indicates that the transverse confinement in the rib waveguides does not adversely 202 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Detector/Camera. Laser Collimator Nd:YAG Sample Neutral Rectangular Cylindrical Density Aperture Lens Filter Figure 4.24 Experimental arrangement used for measurement of the radiation mode propagation angles and outcoupling efficiencies from a GaAs waveguide. 1 2 k O c a > o i t LU c Q. 3 o & 3 o 8 7 6 5 4 3 2L O 0.1 3 |. 2 0.01 1 ------ 1 -------1 ------- '-------1 -------'------- T Theoretical (For saturation grating height of 0.06 pm) M easured; gratings on a slab waveguide (Milled grating depth of 0.1 pm) M easured; gratings on a rib w aveguide array (Milled grating depth of 0.1 pm) * -60 -4 0 -2 0 0 20 Outcoupling Angle (deg) 40 6 0 Figure 4.25 Measured relative radiation mode outcoupling efficiencies and angles for a GaAs/AlGaAs rib waveguide array with 2 /im period surface outcoupling grat ings. affect the outcoupling behavior of the gratings. 203 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4.3.8 Rib Waveguide Array with Surface Gratings Summary In this section, we analyzed the image propagation and outcoupling characteris tics of rib waveguide in densely packed arrays with surface outcoupling gratings. Sources o f nonuniformity in the illumination of external element were considered. We determined that codirectional or contradirectional coupling between rib waveguide modes would be negligible for a 2 fjm period grating. Therefore, we do not expect there to be a significant periodic exchange of power between the modes. The coherent interference of radiation modes at the plane of an externally-mounted detector array was shown to lead to maximum periodic variations in the illumination roughly equivalent to the grating period. Hence the grating period should be smaller than the detector pixel size. The coherent interference of radiation modes from two or more rib waveguides was shown to have a very large beat period compared with the detector pixel size. However, since the radiation mode transverse field profiles are different for each guided mode, it is unlikely that they will fully constructively or destructively interfere across the width of a detector pixel at any given position. Hence, we expect that the integrated intensity from pixel-to-pixel above a given rib waveguide will be sufficiently uniform. The transverse spread of the light outcoupled from the rib waveguides due to diffraction effects makes it necessary to mount external devices in close proximity to the waveguide surface. The associated requirements were calculated for the IOSAR processor and IO correlator. Because of the high density of the IOSAR rib waveguide array, the diffraction effects are large and the external mask and detector array ele ments must be placed within 5 0 //m of the waveguide surface. Two possible approaches were considered that would allow these devices to be mounted without undue crosstalk from the transverse spread of the outcoupled light. The IO correlator, on the other hand, has much greater latitude in the surface detector array spacing. Since the rib waveguide width is large (30 /im), the detector may be placed up to 600 ^m from the surface before suffering from the transverse spread of the outcou- 204 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. pled light. This latitude makes it feasible to mount the detector array on the back side of the waveguide substrate so that it will be illuminated by the higher power substrate radiation modes. We have experimentally demonstrated uniform outcoupling from surface out coupling gratings on rib waveguides over a 5 mm length. The total drop in outcou pled power could be accounted for by propagation losses similar to those measured in rib waveguides without surface outcoupling gratings. This indicates that a very small portion of the light is outcoupled by the gratings, as expected for a grating height of 150 A. Microscopic variations in the outcoupled illumination uniformity were also investigated. Rib waveguide samples with 2 /dm and 4 (dm integrated surface gratings were observed under magnification. Evidence o f periodic non-uniformities on the order of the grating size were observed on the sample with 4 /im gratings. Observed intensity variations in the transverse direction for light outcoupled from 2 /dm period gratings on 8 /dm wide and 28 /dm wide rib waveguides was likely due to the low res olution of the imaging arrangement used to acquire the magnified images. Small ran dom fluctuations observed in the outcoupled light could also be attributed to nonuniform pixel response in the CCD array used to acquire the images. Alterna tively, slight variations in the grating and rib waveguide structures could lead to the observed nonuniformities. There were no observed small-scale nonuniformities that would lead to excessive fluctuations from one detector pixel to the next. Our observa tions and calculations lead us to conclude that uniform illumination of a surface- mounted detector array is possible with the rib waveguide array and grating structures discussed herein, possibly limited by the overall fabrication uniformity that is achieved. The processing sequences for rib waveguides and gratings were transferred to GaAs waveguide substrates. We have fabricated an integrated rib waveguide struc ture with 8 /im rib waveguide widths, 2 /im gap widths, 5000 A gap depths, and 2 /dm period gratings etched 10 0 0 A deep. The cladding radiation modes of the gratings in 205 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. both rib waveguide and planar waveguide regions were observed and the outcoupling efficiencies were measured and compared with theory. The good agreement between measurement and theory indicated that the fabricated structures were extremely uni form, and that the pattern transfer was accurate. In addition, the similarity between the values for the planar waveguide and rib waveguides indicates that the transverse confinement in the rib waveguides does not adversely affect the grating outcoupling properties. 4.4 R eferences D. B. Anderson, “Integrated Optical Spectrum Analyzer: An Imminent ‘Chip’,” IEEE Spectrum, Dec., 22-29, (1978). M. Belanger and G. L. Yip, “A Novel Ti:LiNb03 Ridge Waveguide Linear Mode Confinement Modulator Fabricated by Reactive lon-Beam Etching,” J. Lightwave Technol., LT-5, 1252-1257, (1987). S. V. Burke, P. C. Kendall, S. Ritchie, M. J. Robertson, and P. N. Robson, “Analysis o f Rib Wave-Guide Coupler Filters,” IEE Proceedings-J, 139(1), 59-65, (1992). W. K. Bums and A. F. Milton, “Mode Conversion in Planar-Dielectric Separating Waveguides,” DEEE J. Quantum Electron., Q E -ll(l), 32-39, (1975). J. K. Butler, D. E. Ackley, and D. Botez, “Coupled Mode Analysis of Phase-Locked Injection Laser Arrays,” Appl. Phys. Lett., 44,293-295, (1984). M. Cantagrel, “Comparison of the Properties of Different Materials Used as Masks for lon-Beam Etching,” J. Vac. Sci. Technol., 12(6), 1340-1343, (1975). 206 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. K. C. Chang, V. Shah, and T. Tamir, “Scattering and Guiding of Waves by Dielectric Gratings with Arbitrary Profiles,” J. Opt. Soc. Am., 70(7), 804-813, (1980). K. C. Chang and T. Tamir, “Simplified Approach to Surface-Wave Scattering by Blazed Dielectric Gratings,” Appl. Opt., 19(2), 282-288, (1980). R. J. Deri, E. Kapon, and L. M. Schiavone, “Scattering in Low-Loss GaAs/AlGaAs Rib Waveguides,” Appl. Phys. Lett., 51(11), 789-791, (1987). M. D. Feit and J. A. Fleck, Jr., “Analysis of Rib Waveguides and Couplers by the Propagating Beam Method,” J. Opt. Soc. Am. A, 7(1), 73-79, (1990). K. Handa, S. T. Peng, and T. Tamir, “Improved Perturbation Analysis of Dielectric Gratings,” Appl. Phys., 5, 325-328, (1975). A. Hardy and W. Streifer, “Coupled Mode Theory of Parallel Waveguides,” J. Lightwave Tech., LT-3(5), 1135-1146, (1985). H. A. Haus, W. P. Huang, S. Kawakami, and N. A. Whitaker, “Coupled-Mode Theory of Optical Waveguides,” J. Lightwave Tech., LT-5(1), 16-23, (1987). H. A. Haus and L. Molter-Orr, “Coupled Multiple Waveguide Systems,” IEEE J. Quantum Electron., QE-19, 840-844, (1983). R. C. Hewson-Browne, P. C. Kendall, and D. A. Quinney, “Roughness Scattering into Substrate Radiation Modes o f Rib Waveguides,” IEE Proceedings-J, 136(5), 281-286, (1989). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. T. Hirata, M. Suehiro, M. Hihara, M. Dobashi, and H. Hosomatsu, “Demonstration of a Waveguide Lens MonolithicaUy Integrated with a Laser Diode by Compositional Disordering o f a Quantum Well,” IEEE Photonics Tech. Lett., 5(6), 698-700, (1993). S. Y. Huang and S. H. Lee, “Blazed Grating Couplers on L iN b0 3 Optical Channel Waveguides and Their Applications to Integrated Optical Circuits,” J. Lightwave Technol., LT-4, 1304-1310, (1986). L. D. Hutcheson, Ed., Integrated Optical Circuits and Components, Series on Optical Engineering, B. J. Thompson, Ed., (Marcel Dekker, Inc., New York, 1987). E. Kapon, J. Katz, and A. Yariv, “Supermode Analysis of Phase-Locked Arrays of Semiconductor Lasers,” Opt. Lett., 10, 125-127, (1984). S. Kawakami and H. A. Haus, “Continuum Analog of Coupled Multiple Waveguides,” J. Lightwave Tech., LT-4(2), 160-168, (1986). T. Kubota and M. Takeda, “Array Illuminator Using Grating Couplers,” Optics Letters, 14(12), 651-652, (1989). M. Kuznetsov, “Expressions for the Coupling Coefficient of a Rectangular Waveguide Directional Coupler,” Opt. Lett., 8(9), 499-501, (1983). R. E. Lee, “Microfabrication by lon-Beam Etching,” Semiconductor International, Jan ., 73-82, (1980). Z. M. Mao and W. P. Huang, “Analysis of Optical Rib Wave-Guides and Couplers with Buried Guiding Layer,” IEEE J. Quantum Electron., QE-28(1), 176-183, (1992). 208 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. E. A. J. Marcatili, “Dielectric Rectangular Waveguide and Directional Coupler for Integrated Optics,” Bell Syst. Tech. J., 48,2071-2102, (1969). D. Marcuse, “The Coupling o f Degenerate Modes in Two Parallel Dielectric Waveguides,” Bell. Syst. Tech. J., 50, 1791-1816, (1971). D. Marcuse, “Exact Theory of TE-Wave Scattering from Blazed Dielectric Gratings,” Bell. Syst. Tech. J., 55(9), 1295-1317, (1976). S. Matsui, T. Yamato, H. Aritome, and S. Namba, “Microfabrication of LiNb03 by Reactive lon-Beam Etching,” Jap. J. Appl. Phys., 19, L463-L465, (1980). M. Minakata, “Efficient LiN b03 Balanced Bridge Modulator/Switch with an Ion- Etched Slot,” Appl. Phys. Lett., 35(1), 40-42, (1979). H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). K. Ogawa, W. S. C. Chang, B. L. Sopori, and F. J. Rosenbaum, “A Theoretical Analysis of Etched Grating Couplers for Integrated Optics,” IEEE J. Quantum Electron., QE-9(1), 29-42, (1973). V. Ramaswamy, M. D. Divino, and R. D. Standley, “Balanced Bridge Modulator Switch Using Ti-Diffused LiN b03 Strip Waveguides,” Appl. Phys. Lett., 32(10), 644- 646, (1978). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). 209 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. R. P. Ratowski, J. A. Fleck, and M. D. Feit, “Helmholtz Beam Propagation in Rib Wave-Guides and Couplers by Iterative Lanczos Reduction,” J. Opt. Soc. Am. A, 9(2), 265-273, (1992). B. E. A. Saleh and M. C. Teich, “Optical Coupling in Waveguides,” in Fundamentals o f Photonics, 264-267 (John Wiley & Sons, Inc., New York, 1991). D. R. Scifres, W. Streifer, and R. D. Burnham, “Experimental and Analytical Studies of Coupled Multiple Stripe Diode Lasers,” IEEE J. Quantum Electron., QE-15, 917- 922, (1979). S. Somekh, E. Garmire, A. Yariv, H. L. Garvin, and R. G. Hunsperger, “Channel Optical Waveguides and Directional Couplers in GaAs-Imbedded and Ridged,” Appl. Opt., 13(2), 327-330, (1974). S. H. Song, S. D. Jung, E. H. Lee, and S. G. Lee, “Back-Board Interconnections by Focusing Grating Coupler Arrays,” in the OSA Topical Meeting on Diffractive Optics, (Optical Society of America, 1994). M. S. Stem, “Semivectorial Polarised Finite Difference Method for Optical Waveguides with Arbitrary Index Profiles,” IEE Proceedings-J, 135(1), 56-63, (1988). T. Suhara and H. Nishihara, “Integrated Optics Components and Devices Using Periodic Structures,” IEEE J. Quantum Electron., QE-22(6), 845-867, (1986). M. Takeda and T. Kobuta, “Integrated Optic Array Illuminator: A Design for Efficient and Uniform Power Distribution,” Appl. Opt., 30(9), 1090-1095, (1991). T. Tamir and S. Peng, “Analysis and Design of Grating Couplers,” Appl. Phys., 14, 235-254, (1977). 2 1 0 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. C. S. Tsai, D. Y. Zang, and P. Le, “Acousto-Optic Bragg Diffraction in a LiNb03 Channel-Planar Composite Waveguide with Application to Optical Computing,” Appl. Phys. Lett., 47(6), 549-551, (1985). S. Valette, J. Lizet, P. Mottier, J. P. Jadot, S. Renard, A. Fournier, A. M. Grouillet, P. Gidon, and H. Denis, “Integrated Optical Spectrum Analyser Using Planar Technology on Oxidised Silicon Substrate,” Electron. Lett., 19(21), 883-885, (1983). D. J. Vezzetti and M. Munowitz, “Analysis of Finite Rib Waveguides by Matrix Methods,” J. Lightwave Tech., 8 (8), 1228-1234, (1990). A. Yariv, “Coupled Mode Theory for Guided-Wave Optics,” IEEE J. Quantum Electron., Q E -9,919-933, (1973). B. Zhang, S. Forouhar, S. Y. Huang, and W. S. C. Chang, “C2F6 Reactive lon-Beam Etching of LiNb03 and Nb20 5 and Their Application to Optical Waveguides,” J. Lightwave Technol., LT-2, 528-530, (1984). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Chapter 5 Embedded Lenses in Lithium Niobate Waveguides In this chapter, we discuss the background and state-of-the-art of integrated optical lenses, provide an overview of the embedded lens design and analysis, describe our embedded lens fabrication process, and present the results of embedded lens characterization. In our discussion of the embedded lens design and analysis, we consider general aspects o f the structure, such as waveguide materials and layer thick nesses, and use the specific example of the embedded lenses that we have fabricated to demonstrate key performance issues and to provide a preliminary analysis of this structure in particular. The embedded lens structure specifically considered herein is shown schematically in Fig. 5.1(a) (cross-section) and in 5.1(b) (top view). The structure consists of a lens-shaped recess in a Ti:LiNbC> 3 host waveguide, into which a thin film dielectric waveguide has been deposited. We selected thin films of silicon dioxide (SiC^) and magnesium fluoride (MgF2) to create this embedded waveguide structure. These materials were deposited by electron-beam evaporation. This lens was designed with a 1 cm aperture, a 1 cm focal length, and acircular (non-circular) interfaces using CODE V software as described in Sect. 5.2.1. In the schematic drawing in Fig. 5.1(a), the mode that is shown propagating in the host Ti:LiNb0 3 waveguide is incident upon the embedded waveguide structure. A portion of the incident light couples into the set of embedded waveguide modes. In turn, a portion o f this light couples back into the host waveguide at the second inter- 212 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (a) (b) Figure 5.1 An embedded lens in a titanium-indiffiised lithium niobate waveguide, including schematic diagrams of (a) the cross-sectional view of the lens thin film structure and the relative mode profiles for a mode-depth, d, of approximately 2 /an, and (b) the top view of a single element-aberration corrected f/1 lens design with an aperture size of W= I cm within the thin film overlay region. face. As a result of the effective refractive index difference between the two regions and the curved interface, as shown in Fig. 5 .1(b), the incident light is refracted. In an ideal embedded lens structure, the lens recess would have vertical sidewalls with min imal roughness. In the fabrication of embedded lenses in Ti:LiNb03, however, we have found vertical sidewalls difficult to attain for deeply etched recesses. Typically, the etching procedures lead to tilted sidewalls, such as those depicted in Fig. 5.1(a). The minimum tilt angle that we have achieved in our experiments is 12° from the ver tical. The recess fabrication process also leads to recess sidewalls that have a peak-to- peak roughness of about 1000 A. In the analysis that follows, we attempt to illustrate the effects of these characteristics on the embedded lens performance. The embedded S i0 2 and MgF2 film thicknesses are ideally selected to provide optimal coupling between the different waveguide structures and good confinement in the embedded waveguide. We have found that embedded lens structures with a MgF2 thickness greater than approximately 8000 A formed on LiNb0 3 are unstable and eventually fracture or peel over time. Therefore, we have sought to minimize the nec- 213 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. essary MgF2 thickness to produce a functional lens. Later, we show that a wide range of SiC> 2 and MgF2 film thicknesses can be used without significantly compromising the lens performance. In particular, we show that an embedded waveguide with a 6500 A thick MgF2 layer and a 2 fj.m thick Si0 2 layer would have a theoretical throughput of greater than 90% for an embedded lens recess with smooth, vertical sidewalls. Even with the as-fabricated slanted sidewalls, these embedded lens struc tures exceed the performance obtained by a number of other lens technologies. The issues that led to our selection of this waveguide lens technology and its advantages over previously established waveguide lens technologies are discussed next. 5.1 Review of Waveguide Lens Structures Lenses are basic components in most optical systems, and techniques to fabri cate high performance bulk lenses and lens combinations are well established. In integrated optics, however, the fabrication of a high performance waveguide lens remains a challenge. In our discussion in Chapter 1 we asserted that wide-aperture, low f-number lenses are critical components in the development of advanced IO sig nal processors. A major undertaking in the present research has therefore been to identify and develop a waveguide lens technology that is suitable for these demanding applications. The properties that define a high performance waveguide lens were discussed in Chapter 2 in the context of the IOSAR processor and IO correlator. In summary, these lens properties include a wide-aperture, a low f-number, a high throughput, min imal phase-front distortion, and low scatter. In addition, we require that the lens be manufacturable. Of course it is difficult even for a bulk lens technology to concur rently satisfy all of these requirements, since the highest performance lenses are not easily manufacturable and hence are costly. Integrated optics has the advantage that numerous planar microfabrication tools are available with which to develop a high 214 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. performance waveguide lens technology. Consequently, waveguide lenses can bene fit from the same manufacturability as do electronic semiconductor devices. Several integrated optical lens structures exhibit reasonably high throughputs and near diffraction-limited performance for high f-number operation [Chen, et al., 1977; Yao and Anderson, 1978; Chang and Ashley, 1980]. These lenses are based pri marily on perturbation of the waveguide's effective refractive index either by the use of thin film overlayers [Chang and Ashley, 1980; Valette, et a l, 1982], etching to thin the waveguide, [Chang and Ashley, 1980; Valette, et a l, 1982], or impurity indiffu sion [De Micheli, et a l, 1982; Warren, et al, 1983; Zang and Tsai, 1985; Suhara, et a l, 1986], and typically result in a 2 to 3 percent change in the refractive index. Because of the small refractive index change, these lenses are designed to operate with high f-numbers (f/4 or greater) in order to avoid phase-front distortion due to aberrations. Lens structures such as the geodesic [Chen, et a l, 1977] and Luneburg [Yao and Anderson, 1978] lenses can be made with wide apertures and low f-numbers but are difficult to fabricate, and in the case of the geodesic lens, are subject to exces sive scattering at the interfaces. Two waveguide lens technologies were identified by Rastani that provide a potentially viable approach to the fabrication of high performance lenses [Rastani, 1988]. In the first waveguide lens technology, integrated lenses are fabricated by per turbation of the waveguide refractive index, as mentioned above, through use of the titanium-indiffiision and proton exchange processes. The titanium-indiffused, proton exchanged (TIPE) lens on a Ti:LiNb0 3 waveguide utilizes standard planar microfab rication techniques and has exceptional lensing performance [Zang and Tsai, 1985]. To fabricate this structure on a Ti:LiNb03 waveguide, a lens-shaped opening is cre ated in a Si3N4 masking layer that has been deposited on the waveguide surface. Through a proton exchange process, the refractive index of the exposed lens-shaped region of the waveguide is increased so that it is approximately 0.1 higher than the refractive index o f the waveguide in the masked region. This process results in posi- 215 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tive meniscus, positive focal length lenses. To reduce the operational f-number of these lenses and maintain a wide aperture, several elements can be cascaded so that the number of refractive interfaces is increased [Rastani, 1988]. Experiments were conducted by Rastani to make improvements on this lens fab rication process with the goal of developing a suitable lens structure for advanced IO signal processors [Rastani, 1988]. One process modification consisted of switching the TI and TIPE regions. Since this resulted in a lower refractive index in the lens region than in the planar waveguide region, a negative-meniscus, positive focal length lens was created. In this process, an S i0 2 lens-shaped masking film is deposited on a planar Ti:LiNb0 3 waveguide, and the waveguide is proton-exchanged everywhere except in the lens region. An advantage to this approach is that the S i0 2 layer can be left on the surface since it has negligible effect on the lens properties and will not interfere with prism coupling. With this technique, a functional 1-cm-aperture singlet with fr6 performance and a 75% throughput efficiency, and triplet lenses with f/4 per formance were demonstrated [Rastani and Tanguay, (to be published)]. A second lens technology that was also initially investigated by Rastani is the thin-film coated recess lens [Rastani, 1988]. This lens structure may be fabricated in any host waveguide material by removing a lens-shaped portion of the host waveguide and replacing it with appropriate thin film materials to create a waveguide of different refractive index. Only standard photolithographic patterning, etching, and thin film deposition processes are required to produce such lenses, and very large index changes can be obtained. Others have demonstrated a 2 mm aperture, f/2 lens in a GaAs waveguide with a refractive index change between the GaAs and the embedded thin film waveguide of 1.9 [Minot and Lee, 1990]. This refractive index difference is much larger than that achieved with other waveguide lens technologies. In the present research, we chose to place more emphasis on the development of embedded lenses in Ti:LiNb0 3 waveguides than the TIPE lenses because of the greater potential for wide-aperture, short focal length lenses. The embedded lens has 216 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. greater potential because large refractive index differences are possible, and hence larger refraction angles can be achieved at each interface. In addition to a wide aperture and short focal length, the waveguide lens should also have a high throughput. We show herein that it is theoretically possible to achieve high-throughput embedded lenses in Ti:LiNb03 waveguides with the appro priate choice of embedded waveguide materials and structure. Very high coupling efficiencies (94%) have been shown experimentally by others for embedded waveguides in Ti:LiNb03 planar waveguides [Sopori, et al., 1980]. The technique used by Sopori, et al. resulted in planar/embedded waveguide interfaces that were straight. The recess for the embedded waveguide was created by planar sputtering with a thin cleaved GaAs wafer laid on top the lithium niobate to serve as the straight- edged mask. This technique could not be easily modified to produce curved inter faces. In the work described herein, we attempted to achieve comparably high cou pling efficiencies with the use o f photolithographic and etching techniques to produce arbitrary curved-edge interfaces for the embedded lens recess. In the following two sections, the issues regarding the selection of embedded lens thin films and waveguide structures as well as the design o f the lens shape are considered. For all practical purposes, the analysis of the structure may be divided into two portions. In the first portion of the analysis, the waveguiding nature of the embedded lens is ignored and the structure is treated as a 2-D (flat) lens that can be evaluated by simple formulas and standard ray-trace programs. In this analysis, the refractive index of the embedded thin film waveguide layer is assumed to be the lens refractive index. This approximation is sufficiently accurate to demonstrate the gen eral principles of lens design. A more accurate analysis requires the effective refrac tive index of the embedded waveguide structure to be used as the lens refractive index. In the second portion of the analysis, we ignore the lensing properties of the structure and consider the key waveguiding issues, such as mode coupling across the lens interface. 217 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.2 Geometric Interpretation of an Embedded Lens From the 2-D perspective, the first properties of the embedded lens structure that must be decided upon are the materials and lens shape. We have already estab lished that the host waveguide material under consideration is y-cut lithium niobate with a refractive index of approximately n = 2.2. Since this is a moderate value com pared with the spectrum o f available optical materials, a lens material may be selected with either a relatively higher or lower refractive index. Either a convex lens design with a positive refractive index change or a concave lens design with a negative refractive index change will result in a focusing element. The basic properties for a simple lens design are considered next to gain some insight into the issues relating to the design of an embedded lens. The focal length / of a simple thin lens is related to the refractive indices and curvature of the lens inter faces by the following relationship In this equation, the refractive index of the lens region is nL, the refractive index of the host region is and the radii of curvature of the lens interfaces are /?/ and R2. Since the refractive indices of the embedded waveguide materials that we have cho sen in this work are less than those of lithium niobate, a lens with a concave shape is appropriate to focus a collimated beam or form a real image as is required for the advanced IO processors discussed in Chapter 2. The concave embedded lens shown in Fig. 5.1(b) is shown positioned in the middle of the host lithium niobate waveguide. To establish a sign convention for each lens interface radius o f curvature, consider the left interface of the lens in Fig. 5.1(b). Light traveling from left to right that passes through the interface will be refracted. The medium to the left of the inter face is said to be the incident medium, and the medium to the right of the interface is (5.1) 218 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. said to be the transmitting medium. If the center of curvature of the interface lies in the incident medium then the radius of curvature is taken to be negative, otherwise, it’s positive. With this convention applied to the two lens interfaces shown in Fig. 5.1(b), the left interface has a negative sign and the right interface has a positive sign. The diffraction-limited focal spot size for a uniform-intensity collimated beam of width D directed through the lens is given by l y = J a L , (5.2) n*SD in which Xq is the wavelength of the light in free space and ne^-is the effective refrac tive index of the guided mode. In the case of a TE-polarized mode propagating along the x-axis of a y-cut Ti:LiNb03 waveguide, rigff is approximately ne, or the extraordi nary refractive index of the crystal. An f/1 lens fabricated in this waveguide material would have a focal spot size of approximately 0.3 /zm inside the waveguide and 0.63 /zm in air. The Fresnel reflection coefficient, R, that arises from the relative refractive index change at the lens interfaces can be approximated by the bulk optical formula for a plane wave normally incident on a uniform index discontinuity. This intensity reflection coefficient is given by R From this relation it is apparent that, for a given refractive index change, host waveguide materials with larger refractive indices will yield comparatively lower reflection losses. This relationship is illustrated further in Fig. 5.2. The reflection coefficient for a wide range of lens refractive indices in lithium niobate is less than that of a glass lens in air. Embedded lenses in SiC> 2 and GaAs waveguides may be 219 n L ~ n H n L + n H (5.3) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 0.20 — SiC fe waveguide — LiNb0 3 waveguide —- GaAs waveguide O Bulk glass lens in air 0.15 c © 'o 1 O O 0.10 c o © 0.05 © ec o.oo 1.5 2.0 2.5 3.0 3.5 Refractive Index of Lens Material Figure 5.2 Reflection loss at each interface in an integrated lens structure for differ ent host waveguide materials. realized with very large refractive index differences. For such lenses, the reflection coefficients can be considerably larger. A number of optical thin film materials are listed in Table 5.1 along with their characteristic refractive indices. As can be seen, the majority of optical thin film materials have refractive indices less than that of lithium niobate by a sizeable amount. Consequently, a large variety of embedded lens structures can be fabricated in lithium niobate with reasonably large index differences (0.2 to 0.7) and low reflec tion losses (5% or less per surface). Embedded lenses in gallium arsenide waveguides can have even greater refractive index differences, such as those shown by Minot and Lee [Minot and Lee, 1990] { A n - 1.9), but suffer from large (15%) reflection losses per surface. One possible method to reduce the interface reflection losses is to deposit anti-reflection (AR) coatings on the lens interfaces. Su and Lee employed this method in their experiments on an embedded lens structure in GaAs waveguides, and found an observable increase in throughput compared with similar 2 2 0 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. structures fabricated without an AR coating [Su and Lee, 1994]. The materials listed in Table 5.1 illustrate several interesting possibilities for embedded lens fabrication. For example, the refractive index of some thin film mate rials may be process-controlled over a wide range. This controllability can lead to precisely-tuned embedded-waveguide refractive indices or even graded-index embed ded waveguides. Other than dielectric oxides, nitrides, and fluorides, a wide range of organic polymers could be used in embedded lens structures as well. A thin film S i0 2 layer was chosen for the embedded waveguide material in the present work, since it is easy to deposit and provides a large refractive index change relative to lithium niobate without an excessively high Fresnel reflection loss. In order for light to remain confined in the embedded S i0 2 layer, the materials above (air) and below (LiN b03) should both have lower refractive indices than S i0 2. Since n(L iN b03) > « (S i0 2), a second thin film is required below the lens waveguide mate rial to act as a waveguide barrier layer. This structure is illustrated in Fig. 5.1(a), in which MgF2 (n = 1.38) is used to provide the necessary waveguide confinement. The use o f photolithographic techniques in the fabrication of embedded lenses permits the construction of arbitrarily curved interfaces. This flexibility is useful in the implementation of lenses with aberration corrections. Such lenses may be designed using standard software such as Code V and OSLO. In the present work, two lens designs were generated with Code V software. The first design is a 1-cm- aperture negative meniscus lens such as that shown in Fig. 5.1(b), for which the spherical aberrations have been minimized under the assumption that the lens mate rial is S i0 2 embedded in Ti:LiNb03. This lens was designed to have a 1-cm focal length and therefore is a single-element, aberration-corrected, f/1 lens (designated by SEACF/1). The primary purpose of this design was to investigate and demonstrate large aperture, low f-number embedded lens properties. This design was transferred to a photomask and used in the embedded lens fabrication described later in this chap ter. It also satisfies the requirements for the IO correlator described in Chapter 2 221 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Table 5.1: Thin film materials Material Method Index Si0 2 Electron-beam, Sputtering 1.46 (633 nm) A120 3 Electron-beam, Sputtering 1.63 (550 nm) T i0 2 Electron-beam, Sputtering 2.2 to 2.4 ZnO Sputtering n0 = 1.98, ne = 2.00 Ta20 5 Reactive Sputtering 1.9 to 2.2 Nb20 5 Reactive Sputtering 2.1 to 2.3 S i0 2-Ta20 5 Reactive Sputtering 1.46 to 2.08 Si3N4 PCVD 1.9 to 2.0 SiOxNy CVD 1.46 to 1.54 Coming 7059 glass Sputtering 1.53 to 1.58 (633 nm) MgF2 Electron-beam 1.38 Photoresist (KPR) Spin coating 1.615 PMMA/SAN Spin coating 1.489 to 1.563 After [Hutcheson, 1987; Nishihara, etal., 1989]. (large aperture, f/2 or better performance) and thus may be used as a prototype lens for that processor. The second Code V lens design is a 2.7 cm aperture flat-field doublet that was also made under the assumption that the lens material is Si0 2 embedded in Ti:LiNb03. The purpose of this lens design was to determine whether the available refractive indices for an embedded lens in lithium niobate would be sufficient to yield a short focal length lens structure that satisfies the IOSAR processor requirements for prototype operational values similar to those given in Chapter 2. 2 2 2 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 5.2.1 Single-Element Aberration-Corrected f/1 Lens A single element f/1 embedded lens was designed in collaboration with Gary Scheidegger o f General Dynamics Corporation to determine if the selection of Ti:LiNb03 and SiC> 2 as waveguide materials could lead to a lens with diffraction-lim ited performance. We chose a 1-cm-lens aperture as our benchmark since this corre sponds to the aperture size required for the IO correlator example discussed in Chapter 2. This aperture width is also approximately the size of the largest aperture waveguide lenses previously reported [Wood, et al., 1983]. The shape of this lens is illustrated in Fig. 5.3(a). The initial radii of curvature r of the two lens interfaces were chosen to minimize the lowest-order lens aberrations. This minimization was done under the assumption that the Ti:LiNb03 and S i0 2 mate rials are homogeneous media with refractive indices of 2.23 and 1.5, respectively. Each interface was then made acircular to minimize the higher-order aberrations for an on-axis focus. The acircular interface was specified with respect to the coordinate system shown in Fig. 5.3(b) and by the equation: 2 = j & ----------- j -2 + A y4 + By6 + Cy8 + Dy'°< (5.4) l + V l-(l + * )C „y in which C0 is the interface curvature and is equal to the inverse of the radius r (i.e., C0 ~ Hr), k is the conic constant, and A, B, C, and D are the acircular coefficients. The values of the necessary design parameters that describe each interface are given in Table 5.2. The expected performance of this lens design was determined with the Code V software. The effective focal length is 10 mm with an f-number of f/1.02 and a Petz- val radius of -6.7 mm. The peak-to-valley optical path difference (P-V OPD) is a measure of the deformation of the wavefront after the lens relative to a reference sphere and is used to describe the focal quality of a lens. For a P-V OPD smaller than 223 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1 cm 100 fim n = 2.23 n = 2.23 n = 1.5 (a) Surface Figure 5.3 Shown in this figure is (a) the design for a single-element aberration-cor rected f/1 lens in lithium niobate and (b) the coordinate system used for the definition of the acircular interfaces. 1/4 wave, the focal spot can be considered diffraction-limited. For the lens design described above, the P-V OPD was found to be 9 X 10‘5 waves, which indicates that it is well within the diffraction-limited regime. As will be described in Chapter 7, the SEACF/1 lens design was used in the fab rication of the integrated test modules. With these test modules, we tested the ability of the lens to focus light into a single rib waveguide within an array. For efficient focusing in a single rib waveguide, the focal spot width should be much smaller than the rib waveguide width. As we found earlier, a uniformly-illuminated 1-cm-aper- ture, f/1 Si0 2/MgF2 embedded lens in lithium niobate will have a focal spot size of 0.3 /im for A > = 6328 A. This focal spot size is much smaller than the rib waveguide width used in the integrated test modules (6 /im) and should result in efficient excita tion of a single rib waveguide. In the IO correlator, the light deflected by the surface acoustic wave is imaged onto the rib waveguide array in order to accomplish the signal processing task. The quality of this image is described by the convolution of the lens point spread function 224 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.2: SEACF/1 Lens Design Parameters Symbol Surface I Surface 2 r -13.71718 mm 4.28438 mm k -0.67518 -0.255221 A -0.528026 x IO ' 6 mm"3 0.914978 x 10' 5 mm"3 B 0.125034 X 10"7 mm"5 0.130850 x 10' 6 mm"5 C 0.934063 x 10"9 nun"7 0.139849 x 10' 8 mm"7 D 0.271198 x IO"10 mm' 9 0.548266 x 10' 10 mm' 9 (focal spot intensity profile) with the object field amplitude distribution. In order to prevent the loss of image resolution, the lens focal spot width should be smaller than the smallest feature size in the object. For the example given in Chapter 2, a 60 MHz transducer bandwidth is employed on a LiNbC> 3 substrate. This signal bandwidth in the object plane of the imaging lens combination corresponds to a minimum feature size of 60 iim (acoustic velocity divided by the bandwidth) in the object intensity dis tribution (two rib widths). This feature size is much larger than the focal spot size of the SEACF/1 lens, which implies that this lens design is sufficient for the imaging operation required in the IO correlator. 5.2.2 Flat-Field Doublet The IOSAR processor described in Chapter 2 required a long SAW time aper ture and hence a wide lens aperture in order to process the largest possible range swath. For the example given, a 2.7 cm aperture lens was required to process a 1 km range swath. A major concern in the use of this large element was the degree of field curvature present in the focal plane. Field curvature can lead to significant range error at the rib waveguide array entrance plane. This error can be avoided by two approaches; either the rib-waveguide-array front entrance plane could be modified in 225 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. order to match the Petzval surface of the lens, or a second field lens could be included in the lens design to provide a flat-field image. The second approach introduces more degrees of freedom into the lens design and allows for higher-order aberration correc tion. The design of a flat-field aberration-corrected doublet was also done in collabo ration with Gary Scheidegger of General Dynamics. The final design, shown in Fig. 5.4, consisted of an acircular first interface and circular interfaces for the remain der of the lens doublet. The materials were assumed to be Ti:LiN b03 (n = 2.23) and S i0 2 (n = 1.44) in the host waveguide and lens region, respectively. The second lens element is a positive meniscus with a negative refractive index difference, and hence is a defocusing element. A defocusing element is used in combination with a focus ing element because the signs of the spherical aberrations are opposite, and hence are used to cancel one another. The interface curvatures and spacings for this lens design are given in Table 5.3. A performance analysis was done with Code V software to determine the lens doublet properties. The effective focal length is 4.2 cm with an f-number of 1.56 and a flat field of focus. In Fig. 5.4, three chirped SAW signals are shown positioned 4 mm in front of the lens doublet (at the center of the lens). A SAW transducer with a 425 MHz center frequency and a 100 MHz bandwidth was assumed to be used to pro duce the 3.5 mm long signals. Since the assumed bandwidth for this design is double that considered in the example given in Chapter 2 and the lens focal length is much shorter, the focal spot size of the range-focused image (4 /im ) for this design is roughly half that derived in Chapter 2. The diffracted and undiffracted beams that result from uniform illumination across the entire SAW path are shown in Fig. 5.4 focused by the lens doublet. The diffracted light comes to three distinct focus points along the plane of the rib waveguide array, and the undiffracted light comes to a focus before the rib waveguide array. The undiffracted light is spatially separated from the diffracted light by approximately 1.5 mm. This separation should be sufficient for the 226 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chirped SAW signals Rib waveguide array Focus of undiffracted light ^ -------------------------- 4.2 c m ------------------------ ►! Figure 5.4 Design for a field-corrected lens doublet with a 1 inch aperture showing three chirped SAW signals and the resulting focal positions of the diffracted light on the entrance plane of a rib waveguide array. Also shown is the focal position of the undiffracted light. spatial filtering operation. The point spread function of this lens is presented in Chap ter 7, where it is used in theoretical calculations of the coupling efficiency into a sin gle rib waveguide within an array and into the neighboring rib waveguides. 5.3 Guided-Wave Interpretation of an Embedded Lens The design of an appropriate embedded lens structure for lithium niobate involves the selection of dielectric materials and waveguide layer thicknesses. Throughout each stage in the design, the limitations of the fabrication processes must also be considered. The fabrication limitations that we encountered in our develop ment of the embedded lens are presented below. First, however, we provide a general description of the issues involved in the propagation of a guided wave through an 227 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Table 5.3: Flat-Field Doublet Design Parameters Symbol SI S2 S3 S4 r -30.3976 mm 13.29187 mm 15.03197 mm -27.7431 mm d 0.834 mm 14.760 mm 2.482 mm k -1.866942 A 1.24459 X 10‘6m nf3 B -5.27263 x 10'9mm' 5 C 5.20915 x 10‘13mm' 7 D -2.66915 x 10‘14mm' 9 embedded lens structure. In Fig. 5.5, the embedded lens structure is shown divided into several regions of interest. These include the host waveguide Regions I and VII, the thin film overlayer Regions II and VI, the host/embedded waveguide interface Regions HI and V, and the embedded waveguide Region IV. A mode that propagates in the host waveguide in Region I towards the lens structure will undergo changes as it passes through the suc cessive regions. In Regions I and VII the host waveguide structure consists of a waveguide layer, a cladding layer (air), and a substrate with a refractive index lower than that of the waveguide layer. Depending on the structure, this waveguide may support a single mode or many modes with either TE or TM polarization. As a possible outcome of the lens fabrication process, the host waveguide may have a cladding layer that consists of the embedded waveguide materials in Regions II and VI. If the thin film overlayer has a refractive index higher than that of the host waveguide, the guided mode will spread up into the thin film (see Fig. 5.6(a)). If the refractive index o f the thin film overlayer is lower than that of the host waveguide, the mode will remain confined in the host waveguide but will have a modified mode 228 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. © Overlayers mbedded W aveguide* Host Waveguide Incident Mode - ^ -JB H arrie r Layer - . _•£, 'w ‘ . s- ^Substrate 4 < f “ V » § < ■ ? - . -as^isi Figure 5.5 Schematic diagram of the embedded lens structure divided into several regions for consideration of the guided wave properties. These include: the host waveguide Regions I and VII, the thin film overlayer Regions II and VI, the lens inter face Regions III and V, and the embedded lens Region IV. shape and propagation constant (compared with that in Regions I and VII). For a two-layer embedded lens structure with a thin barrier film such as that shown in Fig. 5.6(b), the light could couple up through the barrier layer and become trapped in the upper film. In Regions HI and V, the light in the waveguide passes through a host- waveguide/embedded-waveguide interface. The Fresnel reflections at the interface are characterized by the refractive index difference between the two waveguide mate rials. The mode coupling that occurs is determined by using the mode overlap inte gral [Nishihara, e ta i, 1989] % = f E l H(x)E[* (x)d x J — oo U fo o ; 2 f ° ° ; 2 f e h (x )d x \ E j l (x )d x J — oo J — oo (5.5) 229 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. (a) t -^-: C *^vT ? '■ * ■ „ ^ -.-. * “ ' v -ts • •^ •.^ •--s y v y * ' .* ' - : * i . . \ ^» . K --** ■ < • -.T IT . • « > • : . « • rf> p ^ r ■ * ’ f v -^ (b) Figure 5.6 Comparison of (a) higher index and (b) lower index embedded lens structures. The lower index structure has small losses associated with leakage through the barrier layer(s). The higher index structure has essentially no mode leak age as long as film deposition onto the host waveguide surface is prevented. In this expression, Elff and EJL designate the electric field distributions of the /th host waveguide mode and the yth embedded lens waveguide mode, and T fo j is the fraction of the optical power transferred between these modes under the assumption of smooth, vertical interfaces. Inter-mode coupling is likely to occur at each interface unless the embedded lens structure is carefully designed to prevent this. In general, inter-mode coupling is not desirable since it will lead to multiple lens focal planes. The occurrence of focus ing at several focal planes is evident if the effective refractive index difference of each mode combination is considered (i.e., n°H- n ° L, n°H- n ' L, n ‘H- n ° L, etc., in which the subscripts and superscripts denote the mode number and the waveguide regions, respectively). 230 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. In Region IV, the embedded waveguide structure consists o f a waveguide layer, a cladding layer (air), and a substrate layer. As previously mentioned, a low-index barrier layer may be required if the substrate refractive index is greater than the embedded waveguide layer. The embedded waveguide may also be multi-mode and support both TE and TM polarizations. For an embedded lens with a barrier layer, light in the guided modes may leak through the barrier and into the substrate at a rate dependent upon the barrier layer thickness (see Fig. 5.6(b)). 5.3.1 Design of the B arrier Layer Thickness The general guided-wave description of the embedded lens in the last section may be reduced in scope with the selection of the single-mode Ti:LiNb(>3 waveguide described in Chapter 3 as the host waveguide, and the S i0 2/MgF2 thin-film combina tion as the embedded waveguide. Since the host waveguide only supports a single mode, the light coupled back into the host waveguide can only excite one mode, regardless of the mode coupling that occurs at the first lens interface. The lens, how ever, can still have multiple focal lengths if several of the embedded waveguide modes are excited. With the proper choice of the S i0 2 and MgF2 layer thicknesses and the lens recess depth, the coupling of light into the higher-order embedded waveguide modes and the transmission of these modes through the embedded lens region can be greatly reduced. To demonstrate this, consider an S i0 2/MgF2 waveguide combination with an S i0 2 layer thickness of 2 ftm . For this choice of thickness, the zeroth order mode depth in the Si0 2/MgF2 waveguide approximately matches the mode depth found for the Ti:LiNb03 single-mode waveguide. The leakage rates through the MgF2 barrier layer for thicknesses from 0.5 to 1.0 fim were calculated for each o f the guided modes using the numerical method for waveguide analysis introduced in Sect. 3.3. The results of these calculations for an operational wavelength of Xo = 6328 A are shown in Fig. 5.7. In general, the leakage rates are lower for larger barrier-layer thicknesses 231 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Si0 2 thickness = 2 pm 100 E o C D 3 c o c o m = 2 3 m = 1 c o as O) as C L 2 C L m = 0 0.5 0.6 0.7 0.8 0.9 1.0 MgF2 Thickness (/im) Figure 5.7 Embedded Si0 2 /MgF2 waveguide leakage rate for each of the modes as a function of the barrier layer thickness. and are higher by approximately an order of magnitude for each higher mode number. The increased attenuation of the higher-order modes is a useful feature of embedded lens structures that employ a barrier layer. Essentially, this feature can be used to filter out light that is coupled into the higher-order modes of the lens. For example, consider the SEACF/1 lens design discussed in Section 5.2.1 with a lens thickness o f 3 mm near the extremes of the lens aperture. If an embedded lens is fab ricated with a 6500 A barrier layer, the attenuation due to leakage through the barrier layer for each of the modes as they pass through that portion of the lens will be 0.9 dB, 6 dB, and 30 dB leading to a 81%, 25%, and 0.1% transmission for the m = 0, 1, and 2 modes, respectively. At the center of the lens, the lens thickness is much smaller (0.1 mm) so the effect is greatly reduced (1 dB for the m — 2 mode). It would be possible to design a lens with a greater width at the center to provide a high degree of mode filtering across the entire lens aperture. In the final design of the embedded 232 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. lens structure, however, the barrier layer should not be made so thin or the width of the lens made so large that there is excessive attenuation of the lowest order mode. Although the mode filtering function described above will increase the amount o f light scattered into the substrate, a much smaller portion of the substrate-coupled light will find its way to the detector array than if it were left in the waveguide. The light will spread out over a much greater volume in the substrate, and a large portion o f the light will be transmitted through the top and bottom of the substrate after a few reflections. As we mentioned in our description of the guided-wave picture of the embedded lens structure, light passing beneath the thin film overlayers (Regions II and VI) can possibly couple up through the barrier layer (MgF2) and into the waveguide layer (S i0 2). The degree of coupling that occurs for an S i0 2 layer thickness of 2 fim was calculated for various MgF2 layer thicknesses and found to be negligible even for bar rier layer thicknesses as small as 1000 A. The case described here for coupling between two well confined modes is very different from the case described above for coupling from the embedded waveguide into the many unconfined modes o f the bulk lithium niobate substrate. As a result, the total attenuation for a guided mode propa gating through the embedded lens structure in Fig. 5.5 is primarily determined by the propagation loss of the thin-film waveguide and the leakage rate through the barrier layer in the embedded lens region. In the SEACF/1 lens design, the propagation distance of a ray through the lens increases for input ray positions farther away from the center of the lens. As a result, the propagation losses will be greater for beams displaced from the optical axis. The lens throughput (not considering reflection and mode-coupling losses) is shown graphically in Fig. 5.8 for different embedded waveguide attenuation constants (due to waveguide attenuation and mode leakage) under the assumption of a collimated beam input to the lens. The throughput is high at the lens center and drops off toward the outer part of the lens aperture. 233 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 0.9 X- c o c /5 < n 0.8 e C O c £ 5 0 .7 0.6 0.5 -4 •2 0 2 4 Position (mm) Figure 5.8 Embedded Si0 2 /MgF2 waveguide lens transmission as a function of beam position in the aperture, for different waveguide attenuation constants (given in dB/cm). In our fabrication process for the embedded lenses, S i0 2 thin films were elec- tron-beam evaporated at room temperature. Such films have a columnar structure and a packing density less than unity [Pulker, 1969]. In earlier work on electron-beam deposited S i0 2 films for bulk waveguide applications, Hutcheson found that S i0 2 waveguides made by this technique have propagation losses of approximately 2 dB/cm [Hutcheson, 1987]. The higher propagation loss of S i0 2 waveguides com pared with the Ti:LiNb0 3 waveguides fabricated for the present work was obvious by the appearance of a relatively brighter guided streak in the S i0 2 waveguide. A barrier layer thickness was selected for the lenses that were fabricated such that the total throughput losses (due to waveguide attenuation and leakage) were acceptable. The barrier layer was made as thin as possible due to fabrication limita tions that will be discussed later. We selected a barrier layer thickness of 6500 A, which leads to a leakage loss o f about 3 dB/cm as determined with the aid of Fig. 5.7. 234 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. The total estimated propagation loss of the S i0 2 embedded waveguide is 5 dB/cm when the attenuation loss is added to the leakage loss. As seen from Fig. 5.8, this total propagation loss results in about a 35% variation in the lens transmittance across the aperture. If the total propagation losses are reduced to 1 dB/cm, there will be less than a 10% variation in the lens transmittance. These values for lens throughput do not include the effects of Fresnel reflections and mode decoupling at the lens interfaces due to mode mismatch. These effects will be taken into account in the next section as we consider in particular the selection of the embedded lens waveguide layer thickness and recess depth. 5.3.2 Optimization of Mode Coupling The mode coupling that occurs at the host/embedded waveguide interfaces will preferentially occur between the lowest order modes o f the two waveguides if both the embedded waveguide thickness and the embedded waveguide offset with respect to the host waveguide are properly selected. In the last section it was shown that for an embedded lens with film thicknesses of 2 /an (S i02) and 6500 A (MgF2), the throughput uniformity for the lowest order embedded lens mode is reasonable (30%) and the higher-order modes are strongly attenuated. We will show in this section that the selection of a 2 /an thick S i0 2 waveguide layer also provides both a reasonable mode-coupling efficiency at the lens interfaces with the single-mode Ti:LiNbC> 3 waveguide and sufficient waveguide confinement. Highly efficient coupling between the host waveguide and embedded waveguide structure is achieved by matching the mode profiles of the lowest order modes in the two structures. The coupling efficiency between the host waveguide and embedded waveguide was calculated using Eq. 5.5 for the lowest order mode in sev eral S i0 2/MgF2 waveguide structures as a function of waveguide offset and for sev eral different Si0 2 layer thicknesses. In these calculations, we assumed that the 235 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. thickness of the MgF2 layer was infinite, the operational wavelength was Aq= 6328 A, and the Ti:LiNb03 waveguide mode profile was that calculated in Chap ter 3. The family of curves that was obtained is shown in Fig. 5.9. Greater than 90% coupling efficiency can be achieved for Si02 layer thicknesses ranging from 1 fim to 2.4 fim as long as the waveguide offset is properly adjusted. This large range of thicknesses allows exceptional flexibility in the lens waveguide design. The optimal thickness of the S i0 2 layer is between 1.4 fjxn and 1.6 fjm (99% mode coupling effi ciency) while the sensitivity to variation in waveguide offset decreases for thicker films. The barrier layer thickness required to provide waveguide confinement in the embedded lens also varies as a function of waveguide thickness. To demonstrate this dependence, the propagation losses of the lowest order mode of a S i0 2/MgF2 embed ded lens due to leakage were calculated as a function of barrier layer thickness for dif ferent S i0 2 thicknesses with the results shown in Fig. 5.10. Evident from this figure is that a thicker barrier layer is required to maintain a low propagation loss for thin S i0 2 waveguide films. Thin film structures with a propagation loss of 3 dB/cm or less can be achieved with total embedded waveguide thicknesses ranging from 1.9 fim to 2.8 fjm for the range of film thicknesses shown in the figure. In the fabrication of embedded lens structures it was found that good quality MgF2 films could only be achieved for thicknesses less than 8000 A. A fair compromise in lens performance with thin film quality can be achieved through selection of a barrier layer thickness of 6500 A and an Si0 2 waveguide layer thickness o f 2 jjm. With this combination, a peak coupling efficiency of 96% is possible with a very slight offset between the host and embedded waveguides (see Fig. 5.9). With this selection of embedded waveguide film thicknesses, the variations in mode coupling efficiency between the single-mode Ti:LiN b03 waveguide and each of the S i0 2/MgF2 modes were calculated as a function of the offset distance of the two waveguide surfaces. The results of these calculations, shown in Fig. 5.11, illustrate 236 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 >» o c 0 3 'o £ 0.9 L U O ) c Q. o 0.8 O 0.7 - 0.6 -0.4 - 0.2 0.0 0.2 0.4 0.6 Waveguide Offset (microns) Figure 5.9 Mode coupling efficiency for various Si02 layer thicknesses (ranging from 1 pm to 2.4 pm) as a function of waveguide structure offset. the drop-off in coupling that occurs with the higher-order modes when the lowest order mode is optimally coupled. From this graph, it is evident that the maximum coupling to the m = 0 mode will occur if the embedded waveguide surface is 500 A above the Ti:LiNb03 surface. The fabrication tolerance in the embedded lens offset is determined by the acceptable amount of light coupled into the higher-order modes (and the associated decrease in the light coupled into the lowest order mode). An error of ±3000 A in the waveguide offset results in less than 2 0 % coupling into any of the higher-order modes. Since this coupling occurs at both boundaries of the lens, the total amount of light coupled into a higher-order lens mode and back into the Ti:LiN b03 waveguide is less than 4% of the initially incident light for this range of offsets. The increased attenuation of these higher-order modes in the embedded waveguide region due to leakage through the barrier layer will decrease this value even further. 237 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. £ C D ■ D Si02 Thickness frzm) w (0 o 1.0 c o O ) co Q. 1.4 2 O. 1.8 2.2 0.1 0.6 0.5 0.7 0.8 0.9 1.0 MgF2 Thickness (/cm) Figure 5.10 Propagation losses of the lowest-order mode of an S i0 2/MgF2 waveguide for various Si02 layer thicknesses and as a function of the barrier layer thickness. The measured mode profile o f the Ti:LiNb03 waveguide (described in Sect. 3.5.2) and the calculated mode profile of the 2.65 /cm thick S i0 2/MgF2 embedded waveguide are shown in Fig. 5.12 for comparison. The mode profiles for these two structures are very similar except for the long evanescent tail of the Ti:LiNb03 waveguide mode that penetrates into the substrate. A slight offset in the depth of the peak intensities of each mode is evident when the profiles are aligned so that the air- boundary side of the profiles overlap. In the measurement of the Ti:LiNb03 waveguide mode profile, the position of the air/waveguide boundary has an uncer tainty o f about ±1 /cm so that the relative position of these peaks is not certain in prac tice. 5.4 Fabrication Limitations and Related Effects In the analysis presented in the previous sections, we assumed ideal embedded 238 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96% Coupling 1.0 n < p N 0.8 (0 B w m = 0 m = 1 m = 2 0 ,c © 1 s ° - 4 ffi Q. 3 ° 0 2 O 0.0 l.O L-1-1— ■ ■ I - »-i i.i I i'f-i ii.j:-ii ‘)'i.J.> ‘i i i I i i i -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Waveguide Offset d (pim) Figure 5.11 Coupling efficiency between the Ti:LiNb03 mode and each of the Si0 2/MgF2 waveguide modes as a function of offset between the waveguide struc tures. These calculations were performed assuming a 2 fim thick Si02 waveguiding layer. lens characteristics such as vertical and smooth-recess sidewalls, homogeneous mate rials, and an abrupt and well-defined transition from the host waveguide to the embedded waveguide. As we describe below, processing limitations require several modifications to this ideal picture of the embedded lens structure. The primary depar ture of the fabricated embedded lens structure from the ideal structure that we have observed experimentally is the tilted lens recess sidewall that develops as a result of the etching process. The tilted sidewall between the host and embedded waveguide structures is shown in Fig. 5.13(a). Both the Fresnel reflection coefficient for a guided mode incident upon the tilted interface and the mode coupling through the interface vary as a function of the tilt angle. In this section, we consider the depen dence of Fresnel reflections and mode coupling across the lens interface on the tilt angle of the interface. We also provide an estimate of the effects of roughness in the lens sidewall, and of the effect of process variations on the lens focal length. Our ear- 239 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 x Waveguide Mode Profile Ti:LiNb0 3 (Measured) - \ . Si02/MgF2 (Calculated) ' 0.8 £ a 3 . 0.6 £ & 0.4 < / > c c. 0.2 0.0 0 1 2 3 Depth (fim) Figure 5.12 Comparison of measured mode intensity profile for the titanium-indif fused lithium niobate waveguide and the calculated mode profile for the S i02/MgF2 embedded waveguide structure with a 2 fim thick waveguiding layer. tier assumption of a homogeneous host waveguide material was useful to simplify the earlier discussion. In reality, LiNbC> 3 is a uniaxial crystal, and hence the waveguide refractive index depends on the polarization state and propagation direction o f the guided mode. In this section, we also investigate the effect of waveguide anisotropy on the lens focal spot size and initial lens design. We conclude this section with an estimate of the total embedded lens throughput given the results obtained from our analysis. 5.4.1 Fresnel Reflections at a Tilted Interface To first order, we assume that the geometrical-optics approximation of the guided mode as a plane wave can be used to determine the Fresnel reflection depen dence on the sidewall tilt. In this approximation, the plane wave can be defined by a ray vector that is perpendicular to the plane-wave phase front. The ray angles are cal- 240 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. (f) = k0A n y tan(y) (a) (b) Figure 5.13 Schematic diagram showing (a) the slanted sidewall in the embedded lens recess, and (b) a thin prism model of the slanted sidewall. culated for each mode using sin6q= N ^/ng, where ng is the refractive index of the waveguide layer, and Ne j j is the effective refractive index of the mode and 6q is the angle of the ray with respect to the surface normals of the waveguide interface. The ray direction of the plane wave propagating in the Ti:LiNb03 waveguide is nearly parallel to the waveguide top and bottom surfaces as shown in Fig. 5 .13(a). A plane wave incident upon an interface that has a tilt angle 7 with respect to the normal of the waveguide surface will be partially transmitted and partially reflected. The reflectivity of a Ti.LiNbCtySiC^ interface with a tilt angle 7 is shown in Fig. 5.14 for TE-polarized light. The reflectivity is fairly small (between 4% and 10%) for sidewall tilts of up to 30°. Thereafter, the reflectivity increases rapidly until the light is totally internally reflected at a sidewall tilt of 41°. The reflected light will remain confined in the host waveguide for small sidewall tilt angles and couple into the substrate for large sidewall tilt angles. 5.4.2 Mode Coupling at a Tilted Interface In the ray-optics approximation, the light transmitted through the interface will be refracted in the vertical plane as shown in Fig. 5.13(a) as a result of the sidewall tilt. We expect that the transmitted light will be coupled most strongly into the 241 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. < b - <I> oc 0.01 0 10 20 30 40 Wall Tilt Angle (degrees) Figure 5.14 Reflectivity at the TirLiNbCtySiC^ interface as a function of lens recess wall tilt angle. embedded waveguide mode that has a similar propagation angle. For large sidewall tilt angles, and thus large refraction angles, the transmitted light may propagate at an angle larger than the TIR condition at the S i0 2/M gF2 interface and subsequently leak from the waveguide. The angle of propagation 6t o f the refracted light in the thin film is shown in Fig. 5.13(a). The value of 0; as a function of the wall tilt angle is com pared with the propagation angles of the S i0 2/M gF2 waveguide modes in Fig. 5.15. To first order, we conclude that a sidewall tilt angle o f less than 15° is needed in order to get significant coupling into the lowest order mode of the embedded waveguide. For this sidewall tilt angle, the refracted beam propagates at an angle close to that of the lowest order embedded waveguide mode. The behavior of the mode coupling through a tilted interface as a function o f tilt angle can be estimated if the interface is modeled as a thin prism that induces a depth- dependent linear phase shift in the incident guided mode. The prism is drawn sche matically in Fig. 5.13(b). We calculated the coupling efficiency, neglecting reflec- 242 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 W 0) 2 85 = —m = 0 O ) < D TJ O ) 5 75 I 70 o 65 — m = 2 ® G C (0 60 E 0) c 55 10 0 20 30 40 Wall Tilt Angle (degrees) Figure 5.15 Refraction angle of the light transmitted into the lens region as a func tion of wall tilt angle. The refracted angle is compared with the waveguide propaga tion angles. tions, of a guided mode in the host Ti:LiNb03 waveguide that was multiplied by the depth dependent phase shift, exp(-1000), into an embedded S i0 2/MgF2 waveguide with a 2 fjm waveguide layer thickness. The results of these calculations for several sidewall tilt angles, shown in Fig. 5.16, indicate that the coupling efficiency decreases for an increase in the sidewall tilt angle, and the waveguide offset that produces the peak coupling for a given wall tilt angle shifts towards a negative offset (embedded waveguide top surface sunken below the host waveguide top surface) as the wall tilt angle is increased. This shift is indicative of the prism effect that we described in the ray-optics analysis; that is, the tilted interface acts as a prism that leads to a linear phase shift in the wave passing through it, which corresponds to a redirection of the wave in the manner described by the refraction of a light-wave incident at an angle upon an interface. If the sidewall tilt is excessively large, the light will tend to couple less efficiently into the lowest order mode of the embedded waveguide. 243 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. >s o c a > t : LU C D C a . 3 O O 1.0 Wall tilt angle 0.8 0.6 0.4 0.2 o .o -2 -1 0 Waveguide Offset {/im) Figure 5.16 Effect of a tilted sidewall on coupling efficiency into the S i0 2/MgF2 embedded waveguide with a 2 ftm thick guiding layer. In the fabrication o f the embedded waveguide structure, the smallest observed sidewall tilt angle was approximately 12° from the normal. In order to anticipate the effect of the sidewall tilt on the coupling into higher-order embedded waveguide modes, the coupling efficiency for this sidewall tilt angle was calculated for these modes. The results shown in Fig. 5.17 indicate that coupling into the higher-order modes has increased substantially. Therefore, a tilted sidewall may lead to a signifi cant fraction of the incident light coupled into the higher order modes. Since the attenuation of the higher order modes is large in the embedded waveguide, light cou pled into these modes will likely not lead to multiple focal planes for the lens, but rather, the overall throughput of the lens will be lowered. 5.4.3 Interface Roughness Scattering Another fabrication-related effect that can lead to embedded lens throughput loss and background noise is the scattering of light at the embedded lens interfaces 244 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 m = 0 — m = 1 m = 2 > . o c .2> o '* = 1 1 1 o> c 0.6 0.4 Q_ 3 O o 0.2 0.0 -2 - 1 0 1 2 Waveguide Offset (/im) Figure 5.17 Mode coupling efficiency from a Ti:LiNb0 3 waveguide into an embedded Si0 2 /MgF2 waveguide with a 12° interface tilt angle. due to interface roughness. This scattering is a major concern in the case of the IO correlator, in which four lens elements are needed and hence there are eight interfaces at which scattering can occur. The effects of interface roughness can be approximated by a uniform sinusoidal corrugation with the mean period and peak-to-valley distance estimated from scanning electron microscope observations of the interface. Light transmitted through the rough interface is modulated in phase. A simple model for diffraction from a uniform phase grating, used herein to approximate the scattering losses, is found in Goodman’s book [Goodman, 1968]. To represent the physical interface as a phase grating, the recess sidewall is approximated as an infinite planar interface between the two materials with refractive index n = 2.2 (TirLiNbC^) on one side and n = 1.46 (Si02) on the other. The interface is assumed to have a sinusoidal undulation in one direction that approximates the predominant amplitude (1 0 0 0 A peak-to-peak) and period (1 fjm) of the roughness observed in fabricated embedded lens recess sidewalls. A plane wave (A = 6328 A in free space) at normal incidence 245 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. on this interface is diffracted by the sinusoidal phase modulation. Most of the trans mitted light remains in the zero-order plane wave and continues to propagate along the same direction as the incident beam. Two diffracted plane waves travel at ±15° with respect to the zero order beam. The intensity ratio o f the higher-order beams to the zero order beam is 0.033. The diffracted light will propagate at a small angle with respect to the transmitted light. If this diffracted light is present at the detector plane in a processor, for example, it will contribute to the background noise. Therefore, we conservatively estimate that lenses fabricated with similar sidewall roughness will exhibit background scattering approximately -15 dB below the signal level. In con sideration of the previous analysis for coupling o f light through an embedded waveguide lens, however, it is reasonable to assume that the transmission of the off- axis scattered light through the lens will be poor. A significant portion of the scattered light will be coupled into the substrate rather than into the rib-waveguide array at the end of the processor. Thus, the light scattered at the interfaces will act primarily as a throughput loss and not a noise source. The exception to this is the scattering that occurs at the last lens interface where no further lens coupling losses or embedded waveguide losses can be incurred before the light reaches the rib waveguide array. 5.4.4 Tolerance Analysis In order to establish the manufacturability of embedded lens structures, it is essential to determine whether a designed lens can be reliably and repeatedly fabri cated with the required performance characteristics. The performance of the advanced 1 0 signal processors described in this thesis will be sensitive primarily to the lens focal length. Therefore, the focal length variation was determined as a func tion of the estimated process variations. The variations of the initial host waveguide’s properties were discussed in Chapter 3. From that analysis, it was determined that the primary change in the char acteristic refractive index of a single-mode waveguide was due to the variation o f 246 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. thickness in the initial titanium deposition layer in that process. In the worst case, the effective refractive index of the single-mode Ti:LiNb03 waveguide will vary by approximately ±0.05% (or n ^ = 2.207 ±0.001) for a variation in titanium thickness from deposition to deposition of ±10%. For the titanium layer thickness used in our fabrication processes described in Chapter 3, this percentage variation corresponds to a ±25 A variation in thickness. This titanium thickness variation is well within our process capabilities. In the manufacturing environment, the degree of control may be somewhat better than this (only a few percent variation). Similarly, it was found that the effective refractive index of the dielectric embedded waveguide is most sensitive to the refractive index of the guiding layer. A variation of ± 1% in the refractive index of the S i0 2 waveguide layer in an Si0 2/MgF2 structure, which is a reasonable value for the electron-beam-deposited films, will lead directly to a ±1% variation in the effective refractive index (ne g = 1.46 ± 0.015). The effective refractive index of the dielectric embedded waveguide is much less sensitive to a variation in the thickness of the waveguide layer. A variation of ±10% in the S i0 2 waveguide layer thickness in the S i0 2/MgF2 structure (2 fim ± 0.2 fim) will lead to a ±0.07% variation in the effective refractive index (ne j f = 1-46 ± 0.001). The effective refractive index of the dielectric embedded waveguide is very insensitive to the refractive index of the barrier layer. For example, a variation of ±1% in the refrac tive index of the M gF2 barrier layer in a S i0 2/MgF2 structure will lead to a ±0.005% variation in the effective refractive index (ne ff~ 1-46 ± 0.00007). The shift in focal length that results from a variation in the refractive index of the host waveguide can be determined by differentiation o f Eq. 5.1 with respect to n The resulting relation for the refractive index variation in the host waveguide is d f nr ditfj -7 = ---- -------- • (5.6) / nL - n H nH 247 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Therefore, for a ±0.05% variation in the host waveguide index due to initial titanium layer thickness fluctuations, the focal length of the 1 cm aperture f/1 lens will vary by ±0.1% (calculated using rtL = 1.46, n H = 2.207, dnff/nfi = ±.Q005). In other words, the position of the focal plane will be shifted by ± 1 0 fim , with a lower host waveguide effective refractive index leading to a longer focal length. Similarly, for the embed ded waveguide, the variation in focal length is given by M- =_________________________________(5.7) f nL - n H nL In this case, for a ±1% variation in the embedded waveguide index due to material fluctuations in the deposited Si0 2 , the focal length of the lens will vary by ±2%. In other words, the position of the focal plane will be shifted by ± 2 0 0 fim, with a higher lens waveguide effective refractive index leading to a longer focal length. O f the parameters considered, this is clearly the parameter that should be most carefully con trolled. 5.4.5 Waveguide Anisotropy A final consideration in the use of y-cut lithium niobate as the waveguide medium for the fabrication of low f-number lenses is the effect of the waveguide anisotropies on the lens performance. In order to evaluate this effect for the SEACF/1 lens design used in the fabrication of embedded lenses, a ray tracing program was developed. This program was written so that the effects o f anisotropy in the refractive index of the host material could be incorporated into the ray trace. The SEACF/1 design was first evaluated with the nominal refractive indices («(Ti:LiNb03) = 2.23, n (S i0 2) = 1.5) under the assumption that the Ti:LiNb03 waveguide is an isotropic medium. The focal spot produced is shown in Fig. 5.18(a) and is a nearly perfect on-axis geometrical focus for a collimated beam input. This result is what we expect since the lens design was optimized for these refractive indi- 248 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 10.02 10.04 10.06 10.08 10.10 Position (mm) 9.72 9.74 9.76 9.78 9.80 Position (mm) (c) 9.66 9.68 9.70 9.72 9.74 9.76 Position (mm) Figure 5.18 Circle of least confusion for the SEACF/1 lens in which the (a) nomi nal design refractive indices are used, (b) measured refractive indices are used, and (c) measured refractive indices are used and waveguide anisotropies are included. 249 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ces. This case was run primarily to test our ray-trace program. The SEACF/1 design was evaluated again (assuming the case of TE polarization as discussed in Chapters 2 and 3) with the experimentally measured refractive indices (n(Ti:LiNb03) = 2.2069, n (S i0 2) = 1.46) under the same assumption. The result is shown in Fig. 5.18(b). In this case focal spot blur 10 /im wide was produced. This increased focal spot width indicates that the lens design is de-optimized for these refractive index values. To obtain a physically complete representation of the SEACF/1 lens focal spot size for a S i0 2 embedded lens in Ti:LiN b03, the measured refractive indices should be used and the waveguide anisotropies should be accounted for in the ray trace program. Consequently, the SEACF/1 design was evaluated with the experimentally measured refractive indices (ne(Ti:LiNb03) = 2.2069, n 0(Ti:LiNb03) = 2.2877, and n(S i02) = 1.46) and with the uniaxial nature of the Ti:LiNb03 waveguide taken into account. The index ellipsoid used in this ray trace was given in Chapter 3 for single-mode Ti:LiNb03 waveguides. The focal spot pro duced is shown in Fig. 5.18(c). The basic behavior induced by the anisotropy is that TE-polarized light propagating at an angle with respect to the crystallographic x-axis experiences a higher refractive index than light propagating along this axis. Hence, a ray displaced from the optical axis that passes though the lens will cross the optical axis closer to the lens than it would in the absence of the anisotropy. The overall effect is that the focal spot blur width is reduced to approximately 4 /xm. This focal- spot blur size is half of the rib waveguide width of the high density arrays (8 jj.m rib waveguide width, 2 /zm gap width). Since the effects o f waveguide anisotropy are clearly significant, they will impact the focal properties of the integrated lenses incorporated in integrated optical systems. To prevent range error in the IOSAR processor, the next generation SEACF/1 lens should be designed with the waveguide anisotropies taken into account. For the IOSAR processor in particular, a lens is required to produce a focused image of light diffracted from a chirped SAW index modulation that is posi- 250 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tioned somewhere within the lens aperture. The important question is whether or not the waveguide anisotropies can taken into account in the lens design such that an aberration-free focus may be obtained for all possible positions o f the SAW modula tion. To address this question, we considered the properties of refraction through an interface in the presence of anisotropy in the host waveguide. The relationship between incident and refracted angles is more complex for an anisotropic host waveguide than for an isotropic host waveguide since the refractive index of the host waveguide medium is a function of the angle of the incident beam. In particular, tran scendental equations describe this relationship; the solutions to which can be found through the use of iterative methods. We implemented such methods into the ray- tracing algorithms described previously so that the effects of waveguide anisotropy could be included. We found that the optimization of lens performance, is compli cated by the focal length dispersion caused by waveguide anisotropies [Zhou and Ris- tic, 1989; Zhou, et al., 1990]. The imaging condition depends on the lens focal length, which in turn depends on the refractive indices of the constituent media. Since these refractive indices are a function of the input field propagation angles, the imaging properties of the lens will vary with the properties of the input field. Consider the imaging conditions in the IOSAR processor. The light diffracted from the SAW modulation is spread over an angular range associated with the spatial frequency bandwidth of the chirped signal. For the example of a 50 MHz bandwidth given in Chapter 2, the diffracted light will be spread over an angular range of ±0.12°. If the SAW modulation is positioned somewhere near the center o f the 2.7 cm aper ture, 10.8 cm focal length (f/4) lens discussed in Chapter 2, the incident and transmit ted fields will propagate primarily parallel to the optical axis (crystallographic jc-axis) and experience minimal waveguide anisotropy. To calculate the focal plane shift due to the presence of anisotropy, the refractive index dispersion relation given in Chapter 3 was examined for this angular range. 251 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The resultant refractive index variation was then inserted into Eq. 5.6 along with the other appropriate parameters (nL = 1.46, n # = 2.207) to determine the focal length shift. The shift in the image plane position is approximately equal to the focal length shift for the IOSAR geometry. In this case, the variation of ±0.12° in incident field angle corresponds roughly to a ±0.075 fim shift in the lens focal plane. Therefore, the focal plane shift is negligible. If the SAW modulation is positioned near the upper or lower extreme of the lens aperture, the focal length shift will be approximately ±4 fim (i.e., the light diffracted from the leading edge and trailing edge of an upchirped SAW modulation is imaged with a lens focal length dispersion of 8 fim). The focal length shift is larger in this case because the refractive index dispersion equation is evaluated at off-axis propaga tion angles of 7° ± 0.25° where the effect of anisotropy is greater. Since the SAW record length (3.5 mm from the example in Chapter 2) is much smaller than the lens focal length ( 10.8 cm), the degree to which the focused image is broadened at the rib waveguide array is negligible (approximately 0.35 cm/10.8 cm x 8 fim = 0.3 /an). For the IOSAR processor example given in Chapter 2, we expect that a lens design optimized for operation in the presence o f waveguide anisotropies will meet the pro cessor requirements of a focal-spot size comparable with the rib waveguide spacing (1 0 fim) across the entire rib waveguide array. For the flat-field lens doublet design introduced in this chapter, the presence of waveguide anisotropies will have a larger effect on the width of the range-focused spot for chirped SAW modulations near the extremes of the lens aperture than the 10.8 cm focal length lens. This is because the focal length of this lens combination (4.2 cm) is much shorter than the focal length of the lens discussed above and also because the assumed SAW bandwidth is larger (100 MHz) than assumed above. Still, this effect is tolerable (1.25 /an maximum broadening of the focused image) for a lens combination optimized for operation in the presence of waveguide anisotropies. 252 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.4.6 Estimation of Embedded Lens Structure Throughput The embedded lens throughput was estimated with the cumulative effects of reflection losses, mode coupling losses, waveguide attenuation losses, and scattering losses taken into account as a function of beam position within the lens aperture. We previously considered only the reflection arising from a beam incident on the lens boundary at normal incidence. However, for off-axis beams, a larger reflectivity results from an oblique angle of incidence with the lens interface, as shown in Fig. 5.19. These curves were calculated under the assumption that a TE-polarized waveguide mode was incident on the S i0 2/Ti:LiN b03 interfaces in the lens structure and that all transmitted light was coupled into the embedded waveguide regardless of any rotation of the polarization state caused by angled refraction through the tilted lens interfaces. Consider the case of the embedded lens structure described at the beginning of this chapter with a 3 /xm etch depth, 12° sidewall tilt, 2 fim Si02 layer thickness, and 0.65 fim MgF2 layer thickness in a Ti:LiNb03 waveguide. Furthermore, assume that the recess sidewall roughness is 1000 A peak-to-peak. The loss per surface due to scattering is 3% (from Sect. 5.4.3), whereas that due to mode coupling losses is 43% (approximated from Fig. 5.17 for a 3500 A waveguide offset). Since the reflectivity and propagation losses vary as a function of aperture position, the total lens transmit tance is shown in Fig. 5.20 as a function of position within the lens aperture and is about 28% at the lens center: 0.572 (mode coupling) x 0.972 (scattering) x 0.99 (prop agation losses) X 0.962 (reflection losses for both interfaces). This estimated trans mittance will be compared with the measured transmittance in Sect. 5.6, where the experimental characterization of the lens is discussed. 5.5 Fabrication of Embedded Lenses in Lithium Niobate In this section, we discuss the development of the key fabrication processes 253 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 1.00 W all tilt angle 0.98 ffi o c (0 33 e (0 c C O W h- ■10° 0.96 ■ 1 5 ° 0.94 0.92 0.90 -4 0 -2 2 4 Position (mm) Figure 5.19 Transmittance as a function of position within the aperture of an embedded S i0 2/MgF2 lens structure in a Ti:LiN b03 waveguide for different side wall tilt angles not accounting for scattering, mode-coupling, and propagation losses. required for embedded lens fabrication, as well as the final process sequence that pro duced lenses with the best performance. We experienced great difficulty in the fabri cation of a deep embedded lens recess in Ti:LiN b03 with the vertical and smooth sidewalls necessary for optimal lens performance. A large number of experiments were attempted to achieve these ideal qualities. Our most successful process fell just short of this goal with a recess sidewall that was 12° tilted from the normal. This sidewall tilt is the origin of most of the fabrication-related issues pointed out in the theoretical overview. The performance values achieved with these embedded lenses, however, are sufficient for the continued development o f advanced IO signal proces sors despite the shortcomings of the recess fabrication process. The effects o f mask shrinkage, mask faceting, and lithium niobate redeposition associated with ion-beam etching a deep recess in lithium niobate were discussed in Chapter 3. These effects contribute significantly to increased sidewall tilt in the fabri- 254 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 Fresnel reflections 4 dB/cm propagation loss Total lens transmittance ( O c 2 0.4 h- 0.2 0.0 -4 -2 0 2 4 Position (mm) Figure 5 JO Total transmittance of an embedded Si0 2 /MgF2 lens structure in a Ti:LiNb03 waveguide, including scattering and reflection losses from both surfaces as well as mode mismatch and embedded lens propagation losses. cation of deep recesses; thus, to minimize that sidewall tilt, the embedded lens recess should be made as shallow as possible. Likewise, for efficient coupling between the host and embedded waveguides, the host Ti:LiN b03 waveguide should not extend any deeper into the substrate than the preferred embedded lens recess depth. This limitation was one o f our primary concerns in our effort to produce shallow Ti:LiNbC> 3 waveguides as discussed in Chapter 3. Another reason that the host waveguide should be shallow is that it is difficult to deposit a thick dielectric embed ded lens structure. The majority of the optical power in the single-mode Ti:LiNb03 waveguide is confined to within 2 fim of the surface; hence, the total thickness of the embedded lens structure need be only 3 fim or less. The development of the key fab rication processes discussed next was aimed at producing recess depths approxi mately 3 fim deep in LiN b03 and thin film structures with a total thickness o f approximately 3 fim . 255 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5.5.1 Development of Key Fabrication Processes In Chapter 4, we discussed the primary issues of component fabrication by pho tolithographic and etch techniques that limit the component definition. We investi gated several processes in order to address issues of mask selectivity, mask shrinkage, bevel formation, and redeposition. Next we present a brief description of each pro cess that was attempted in order to produce the best possible embedded lens recess. One of the first processes attempted was a straightforward extension of the pro cesses used for the fabrication of the rib waveguide array structures. This process consisted of the use of S 1400-33 photoresist with a 2.2 /ma thickness to mask the sample during an argon-ion beam milling process. The ion-beam milling process was conducted at normal incidence under conditions similar to those described for the rib waveguide fabrication. This process, however, resulted in recesses from 1 to 2 fim deep and with a sidewall tilted approximately 40° from the normal to the waveguide surface. This sidewall tilt is near the total internal reflection condition described ear lier in the theoretical overview. As a result, the light incident upon the integrated lens is reflected into the waveguide substrate. The reason for the large sidewall tilt is related to the low etch selectivity of the photoresist, its small thickness compared with the etch depth, the quick formation of a bevel in the photoresist, and redeposition. During the ion-beam milling process, a 45° bevel would quickly form in the comer of the mask and reach the sample by the time the etch depth was approximately I fjxn. At this point, the photoresist mask would rapidly shrink, expanding the embedded lens in size and resulting in the large recess sidewall tilt. In order to reduce the effects of bevel formation and mask shrinkage, a thicker photoresist was selected. In this case, S 1400-37 photoresist was used with a thickness of 4 fim. Slight improvements observed in the sidewall tilt were observed, which can be attributed to the additional photoresist thickness. However, the initial pattern defi nition in this thicker photoresist resulted in a photoresist profile that was slightly tilted 256 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. by about 15°. This initial mask sidewall tilt is directly transferred into the substrate material since the selectivity of the photoresist mask is near unity. The added effects of redeposition limited the sidewall tilt angle to no better than approximately 25° from the waveguide normal. Several of the difficulties associated with the use o f photoresist as an etch mask for deep etching in LiNbC>3 were identified after these experiments and alternative approaches were considered. The two primary effects that had limited the results of the argon-ion milling process and the photoresist masking technique were the low mask selectivity and the substrate redeposition. Through the use of reactive ion etch ing (RIE) techniques, others have reported increased mask selectivities for both pho toresist [Zhang, et al., 1984; Huang and Lee, 1986] and metal mask materials [Matsui, et al., 1980; Jackel, et al., 1981; Chung, et al., 1986]. In addition, reactive ion etching techniques are known to reduce the effects of redeposition, since the etched substrate material combines with the reactive ions to form gaseous compounds that are pumped away by the vacuum system [Jackel, et al., 1981]. Others have reported selectivities as large as 10; 1 with a Cr or NiCr mask on LiNb03 for RIE conducted with CF4 [Chung, et al., 1986] and CF4/0 2 mixtures [Jackel, et al., 1981]. We attempted to duplicate these results so that this method could be applied to the embedded lens fabrication process. According to the results obtained by others, a 3 ftm deep recess could potentially be etched using a metal mask layer of about 0.5 fan in thickness. This metal mask can be easily fabricated by elec- tron-beam deposition and photolithographic lift-off techniques. Many samples were prepared with Cr and NiCr masks and etched under a variety of conditions in an Oxford Instruments RIE chamber with a parallel plate configuration. However, the required conditions to produce a high selectivity could not be established in this sys tem. The maximum selectivities observed were between 3:1 and 4:1 for an etch depth of about 1 fim. The observed sidewall tilt angle in this case was approximately 20° from the normal to the waveguide surface. 257 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. There were a number of limitations to this process that prompted us to seek another approach. The first limitation is the observed formation of a steep bevel in the mask materials. Despite an initial mask profile that was nearly vertical, a steep bevel will quickly lead to a tilted mask sidewall and subsequently, mask shrinkage. We believe that mask shrinkage is responsible for the large sidewall tilt angle despite the improvement in selectivity gained over the use of a photoresist mask in the ion- beam-milling process. The second limitation is that to obtain these high selectivities, the etch rate is very low. This results in a long process-development cycle. The third limitation that we observed is non-uniformity in the etch depth across the lens pattern. We found that a lens-shaped opening in a metal mask on LiN b03 had a higher etch rate in the larger openings in that mask than the narrow openings (such as that at the lens center). Mask-shape dependent etching has been known to occur in wet chemi cal etching. In such cases, the rate at which fresh chemical etchant can reach portions of material exposed through a mask is dependent on the distance from the mask edge. A similar chemical-related effect may have occurred during our RIE experiment. Although the RIE process had the potential for high mask selectivities, the shortcomings of the process that were encountered compelled us to identify an argon- ion beam etching process other than the photoresist-mask process described above that would produce better results. A better selection o f etch mask was our primary concern. Others have established ion-beam etch rates for a number of materials that can be used as an etch m ask [Cantagrel, 1975; Somekh and Casey, 1977]. The only material with an etch rate substantially lower than that of L iN b03 is carbon (about one fourth that of LiN b03). A two-step mask fabrication technique that can be used to create a carbon mask with smooth and vertical sidewalls has been previously reported [Cantagrel, 1975]. Therein, the effects of bevel formation in the carbon mask were shown to be small. This process consists of an initial deposition of a car bon layer onto the sample surface. A titanium metal layer is defined with the desired pattern on the surface of the carbon by the photoresist lift-off technique. The pattern 258 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. is then accurately transferred with a vertical sidewall into the carbon layer by oxygen- ion etching. A vertical sidewall is achieved since the titanium mask selectivity with respect to the carbon is very high (40:1). Next, the pattern is transferred into the LiN b03 through the carbon mask by argon-ion beam etching. We attempted to repeat this procedure and apply it to the fabrication of embed ded lenses. A number of difficulties were encountered in this process. To produce a recess depth of 3 /tm for the embedded lens structure, we estimated that the required carbon layer thickness is approximately 1.5 //m. A carbon layer of this thickness was deposited onto several samples by sputtering from a ring target. However, we found that the films did not adhere well everywhere to the L iN b03 surface. In regions of the sample where the films did adhere, a titanium film was deposited and photolitho- graphically patterned. Pattern transfer into the carbon layer by oxygen-ion milling resulted in smooth, vertical carbon mask sidewalls. However, a large amount of car bon residue remained in the etched region that could not be removed by continued oxygen-ion etching. Some regions that were free o f residual carbon were observed after the subsequent argon-ion milling run. The recess sidewall in these regions appeared nearly vertical. However, carbon redeposition could be observed on the sidewalls of the recess as a thin layer of material extending downward from the car bon mask when the recess was observed in cross-section. Despite the vertical-look ing recess sidewall, the lithium niobate sidewall underneath the redeposition was actually tilted. The range of lithium niobate sidewall tilts ranged from very slight to greater than 20°. Furthermore, the carbon mask edge developed large grooves and did not lead to a smooth lithium niobate recess sidewall. Because of the number of difficulties encountered in the carbon masking tech nique, that approach was abandoned and instead we pursued a technique that utilizes a metal mask during the argon-ion beam milling process. It had been previously reported by others that the etch rate of certain metals such as chromium and titanium is reduced during argon-ion beam milling by the introduction of a low oxygen back- 259 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. ground pressure in the vacuum chamber [Cantagrel, 1975; Somekh and Casey, 1977]. The etch rate is reduced since the metal surface is continually oxidized during the etching process and the oxygen layer protects the metal film from the ion beam. In the case of titanium, the etch rate is lowered by as much as a factor of eight. Since the initial selectivity of titanium on LiN b03 is nearly unity, the expected selectivity with this process is approximately 8 :1. However, there is also a decrease in the etch rate of the L iN b0 3 associated with the introduction of oxygen into the chamber. Hence, the selectivity is improved to only 4 :1. This selectivity is comparable with the selectivity obtained in the RIE experiments with Cr and NiCr masks. Both chromium and titanium masks were investigated for use in this process. Either material may be deposited by electron beam evaporation and the results of the ion beam etching (IBE) process with either material are similar. However, the use of titanium as the mask material was found preferable to chromium for several reasons. The primary disadvantage with a chromium mask is that the electron-beam deposited films have very high stresses. We found that chromium films with a thickness greater than approximately 0.5 fim tended to fracture and peel during the deposition proce dure. The deposition process is also slightly more difficult than it is for titanium since chromium sublimes when heated by the electron-beam. The chromium source mate rial cannot be melted into a uniform mass in the hearth pocket as can titanium. As a result, the deposition is not as stable or efficient in the use of the source material. The third disadvantage with the use o f a chromium mask compared with titanium is that the metal mask removal after the ion beam milling process is not as thorough as it is for titanium. To strip the chromium from the lithium niobate waveguides, we used a commercial chemical chromium etchant normally used to etch through chromium on photolithographic masks. Under close inspection with a microscope of L iN b03 sam ples from which chromium had been stripped, residual traces of chromium could be identified that could not be removed regardless of the etch duration. Titanium masks, on the other hand, were easily deposited to thicknesses as large as 4 fim without any 260 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. sign of peeling. These films could also be quickly and completely stripped from the LiNb03 surface with dilute HF acid at room temperature. In addition, there was no observable trace or modification of the LiNb03 surface through exposure to the HF acid even for pure concentrations and long etch durations (over 24 hours). In conjunction with the use of titanium as an etch mask for this process, we adopted a photoresist lift-off process suitable to pattern films up to as thick as 10 fim [Zappella, 1992]. The process consists of a double application of a thick photoresist (7 /xm per application) and an exposure sequence that results in an undercut photore sist profile after the photoresist development. The application of this technique was critical to the fabrication of the embedded lens recess since a titanium film thickness greater than 1 /xm was required to prevent the deleterious effects caused by the forma tion of a bevel in the mask sidewall. A number of argon-ion beam milling experiments were conducted with tita nium-masked samples and with various oxygen background pressures to determine the optimal conditions for high selectivity etching. It was found that an Ar:0 2 gas mixture of 7:2 in flow rate produced the peak etch selectivity of 4:1 for normal inci dence ion beam milling. The smallest sidewall tilt angle observed by this process was 17° for a 3.2 //m recess etch depth. These results and the milling conditions are sum marized in Table 5.4. Table 5.4: Comparison of Etching Processes Sample A r:0 2 flow rate Incidence angle Selectivity Sidewall Angle Lens 8a 7:2 0 ° 4:1 17° Lens 16, 17,55b 2 0:1 2 0° (rotation) 3.5:1 12° a. 5 seem Ar, 750 V beam voltage, 1 mA/cm2 beam current density b. 2 seem Ar, 750 V beam voltage, 0.3 mA/cm2 beam current density 2 6 1 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. The selectivity obtained by this method is comparable with that obtained by the RIE process; therefore, the recess sidewall improvements can be attributed to the for mation o f a thick metal mask with steep sidewalls. In addition, the etch depth across the lens recess was uniform for the ion beam etching process unlike what was observed for the RIE process. Although we anticipated that the oxides on the surface of the titanium mask may be etched away faster than new oxides could be formed at high etch rates, we did not experience a drop in the selectivity for large ion-current densities. As a result, we were able to obtain much higher etch rates with a compara ble mask selectivity for this process compared with the RIE process. This process is still limited, however, by redeposition of sputtered material onto the recess sidewalls. A further improvement was made in this process by ion-beam milling the sam ple at off-normal incidence and with rotation. For a particular sample tilt angle, the amount o f redeposition onto the recess sidewall can be minimized [Lee, 1980]. A secondary effect of ion-beam etching at normal incidence called trenching can also be eliminated by milling with a tilt angle. Trenching arises from argon-ion reflection from the recess sidewalls that leads to an increased etch rate at the base of the side wall. To perform ion-beam etching at an angle, with rotation, and with water cooling, a special water-cooled rotation mount is required to hold the sample. Such a mount was available to us in the Technics RIBE system described in Chapter 3. Once the oxygen flow rate required for maximum titanium mask selectivity was recalibrated in this system, the recess sidewall profile was studied as a function of incidence angle from 0° to 30° in tilt angle. A sample tilt angle of 20° produced the steepest sidewall with minimum trenching. The achieved sidewall tilt was approximately 12°. The parameters used in the argon-ion beam milling process are summarized in Table 5.4 and described in more detail in the next section. This recess sidewall tilt of 12° is most likely the minimal tilt that can be achieved by this process without a significant process change, such as the addition of another reactive gas. To obtain these results, a significant amount of effort was required that detracted 262 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. from our ability to make advances in other important areas of the research described herein such as the integration o f surface mounted devices and development of waveguide spatial filters or beam blocks. A wide-aperture, short focal length lens is a crucial building block in the development of advanced 1 0 signal processors, however, and it was considered to be o f little worth to continue the investigation into these other areas in the absence of such a device. There are other issues in the development of the embedded lens that require fur ther attention, including the sidewall roughness and embedded thin-film waveguide propagation losses. The observed sidewall roughness that results from the final argon-ion beam milling process described above is not severe (an average of 0.1 jtm peak-to-peak roughness with a a predominant periodicity of 1 Jim) and is comparable with that observed in the GaAs embedded lens fabricated by Minot and Lee [Minot and Lee, 1990]. However, our earlier calculations indicated that this roughness can lead to scattering levels higher than desired in the IOSAR processor. The second critical task in the fabrication of an embedded lens structure is the deposition of the thin-film waveguide structure. A number of reasons were given above for the selection of an S i0 2/MgF2 thin-film combination as the embedded waveguide structure. From our previous experience, electron-beam deposition of Si0 2 and MgF2 films at room temperature is a simple and stable process that produces high quality, durable films. The drawback of this deposition technique is the high loss associated with guided wave propagation in the thin-film structure. Sputtered thin films, on the other hand, produce high quality films with low waveguide losses [Hutcheson, 1987]. However, for the equipment that was available in our labs, we observed that the material that is sputtered (in which case the sample is within 10 cm of a source that is over 10 times larger in diameter than the length of the lens pattern) is not as directional as is the evaporated material in electron-beam deposition (in which case the sample is more than 50 cm away from a source that is approximately the size of the lens pattern). As a result, the recess sidewalls can be coated with a film 263 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. layer nearly as thick as that on the sample surface by the sputtering process. The impact on the recess sidewall on the coupling efficiency through the lens interface of a thick barrier layer film is unknown. However, a beam that is coupled through this region will not have waveguide confinement for a short propagation distance (0.5 fim to 1 fim) and may partially decouple from the waveguide. To avoid this possibility, we sought to minimize step coverage for the embedded lens structure, and hence chose to focus on the electron-beam deposition technique. In the electron-beam dep osition technique, the flux of evaporated material emanates from a point source and is highly directional for large sample distances (30 cm in this case). Hence, the sidewall coverage is minimal. Although S i0 2 and MgF2 are stable materials when electron-beam deposited separately, we found that the combination of these films would peel for thicknesses close to those required for the embedded lens. Therefore, we sought to identify the factors that led to the poor film adhesion properties. Our initial thin film structure consisted of a MgF2 and S i0 2 film combination with a single layer of each film deposited with total thicknesses of approximately 3 fim into the recess etched in the Ti:LiNb0 3 waveguide. These films fully detached from the LiN b03 surface, which led us to believe that the poor adhesion was at the LiN b03/MgF2 interface. We noted that the same thin film structure deposited onto a glass slide adhered very well. In addition, in another experiment we deposited a sin gle S i0 2 layer onto a LiN b03 substrate with a thickness greater than 3 fim and observed no adhesion problems. We concluded that a thin layer of S i0 2 deposited into the lens recess prior to the MgF2 layer could potentially enhance the MgF2 adhe sion. Consequently, we developed a deposition sequence that produces stable waveguide structures that can function satisfactorily as embedded lens waveguides. This sequence consists of 500 k layer of S i0 2 deposited first into the etched recess, followed by the MgF2 barrier layer deposition. We found that this initial thin layer improved adhesion of the MgF2 to the Ti:LiN b03 surface. 264 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. This modified sequence was then used to produce embedded lens samples with barrier layer thicknesses approximately 1 f m thick and S i0 2 waveguide layer thick nesses o f 2 /im. However, for a MgF2 film more than approximately 8000 A thick, we found that fractures would appear in the film over a period of months. Two possi ble causes of this long-term film degradation were considered. The first cause may be induced film stresses due to thermal cycling of the film from the daily swing in ambi ent temperature (approximately 5° C). The average coefficient of linear thermal expansion for Si0 2 and MgF2 is approximately the same (1.1 x 10' 5 K '1 ); hence, we don’t expect film stresses to develop within the thin film structure due to thermal cycling. The coefficient of thermal expansion for LiNbC>3, on the other hand, is 4.1 X 10' 6 K‘l along the z-axis and 1.5 x 10' 5 K' 1 along the x-axis. If the fracturing were due to thermal cycling, the fracture lines would most likely be preferentially ori ented along one of the crystal axes. We did not observe such preferential orientation, however. The second cause of the film failure may be due to absorption of moisture within the thin films. It is well known that MgF2 films deposited onto samples at room temperature by electron-beam evaporation are porous (packing density of 0.72 [Pulker, 1969]), will absorb moisture, and will undergo a change in film stresses when moisture is absorbed [Ennos, 1966; Pulker, 1969]. By phase contrast microscopy, we have observed moisture absorption in our embedded waveguide structures by the appearance over time of rings near defect points in the film surface. This effect has been also observed by others [Gee, et al., 1985], We found that the degradation of embedded lens films was slowed for samples stored carefully in a desiccator. The sensitivity of the thin films to air led us to conclude that the thin film fracturing was due to stress changes caused by moisture absorption. Films of Si0 2 deposited onto samples at room temperature by electron-beam evaporation are also porous (packing density of 0.9 [Pulker, 1969]) and will also absorb moisture [Martin, et al., 1983]. To increase the packing density of these thin films, we attempted to deposit the embedded film structure onto heated LiNb0 3 sub- 265 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. strates [Pulker, 1969]. The substrate temperature was varied between 100° C and 300° C. All of the deposited structures in this range of temperatures resulted in thin films that adhered very well and were moisture resistant but that also had slight frac ture lines parallel to the LiNb0 3 z-axis. The fractures were apparently due to the ther mal expansion mismatch between the thin films and the substrate. In addition, we found that MgF2 layers deposited at temperatures in excess o f 150° C onto Ti:LiN b03 waveguides disrupted the waveguide characteristics. This disruption was observed when we attempted to propagate light from a He-Ne laser through a region o f the waveguide with an overlayer of MgF2. The guided streak completely disappeared where it entered the overlayer region and no light emerged from the other end of the region (about 1 cm long). The measured refractive index of the high-temperature MgF2 film was 1.38 and did not appear absorptive. Hence, there was no reason to expect that the Ti:LiNb03 mode would be disrupted by this overlayer. We found that the waveguide properties were restored by the removal o f the MgF2 overlayer. Our only explanation for this behavior is that stresses from the MgF2 film distorted the Ti:LiNb03 waveguide structure. This behavior was not observed for films deposited onto Ti:LiNb03 at room temperature. In light o f the difficulties associated with heated substrate deposition, we chose to adjust our room temperature process to elim inate the long-term film degradation. We observed that embedded waveguide structures with a MgF2 layer thickness less than 8000 A did not fracture over time. In addition, we observed that embedded waveguide structures could be deposited with a MgF2 layer thickness greater than 1 fim if the MgF2 deposition was interrupted halfway through with a thin 500 A layer of S i0 2. We suspect that a thick MgF2 layer fails because of film stresses and that the S i0 2 layer acts as a stress-relief layer. In our final embedded lens design, these two observations were accounted for by conservatively limiting the total barrier layer thickness to 6500 A, of which 500 A is the thickness of the interleaved Si0 2 layer. A 2 fim thick S i0 2 waveguiding layer was then deposited on top of the barrier layer. A 266 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. thin 500 A cap layer of MgF2 was deposited on top of the waveguiding layer initially with the goal to provide a barrier for moisture penetration into the thin films, although, as we later learned and discussed above, these films are somewhat sensitive to moisture. A Ti:LiNb03 witness sample was prepared with an argon-ion beam etched recess and overcoated with this thin film structure. A cross-sectional view of the thin films deposited into this recess is given in Fig. 5.21. The photographs shown were taken on a Cambridge S360 scanning electron microscope (SEM) of opposite side walls. The recess etch depth and thin film deposition sequences for four different embedded lenses that were used for characterization of the lens properties are summa rized below in Table 5.5. All of these structures were fabricated using the SEACF/1 lens design described in Section 5.2.1. 5.5.2 Embedded Lens Processing Sequence The general embedded lens fabrication sequence is shown in Fig. 5.22 and con sists of the deposition of a 4 jum thick titanium layer onto the Ti:LiNb03 waveguide surface, definition o f a lens-shaped opening in the metal mask by photolithographic lift-off techniques, lens recess fabrication, metal mask removal, and deposition of the embedded waveguide thin films. The process details for the formation of the lens recess and the deposition of the thin film waveguide materials are described next. Embedded lens structures were fabricated on Ti:LiN b03 waveguides prepared by the process described in Chapter 3. These waveguides were sliced into 1 inch square samples, scribed, cleaned, and dehydration baked as described in Section 4.2.1 in preparation for photolithography. A positive type photoresist called P4620, supplied by Hoecht Celanese, was used in a process that allows for the lift-off of very thick metals. A layer of this pho toresist was first spin-coated onto the samples to a thickness of 7 fdm (4 krpm spin 267 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 5.21 Deposition of dielectric materials into an etched recess. The sidewall tilt and Si02 /MgF2 structure are shown. speed for 30 seconds) and baked on a hot plate at 115° C for 2 min. Each sample was blanket-exposed for 15 seconds on a mask aligner with the exposure intensity set at 7.5 mW/cm2 at a wavelength of 365 nm. The samples were then left standing for a minimum of two hours before the next step in the process. This waiting period was to allow gases trapped in the photoresist sufficient time to escape from the film. This step was followed by a second spin-coated layer (6 krpm for 30 seconds) o f the same photoresist to bring the total layer thickness up to approximately 14 /Jm. The second spin rate is high so that the additional layer is not thicker than the first (which is the observed tendency when the same rate is used). If the second layer is made too thick, some accuracy is lost in the pattern transfer. The sample was then baked again at U5°C for 2 min. If the sample was not allowed to stand after the blanket exposure, the gases trapped in the first photoresist film would have formed bubbles in the upper film during the second hot-plate bake. The top photoresist layer was then pattern- exposed through a light-field photomask that contained the lens pattern. The pattern exposure was performed at the same intensity settings for a 40 second duration. The sample was then developed in Shipply 42IK developer for 40 seconds. During the development o f this positive photoresist, the exposed regions of the photo resist were removed. In the two-layer, two-step exposure process described above, the photoresist was exposed everywhere except the top 7 fim in a lens-shaped region. 268 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Patterning of lens with 14 pm thick resist I Oxygen reactive ion etching I Deposition of 4 pm thick titanium I Argon-ion moling with introduced oxygen Removal of remaining titanium with HF acid Contact mask prior to deposition I Lift-off of excess titanium Deposition of MgF2 and SiOj Figure 5.22 Embedded lens processing sequence. The photoresist in this region only will therefore not be developed and it will mask the exposed photoresist below it from the developer solution. However, developer can reach the photoresist under the lens-shaped region from the edge of the pattern. Consequently, during the development sequence the photoresist is removed from the sample everywhere except in the region that defines the lens. The edges of this lens shaped terrace of photoresist are undercut due to the partial development of the lower photoresist layer. A photomask with the SEACF/1 lens design defined on the surface was used in the fabrication of embedded lenses. This photomask was an iron oxide duplicate mask supplied by Photoscience Corp. Another mask with the SEACF/1 lens design and fiducial alignment patterns defined on the surface was used in the fabrication of the lenses on the test modules. This photomask was a chromium-on-glass mask writ ten by electron-beam lithography and supplied by Army Research Laboratories in 269 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5.5: Summary of Embedded Lens Fabrication in Lithium Niobate Waveguides Sample LENS-1 LENS-8 LENS-16, 17,55 Etched Depth 3 jjm 3.5 fim 3.1 fim Adhesion layer (S i02) - 350 A 500 A Barrier Layer (MgF2) 0.8 fim 1.1 fim 0.3 ftm Inter-Barrier Layer (S i0 2) - - 500 A Barrier Layer (MgF2) - — 0.3 fim Waveguide Layer (S i0 2) - 2.0 fim 2.0 fim Waveguide Layer (Photoresist) 2.2 fim — - Cap Layer (MgF2) - 500 A 500 A Total Structure Thickness 3 fjin 3.2 fim 2.7 fim Adelphi, MD. The fiducial markings are needed for the alignment of the lens with the rib waveguides in the fabrication of the integrated test modules. Once the samples were patterned and before the deposition of the titanium mask, the samples were lightly ashed in an Oxford Instruments Reactive Ion Etching Chamber with an oxygen plasma. This was done in order to remove any residual pho toresist from the sample surface and thereby improve the titanium adhesion. The oxy gen pressure and RF power were 200 mTorr and 150 W, respectively, and the ashing duration was 15 seconds. The titanium mask was deposited by electron-beam evaporation in a Balzers B AK-640 Box Coater with the samples fixed at a distance o f about 30 cm directly above the source. This large throw (source-to-sample distance) led to a very sharp mask edge after the excess metal was lifted off. A sharply-defined mask edge is nec essary for an accurate pattern transfer into the lithium niobate during the ion-beam 270 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. milling process. The remaining titanium adhered well to the lithium niobate surface. Photoresist residue was removed from the sample by rigorously cleaning the surface with a cotton swab soaked with acetone. A final cleaning with a cotton swab soaked with dilute soapy water followed by a de-ionized water rinse resulted in the lowest level of surface contamination. Occasionally, small pinholes in the titanium mask would appear where there was poor adhesion due to dust particles or unremoved pho toresist residue below the titanium film. To safeguard against etching o f the waveguide surface through any pinholes that may have formed, the sample surface was coated with colloidal graphite in regions away from the lens. The colloidal graphite was allowed to dry thoroughly before the sample was milled. The best recesses were formed by argon-ion beam etching with an oxygen back ground pressure, with a sample tilt of 20° from normal incidence, and with rotation during etching. This etching was done in a Technics RIBE system equipped with a 5.5 cm Technics ion source and a Commonwealth ID2500 ion drive. The distance between the gun and the sample in this system is approximately 20 cm. The introduc tion of oxygen at a flow rate of about 5% of the argon flow rate (5 seem) enhanced the substrate-to-mask etch selectivity and led to improved pattern transfer into the lithium niobate as discussed in Sect. 5.5.1. The recess was milled to a depth of approximately 3.0 /un, after which the sample was taken from the chamber and cleaned with isopro panol in order to remove the remaining colloidal graphite from the sample. The tita nium mask was removed from the sample with dilute hydrofluoric acid. The sample was carefully cleaned with a cotton swab soaked with acetone, and then with a cotton swab soaked with dilute soapy water followed by a de-ionized water rinse. A sequence of thin films comprising interleaved layers o f Mg?2 and SiC>2 were then deposited as discussed in Section 5.5.1 by electron-beam evaporation at room temperature. Portions of the waveguide were physically masked with aluminum foil as indicated in Fig. 5.22 to allow for prism coupling. These films were deposited in the Balzers system using a multiple pocket hearth so that it was unnecessary to break 271 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. vacuum during the deposition sequence. 5.6 Experimental Characterization o f Em bedded Lenses In this section, we present the results o f the embedded lens characterization. The embedded lens structures were evaluated by prism-coupling light from a He-Ne laser into the waveguide and directing it through the lens structure. The lens focal length, throughput, and focal spot size were determined. The focal length of the S i0 2/MgF2 embedded lens was determined for several o f the samples described in the previous section. To accomplish this measurement for each lens sample, a narrow (approximately 300 fim wide) beam from a 15 mW He-Ne laser (A> = 6328 A) was prism-coupled into the waveguide and directed through the lens at different transverse displacements from the optical axis. The refraction of the beam at each lens interface was indicated by the path o f the guided streak. This streak is very faint in the host waveguide due to the high quality of the Ti:LiNb03 waveguide, and is much brighter in the embedded lens films due to higher waveguide scattering. A photograph was taken of the lens sample with multiple exposures that correspond to the different input beam positions. The distance from the center of the lens to the point where the refracted beams cross was then measured from the photo graph with the image magnification taken into account. Shown in Fig. 5.23 is a mul tiple exposure photograph taken of Lens-16 with the narrow beam of light directed through the lens at a different transverse displacement from the optical axis for each exposure. The power coupled into the waveguide was approximately 100 to 200 fiW. Each exposure was several seconds long in order for the faint light to fully expose the photographic film. The beams were evenly spaced over approximately 4 mm at the input of the lens and converged to a focal spot approximately 1 cm beyond the lens center. This is the focal distance that we expected for this lens design. In order to confirm that the light after the lens was indeed propagating in the waveguide layer 272 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 5.23 Multiple exposure photomicrograph of a 1-cm aperture MgF2 /S i02 lens embedded in a titanium-indiffused lithium niobate waveguide (sample Lens-16). For each exposure the narrow beam of light (A = 6328 A) incident on the lens from the left was shifted in the transverse direction so that the focusing effect of the lens would be evident. The resulting focal plane is visible about 1 cm to the right of the lens center. and not in the substrate, a second outcoupling prism was placed after the lens. The waveguide m-line output from the prism could be easily observed. In addition, the direction of the outcoupled beam could be steered by changing the position of the nar row beam within the lens aperture. In order to demonstrate an embedded lens with a different thin-film structure, this same lens design was used with a photoresist waveguide layer and MgF2 lower barrier layer (sample Lens-1 described in Table 5.5). From Table 5.1 the refractive index of photoresist is approximately n = 1.6. For a photoresist/MgF2 embedded lens in a Ti:LiNb03 waveguide with the SEACF/1 lens design, the expected focal length is 273 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. / = 1.3 cm. A multiple exposure photomicrograph, taken under conditions similar to that described for sample Lens-16 described above, is shown in Fig. 5.24. The observed focal length also matches the expected focal length of 1.3 cm for this embedded lens. To measure the S i0 2/M gF2 embedded lens focal spot size, a lens sample was prepared with a polished end-facet so that light directed through the lens would come to focus outside of the waveguide. The focal spot was then scanned with the SpotScan beam profiler described in Chapter 3. The focal spot-size measurement was performed on sample Lens-8, which was described in Table 5.5. This lens sample was prepared with a polished end-facet after the embedded lens recess fabrication and before the thin film deposition. The end of the sample was cut with a wire-saw approximately 9.5 mm beyond the center of the lens and polished according to the process described in Chapter 3. The emission from a 250 mW laser diode (SDL-7420-C, X = 670 nm) was used to illuminate the lens for the focal spot measurement. This laser source was used rather than the 15 mW He-Ne laser since the total throughput loss of the spatial filter, collimating optics, prism coupler, waveguide lens, and waveguide output facet was too high to provide a clear signal for the SpotScan beam profiler with the low-power He-Ne laser. Using the laser diode, the total power that reached the SpotScan device was approximately 10 pW . The laser diode source has a 1 p m x 100 p m emitting area and a spectral width of 20 nm. There are multiple transverse spatial modes along the wide dimension of the laser aperture. However, the output illumination along this transverse direction appeared uniform. The laser emission was collimated, prism- coupled into the waveguide, and directed parallel to the optical axis of the lens such that the center 3/4 cm of the lens aperture was filled. The portion of the lens aperture that could be uniformly illuminated was limited by the prism-coupling uniformity across the 1 cm wide prism. The output power of the laser could be adjusted to pro vide adequate illumination of the SpotScan device. The measured focal-spot intensity 274 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 5.24 Photomicrograph of a MgF2/Photoresist lens embedded in a tita- nium-indiffused lithium niobate waveguide (sample Lens-1). The dark lines on the scale in the lower region of the photograph are separated by 1 mm. The lens is illuminated by two thin guided beams (A = 6328 A) from the left, and the resulting focal plane is visible about 1.3 cm to the right of the lens center. profile is shown in Fig. 5.25, and has a full-width at half-maximum (FWHM) of 2.0 fim. The highest measured sidelobe level is -11 dB down from the peak intensity level. According to Eq. 5.2, a 2 jmn focal spot size measured in air corresponds to a 0.9 fim focal spot size inside the waveguide. From our discussion earlier for a uni formly filled f/l lens, the expected diffraction-limited focal spot size is 0.3 fim. The reason for the larger observed focal spot size is due to the 3/4 lens fill and non-uni form illumination profile that is accentuated by a drop-off in lens transmittance toward the extremes of the aperture. A drop in transmittance such as this would result in an effectively smaller aperture for the lens and hence, a larger focal spot size. From the discussion in the first part of this chapter, it is likely that lens aberrations also contribute to the larger focal spot size. In the experiments described above, it was evident through direct observation of the guided streak that the lens throughput drops away from the center of the lens. We 275 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 J3 c Z3 0.6 FW HM = 2 microns & « 0.4 e 0.2 0.0 •7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 Distance (/rm) Figure 5.25 Focal-spot intensity profile of a 1-cm aperture f/1 embedded lens fabri cated in a titanium-indiffused lithium niobate waveguide. identified a few attributes of the embedded lens structure in the theoretical overview that lead to this type of behavior. In order to quantify this effect in our fabricated samples, a number of throughput measurements were performed. A simple way to measure the lens throughput is to prism-couple light into the host waveguide before the lens, prism-couple the light out of the waveguide after the lens, and then compare the input and output powers. This measurement requires the prism coupling efficien cies to be taken into account. These efficiencies, however, are not easily controlled and may vary across the prism aperture. The approach that we used to estimate the lens throughput was to measure the transmittance of the lens at several positions within the aperture and then take the average value of the transmittance function across the aperture. To measure the trans mittance of the lens at a particular position within the lens aperture, a 15 mW He-Ne laser was focused to a narrow beam and prism-coupled into the waveguide as was 276 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. done in the focal length measurement described above, so that the guided streak before and after the lens could be clearly identified. The relative optical power cou pled into the waveguide was determined through measurement o f the brightness of the guided streak before the lens. The optical power transmitted through the lens was measured relative to the input power either through measurement o f the brightness of the guided streak after the lens or through direct detection of the optical power emit ted at a polished waveguide edge. For these measurements, the waveguide samples were assumed to have fairly uniform scattering characteristics across their surface, which is what we have observed in practice. This transmittance measurement was performed on sample Lens-8 for a beam directed through the center of the lens. An optical fiber with a 50 film core diameter was used to probe the surface-scatter intensity before and after the lens. To probe the surface scattering, the cleaved fiber end was positioned in close proximity to the waveguide surface. The light coupled into the fiber was detected at the output end of the fiber with a silicon detector head connected to a trans-impedance amplifier and a lock-in amplifier. The input laser beam was chopped to provide a reference for the lock-in detection. The entire fiber-probe system was calibrated for linearity. The transmittance o f sample Lens-8 was determined at the lens center to be 32% ± 7% by this method. The transmittance was then measured as a function of the beam position within the aperture o f sample Lens-8. As before, a narrow beam was coupled into the waveguide and directed through the lens. In this case however, a CCD detector array rather than a fiber probe was used to detect the surface scattering o f the input beam. To accomplish this, the guided streak was imaged onto the CCD detector array. The CCD response was digitized, calibrated for linearity, and measured along the guided streak. The relative amount of power coupled into the waveguide for different input beam positions was determined from the CCD response. We found that this method was less complicated than the fiber probe method. Since the guided streak after the 277 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. lens was too dim for the CCD array to detect, the relative transmitted intensity was measured at the polished end of the waveguide after the lens. The planar waveguide end-emission was measured with a detector placed close to the waveguide edge so that end-emission illuminated the detector surface. The angular dependence o f the Fresnel losses at the polished Ti:LiNb03/air interface for different input beam posi tions was taken into account. The results of the transmittance measurements are shown in Fig. 5.26. For beam offsets of 3.5 mm or more, the angle of refraction o f the beam transmitted at the polished waveguide edge was greater than 45°. The photode tector that was used could not be placed in close enough proximity to the waveguide edge to capture all of the end-emission for these large refraction angles. As a result, the lens transmittance could be accurately measured only for beam offsets o f 3 m m or less. It is apparent that the central 4 mm portion of the lens contributes most to the total throughput o f the lens. This aperture size is approximately the size required for a lens with a 1 cm focal length to produce the 2 fim focal spot size mentioned above. The relatively low transmittance of the embedded lens on sample Lens-8 can be attributed primarily to the sidewall tilt. As described earlier, the sidewall tilt was improved from 17° on sample Lens-8 to 12° on several other samples through a mod ification of the recess fabrication process. The transmittances of two o f these improved embedded lens structures were also measured. The transmittance at the center of the lens on sample Lens-16 was measured through a comparison of the intensity of the surface-scattered light of a narrow guided beam before and after the lens. This surface scattering is evident in a photograph of this lens as shown in Fig. 5.27. This photograph was taken with an extended expo sure (4 seconds) to produce a visible image of the faint guided streak on the film. The observed improvement in transmittance of this lens compared with sample Lens-8 is probably due to the improvement in sidewall tilt. Since the guided streak after the lens was sufficiently bright for detection with a CCD detector array, we measured the transmittance o f this lens simply by imaging the guided streak onto a CCD detector 278 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 - * c © 2 Lens-8 O Measured data Polynomial fit c £E 1 0 3 2 4 5 Offset From Optical Axis (mm) Figure 5.26 Measured transmittance of embedded lens sample Lens-8 as a function of the ray position within the aperture. array. The calibrated CCD detector response to the guided streak intensity before the lens was compared with that after the lens to determine the lens transmittance. The measured transmittance for this sample was approximately 65% ± 5%. Nonunifor mity in contact of the incoupling prism with sample Lens-16 prevented a measure ment of the lens transmittance at different beam positions within the lens aperture. Nonuniform contact such as this results from a waveguide surface that is slightly bowed, which is occasionally the case for substrate sections diced from near the wafer edge. An embedded lens was fabricated on another sample, Lens-55, with the same processes parameters and structure of the embedded lens on sample Lens-16. The transmittance as a function of the beam position within the aperture was measured on this sample. The measurement was performed in the same manner as that described above for sample Lens-16 except that the beam was offset from the optical axis at 279 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 5.27 Photomicrograph of sample Lens-16. This is a 1-cm aperture MgF2/SiC> 2 lens embedded in a titanium-indiffused lithium niobate waveguide. A narrow beam of light was directed through the lens center, and the intensity of the sur face-scattered light was measured before and after the lens to determine the transmit tance. several positions within the lens aperture (see Fig. 5.20). A drop-off in lens transmit tance for offsets greater than approximately 3 mm resulted in a transmitted beam with surface-scattered light too faint to be detected by the CCD camera. The result of this measurement is shown in Fig. 5.28. Notice that the throughput is fairly uniform and high (60%) for the center 2 mm portion of the sample and then drops off rapidly towards the extremes of the aperture. The high transmittance measured at the lens center and quick drop in transmit tance away from the lens center is in contrast to our theoretically predicted transmit tance across the lens aperture, which was fairly uniform across the entire aperture and much lower overall. These differences may be attributed to two factors. First, the 280 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lens-55 0.8 © O c as c 'E V) 0.6 c 2 0.4 0.2 0.0 -4 -2 0 2 4 Offset from Optical Axis (mm) Figure 5.28 Measured transmittance of embedded lens sample Lens-55 as a func tion of the ray position within the aperture. high measured transmittance at the center o f the lens suggests that our model of the slanted sidewall as a thin prism does not accurately predict the absolute transmittance for a given wall tilt angle. The trends that were indicated by that model (i.e., the decrease in coupling, the shift in optimal waveguide offset, and the increased cou pling into the higher-order modes for larger tilt angles) are intuitively sound and may still be valid although we have no experimental data to confirm this. Second, the drop in the measured throughput towards the extremes of the lens aperture may be due to losses incurred by changes in polarization due to the oblique angle of incidence of offset beams on the tilted sidewall interface. This implies that an offset beam of TE polarized light incident on the first interface of the embedded lens couples into both TE and TM modes in the lens region. M ost of the TM polarized light will be lost at the second interface since the Ti:LiNbC> 3 waveguide does not support TM polarized modes. This effect was not included in the theoretical analysis of total lens through put. 2 8 1 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. The average throughput of the center 4 mm of the lens (Lens-55) is approxi mately 50%. If the center 6 mm of the lens is considered, the average throughput drops to about 43%. Thus, these lenses can function effectively at f/2 with a device insertion loss of approximately 3 dB. 5.7 Sum m ary In this chapter, we introduced an embedded waveguide lens in Ti:LiNbC> 3 that utilizes a very large refractive index difference (0.7) compared with other waveguide lenses in this material, and that may be fabricated by planar fabrication techniques. The embedded lens structure design issues were discussed from both ray-optics and wave-optics standpoints. Two lens designs were developed with Code V software for the IOSAR processor and 10 correlator. It was shown that the respective system requirements for aperture size and focal length could be satisfied with the material parameters o f the S i0 2 /MgF2 embedded lens in Ti:LiNb0 3 . Embedded lens structures with higher and lower lens refractive indices com pared with the host waveguide refractive index were considered. Low refractive index lenses have a useful feature in that higher-order modes excited in the lens waveguide region will be preferentially attenuated by leakage through the barrier layer. This preferential attenuation reduces the spurious noise in the transmitted sig nal by filtering out the light coupled into the higher-order modes. In the design o f our lens with a low-refractive-index embedded waveguide, the minimum barrier layer thickness was determined by the maximum attenuation of the lowest order mode that was acceptable due to leakage through the barrier layer. The MgF2 film thickness selected was 6500 A. We found that for an ideal embedded lens recess with smooth vertical sidewalls, the embedded lens with an S i02 layer thickness of 1.5 fiva. has a 99% mode-coupling efficiency through each lens interface. Variation of the waveguide thickness by 282 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ±0.5 /an and waveguide offset by ±0.15 /an from the optimized offset for a given film thickness leads to coupling efficiencies that are still in excess of 90%. These wide ranges of variation are easily satisfied in practice. We selected an Si02 film of thick ness of 2 fjin for our embedded waveguide fabrication since its combination with a 6500 A MgF2 barrier layer results in sufficiently low leakage of the lowest order mode. If this lens structure is offset in depth from its optimal position with respect to the surface o f the host waveguide by as much as ±0.3 /an, there will be less than 20% coupling of light into the higher-order lens waveguide modes. Consequently, a large degree of variation in the etch depth may be tolerated in the fabrication of such an embedded lens. A number of fabrication limitations were considered that affect the embedded lens performance. The primary limitation in these structures is the tilted recess side wall. We suggested theoretically that the coupling efficiency into the lowest order embedded waveguide mode will drop and the coupling efficiency into the higher- order modes will increase with a larger sidewall tilt angle. A second potential limita tion in the embedded lens performance is interface scattering. We estimated that the scattered power from the observed sidewall roughness of fabricated embedded lens structures will be approximately -15 dB below the peak power. The third limitation in the fabricated embedded lenses is the variation in lens focal length due to variation in the embedded S i0 2 film refractive index. A ± l% variation in this refractive index can lead to a focal length shift of ±200 /an in a lens fabricated with the SEACF/1 lens design. The effect o f waveguide anisotropies on embedded lens performance was inves tigated. We performed ray-trace calculations for the SEACF/1 lens design in the pres ence of waveguide anisotropies. We found that the focal spot size for f/1 operation of this lens was broadened to about 4 /an. The potential to account for waveguide anisotropy in the design of the IOSAR processor imaging lens was considered. We concluded that to an extent, the effects could be accounted for in lens design software. 283 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. However, there will always be some focal-length dispersion that is dependent upon the input field. For the IOSAR processor, we concluded that the focal length disper sion would not significantly affect the processor performance. Embedded lenses were fabricated with a 6500 A thick MgF2 barrier layer and a 2 fim thick S i0 2 layer in a 3.1 }im deep etched recess with a 12° sidewall tilt angle. It was shown that the average throughput of the center 4 mm portion of the lens is approximately 50%, and the throughput of the center 6 mm portion of the lens is about 43%. Thus, these lenses can function at f/2 with a device insertion loss of approximately 3 dB. In addition, it was shown that focal spot widths of 2 fan external to the waveguide can be achieved. This focal spot width measured outside of the waveguide corresponds to a focal spot width inside the waveguide of approximately 0.9 fan. Although this is greater than the diffraction-limited focal spot size, it is suffi cient for efficient focusing into densely packed rib waveguide arrays, as will be shown in Chapter 7. For the realization of the full potential of embedded lens structures in Ti:LiNb03 waveguides, the primary issue that should be further addressed is the recess sidewall tilt. We consider this to be the main complication in the analysis of the lens performance. The issues of recess sidewall smoothness and thin film waveguide development are secondary to this concern. A number of different embed ded waveguide materials and structures may provide advantages over those selected for the present work. Such advantages could include lower propagation losses, greater thin film longevity, and single mode behavior. 5.8 References M. Cantagrel, “Comparison of the Properties of Different Materials Used as Masks for Ion-Beam Etching,” J. Vac. Sci. Technol., 12(6), 1340-1343, (1975). 284 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. W. S. C. Chang and P. R. Ashley, “Fresnel Lenses in Optical Waveguides,” IEEE J. Quantum Electron., QE-16(7), 744, (1980). B. U. Chen, E. Marom, and A. Lee, “Geodesic Lenses in Single-Mode L iN b03 Waveguides,” Appl. Phys. Lett., 31(4), 263-265, (1977). P. S. Chung, C. M. Horwitz, and W. L. Guo, “Dry Etching Characteristics of L iN b03,” Electron. Lett., 22, 484-485, (1986). M. De Micheli, J. Botineau, P. Sibillot, D. B. Ostrowski, and M. Papuchon, “Fabrication and Characterization of Titanium Indiffused Proton Exchanged (TIPE) Waveguides in Lithium Niobate,” Opt. Commun., 42, 101-103, (1982). A. E. Ennos, “Stresses Developed in Optical Film Coatings,” Appl. Opt., 5(1), 51-61, (1966). J. R. Gee, I. J. Hodgkinson, and H. A. Macleod, “Moisture-Dependent Anisotropic Effects in Optical Coatings,” Appl. Opt., 24(19), 3188-3192, (1985). J. W. Goodman, Introduction to Fourier Optics, (McGraw-Hill, San Francisco, 1968). S. Y. Huang and S. H. Lee, “Blazed Grating Couplers on LiNb03 Optical Channel Waveguides and Their Applications to Integrated Optical Circuits,” J. Lightwave Technol., L T -4,1304-1310, (1986). L. D. Hutcheson, Ed., Integrated Optical Circuits and Components, Series on Optical Engineering, B. J. Thompson, Ed., (Marcel Dekker, Inc., New York, 1987). J. L. Jacket, R. E. Howard, E. L. Hu, and S. P. Lyman, “Reactive Ion Etching of LiN b03,” Appl. Phys. Lett., 38,907-909, (1981). 285 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. R. E. Lee, “Microfabrication by Ion-Beam Etching,” Semiconductor International, J a n ., 73-82, (1980). P. J. Martin, H. A. Macleod, R. P. Netterfield, C. G. Pacey, and W. G. Sainty, “Ion- Beam-Assisted Deposition o f Thin Films,” Appl. Opt., 22(1), 178-184, (1983). S. Matsui, T. Yamato, H. Aritome, and S. Namba, “Microfabrication of LiNb03 by Reactive Ion-Beam Etching,” Jap. J. Appl. Phys., 19, L463-L465, (1980). M . M. Minot and C. C. Lee, “A New Guided-Wave Lens Structure,” J. Lightwave Technol., 8(12), 1856-1865, (1990). H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). H. K. Pulker, “Characterization of Optical Thin Films,” Appl. Opt., 18(12), 1969- 1977,(1969). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). K. Rastani and A. R. Tanguay, Jr., “Large Aperture Negative Meniscus Singlet and Triplet Lenses with Positive Focal Lengths Developed on L iN b03,” (to be published). S. Somekh and H. C. Casey, Jr., “Dry Processing of High Resolution and High Aspect Ratio Structures in GaAs-AlxGa,.xAs for Integrated Optics,” Appl. Opt., 16(1), 126- 136, (1977). 286 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. B. L. Sopori, C. M. Phillips, and W. S. C. Chang, “Efficient Optical Waveguide Coupler,” Appl. Opt., 19(5), 790-801, (1980). T. J. Su and C. C. Lee, “An Embedded Waveguide Lens with Anti-Reflection Layer,” IEEE Photonics Tech. Lett., 6(1), 89-91, (1994). T. Suhara, S. Fujiwara, and H. Nishihara, “Proton-Exchanged Fresnel Lenses in Ti:LiNb03 Waveguides,” Appl. Opt., 25(19), 3379-3383, (1986). S. Valette, A. Morque, and P. Mottier, “High-Performance Integrated Fresnel Lenses on Oxidised Silicon Substrate,” Electron. Lett., 18(1), 13, (1982). C. Warren, S. Forouher, W. S. C. Chang, and S. K. Yao, “Double Ion Exchanged Chirp Grating Lens in Lithium Niobate Waveguide,” Appl. Phys. Lett., 43(5), 424, (1983). V. E. Wood, J. R. Busch, D. T. Moore, C. B. Wooley, and W. H. Southwell, “Rectangular Luneburg-Type Lenses for Integrated Optics,” Opt. Lett., 8, 226, (1983). S. K. Yao and D. B. Anderson, “Shadow Sputtered Diffraction-Limited Waveguide Luneburg Lenses,” Appl. Phys. Lett., 33(4), 307, (1978). D. Y. Zang and C. S. Tsai, “Single-Mode Waveguide Microlenses and Microlens Arrays Fabrication in LiN b03 Using Titanium Indiffused Proton Exchange Technique,” Appl. Phys. Lett., 46(8), 703-705, (1985). P. I. Zappella, “Method of Forming Detector Array Contact Bumps for Improved Lift Off of Excess Metal,” United States Patent Number 5091288, (1992). 287 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. B. Zhang, S. Forouhar, S. Y. Huang, and W. S. C. Chang, “C2F6 Reactive Ion-Beam Etching of LiNb03 and Nb20 5 and Their Application to Optical Waveguides,” J. Lightwave Technol., LT-2, 528-530, (1984). W. Zhou, G. Ji, and V. M. Ristic, “A Design of Waveguide Collimating Lenses in the Presence of Anisotropic Aberrations,” J. Lightwave Tech., 8(8), 1187-1191, (1990). W. Zhou and V. M. Ristic, “Anisotropic Aberrations in Planar Waveguide Lenses,” IEEE J. Quantum Electron., QE-25(4), 749-754, (1989). R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 6 Surface Acoustic Wave Modulators In Chapter 2, two hybrid-bulk/integrated optical processors were introduced that included surface acoustic wave transducers as signal input devices. These devices were selected because they provide a convenient and efficient means for transforma tion o f the information carried by the electrical signal into a format that can interact with an optical signal. This interaction is accomplished via the acoustooptic (AO) effect. In this chapter, we review the surface acoustic properties of lithium niobate waveguide substrates and introduce a transducer design capable of 100 MHz band width modulation. Several of these transducers have been fabricated for this thesis work, and are described below. Key fabrication details are provided, as well as the results of electrical and optical testing. 6.1 O verview of Inter-Digital SAW Transducers Surface acoustic waves may be generated on a piezoelectric solid such as lith ium niobate by the application o f a voltage to a metal film interdigital transducer (IDT) [Campbell, 1989]. Generally, EDTs are arranged in pairs as shown in Fig. 6.1. The first IDT acts as an input device and is driven with a time varying signal that is then piezoelectrically converted into a mechanical surface acoustic wave. The input signal from the first EDT propagates along the crystal and is received by the second transducer, which converts the mechanical surface acoustic wave back into an output signal. To perform optical modulation on y-cut Ti:LiNb03 waveguides, the IDTs are 289 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ti:LiNb03 Waveguide Exploded view of interdigital transducer Figure 6.1 SAW transducer designs for operation on titanium-indiffused lithium niobate waveguides. The transducer finger width is the critical dimension, CD, and is approximately 2 jlm. oriented with the SAW propagation direction along the crystallographic z-axis, for reasons described in Chapter 3 and summarized in Sect. 6.1.1 below. The surface acoustic wave propagates across the crystal surface perpendicular to the x-axis so that acoustooptical interaction with an x-propagating guided optical wave is possible. This type of interaction is classified as coplanar, noncollinear AO diffraction. This simply indicates that the interaction between the acoustical and optical waves takes place in the same plane, but that the respective wave vectors are at an angle with respect to one another. This type of interaction enables the performance of the signal processing functions described in Chapter 2. Lithium niobate wafers cut for this geometry are referred to as YZ-LiN b03 substrates. The physical relations that describe SAW generation and propagation on lithium niobate are fairly complex and have been thoroughly investigated [Kim and Tsai, 1976; Slobodnik, 1976; Tsai, etal., 1976; Nguyen and Tsai, 1977; Tsai, 1979; Tsai, 1990]. The goal herein was to intro duce a simple IDT design with a useful operational bandwidth for use in the fabrica tion of integrated test modules. Consequently, a brief introduction to the basic 100 am 290 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1852 principles involved in SAW generation and AO interactions as it pertains to the EDT design are given below followed by the fabrication processing sequence and the results of the device characterization. 6.1.1 C rystal Orientation and Physical Properties Several reasons were given in Chapter 3 for the selection of YZ-LiNb0 3 sub strates for integrated optical signal processing applications. These included efficient electromechanical coupling, wideband operation, low SAW propagation losses, and a long collimation length for the surface acoustic wave. To further illustrate the value of this selection, the electromechanical coupling coefficient K2, the SAW velocity, and the SAW propagation loss for YZ-lithium niobate are compared in Table 6.1 with other waveguide materials that support surface acoustic waves. The quantity K2 is defined in terms of the piezoelectric parameters K2 = (6-1) ec in which e is the piezoelectric constant, £ is the dielectric permittivity, and c is the elastic constant. This is a relative measure of the ability of a given piezoelectric to efficiently convert an applied electrical signal to the mechanical power associated with a surface acoustic wave. As can be seen, the value of K 2 is relatively high for lithium niobate compared with other materials, whereas the surface acoustic velocity for YZ-LiNb03 is compa rable with most of the other materials. For those materials whose acoustic velocity is significantly lower such as GaAs and T e02, the acoustic propagation losses are much greater. This trade-off in SAW properties tends to limit the useful SBWP in such materials. Although K2 for GaAs is rather small, useful SAW operation may be per formed with this material by using a highly piezoelectric ZnO layer in the IDT region to enhance the coupling of the electrical signal into a surface acoustic wave. The 291 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6.1: SAW Parameters of Selected Piezoelectric Substrates Material Crystal SAW Velocity K2 r Cut Axis (m/s) (%) dB/cmGHz2 Quartz ST X 3158 0.11 LiNb0 3 Y Z 3488 4.5 0.15 LiN b03 128° X 3992 5.3 Bi12G e02o 110 001 1681 1.4 LiTa03 Y z 3230 0.72 0.1 GaAs <001> (110) <2841 <0.06 30 Te02 620 90 As2S3 2600 170 Ta05/S i0 2/Si -3600 -190 Si3N4/S i0 2/Si -5100 -44 ZnO/Si02/Si -3900 -32 7059/SiO2/Si -4000 -16 after [Nishihara, 1989; Campbell, 1989; Tsai, 1990] acoustic velocity for T e02 given in the table is that of the bulk acoustic wave and not that of the surface acoustic wave. 6.1.2 Guided O ptical Wave/SAW Acoustooptic Interaction When an IDT on a piezoelectric surface is driven by an applied voltage signal, a traveling wave is generated that consists of both a moving index grating and a moving surface corrugation. The acoustooptical modulation of an optical guided mode inci dent upon the surface acoustic wave is the result of diffraction of the light from these periodic structures, with the dominant contribution coming from the index grating [Tsai, 1979]. The diffraction may either be of the Raman-Nath type or the Bragg type 292 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. depending, as in bulk AO interactions, on the magnitude of the AO parameter Q = -----^ • (6.2) neff in which Xq and A are the wavelengths of light in free space and o f the acoustic wave, respectively, L designates the aperture of the acoustic wave, and ne ^ is the effective refractive index of the waveguide. Bragg type diffraction is desired since it provides a larger modulation bandwidth and dynamic range [Tsai, 1979], and is achieved when Q is greater than or equal to An. We consider the case of isotropic Bragg diffraction, for which there is no mode or polarization conversion. Our selection of substrate orientation and mode structure of our Ti:LiNb0 3 waveguides (i.e., isotropic Bragg diffraction is the only AO interac tion possible since there are no other waveguide modes) has ensured this type of Bragg diffraction. For other orientations, acoustooptic interactions in lithium niobate can lead to anisotropic AO Bragg diffraction, in which the diffracted light propagates in a different waveguide mode or with a different polarization than that of the incident light. Under these conditions, light incident upon a periodic SAW perturbation will diffract efficiently at an angle known as the Bragg angle, 0B, that can be determined by momentum conservation considerations. The deflection angle, a , is twice the Bragg angle [N ishihara, et al., 1989] and is given by a = 20 B = 2 sin - l Xq _ _ Xg_ _ X p f , (6.3) 2 n A nA n v in w hich/and v are the SAW frequency and velocity, respectively. The temporal frequency bandwidth over which acoustooptic interactions have a useful diffraction efficiency is limited by both the excitation bandwidth of the trans- 293 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. ducer and the bandwidth o f the acoustooptic interaction. The excitation bandwidth is related to the characteristic impedance of the IDT design. The bandwidth is maxi mized at the optimum number o f finger pairs, N Topt. The relative SAW-excitation bandwidth RBWSA ^relative with respect to the center operation frequency,./) [Nishi- hara, et al., 1989] is given by 2 ^ = - p - . (6.4) / N Topt in which the T in Afj signifies ‘transducer’. This equation implies that for a given SAW-excitation bandwidth, 2AfT , the optimum number of finger pairs increases for larger center frequencies. In practice, when a SAW transducer is fabricated, it is oriented with respect to the optical axis such that the light will be incident upon the surface acoustic wave at the Bragg angle for the central frequency of the IDT. The angular sensitivity of the Bragg diffraction determines the acoustooptic bandwidth. The relative acoustooptical interaction bandwidth RBWAo(a.gain, relative with respect to the center operation fre quency, j) [Nishihara, etal., 1989] is given by RBW,0 s 2 = 1® . (6.5) AO f Q The overall modulation capabilities of the acoustooptic modulator will then be set by the smaller of the two quantities in Eq. 6.4 and Eq. 6.5. The time bandwidth product is a measure of this capability and is given by the product of the overall bandwidth, Af, and the time aperture, r, of the acoustooptic interaction region T • 2Af = — A f’ (6- 6) v 294 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. in which D is the width o f the optical beam. This product is equivalent to the number of resolvable spots that can be obtained in the far-field as a result of deflection of a light beam by the surface acoustic wave within the overall bandwidth o f operation. 6.1.3 SAW Transducer Design Two simple IDTs were designed with the system issues described in Chapter 2 and the performance issues discussed above taken into account. The first design, referred to as EDT1. Because it has the largest accessible bandwidth and most uni form response of the two designs, it is the primary design employed in the experi ments in this work. A photomicrograph of this transducer is shown in Fig. 6.1. The second design, EDT2, was designed with a center-frequency of about 500 MHz, which was later found to be at the limit of our signal generator; thus, the full bandwidth of these transducers could not be accessed in our optical modulation experiments. Some electrical characterization and description of this second design follows in the remain der of this chapter, but the primary focus is on IDT1. IDT1 was designed with 4 finger pairs and has an input line impedance near 50 Q. The center frequency of operation defines the required finger width, CD, of 2.12 /im, which is given b y / = v/(4CD) =411 MHz. The excitation bandwidth given by Eq. 6.4 is 2 A fj= 103 MHz. For modulation on YZ-LiNb03 with X0 = 6328 A, an interaction length of L = 1 mm, and an effective refractive index of ne jj= 2.2069, the AO Q-factor for this design is Q = 25 as given by Eq. 6.2. Therefore, this transducer operates well into the Bragg regime and has an acoustooptic bandwidth given by Eq. 6.5 of 2AfA0 = 164 MHz. The overall bandwidth is thus limited by the excitation bandwidth. For operation at the center frequency, the defection angle, given by Eq. 6.3, is 1.94° and the deflection angle range is 0.485°. 295 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 6.2 Surface Acoustic Wave Transducer Fabrication The surface acoustic wave transducers studied in the present work were fabri cated by standard photolithographic and thin film processes. For the purpose of char acterization of the transducer electrical and acoustooptical modulation properties, Ti:LiNbC> 3 substrates were diced into 2 cm x 2 cm samples, scribed, cleaned, and dehydration-baked as described in Chapter 4 prior to photolithography. Negative photoresist type AZ5214-IR obtained from Hoecht-Celenese was then spin-coated onto the samples at 5 krpm for 30 seconds. The coated samples were pre-exposure baked on the hot plate for 60 seconds at 115° C to bake out excess solvents. Pattern exposure was done for 3 sec (4.15 mW/cm2 at 365 nm) using a light-field mask sup plied by Army Research Laboratories. The sample size selected is sufficient to hold five transducer pairs oriented so that the SAW propagation will take place along the direction of the crystalline z-axis. The transducers in each pair face one another and are situated on opposite sides of the sample so that an acoustic wave launched from one transducer will propagate a distance of 1 cm along the sample surface before it is received by the opposite transducer. The samples were post-exposure baked for 60 sec at 115° C and then blanket-exposed for 30 sec. on the same aligner (4.15 mW/cm at 365 nm) to complete the image reversal process. This two-step exposure process reverses the polarity of the photoresist so that processing in 300 MIF devel oper results in the dissolution of photoresist in the transducer regions and an undercut profile along the edges of the remaining resist. Once the transducers were photolithographically defined, the samples were treated to an oxygen plasma sustained at 150W RF and 200m torr pressure for 10 seconds in order to remove all residual photoresist from the exposed regions of the lithium niobate, thereby improving the metal film adhesion. Aluminum was used as the transducer metal and was deposited in the Balzers BAK-640 boxcoater. Prior to the deposition, the samples were exposed to a glow discharge process at approxi mately 20 mtorr nitrogen pressure. An aluminum thin film of thickness 3500 A was 296 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. deposited by electron-beam evaporation. The samples were removed from the cham ber and the excess metal was lifted-off by photoresist removal with acetone. It was found that the aluminum transducers adhered sufficiently well to the lithium niobate to be gently cleaned with a cotton swab soaked with acetone. The sample was then cleaned with a cotton swab soaked with dilute soapy water, rinsed in DI water, and blown dry with nitrogen. A photomicrograph of a fully fabricated transducer is shown in Fig. 6.2. Following this process for the transducer fabrication, better than 90% yield of the transducers was not difficult to achieve. The finished transducer samples were then mounted in chip carriers using either black rubber epoxy or clear crystal bond. The crystal bond had two advantages: (1) the mounted sample could easily be removed and remounted by dissolution o f the crystal bond with acetone, and (2) the crystal bond material does not deform as easily during the prism mounting process as does the rubber epoxy. Hence prism coupling to crystal bonded samples was somewhat easier. The mounted transducers were then wire bonded to the chip carrier leads using a wedge-bonding apparatus and 0.1 mil aluminum wire. The transducers with 3500 A thickness could be easily and repeatedly bonded without failure, whereas thinner metal layers often resulted in detachment of the pads. The wire-bonded chip carrier was then attached to an aluminum base plate with double-sided tape, and the leads from SMA connectors were soldered to the external chip carrier leads as shown in Fig. 6.3. In the preparation and mounting of the SAW transducer samples, care was taken to leave clear enough space to accommodate prism incoupling and outcoupling at the opposite ends of the waveguide. In some cases, the ends of the chip carriers were milled off prior to mounting the sample in order to provide extra space for the prisms. Once prepared, the entire base-plate fixture could be inserted into a prism coupling mount and fixed down to the mount with a pair of miniature C-clamps. The acous tooptic measurements described below were then taken. 297 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.2 Surface acoustic wave transducer on a lithium niobate waveguide show ing six finger pairs and a 100 /im X 100 /im bonding pad of aluminum metal. 6.3 Surface Acoustic W ave Transducer Electrical Testing The scattering parameters for SAW transducers of two different designs were measured at the Army Research Laboratories with an HP Network Analyzer. As pre viously described, identical transducers were fabricated to act as an equivalent trans mitter/receiver pair. Under these circumstances, the two-port network input and output admittances are S /j and S22, respectively. The transfer admittance, S I2 = S21, consists of both a magnitude and phase part [Campbell, 1989]. The magnitude fre quency response of the transfer admittance is measured to characterize the operation bandwidth of the transducer pair. The 3-dB bandwidth for the SAW that modulates the optical signal corresponds to the 6-dB bandwidth o f the transducer pair magnitude transfer admittance since the frequency response of both the transmitter and receiver should be considered in the electrical measurement. These samples were mounted for electrical and optical characterization as 298 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 6.3 Photograph of transducer fixture for optical beam deflection experiments with three wire-bonded transducer pairs. shown in Fig. 6.3. In order to perform the electrical characterization for each sample, the network analyzer output channel was attached to the SM A connector of one of the transducers. The SMA connector for the opposite transducer in the pair was then con nected to a spectrum analyzer in order to monitor the transmitted signal. A portion of the electrical power used to drive the input transducer is coupled into a surface acous tic wave. For the simple transducer design described in Section 6.1.3, a surface acoustic wave is launched in two directions; hence, the acoustic power that reaches the second transducer is only half of the total coupled acoustic power. Through the piezoelectric effect, the acoustic wave incident upon the second transducer is con verted back into an electrical signal. The amplitude frequency response with fre quency is shown in the two graphs in Fig. 6.4, which represent the frequency- dependent conversion efficiencies of two transducer designs. The designed and mea sured frequency response characteristics are shown for comparison in each graph, and 299 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. CD C O 05 O C M C O .594 MHz 270 M H z! ! / ......-51.6 dB!........... ...LU../. -V.... : -45.1 dB --/A X / 377 MHz..........I 487 MHzl...\h................... -32.6 dB -32.6 dB Designed Measured Center Frequency 411 428 Bandwidth 103 110 o> (0 0 5 o C M C O I _____ I _____ I _____ I _____ I I I - I 200 250 300 350 400 450 500 550 600 650 Frequency (MHz) o -10 -20 -30 -40 -50 -60 -70 -80 -90 -100 T " m E i z 490 MHZ -22.3 dB --51.2 dB .........j....... ^N T -55.() dB / i \ X V / 440 MHz I 539 mHz \ r -28.2 dB -28.2 dB Designed Measured 7 Center Frequency 495 489 i Bandwidth 82 99 300 350 400 450 500 550 Frequency (MHz) 600 650 Figure 6.4 Frequency response of transducer IDT1 (top), and IDT2 (bottom). The transducer bandwidths are approximately 100 MHz for each design, but the central frequencies differ from one another. 3 00 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. are in good agreement. Another electrical measurement that was performed on these samples was a group delay measurement. This measurement determines the signal transit time from one transducer to another and should be equal to the relative separation of the trans ducers divided by the SAW propagation velocity. For the 1 cm separation of the transducers and a SAW velocity of 3488 m/s, the group delay should be approxi mately 2.86 //sec. The actual measured value was found to be in good agreement with this calculated value. Hence, the signal that reaches the opposite transducer in the S2i scattering parameter measurement has the transit time expected for the surface acoustic wave and can be discounted as a bulk acoustic wave propagating in the sub strate. 6.4 Surface Acoustic W ave O ptical M odulation Characterization of the optical modulation properties of the SAW transducer (IDT1) designed and fabricated as described in Sect. 6.1.3 and Sect. 6.2 was per formed by prism coupling a collimated beam (Xq = 6328 /zm) into the waveguide and directing it through the SAW path at or near the Bragg angle, and subsequently prism outcoupling the undiffracted and diffracted light. Diffraction efficiencies were calcu lated as the ratio of diffracted light to the total of diffracted and undiffracted light. A MicroDot Signal Generator was used to drive the input transducer at a constant fre quency, tunable within the range of 200-500 MHz. The signal received by the output transducer was monitored on an HP Model 54503A high frequency digitizing oscillo scope. The prism coupling mount was attached to a stage with a tilt adjustment so that the propagation direction within the waveguide could be Bragg matched to the SAW signal. Error in the measured diffraction efficiencies presented below is due primarily to nonuniformity in the contact of the coupling prisms. For this reason, the prisms 301 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. were placed as close as possible to the acoustooptic interaction region; thus, the zero and first order diffracted beams outcouple at nearly the same position within the aper ture of the outcoupling prism. On the incoupling side, maintaining the same incou pling position is not as important since the diffraction efficiency is normalized by total power. Including prism coupling variations, variations in the collection of light on the surface of the silicon detector, and variations in the beam coupling angle, the experimental error in the diffraction efficiency measurements is at best 5%. In order to determine the stage tilt position that corresponds to a particular beam propagation direction inside the waveguide, the diffraction efficiency was measured for various stage tilt positions at three different operating frequencies for the IDT. This measurement produced three curves that represent the angular Bragg response of the three different period SAWs that were generated. The peak response of each curve was assumed to correspond to the perfect Bragg matching condition so that from Eq. 6.2, the internal propagation angle for the incident beam with respect to the SAW orientation could be calculated for each drive frequency. With these three known internal angles, the internal angles for all other stage tilt positions were calcu lated by a linear fit. The measured diffraction efficiencies are plotted in Fig. 6.5 as a function of the internal angle calculated by this approach. The diffraction efficiencies are plotted as the log-ratio o f the diffracted light power divided by the total incident power at each internal beam angle. Notice that the peak diffraction values for two of the curves are lower than the peak value for the third curve. The lower peak values are due to a lower electrical signal conversion efficiency of the IDT for a given input power away from the IDT center frequency. The width of the response curve at the center frequency was found to be 0.5° to the -3 dB points in diffraction efficiency. The internal beam deflection angle was measured as a function o f frequency as shown in Fig. 6.6. This measurement was performed with a 5 mW collimated He-Ne beam that was coupled into and out of the waveguide as described above. To measure the internal deflection angle, the input beam was Bragg-matched to the center fre- 302 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. = 380 MHz = 428 MHz = 480 M Hz -10 -20 O ) O -30 -40 • - 0.0 0.5 1.0 1.5 2.0 Internal Angle of Incidence Relative to the Optical Axis (degrees) Figure 6.5 Optical diffraction efficiency as a function of input beam angle mea sured from the optical axis for three different drive frequencies within the bandwidth oflDTl. 2.4 c n C D £ O ) C D TJ — Theoretical - o Measured 2.2 o o> < 2.0 c o o C D 0) Q Center frequency (428 MHz) C O c 0 c 400 420 360 380 440 460 480 500 Frequency (MHz) Figure 6.6 Internal deflection angle as a function of frequency for the SAW trans ducer design described in the text. 303 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. quency o f the IDT. Both the undifffacted and diffracted beams were visible over the entire SAW bandwidth. The internal deflection angle was calculated from the mea sured separation of the beams at a distance of about one meter from the SAW position with the refraction that occurs at the waveguide-prism and prism-air interfaces taken into account. By this approach, the distance between the two beams at an intersecting plane one meter from the SAW could be determined within about half a millimeter, which corresponds to an error of about 0.01° in the internal measured angle. The the oretical angles were calculated with the following equation, which takes into account the mismatch in input angle from the frequency dependent Bragg angle [Rastani, 1988] in which 0Bc is the Bragg angle for the central frequency, 0Bj is the Bragg angle for the operating frequency, and Qj is the angle between the deflected beam and the opti cal axis, so that the total beam deflection angle is @ Bc-0d- Note that Eq. 6.2 also holds even if the input beam is not at the Bragg angle that corresponds to the operating fre quency, and can therefore also be used to calculate the diffraction angle. With a same experimental configuration, the acoustooptic diffraction efficiency was measured over the frequency bandwidth o f the SAW transducer when the input optical beam angle was Bragg matched to the transducer’s central frequency. Since the output power of the signal generator varies slightly with frequency, this data was normalized to a constant transducer input drive power over the tested frequency range. The results of this measurement shown in Fig. 6.7 indicate that the drop-off in diffraction efficiency is approximately -3 dB over the 100 MHz bandwidth. These results imply that the diffracted beam intensity at a given drive power is not indepen dent of frequency over the entire modulation bandwidth. Certainly, there is an appli (6.7) 304 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. cation-related trade-off between the tolerable non-uniformity in this response and the required operational bandwidth of the AO interaction. Finally, the diffraction efficiency was measured as a function of input power for three different drive frequencies while Bragg matched to the central frequency. The results are shown in Fig. 6.8. The expected diffraction efficiency dependence is sinu soidal with drive power with near linear response near the 50% diffraction efficiency level. At drive powers less than 1.3 W, the signal generator output was not stable while the transducer could not withstand powers of greater than 2 W. These limita tions prevented us from observing a clear sinusoidal behavior at the extremes o f the diffraction efficiency range. The non-uniform behavior between the three curves may be due to non-linear response of the signal generator to power reflected from the transducer as the drive power is increased. The drive powers used for the measure ment are comparable with those obtained by others for a similar transducer design on YZ-LiNbC>3 substrates [Tsai, 1990]. The electrical insertion loss of this transducer design, shown in Fig. 6.4, is high (-26 dB) and the efficiency may be improved with the use of impedance matching circuits. 6.5 Summary In this chapter, we introduced the principles behind surface acoustic wave mod ulation in piezoelectric materials and considered in particular YZ-lithium niobate sub strates. This material was compared with some other promising waveguide substrate materials and found to have equivalent, and in some cases, favorable properties. The issues related to the design of a simple SAW IDT were discussed and the design for a 100 MHz bandwidth IDT was given. Several transducers were fabricated using this design, electrically tested, and optically tested. The measured AO interaction band width was found suitable for the 10 signal processors discussed in this thesis. 305 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. m T3 C L CL O ) O -5 - -10 - -15 - O O o o o Center frequency (428 MHz) -20 _L 350 400 450 Frequency (MHz) o o o o 500 Figure 6.7 Diffraction efficiency with frequency when Bragg matched to the center frequency of EDT1. - A - 400 MHz - 0 - 430 MHz - 3 - 460 MHz p 0.8 O O X I ' 0.6 ( D ’o 0.4 c o o (0 u o.o 14 15 13 16 17 18 19 20 Transducer Input Power Setting (x100 mW) Figure 6.8 Diffraction efficiency as a function of transducer drive power for three different drive frequencies. 306 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 6.6 References C. Campbell, Surface Acoustic Wave Devices and Their Signal Processing Applications, (Academic Press, Inc., Boston, 1989). B. Kim and C. S. Tsai, “High Performance Guided-Wave Acoustooptic Scanning Devices Using Multiple Surface Acoustic Waves,” Proc. IEEE, 64,788-793, (1976). L. T. Nguyen and C. S. Tsai, “Efficient Wideband Guided-Wave Acoustooptic Bragg Diffraction Using Phased Surface Acoustic Wave Array in LiNb03 Waveguides,” Appl. Opt., 16(5), 1297-1304, (1977). H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). A. J. Slobodnik, “Surface Acoustic Waves and SAW Materials,” Proc. IEEE, 64(5), 581-595,(1976). C. S. Tsai, “Guided-Wave Acoustooptic Bragg Modulators for Wide-Band Integrated Optic Communications and Signal Processing,” IEEE Trans. Circuits Syst., CAS- 26(12), 1072-1098, (1979). C. S. Tsai, Ed., Guided-Wave Acousto-Optics, Springer Series in Electronics and Photonics, D. H. Auston, etal., Eds., Vol. 23, (Springer-Verlag, New York, 1990). C. S. Tsai, M. A. Alhaider, L. T. Nguyen, and B. Kim, “Wide-Band Guided-Wave Acoustooptic Bragg Diffraction and Devices Using Multiple Tilted Surface Acoustic Waves,” Proc. IEEE, 64(3), 318-328, (1976). 307 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 7 Integrated Optical Test Modules In Chapters 3-6 we presented results on the fabrication and characterization of the waveguide and key components for the advanced 10 signal processors. We dem onstrated dense rib waveguide arrays fabricated on low-loss single-mode Ti:LiNb03 waveguides, large aperture lenses with a 1 cm focal length and a 2 fim focal spot size with throughput of 50% over a 5 mm aperture, and SAW transducers with an acous tooptic modulation bandwidth of 100 MHz. In this chapter, we present the results of the fabrication and characterization of integrated optical test modules that consist of these devices integrated onto a common substrate. This test module integration is the first step towards the integration of the more complex advanced 10 signal processors. To complete the discussion of our research results, this chapter concludes with our assessment of the component development and integration results and the feasibility of full IOSAR processor or 10 correlator integration. The test module geometry is shown in Fig. 7.1. This configuration is similar to that of 10 spectrum analyzers investigated by others [Anderson, 1978; Bamoski, et al., 1979; Mergerian, et al., 1980; Mergerian, etal., 1983; Valette, et al., 1983]. How ever, a rib waveguide array rather than a detector array is situated with its entrance plane in the back focal plane of the waveguide lens. A collimated beam of light, cou pled into the module with a prism, is focused by the lens into a single rib waveguide. The transducer situated before the lens is used to deflect the guided beam at an angle that is dependent upon the spatial frequency of the surface acoustic wave. Through adjustment of the SAW transducer drive frequency, the input light is selectively 308 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Figure 7.1 Schematic diagram of integrated test module on a Ti:LiNb0 3 waveguide with integrated transducer, lens, rib waveguide array, and alignment fidu- cials. focused into different rib waveguides within the array. Two specific functions of this test module were important to demonstrate and evaluate. The first function is the excitation of a single rib waveguide with the waveguide lens and no transducer input signal. From our previous experimental mea surement of the embedded lens focal spot size, this should be feasible for a 10 /an rib waveguide pitch. The actual demonstration of this function will indicate the compati bility of the component fabrication processes and the combined performance of two integrated components. The second function is the selective excitation of different rib waveguides in the array by adjustment of the transducer input drive frequency. Dem onstration of this function will indicate the capacity of the lens and rib waveguide array to handle variable inputs. In these experiments, the rib waveguide excitation is observed at the polished end of the rib waveguide array. The quality of the lens focus can in part be evaluated based upon the observed level of light in the neighboring rib waveguides. However, if the rib waveguides have a significant coupling constant, the mechanism responsible for excitation of the neighboring rib waveguides will be unclear. Since a rib 309 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 644975 waveguide array with 8 fim wide rib waveguides, 2 fim wide gaps etched 5000 A deep, and a 1 cm propagation distance was found in Chapter 4 to lead to coupling crosstalk greater than -2 0 dB, a rib waveguide array with 6 ftm wide ribs waveguides, 4 fim wide gaps etched 5000 A deep, and a 5 mm propagation distance was used in these experiments to reduce the rib-to-rib coupling even further. Although the general geometry of the test module has been investigated previ ously, the combination o f a large aperture, low f-number embedded lens with a high density rib waveguide array and a SAW transducer represents the first implementation of this degree of integration to the best of our knowledge, and is therefore a unique feature of the present work. In this chapter, the geometric considerations of the test module are presented. The incident field coupling into a rib waveguide array is then calculated for the test modulator as well as for the IOSAR processor. We then describe the integration of the test modules and the results o f the two test module experiments. Finally, we conclude with an evaluation of the advanced IO signal pro cessors based on these integrated test module results. 7.1 Test M odule G eom etric Considerations A rib waveguide array with a 10 fim pitch from one rib to the next was used to demonstrate the selective focusing capability of the lens. For larger rib waveguide pitches, the number of accessible channels will be reduced because of the limited deflection range of the transducer. The SAW transducer used in the fabrication of the integrated test modules was described in Chapter 6. It functions with a 100 MHz bandwidth and a central frequency of 430 MHz. The internal deflection angle for this design was also theoretically calculated and experimentally measured. The central deflection angle is approximately 2° and varies by ±0.3° over the modulation band width. The expected shift in the focal spot position over this range of angles is Ax = ftm (A d D) = 100 fim for a 1-cm focal length lens. Therefore, it is expected that 310 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the focused beam may be steered over the width o f approximately 10 rib waveguides in the focal plane of the lens. The position of focus of the zero order is separated from the diffracted orders by about a distance of 350 fim. The ability of the lens to focus light into a single rib waveguide is strongly dependent upon the relative placement of the lens and rib waveguide array compo nents. If the rib waveguide array entrance plane is displaced a small distance ±£ from the lens focal plane then the associated width of the region illuminated by the lens will be eD lfio t a given input beam width D and lens focal length/. For a 1 cm aper ture f/1 lens, a placement error of e = 50 fim will lead to the illumination of 5 rib waveguides. A relative placement error for the rib waveguide array and lens can result from either variation in the lens focal length or inaccuracy in the photolitho graphic alignment of the devices. Photolithographic alignment can be controlled to within ±1 fim, whereas variation in the lens focal length can be much greater than this. We determined in Chapter 5 that the primary process variation that leads to a change in focal length in the embedded lens was due to refractive index variations in the embedded thin film waveguide layer. For a 1% variation in the refractive index of this layer, the focal length o f the lens will shift by 200 fim. If a 1 cm aperture, f/1 lens with this variation in the position of the focal length is placed exactly 1 cm in front of a rib waveguide array with a 10 fim rib waveguide pitch, then 20 rib waveguides will be illuminated. In order to characterize a module in which such a sizeable error has occurred in the relative component placement or in the lens focal length, as is the probable case, a slightly focused or defocused beam can be coupled into the waveguide rather than a collimated beam in order to shift the position of the integrated lens focal plane. Con sider the geometry shown in Fig. 7.2, in which an external lens is used to focus a col limated beam a distance D 0 in front of an integrated lens of aperture width WL. The light is refracted when it is prism coupled into the waveguide so that the angle of divergence is slightly changed. As a result of the refraction, the light incident upon 311 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Coupling Prism D 0 D , Figure 7.2 Geometry for focal length compensation with an external lens. The external lens (not shown) is used to focus a collimated beam a distance D 0 in front of the integrated lens. The external focus is effectively a distance D , away from the lens as a result of the refraction of the coupled beam at the prism/waveguide boundary. The position of the focal spot is given by the imaging condition. the lens effectively emanates from point source a distance D t before the lens. This distance can be determined geometrically and is given by D f = U n 2 H - l ) w L2 + n 2 HD Z , (7.1) in which it has been assumed that the incident medium is air (refractive index = 1) and nH is the refractive index of the waveguide. The distance that the light is focused after the lens is then given by the imaging condition (7.2) in which D 2 is the Iens-to-image distance a n d /is the lens focal length. Hence, the focal distance can be shifted through adjustment of the input beam divergence. Two other approaches may be used to correct a misfocused lens. The first approach would be simply to strip the thin film materials from the lens recess and re deposit slightly different thin film layers in an attempt to compensate for the focal 312 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. length error. The second approach would be applicable if the lens focuses before the rib waveguide array (i.e., the relative index change is too large). A high index film such as Si3N4 (n = 2.0) deposited on the embedded lens waveguide surface will cause an increase in the effective refractive index of the guided mode. For the 1 cm aper ture, 1 cm focal length S i02 /MgF2 embedded lens structure discussed in Chapter 5, a 1000 A overlayer of Si3N4 will lead to a 0.25% change in the effective refractive index of the lowest order waveguide mode and subsequently to a 0.25% change in the focal length. For a 1 cm focal length lens, this change corresponds to a 25 fim focal plane shift. If this corrective film is made too thick, the position of the embedded waveguide mode may be shifted upwards and result in a drop in the coupling effi ciency through each lens interface. In addition, a thick overlayer can lead to a failure of the thin film structure to adhere within the lens recess. These limitations make it impractical to correct a focal length error o f 200 fim by this method. In the selective focusing experiment, the beam input into the lens is no longer necessarily parallel to the optical axis since it must be tilted to pass through the sur face acoustic wave at the Bragg angle. Even if the relative component spacing is accurate for an on-axis focus, tilted beam inputs may not focus at the rib waveguide array due to lens field curvature effects. In this case, an external lens is needed to compensate for this misfocus as described above. When the input beam is directed through the surface acoustic wave at the Bragg angle and the external lens is positioned to optimize the focus of the diffracted order into a rib waveguide, the undiffracted order will not be focused at the entrance o f the rib waveguide array and will illuminate several rib waveguides. For the SEACF/1 lens design presented in Section 5.2.1, we found that the Petzval radius was -6.7 mm. Thus, for a 2° surface acoustic wave deflection angle and this Petzval radius, the geo metrically determined width of the region illuminated by the undiffracted light at the rib waveguide array is approximately 10 fim. Since the test module does not have a zero-order stop, our concern is that the zero-order light does not overlap the diffracted 313 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. light for the 1-cm aperture lens. As we found above, the distance between the zero- order and first order beam focus is nearly 350 fim, hence the undiffracted light is not a concern in this case. 7.2 Focusing with an Embedded Lens into Rib Waveguides In this section, we present our calculations of the coupling efficiency of an input (optical) field into and array of rib waveguides for two different cases. The first case corresponds directly to the test module geometry discussed above, in which a beam that is focused to a 0.9 fim focal spot width is incident on the 6 fim wide rib waveguides. This analysis provides us with key information regarding the coupling efficiency obtained in the test modules. In addition, analysis of the incoupling effi ciency as a function of incoupling angle provides information regarding the angular selectivity o f the rib waveguide structure. The second case corresponds to the IOS AR geometry, in which we assume that a flat-field lens doublet (introduced in Chapter 5) is used to focus light diffracted from a chirped surface acoustic wave into a rib waveguide array with 8 fim wide rib waveguides and 2 fim wide gaps. The amount of light coupled into the neighboring rib waveguides is also calculated to provide us with an estimate o f the crosstalk level that results from the input-field side-lobe coupling. In order to calculate the input-field coupling efficiency in the test modules, the rib-waveguide-array field distributions were first calculated. These calculations were accomplished within the theoretical framework presented in Chapter 4. The solution for the eigenfunction equation along the vertical dimension was included as a part of that discussion. In order to solve the eigenfunction equations for the transverse dimension, we assumed the gap region to have a refractive index equal to that derived from the rib-to-rib coupling measurement of the rib waveguide structure with a 5000 A gap depth (ng = 2.199). With this assumption, we found that the 6 fim wide structure supports 6 transverse guided modes. 314 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. In the selective focusing experiment, a SAW transducer is used to deflect the guided beam. Since the angle of incidence of the focused beam on the rib waveguide entrance face will deviate by ±0.3 degrees as the transducer is tuned over a 100 MHz bandwidth, the coupling efficiency was studied as a function of angle. In these test modules, the SAW transducers were aligned along the crystallographic z-axis. In addition, the lens and rib waveguide components were lined up along the crystallo graphic x-axis. Therefore, when the input optical beam is Bragg matched to the SAW center frequency, the deflected beam will propagate at an angle a few degrees off the optical axis. In our coupling efficiency calculations, we considered incidence angles from normal incidence up to 8° off axis, which is significantly larger than we would expect to occur in practice even if there were some angular misalignment o f the SAW transducer. The coupling efficiency is given by the mode overlap integral [Nishihara, et al., 1989], in which we adopt the coordinate system employed in Chapter 4 (x signifies the lateral direction across the entrance of the rib waveguide array, and z signifies the direction along the rib waveguide array). In this equation Ex is the field distribution of the mth mode in the nth rib waveguide in the array and E^ist is the electric field profile o f the embedded lens within the focal spot. The measured focal spot intensity profile of the embedded lens structures pre sented in Chapter 5 may be closely approximated by a Gaussian function with a FWHM focal spot size of 2 jim in air and 0.9 /xm in the Ti:LiNb03 waveguide (calcu lated). For small angles of incidence, the complex-valued field distribution for the focused beam at the rib waveguide entrance face may be expressed as T E ™ ( x - n W ) E jisl( x - x a) d x ' |« r O Q # (7.3) 315 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. E dist(x ) = exP x cos 8i \ 2 d0 iP 0x s m 6 i (7.4) in which 0[ is the angle of incidence, d0 is the focal spot width to the 1 fe point in amplitude, and (3 0 is the propagation constant o f the guided mode in the planar waveguide. For the single-mode Ti:LiNb0 3 waveguides, the propagation constant is /30 = 21.91 /a n '1 . Also, for a Gaussian focal spot with a FWHM in intensity of 0.9 /an, the distance between the l/e points in amplitude d0 = 0.76 /an. With these values, Eq. 7.3, Eq. 7.4, and the expressions for the electric field distributions, the coupling efficiencies were calculated for a focused beam centered in a 6 /an wide rib waveguide for various beam angles. The results o f these calculations are shown in Fig. 7.3. Since the focal spot width is much less than the rib waveguide width, sev eral modes are excited. Near normal incidence, the light couples predominantly into the even-number modes with approximately 34% of the optical power coupled into the lowest order mode. As the beam angle is increased the coupling into the odd- numbered modes increases. A corresponding decrease of light coupled into the first two even-numbered modes occurs such that the total coupling efficiency summed over all of the modes is greater than 90% only for input beams tilted up to 2° off-axis. For larger angles, the total coupling efficiency drops off more rapidly with angle. The total coupling efficiency is less than 50% for angles larger than the acceptance angle of this rib waveguide structure (7°). Over the range of beam deflection angles produced by the SAW transducer, the variation in coupling efficiency is very small as long as the light is close to normal incidence on the rib waveguide entrance plane. Therefore, we expect this rib waveguide array structure to function fairly well in the test modules. In addition, there is a good correlation between the acceptance angle, as calculated from the prin ciple of total internal reflection, and the angle where the coupling efficiency has sig- 316 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. >* o c © o 5 1 = L U O ) c Q. 3 O o 1.0 0.9 0.8 0 .7 Total m = 0 m = 1 m = 2 m = 3 m = 4 m = 5 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 1 2 3 4 5 8 6 7 Incidence Angle (degrees) Figure 7.3 Calculated coupling efficiency for a beam focused into a 6 {tin wide rib waveguide by an embedded lens as a function of incidence angle. Near normal inci dence, the first three symmetric modes are predominantly excited. The coupling into the anti-symmetric modes increases with larger angular offset. nificantly dropped, as calculated in this mode coupling analysis. Next, therefore, we consider the coupling efficiency of the optical field distribu tion formed by light diffracted from a SAW chirp signal and directed through the flat- field lens doublet into a rib waveguide array with 8 fim. wide rib waveguides and 2 {tin wide gaps. The purpose of this calculation is to compare the amount of light coupled into the central and neighboring rib waveguides in the IOSAR configuration for a range- focused spot. The flat-field lens doublet has a focal spot size nearly half that o f the f/4 lens given in the IOSAR processor example o f Chapter 2. Since the focal spot size is smaller, there should be less side-lobe coupling into the neighboring rib waveguides than with the f74 lens. The operation of the flat-field lens doublet was evaluated with Code V software under the assumption that the input light (A = 6328 A) was dif- 317 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. fracted from a 100 MHz bandwidth SAW chirp. The SAW chirp was modeled to act as a lens a focal length and position related to the chirp rate as was discussed in Chap ter 2. As a result of this analysis, the spot profile of the light in the focal plane o f the system was found to have a focal spot width of approximately 4.5 /im. The calculated intensity profile is shown in Fig. 7.4. The profile has the characteristic sine2 shape expected for a uniformly illuminated lens. The coupling efficiency of this light focused into a rib waveguide array (8 [im wide rib waveguides, 2 fim wide gaps) was calculated for normal incidence by using Eq. 7.4. The field distributions for these rib waveguides were derived in the same man ner as was done for the test module case described above. This analysis was per formed twice with a different assumed value for the gap refractive index in each case. In the first analysis, the gap refractive index was assumed to be ng = 2.199, which is the value derived from the rib-to-rib coupling calculations for a 5000 A deep gap region. In the second analysis, the gap refractive index was assumed to be ng = 1.9, which corresponds to the value obtained through consideration of the mean dielectric permittivity in the gap region for a 5000 A gap depth. The mean dielectric permittiv ity under these assumptions was calculated in Chapter 4. The values obtained for the coupling efficiency in both analyses were nearly identical. The total amount o f power coupled into the central rib waveguide for both cases is 90% (88% of which is cou pled into the lowest order mode) and 2% of the power is coupled into each o f the nearest neighboring rib waveguides. The presence of light in the neighboring rib waveguides indicates that there may be crosstalk on the order of -17 dB even though the rib waveguide width is approximately twice the focal spot width. This crosstalk is due to the input-field coupling considerations alone and does not include the rib-to-rib coupling. Therefore, it may be unnecessary to reduce the crosstalk due to rib-to-rib coupling much beyond -2 0 dB in the design of the rib waveguide array since the side- lobes of the input-field will introduce crosstalk above this level. The primary method to reduce the side-lobe level of the focused beam is to 318 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 10 Log(PSide/ P total) = -17dB 0.6 0.4 0.2 ri = 0.02 /rj = 0.9 \ r\= 0.02 : u — ; — l j — ; — 0.0 -20 -10 0 10 20 Position (microns) Figure 7.4 Calculated coupling efficiency for a beam focused into an 8/im wide rib waveguide by a chirped SAW signal and lens doublet combination. The calculated coupling efficiency into the neighboring rib waveguides separated by 2 fim gaps from the center rib waveguide is also shown. shape the intensity distribution of the light that enters the optical system. In the IOSAR processor, the shape of the intensity profile across the 2.7 cm lens aperture is not really and adjustable parameter since the light directed towards the rib waveguide array is diffracted from a 3.5 mm wide SAW modulation at an arbitrary position within the 2.7 cm aperture. Rather, the SAW modulation depth should be varied across the record length so that the light diffracted from a single SAW chirp is shaped across the width of the modulation. To accomplish this modulation, however, the ini tial radar pulse produced by the radar system should be shaped with the IOSAR pro cessor requirements in mind. 7.3 Component Integration Sequence The fabrication o f the test modules consists of five steps: fabrication of the 319 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ti:LiNb03 waveguide, argon-ion beam etching of the rib waveguide structure, waveguide end-facet preparation, argon-ion beam etching of the lens recess, and the thin film deposition of the transducers and embedded lens waveguide layers. After the module fabrication, there are several more steps that are required to package and mount the module for guided-wave excitation and characterization. To package the test module we must bond the module to a chip carrier, wire-bond the internal carrier leads to the transducer pads, mount the chip carrier onto an aluminum holder, solder the SMA connectors to the external chip carrier leads, attach the SMA cables, and mount the aluminum holder onto the prism coupling holder. In the first processing step, the waveguide formation and substrate preparation procedure (described in Chapter 3) was used to prepare 1 in. X 1 in. low-loss Ti:LiNb03 waveguides for test module fabrication. Before the component fabrication and between each step in the fabrication sequence, when possible, light was incoupled and outcoupled from the planar waveguide to determine whether the fabrication pro cess had affected the waveguide performance. In the second processing step, the rib waveguides were patterned and etched near one edge of the waveguide as described in Chapter 4. The photomask that was used consisted o f the rib waveguide array pattern as well as fiducial patterns needed for the alignment of the integrated lens to the rib waveguide. The rib waveguide array pattern on the photomask was designed to produce 8 fim wide rib waveguides with 2 fim wide gaps. However, slight over-exposure and over-development of the photo resist resulted in gap regions in the photoresist that were 4 ftm wide. Although this pattern is not o f the original design, the wider gap region is beneficial for the test module experiments for which we would like to minimize the rib-to-rib coupling. The fiducial patterns were transferred into the waveguide surface during the rib waveguide etching process so that the embedded lens recess could be accurately aligned to the rib waveguides in a later processing step. The end of the rib waveguide array near the sample edge was then sliced off 320 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. with a wire saw, about 1 mm into the sample, in order to remove the nonuniform por tion of the array affected by the photoresist edge-bead. This end of the sample was end-polished to permit observation of the rib waveguide end-emission during the module characterization. The end-polishing sequence was described in Chapter 3. Wafer-dicing tape was attached to the waveguide surface to protect it from becoming damaged. Since there is a small risk of damage to the sample with all of the mechan ical handling required in the polishing sequence, a minimal amount of processing is performed prior to this step. If a sample is clamped too tight in the polishing mount it is likely to break. Previously we have observed fractures in the sample that occur where the back-side was scribed with an identification mark. The wafer dicing tape used during the polishing process is destructive to thin films deposited on the surface of the waveguide when it is removed. Therefore, no thin films are deposited prior to the polishing sequence. The disadvantage of end-polishing the waveguide early in the integration sequence is that there is an increased risk that the carefully prepared edge will be damaged during the subsequent sample processing. With care in the later pro cessing steps, this has been not found to be a major concern. After the rib waveguides were fabricated and end-polished, the sample was coated with a thick titanium layer for the embedded lens recess fabrication as described in Chapter 5. We found that the rib waveguides remain unaffected by the deposition and removal o f the titanium films. The polished waveguide edge was next coated with colloidal graphite and allowed to dry before the lens recess was etched. The sample was coated in this manner to mask the edge from exposure to the argon- ion beam. Since the removal of the titanium mask layer requires the use of a dilute hydrofluoric acid mixture, the processing sequence is arranged to avoid the fabrica tion of surface structures that are sensitive to this etchant before the embedded lens recess formation. In addition, after the titanium films are removed, the surface of the sample is rigorously cleaned with a cotton swab soaked with dilute soapy water. Therefore, the fabrication of fragile structures should also be avoided before this step 321 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. in the process. Since the titanium layer is patterned by the lift-off technique, the lens mask can be photolithographically aligned to the rib waveguide array. Fiducial marks on the lens photomask can be aligned with the fiducial marks that were etched into the waveguide surface along with the rib waveguide array. In the next process step, the aluminum transducers were fabricated. Since the waveguide substrate consists only of etched features and a polished edge, there are few process or component incompatibility issues involved in this step. The trans ducer fabrication was carried out as described in Chapter 7 with the alignment of the transducer performed relative to the embedded lens recess. On the photomask that was used to pattern these devices, the device spacing permitted approximately three transducer pairs to be fabricated between the embedded lens recess and the waveguide edge. As the last step in the device integration sequence, the embedded lens thin films were deposited. The process used was the same as that described in Chapter 5, in which the majority of the Ti:LiNb0 3 surface was physically masked with aluminum foil to prevent deposition onto the other devices and the prism coupling region. The dielectric thin-film deposition was arranged as the last step in the processing sequence since it is generally suggested not to perform wet processing steps after the dielectrics have been deposited [Nishihara, et al., 1989]. Subsequent wet processing is avoided because of the moisture sensitivity of most thin films, which can result in a change in the thin film properties or detachment of the thin films altogether. After the integrated module was fabricated, it was mounted for characterization purposes as illustrated in Fig. 7.5. The mounting process described in Chapter 6 that was used to prepare transducer samples for optical characterization was also used for the test modules with only two differences. The first difference is that during the mounting of the test module to the chip carrier, additional care was taken to minimize heating of the sample in order to reduce the risk of damage to the thin film dielectrics on the sample surface. The first step of the mounting process involves the attachment 322 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. © © Figure 7.5 Schematic diagram of an integrated test module showing (a) SAW trans ducers, (b) thin film overlayers, (c) embedded lens, (d) rib waveguide array, (e) Ti:LiNb03 waveguide, (0 chip carrier, (g) SMA connectors, and (h) aluminum base. o f the waveguide substrate to a chip carrier using crystal bond adhesive. This mate rial is solid at room temperature and must be heated to approximately 100° C before it liquefies and becomes tacky. In order to minimize the heating during this process, the chip carrier was heated on the hot-plate until the crystal bond material melted on the carrier surface. The carrier was then removed and allowed to cool for one minute before the module was attached. With this process, the embedded lens thin films remained unharmed. The second difference between the mounted transducer samples and the mounted test modules is that the test modules were placed off center towards 323 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. one edge of the chip carrier so that a microscope objective could be used to observe the rib waveguide end-emission during the module characterization process. 7.4 Description of Fabricated Modules Two integrated test modules were fabricated with the process sequence described above. The first test module was used to demonstrate focusing into a single rib waveguide. It consists of a 1000 element rib waveguide array (6 ftm wide ribs, 4 ftm wide gaps, and a 5000 A gap depth) and an embedded lens structure (sample Lens-16) with a recess depth of 3.1 ftm placed 0.97 cm from the front edge of the rib waveguide array. This distance was purposely made less than 1 cm since the ellipso- metrically measured refractive indices of the host waveguide and embedded waveguide materials were slightly lower than the values used in the lens design. A lower embedded lens refractive index implies that the focal length will be shorter than the nominal design focal length. The embedded lens recess was formed using sub strate tilt and rotation during the argon-ion beam milling process as described in Chapter 5 and summarized in Table 5.4. The thin film waveguide consisted of a 2 ftm thick S i0 2 waveguiding layer with a 0.65 ftm thick MgF2 barrier layer as summarized in Table 5.5. The functionality o f the rib waveguide array was verified by prism cou pling before the fabrication of the embedded lens recess. The second module was identical to the module described above in all manners except that the transducers were also fabricated on the waveguide. This test module was used to demonstrate selective focusing into rib waveguides by adjustment of the SAW transducer drive frequency. The module was mounted on a chip carrier so that the transducers could be wire-bonded. The lens recess milling and waveguide deposi tions for this module (sample Lens-17) are also summarized in Tables 5.4 and 5.5, respectively. 324 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 7.5 Excitation of a Single Rib Waveguide With the first test module containing Lens-16, we sought to demonstrate the combined performance of the embedded lens and rib waveguide array in a Ti:LiNb03 host waveguide. As described earlier, our intent was to use the waveguide lens to focus a wide beam of light into a single rib waveguide within an array of rib waveguides. Excitation o f a single rib waveguide should be possible since the embedded lens focal spot size (0.9 fim) is significantly less than the rib waveguide width (6 /mi). The incoupling calculations presented above indicated that as much as 95% of the light incident on the rib waveguide array will be coupled into a single rib waveguide if it is properly positioned in the middle o f the rib waveguide. In Chapter 5, the measured side-lobe level of the focused spot from an embedded lens was -11 dB below the peak intensity level. In addition, the crosstalk due to rib-to-rib coupling should be more than -2 0 dB down from the peak intensity level of the excited rib waveguide. Hence, we expect that little light will be present in the rib waveguides that neighbor the excited rib waveguide. In our experimental set-up for characterization o f the first test module (sample Lens-16), we used a prism coupling mount, a 15 mW He-Ne laser, a spatial filter, col- limation optics, and an additional cylindrical lens to adjust the waveguide lens focal plane position as described earlier. The laser was linearly polarized and oriented so that the TE-polarized waveguide mode could be excited. A 40x microscope objective was used to image the end emission of the rib waveguides onto a CCD array placed approximately 20 cm away from the shoulder of the objective. Since the coupling prism is only 1 cm wide, we were able to illuminate only about 0.75 cm of the lens uniformly. Without the external adjustment lens in place, the collimated beam was prism coupled into the waveguide and the lens was illuminated. We observed the rib waveguide end-emission and found that approximately 15 consecutive rib waveguides were excited to various degrees. The excitation of several rib waveguides indicated that there was some error in the relative placement of the rib 325 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide array and the embedded lens compared with the focal length o f the lens. The number of rib waveguides that were excited indicated that the error was on the order o f 300 fim, since the lens functions at approximately f/2 and a 150 fim region was illuminated. The cylindrical lens was then placed after the spatial filter and before the colli mating lens. Through translation o f the cylindrical lens position, the beam could be focused to a line anywhere from 20 cm before the prism coupling mount to several meters after the mount. The position o f the external cylindrical lens and the tilt of the prism coupling stage were adjusted to provide optimum coupling into a single rib waveguide. A calibrated line scan of the rib waveguide end-emission that was acquired with the CCD camera is shown in Fig. 7.6. This optimal focusing was obtained when the cylindrical lens was placed such that the beam focus position occurred 12 cm before the integrated lens and approximately 0.6 cm of the integrated lens aperture was illuminated. With the refractive index of the host waveguide (n = 2.2), height of the aperture (H = 0.3 cm), and distance of the external focus (D0 = 12 cm) inserted into Eqs. 7.1 and 7.2, the shift in the focal position inside the sample was found to be approximately 400 fim. Since the beam coupled into the waveguide was a diverging wave, the focal plane was shifted 400 fim away from the lens. The direction of this shift indicates that the focal length of the lens was shorter than expected, and hence the refractive index of the embedded waveguide material was probably lower than the expected value (i.e., n = 1.46) by about 2%. Excitation of a a single rib waveguide is clearly evident in Fig. 7.6 with the sig nal in the nearest-neighbor waveguides approximately 10 dB below the peak level of the central rib waveguide. To either side of the central rib about 60 fim to 100 fim away there is some guided wave excitation about 6 dB down from the level of the central rib. The presence of this light indicates that there may be scattering some where within the module. A likely place for this scattering to occur is at the embed ded lens interfaces or from within the embedded waveguide film. Alternatively, this 326 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.0 0.8 to c = J . 0.6 ■ ( / ) c 0.4 c 0.2 -100 -50 0 50 100 Position (microns) Figure 7.6 Measured end-emission from a rib waveguide array in which light is focused by an embedded lens into a single rib waveguide. scattering may be from light incident on the etched gaps separating the rib waveguides in the rib waveguide array. 7.6 Selective Focusing into Several Rib Waveguides In the same manner as described above, the second test module containing Lens-17 was used for the selective focusing experiment. As described previously, this test module was prepared with transducers and placed into a chip carrier so that the input beam could be modulated by driving the transducers. In this case the sample was mounted onto an aluminum holder and inserted into the prism coupling arrange ment with the SMA cables attached. All three transducer pairs on this sample had excellent pattern definition without any observable defects in the metal lines. In addi tion, all of the IDT pads were successfully wire bonded. A constant frequency electri cal signal was applied to the input transducer and monitored at the output transducer 327 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. with a high frequency oscilloscope. The presence of a signal at the output transducer indicated that a surface acoustic wave was successfully launched across the crystal. With the power to the transducer turned off, a beam was coupled into the waveguide. The external lens position and prism-coupling-stage tilt angle were adjusted to provide optimum coupling of the light focused by the embedded lens into a single rib waveguide. The results were similar to those given above for sample Lens-16. With the power to the transducer increased to provide approximately 50% diffraction efficiency and the drive frequency set to the transducer center band fre quency, the beam was clearly diffracted to another portion of the rib waveguide array. The external lens position and prism coupling stage tilt were then adjusted in order to optimize the focus for the diffracted order. Through variation of the signal generator frequency over a portion of the transducer bandwidth, we were able to steer the focal spot across approximately six of the rib waveguides as shown in Fig. 7.7. At some transducer input frequencies, a single rib waveguide was clearly excited. When the transducer input was adjusted so as to excite the third rib waveguide of the six, a fair amount of light was observed in nearby rib waveguides. The fourth rib waveguide failed to light up whatsoever. We suspect that a defect somewhere along the fourth rib waveguide is responsible for decoupling the light. The position of the excited rib waveguide was plotted as a function of drive fre quency and compared with the expected beam translation as shown in Fig. 7.8. The measurements are in excellent agreement with the theoretical calculations. This agreement indicates that the components are working properly together and that there is no ambiguity in the observations of Fig. 7.7. 7.7 IOSAR Processor Assessment In the work presented in this thesis, we have successfully established Ti:LiN b03 waveguide fabrication parameters that yield waveguides with properties that are suit- 328 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 0.12 0.10 c o 0.08 0.06 0.04 0.02 0.00 0 10 20 30 40 50 60 70 Position (microns) Figure 7.7 Measured end-emission from a rib waveguide array in which individual rib elements are selectively excited by a transducer and embedded lens (lens-17) com bination integrated on a Ti:LiNb0 3 waveguide (second test module). Each curve cor responds to a different transducer input frequency. able for advanced IO signal processors. The waveguides have a low propagation loss of 0.41 dB/cm, support a single TE-polarized mode, and provide good confinement of the mode within 2 fjxn. of the waveguide surface. In addition, it was shown that the waveguide process variations can be tightly controlled as long as the diffusion proper ties of each substrate wafer are determined prior to the processor fabrication. The primary limitation of the LiN b03 waveguide substrate in the IOSAR pro cessor application is the wafer dimension. At present, y-cut LiN b03 wafers are avail able in 4” (10 cm) diameter wafers. However, the required substrate wafer diameter for the processor example given in Chapter 2, which includes beam expansion and collimation lenses, was 7.7” (19.5 cm). As designed, the example IOSAR processor as given in Chapter 2 is not presently feasible. One alternative approach is to use integrated beam turning mirrors to fold the optical path within the wafer and provide a 329 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 440 N X 2 420 >* o c d > 3 O' < D w U- • Measured Calculated 400 380 30 0 10 20 40 50 60 70 Position (microns) Figure 7.8 Comparison o f measured frequency for maximum excitation of individ ual rib waveguides in an array to the calculated beam deflection with frequency. greater path length. A second alternative approach is to eliminate the need for the waveguide collimation elements altogether and use grating couplers to incouple an optical beam formed by external optics, and thereby reduce the space requirements. A third alternative approach is to scale down the integrated lenses (aperture size and focal length) so that they will fit on a single substrate and thus give up some of the range swath of the processor. A fourth alternative approach is to design the waveguide lenses with shorter focal lengths and retain the aperture size, as we have done with the flat-field lens doublet design presented in Chapter 5. The lens doublet has a combined 4.2 cm focal length and operates at f/1.56. If the beam forming lenses are designed with the same focal length, it is feasible that the entire processor could be fabricated on a 4” wafer without beam folding techniques or compromise of the processing capacity. We have successfully fabricated and characterized SAW modulators suitable for 330 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. the IOSAR application with previously established methods [Tsai, 1990]. We have shown that these modulators operate properly when integrated with other waveguide components. Although we demonstrated fairly high diffraction efficiencies, to obtain a linear diffraction efficiency with drive power, SAW modulators are not operated at full efficiency. Consequently, they will impart an optical power loss of approximately 3dB . Uniform, high density rib waveguide arrays (8 fim rib waveguide widths, 2 /xm gaps widths, 1 /xm gap depths, 1 cm long, and 660 elements) were previously demon strated by Rastani [Rastani, 1988] on a multi-mode Ti:LiNbC>3 waveguide. We have successfully fabricated similar structures with different gap widths, gap depths, rib waveguide lengths, and array sizes. The ability of these structures to accurately and efficiently dissect and disseminate the focused range image to the external mask and CCD arrays is subject to an inherent trade-off in the structure design. We have found that in order to accurately and efficiently dissect the incident field distribution, the rib waveguide gaps must be narrow and deep. At the same time, we have found that a deep gap region is associated with higher propagation losses that will lead to non-uni form outcoupling along the rib waveguide length. Both the dissection and propaga tion requirements can be satisfied if the rib waveguide sidewall roughness, responsible for the propagation losses, can be minimized in the fabrication process. At present, there are no established technique available for LiN b03 with which this can be achieved. Perhaps the most likely approach would be to perform an isotropic plasma etch after the ion-beam milling process to smooth out the sidewall roughness transferred from the mask. In the absence of an established technique, the design o f the rib waveguide array should consist of a compromise between performance factors. We learned in our inci- dent-field coupling analysis earlier in this chapter that the range-focused image formed by the flat-field lens doublet will result in a side-lobe crosstalk level of -17 dB. Therefore, we need not over-design the rib waveguide array structure to pro- 331 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. vide rib-to-rib coupling much lower than this value. We found that the crosstalk level between 8 fJin wide rib waveguides with 2 fim gaps etched 5000 A deep over a 1 cm propagation length was approximately -16 dB. In addition, the propagation loss of this structure was only 1.6 dB/cm. To first order, this may be an acceptable design since this low attenuation constant will permit fairly uniform outcoupling over a 1-cm rib waveguide length. We consider this design below in our assessment of the IOSAR processor feasibility since we have experimental data to describe the crosstalk and propagation losses. The selection of this design is made in spite of theoretical evi dence that the crosstalk values may be higher for the higher order modes in the rib waveguides than for the lower order modes. To compensate for this trend, a rib waveguide array design may be considered with slightly deeper or wider gap regions. If this adjustment is made, the present assessment of the IOSAR processor should still hold. We have successfully demonstrated nearly uniform grating outcoupling from both planar waveguides and rib waveguides over a large propagation distance. This was accomplished for both 2 fim and 4 /tm period gratings with a very small grating height (less than 250 A). We have theoretically and experimentally analyzed the out coupling uniformity and determined that there will be insignificant fluctuations in the illumination of the surface mounted devices due to coherent interference effects as long as the grating period is small compared with the pixel size. Grating theory indi cates that the grating outcoupling is sensitive to slight fluctuations in the grating height for modulation depths smaller than the saturation depth. We observed slight random fluctuations in the outcoupled light from the gratings, although a direct corre lation was not made with grating height nonuniformity. We did not observe any non- uniformities that would hinder the performance of the IOSAR processor. The primary nonuniformity is due to the propagation losses of the rib waveguide array, which causes a 30% decrease in outcoupled power over a 1 cm propagation distance. From calculations of the grating outcoupling efficiencies, it was determined that 332 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. approximately 10% of the total outcoupled light is coupled into the cladding (air) region and the rest is coupled into the substrate. The light directed into the substrate is a large source of loss in the IOSAR processor. The substrate light can also contrib ute to background noise at the CCD array unless the reflections of the substrate modes from the substrate lower surface are eliminated. Due to the narrow width of the rib waveguides, the surface mounted devices must be placed in very close proximity (50 fim) of the rib waveguide arrays. Two possible approaches to mount both the Doppler phase history mask and CCD detector array in close proximity of the surface were suggested in Chapter 4. However, this is a complex issue and requires further investigation. We have successfully demonstrated large aperture, short focal length (f/2) embedded waveguide lenses. The lens insertion loss of 3 dB is comparable with other waveguide lenses. The measured focal spot size of the lens is 2 ^m (0.9 /zm inside the waveguide). We have determined that the waveguide anisotropies present in the Ti:LiNbC> 3 waveguide do not pose a problem for the lens performance in the IOSAR processor as long as the anisotropy is taken into consideration in the initial lens design. Although a lens with a 2.7 cm aperture has not yet been fabricated, the pro cesses developed for the 1-cm aperture lens should be effective in scaling to larger sizes. The greatest concern in the fabrication of such a large lens is the etch depth uniformity, since the region of uniformity (±5%) is only 3 cm in diameter in our ion- beam etching system. However, other systems are available with larger ion-beam guns that provide uniformity over 10 cm diameter regions or larger. The embedded lens was successfully integrated and tested with other waveguide components, as described earlier in this chapter. In the test module experiments, the lens was used to focus light into a single rib waveguide as well as to selectively focus light into individual rib waveguides with the aid of a SAW modulator. To accomplish this function, an external lens was required to compensate for error in the focal length of the integrated lens. In a fully integrated IOSAR processor, this type of compensa- 333 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. tion is not possible. Therefore, in order for the fabrication of a functional IOSAR pro cessor to be feasible, the embedded lens fabrication process must be refined until the refractive indices of the constituent materials can be accurately controlled. We identi fied the refractive index variation of the electron-beam deposited Si0 2 waveguide layer as the primary contributor to the lens focal length variations. There are other processes such as sputter deposition and ion-assisted electron-beam deposition that can be used to stabilize the S i0 2 film refractive index for more repeatable results. With proper precautions taken and with a refined process, the embedded lens struc tures with repeatable properties should be feasible. In the test module experiment aimed at excitation of a single rib waveguide with the embedded lens, a substantial amount of light was observed in rib waveguides that were approximately 6 to 10 rib waveguide widths away (60 to 100 fim) on either side of the excited rib with a peak intensity about -6 dB down. This light is most likely due to scattering either in the embedded lens thin film waveguide or at the embedded lens interfaces. This level of scattering is a cause for concern in the IOSAR processor since it is much higher than for other forms of noise or crosstalk that we have consid ered. The use of sputtered thin films in the fabrication of the embedded lens may help to reduce this scattered light, since sputtered films have lower in-plane scattering lev els (as a result of fewer voids) than electron-beam evaporated films. If the scattering present in the side rib waveguides originates from the embedded lens interfaces, then it may be difficult to find an etching process that produces smoother lens sidewalls without a compromise in the sidewall tilt angle. For the IOSAR processor example given in Chapter 2, the total processor length is approximately 19.5 cm including the 1-cm rib waveguide length. The associated waveguide propagation losses are 7.8 dB (18.5 cm X 0.42 dB/cm) for the planar waveguide regions and 1.6 dB (1 cm x 1.6 dB/cm) for the rib waveguide region. For a laser diode coupled to the planar waveguide, the associated loss is on the order of 1 dB under the assumption a that GRIN lens is used to provide the necessary mode 334 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. matching to the polished waveguide edge. A high throughput lens with a small refractive index change (thin film overlay lens or TIPE lens) can be used as the beam expanding lens since the required lens f-number is not very low. In this case, the insertion loss of this lens is approximately 1 dB. If embedded lenses are used both for collimation and imaging, the associated loss will be about 6 dB (3 dB per lens). As mentioned earlier, the SAW modulator will incur a loss of approximately 3 dB due to light lost into the undiffracted order. The coupling efficiency into the rib waveguide array is roughly proportional to the rib fill factor (0.8 for the rib waveguide design discussed above). Hence, there will be a 1 dB loss associated with the incident-field coupling. Finally, it is assumed that all of the light is coupled out of the rib waveguides by a grating with a varied depth so that the loss associated with the grat ings is only the 10 dB associated with the power division between the cladding and substrate radiation modes. The total system loss is therefore expected to be 31.4 dB for the IOSAR processor example given in Chapter 2. If a low f-number beam expan sion lens is used in conjunction with the flat-field lens doublet design in the IOSAR processor such that all of the elements will fit on a single 4” wafer, then the total pro cessor losses will be reduced by approximately 1 dB (minus 4 dB for the shorter prop agation distance, plus 3 dB for the extra lens element in the doublet). The CCD array detects and integrates the output of the planar portion of the pro cessor over several radar cycles. Commercially available CCD arrays have a dynamic range of approximately 72 dB and megapixel array sizes. For example, the Kodak KAF-1600 series image sensor operates in full-frame mode and has an array size of 1552 X 1032 pixels with a 9 fm l x 9 /mi pixel size, an 85,000 electron well size, a 100% fill factor, 72 dB of dynamic range, and a 32% quantum efficiency at 650 nm. Similar CCD arrays can operate in the time-delay and integrate (TDI) mode rather than the full-frame mode, as is necessary in the IOSAR processor. In order for the entire dynamic range o f the CCD array to be utilized, the laser diode must supply enough power to overcome the system losses and yet have enough power left to satu- 335 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. rate the CCD array. The amount of power required by the laser to fully saturate the CCD array can be calculated as follows. For the CCD array example given above, only 260 of the 1552 array columns will be aligned above rib waveguides. The total number of electrons in this 293 x 1032 pixel region is 2.57 X 101 0 at saturation. The optical energy required per electron is 9.8 X 10-19 J, in which the quantum efficiency has been taken into account. Therefore, the total required energy incident on the CCD array should be about 2.5 x I O '8 J. Since there are approximately 1000 elements along each rib waveguide, the saturation energy per integration step is 2.5 x 10_11 J. In Chapter 2 we estimated that the maximum limit in the laser pulse duration is 20 ns if range smearing is to be avoided. In consideration of the energy required to saturate the CCD array, the optical power supplied to the CCD array throughout this time duration should be 1.3 mW. For this to be possible with the processor throughput losses (31.4dB), the peak power requirement of the pulsed laser diode is approxi mately 1.8 W. The derating factor for pulsed mode operation is calculated from the ratio of the laser pulse duration (20 ns) to the pulse repetition period (1 ms) that yields 2 x 10'5. The average power requirement for this processor then is 36 /rW. This power level can be attained with currently available visible laser diodes. The power requirements will be slightly lower for the processor design that utilizes the flat-field lens doublet design because of the reduced processor throughput loss in that case. Therefore, on a single lithium niobate substrate, the IOSAR design discussed in Chapter 2 with a 50 MHz bandwidth SAW transducer, multiple integrated waveguide lens components for forming and imaging a 2.7 cm wide beam, a densely-packed rib waveguide array, low efficiency uniform outcoupling gratings, a surface-mounted mask, and a CCD detector array should be feasible. This module is capable of pro cessing 264 range resolution elements in a radar return signal that covers a 1 km ground swath and an unlimited number of azimuth resolution elements for continuous operation using a 650 nm laser diode with a low average power requirement of just 36 ^tW. 336 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. 7.8 IO Correlator Assessment The component requirements for the IO correlator are somewhat relaxed in comparison to those for the IOSAR processor. The Ti:LiNb03 waveguides and SAW transducers described earlier are fully suitable for this application. In addition, the processor can be fabricated on presently available 4” diameter y-cut LiN b03 wafers as long as the lenses perform slightly under f/2. The rib waveguides required for the IO correlator are low-density in comparison to those required for the IOSAR processor. In addition, uniform outcoupling gratings are not required over a long distance (less than 1 mm) in this case. Even with shallow or narrow rib waveguide gaps, the rib-to-rib coupling will be small over this short dis tance. Because these requirements are not very severe, the present state of develop ment of the rib waveguide and grating structure is sufficient for the IO correlator. Others have demonstrated efficient grating outcoupling over very short dis tances for efficient waveguide-to-detector coupling. For 2 firn period gratings, Rast ani calculated that approximately 64% of the light can be outcoupled over a 1 mm propagation distance if the grating height is equal to the saturation grating height [Rastani, 1988]. However, linear detector arrays are not available with such a large detection region per pixel. To make maximum use of the outcoupled light, a grating frequency chirp can be used to focus the light onto a linear detector array. In order for the light to come to focus, the detector should be placed a small distance away from the waveguide surface above the outcoupling gratings. Since the transverse spread of the light outcoupled from the rib waveguides is small in this processor, the detector may be placed as far as 600 /im above from the waveguide surface. The grating cou plers in the IO correlator suffer the same drawback as those in the IOSAR processor in that only 10% of the outcoupled light is directed into the cladding region. How ever, since the transverse spread of the outcoupled light is so large, it is possible that the array can be mounted on the backside of the substrate in order to capture the 337 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. higher-power substrate radiation modes. In order to realize the IO correlator example given in Chapter 2, waveguide lenses with a 1 cm aperture and f/1.6 or better performance are required if all of the components are to fit onto a 4” substrate. With some slight improvements in the embedded lens structure, this performance should be possible. Alternatively, the aperture size of the lenses may be designed slightly smaller (0.8 cm) at the expense of processor resolution, so that the f/2 embedded lenses fabricated in the present work can be used to assemble the correlator onto a single 4” substrate. Since there are at least 3 embedded lenses in the correlator, a number of diffi culties arise. First, the 3 dB insertion loss per lens can become quite costly in terms of the total system throughput. Second, the tolerance required to align 3 lens ele ments to one another as well to the rib waveguide array can pose difficulties if there are significant focal length variations. Third, the presence of scattering discovered in the test module experiments may be compounded in a processor with three lenses and lead to severe degradation of the final output image. In the same manner as for the IOSAR processor, the total system losses were calculated for the IO correlator. The total processor length in this case is approxi mately 10 cm assuming f/1.6 lenses are used after the SAW transducer, a spacing of 0.25 cm of the SAW path from the collimating lens, and a 0.25 cm long rib waveguide array used to dissect the in-plane image before outcoupling. The contributions to the waveguide losses are: waveguide propagation distance (4 dB), laser/waveguide cou pling (1 dB), beam expanding lens (1 dB), three embedded lens elements (9dB ), SAW diffraction (3 dB), rib waveguide coupling (0.3 dB), rib waveguide propagation (0.4 dB), grating coupling power division (10 dB), and grating efficiency (2 dB). The total losses for the IO correlator are thus 30.7 dB. A commercially available linear detector array such as the Kodak KLI-2103 Image Sensor has an array size o f 2098 pixels with a 14 fim x 14 fim pixel size, a 300,000 electron well size, 72 dB of dynamic range, and a sensitivity o f 4.27 x 10'1 9 J/electron at 650 nm. As a result of 338 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. the detector pitch, there will be approximately 2 pixels per rib waveguide across the 333 rib waveguides. The optical energy required to saturate this array is 2.56 X 10'1 3 J/rib waveguide or 8.5 X 10'1 1 J for the array. In consideration of the processor throughput losses, the input energy required to saturate the detector array is 1 x IO'7 J. For a 1 ms integration period, the required input optical power is 0.1 mW. This optical power can be easily satisfied with presently available laser diodes. For this integration time, the time-bandwidth product of the processor would be approxi mately 6 X 105, which is comparable with the integration time and results obtained by Liao [Liao, et a i, 1982]. From the preceding analysis, the complex correlation of two 15 MHz input sig nals should be feasible using the IO correlator design discussed in Chapter 2. This correlator, completely integrated onto a Ti:LiN b03 waveguide, would consist of a 60 MHz bandwidth transducer, multiple integrated waveguide lenses for forming and imaging a 1 cm wide beam, a 1 cm wide rib waveguide array with 333 elements, highly efficient surface outcoupling gratings for detector coupling, and a surface- mounted high dynamic range linear detector array. The source laser diode, with an output wavelength o f 650 nm, would need an output power of only 0.1 mW to enable this processor to perform with a time-bandwidth product of 6 x IO5. Although the IOSAR processor and IO correlator examples discussed above can be accomplished using available lasers, there are other considerations that should be taken into account to evaluate the total processor performance. One such consider ation is the effect of light lost by scattering mechanisms. The light that is scattered from the waveguide goes primarily into the substrate and can work its way onto the detector array. Unless care is taken to prevent this background light from reaching the detector array, the signal-to-noise ratio of the processor will be degraded. The two primary sources of scattering loss in the waveguides are the outcoupling gratings and the waveguide embedded lenses. The situation with the gratings can be improved if the detector array can be mounted onto the substrate side of the sample rather than the 339 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. waveguide side. Backside mounting may be acceptable in the case of the correlator where there is a large collimation length for the outcoupled beam (that is even longer in the higher index substrate material), but is not practical in the case of the IOSAR processor. Alternatively, techniques such as blazing o f the grating profile can be used to suppress coupling into the substrate modes. Improvements in the embedded lens will greatly improve the feasibility of these processors. In particular, increased lens throughput and decreased scattering levels may be critical to the performance of fully integrated processors. The embedded lens fabrication processes must be made stable enough to produce lenses with repeatable focal properties. These are the primary improvements in the components that should be made in order for the IOSAR proces sor and IO correlator to be feasible. 7.9 References D. B. Anderson, “Integrated Optical Spectrum Analyzer: An Imminent ‘Chip’,” IEEE Spectrum, Dec., 22-29, (1978). M. K. Bamoski, B. U. Chen, T. R. Joseph, Y. M. Lee, and O. G. Ramar, “Integrated- Optic Spectrum Analyzer,” IEEE Trans. Circuits Syst., Cas-26(12), 1113, (1979). K. Y. Liao, C. C. Lee, and C. S. Tsai, ‘Time-Integrating Correlator Using Guided- Wave Anisotropic Acousto-Optic Bragg Diffraction and Hybrid Integration,” in the 1982 Topical Meeting on Integrated and Guided-Wave Optics, Technical Digest WA4-1 to 4, IEEE Cat. No. 82CH 1719-1724, (Pacific Grove, Ca., 1982). D. Mergerian, E. C. Malarkey, and R. P. Pautienus, “High Dynamic Range Integrated Optical RF Spectrum Analyzer,” in Proc. 4th International Conference on Integrated Optics and Optical Fiber Communication, Paper 30B3-b, (Tokyo, Japan, 1983). 340 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. D. Mergerian, E. C. Malarkey, R. P. Pautienus, J. C. Bradley, G. E. Marx, L. D. Hutcheson, and A. L. Kellner, “Operational Integrated Optical R. F. Spectrum Analyzer,” Appl. Opt., 19(18), 3033-3034, (1980). H. Nishihara, M. Haruna, and T. Suhara, Optical Integrated Circuits, (McGraw-Hill Book Company, New York, 1989). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). C. S. Tsai, Ed., Guided-Wave Acousto-Optics, Springer Series in Electronics and Photonics, D. H. Auston, etal., Eds., Vol. 23, (Springer-Verlag, New York, 1990). S. Valette, J. Lizet, P. Mottier, J. P. Jadot, S. Renard, A. Fournier, A. M. Grouillet, P. Gidon, and H. Denis, “Integrated Optical Spectrum Analyser Using Planar Technology on Oxidised Silicon Substrate,” Electron. Lett., 19(21), 883-885, (1983). 341 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Chapter 8 Conclusions 8.1 Summary Several different aspects of integrated optical technology and hybrid bulk com ponent integration have been considered herein in our pursuit to establish a flexible approach to the realization of advanced integrated optical signal processing tasks. Foremost among these considerations is the use of rib waveguide array components with surface outcoupling gratings to expand information processed in the plane of the waveguide and to couple this information upward to surface mounted devices for fur ther computations to be performed. This waveguide component is well suited for some integrated optical signal processing tasks, such as SAR image formation, for which a two dimensional signal output is required. For other tasks, such as integrated optical correlation, rib waveguide arrays with surface outcoupling grating structures provide a flexible and manufacturable means to couple in-plane guided-wave infor mation to an external detector array. Rib waveguide arrays designed for use in the IOSAR processor (8 fJ.m wide rib waveguides, 2 /an wide gaps, 5000 A gap depths, 5 mm long elements, and 1000 ele ments) and in the IO correlator (28 /an wide rib waveguides, 2 /an wide gaps, 5000 A gap depths, 5 mm long elements, and 333 elements) were fabricated on single mode Ti:LiNb03 waveguides by ion-beam milling. For each application, the rib waveguide widths were chosen to correspond to the number of processor resolution elements, the total width of the in-plane image, and the spacing of the CCD detector elements. The fabrication processes used were first established by Rastani [Rastani, 1988], and have 342 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. been modified in the present work to fabricate devices with greater uniformity over larger areas. Possible sources of crosstalk present in rib waveguide arrays have been identi fied and investigated. The side-lobe overlap of a range-focused spot with the neigh boring rib waveguides was calculated theoretically for a spot size (FWHM) that was roughly half of a rib waveguide width. The level of light present in the neighboring ribs was found to be approximately -1 7 dB. In order to determine the crosstalk between rib waveguides due to rib-to-rib coupling, the rib waveguide array designed for use in the IOSAR processor was examined as well as variations of this structure with different gap widths, gap depths, and rib waveguide lengths. Wider and deeper gaps were found to result in less cross-coupling between neighboring ribs. Rib waveguides fabricated in Ti:LiNb0 3 with a gap depth of 5000 A, gap width of 2 /an, and length of 5 mm were shown to lead to crosstalk o f-1 6 dB. This isolation is com parable with the crosstalk due to side-lobe overlap of a range-focused spot with the neighboring rib waveguides. We concluded that gap depths greater than 0.5 /an will lead to lower rib-to-rib coupling, but will also lead to higher rib waveguide scattering losses. The scattering that arises from rib waveguides with such deeply etched gaps will greatly decrease the outcoupling uniformity along the rib waveguide length as demonstrated herein. We also postulated that the scattering from the rib waveguide side walls will lead to increased crosstalk at the external detector array. Large area surface outcoupling gratings (1.2 cm x 1 cm) with a 4 /an period and 150 A etch depth were fabricated on planar Ti:LiNb03 by argon ion beam milling, and nearly uniform outcoupling over a 1 cm length was demonstrated. By the same process, several large area-surface outcoupling gratings with either 2 /an or 4 fan periods and a uniform depth anywhere between 150 and 250 A were fabricated on rib waveguide arrays. For grating depths of 150 A and 250 A, it was calculated that 10% and 50% of the light in the rib waveguides could be outcoupled over a length of 1 cm, respectively. The total outcoupling efficiency was shown to be much more sensitive 343 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. to the grating etch depth than to the grating aspect ratio for the case of 2 fim period gratings. For a particular IO signal processing application, we found that grating peri ods smaller than the expected detector pixel detector size are preferable to reduce variation in the outcoupled light from pixel to pixel. We have evaluated such integrated waveguide components fabricated in waveguide materials other than LiNb0 3 . The processing sequences for rib waveguides and gratings were also applied to GaAs/AlGaAs waveguides. As was described herein, we have integrated a rib waveguide array (8 fim wide rib waveguides, 2 fim wide gaps, 5000 A gap depths, 660 elements, 1 cm length) on a 1 fim thick GaAs waveguide with surface outcoupling gratings (2 fim period, 1 mm long, 1000 A deep) over half of the rib waveguides. The cladding radiation modes of the gratings in both rib waveguide and planar waveguide regions were observed and the outcoupling efficiencies were measured and compared with theory. Good agree ment between measurement and theory indicates that the fabricated structures were extremely uniform and that the pattern transfer was accurate. In addition, the similar ity between the efficiencies measured for the gratings on planar waveguide and rib waveguide regions indicates that transverse confinement in the rib waveguides does not adversely affect the grating outcoupling properties. The requirements for surface mounted devices such as amplitude masks and 2-D CCD arrays were not considered in depth herein. However, their required prox imity was determined to first order based upon the transverse spreading (due to dif fraction) of the light outcoupled from the rib waveguides. In the case of the IOSAR processor, it is necessary to mount the surface devices in very close proximity (50 fim) to the rib waveguide arrays because of the small rib waveguide widths (8 /an). Possible methods to include the azimuth mask in this small spacing were dis cussed, including a thin film deposited mask on the rib waveguide or on the CCD array, and the use of a cylindrical micro-lens array to alleviate the spacing require ment. For the correlator application, the rib waveguides are wider (28 fim) than in the 344 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. case o f the IOSAR processor (8 fim) and hence a spacing up to 600 fim can be toler ated. Another critical waveguide component required for the fabrication of advanced IO signal processors is a large aperture, short focal length lens. Lenses with short focal lengths are required so that more components may be integrated onto a single waveguide substrate. Large lens apertures permit a greater number of resolvable spots in the signal processing operation. We addressed this need by the development o f large 1-cm aperture embedded waveguide lenses with a 1 cm focal length in single mode Ti:LiNb03 waveguides. The lens structure consists of a 3 fim deep ion-beam- milled recess with a back-deposited S i0 2/MgF2 waveguide. The theoretically calcu lated throughput of this structure is in excess of 90%. The highest measured transmit tance at the center of the lens was approximately 70%. From measurements of the lens transmittance across the lens aperture, it was calculated that with the center 5 mm o f the lens illuminated the lens throughput is slightly under 50%. The cause of the lower throughput compared with that theoretically calculated is believed to be related to a slight residual recess sidewall tilt (12°) and to the high propagation losses in the electron-beam-deposited S i0 2 layer. These throughput losses become severe for light incident on the extreme ends of the lens aperture. The focal spot size of an embedded lens was measured external to the waveguide, and was found to be 2 fixn FWHM with a side-lobe level o f -11 dB. This measurement corresponds to a focal spot width of 0.9 fim inside the waveguide. The lens performance is nearly diffraction-limited and more than sufficient for focusing efficiently into individual rib waveguides. Mode coupling calculations for this focal spot width indicated a peak incoupling efficiency o f 96% for coupling into a 6 fim wide rib waveguide. In order to take full advantage of the high index changes available with embed ded lenses (0.7 in the case of S i0 2/MgF2 in Ti:LiNb03), aberration-corrected lenses were designed. The 1-cm aperture lens design used in the lens fabrication and testing described above was aberration-corrected and consists of acircular-shaped interfaces. 345 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. Likewise, a lens doublet with flat-field performance was designed for use in the IOSAR processor and consists of a 2.7 cm aperture and a 4.2 cm focal length (f/1.56). The effects of process variations in the fabrication of the embedded lens struc tures on lens performance were investigated. The lens focal length was found to be sensitive to deviations in the embedded lens waveguide refractive index from the design value. This result indicates the necessity for tight process control during embedded-lens thin film deposition. In addition, it was determined that the effects of waveguide anisotropy, which were not included in the lens design, would have only a slight effect in our fabricated lens structures. These anisotropies can be accounted for in the initial lens design so that they are reduced to reasonable levels for use in IO sig nal processors. In order to determine the potential of these waveguide components for integra tion into a complete integrated optical processor, two test modules were fabricated. These modules consisted of a 1-cm aperture embedded lens and a rib waveguide array (6 fim wide rib waveguides, 4 fim wide gaps, a 5000 A gap depths, 5 mm length, 1000 elements) on a single mode Ti:LiNbC> 3 waveguide with a polished end-facet for observation of the rib waveguide end-emission. One of the modules also included SAW transducers (425 MHz center frequency, 100 MHz bandwidth) for beam steer ing. Several processing runs were performed before the final test modules were fabri cated to determine the processing sequence best suited for device compatibility. The ability of the embedded lens to focus light into a single rib waveguide and the ability of the embedded lens in combination with the SAW transducer to selec tively focus light into different rib waveguides were investigated with these modules. With a slightly diverging beam prism-coupled into the test module without the SAW transducers, a single rib waveguide within the 1000 element array could be excited. Light was observed in nearby rib waveguides approximately 6 to 10 rib waveguide widths away (60 to 100 fim ) on either side of the excited rib. The peak intensity of this light was about -6 dB below the intensity in the excited rib. The presence of this 346 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. much light in these other rib waveguides is possibly due to scattering in the waveguide or integrated lens structure. The light in the nearest rib waveguides was at the -10 dB level. With a slightly diverging beam prism-coupled into the test module with the 100 MHz bandwidth SAW transducer, selective focusing was demonstrated across 6 rib waveguides by adjustment of the transducer drive frequency. Light scat tered into nearby rib waveguides was apparent in this experiment as well. The component developments and analysis presented herein show promise and suggest the feasibility o f the proposed advanced IO signal processors. The rib waveguide and grating fabrication uniformity is excellent, and the means for control ling crosstalk and uniformity are now known. The embedded lens structure can be completely fabricated using standard photolithographic, ion-beam milling, and thin film deposition techniques. For f/2 operation, the lens throughput is reasonable (-50% ) and the achieved focal spot size is much smaller than the required rib waveguide widths. The performance demonstrated with this embedded lens structure exceeds that of any other waveguide lens in LiNb03 that uses planar fabrication tech niques. We have shown that these devices can be integrated onto the same substrate with SAW modulators, and their combined operation has also been demonstrated. In addition, rib waveguide arrays with surface outcoupling gratings have been fabricated in GaAs waveguides. Embedded lenses in GaAs waveguides have been previously fabricated by others [Minot and Lee, 1990], which makes this also a potentially useful waveguide material for advanced IO signal processors. 8.2 Future Research Directions To improve upon the feasibility of the advanced integrated optical signal proces sors discussed herein, several issues regarding the requisite waveguide components should be pursued further. With these improvements, IOSAR processors and IO corr elators with performance very close to that specified in the examples o f Chapter 2 347 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. may be realized. Further efforts to fully realize these IO signal processors is a worth while pursuit, since they fill a need in certain application areas that cannot be appro priately addressed by electronic or bulk optical approaches. One of the primary issues to consider pertains to the rib waveguide array with surface outcoupling gratings. Although this structure adds flexibility in the processor design as well as an extra processing dimension, the drawback at present is that a rel atively small portion o f the grating outcoupled light is actually directed towards the surface elements compared with the light directed into the substrate. For example, in Chapter 4 we found that for a 2 /an period grating only 10% of the total outcoupled light will directly impinge upon the surface mounted device. The light directed into the substrate can potentially be reflected or scattered back towards the detector array effectively and increase the amount of background noise. It was shown in Chapter 4 that, the power division can be improved to -4 dB for a grating period of 1.54 /im. The power division can be improved to -3 dB (i.e., a 50/50 split of the power into the cladding and substrate regions) for submicron grating periods. To enhance the cou pling into the cladding beyond this level, blazed grating techniques should be consid ered. The second issue of concern pertains to the observations made with the test modules. The high levels of light observed in the test module rib waveguide array 60 to 100 /an to either side of the excited rib waveguide can be deleterious to the signal processing function in a fully integrated processor. The cause of this scattered light should be identified and eliminated if advanced integrated optical signal processors are to operate with high signal-to-noise ratios. A likely cause of this scattering is non- uniform embedded lens interfaces. From the measured sidewall roughness (0.1 /an peak-to-peak), the level of scattered light was estimated to be about -15 dB, which is much lower than what was observed. However, some contribution to this scattering may be due to index inhomogeneities arising from stresses in the thin films near the embedded lens interface. The use of different thin film deposition techniques or dif- 348 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ferent embedded lens recess etching techniques should be considered and investigated if the lens is identified as the cause of the scattering. The third issue pertains to the overall low throughput o f the lens compared to its potential throughput. The observed throughput loss may be due to the recess side wall tilt, the lossy embedded waveguide films, or the mode coupling losses that arise from a relative height mismatch in the waveguide structures. The primary attribute of the fabricated embedded lens structure that complicates the theoretical evaluation of the lens throughput is the sidewall tilt. If methods are established to improve on this feature, comparisons between the theoretical models and the actual performance will become much more meaningful and the throughput should be improved substantially. Although a modest level of device integration has been successfully demon strated, integration issues should be further explored. Test modules should be fabri cated that include surface outcoupling gratings in addition to the other waveguide components. In more advanced stages of device integration, detector arrays should be aligned above the rib waveguide arrays. In this way, the issues involved in the inte gration of external devices, such as proximity coupling and device registration with the rib waveguide array, can be explored. The issues surrounding laser-to-waveguide coupling and subsequent beam expansion should be investigated before fully fabricated advanced integrated optical processors are realized. At this level of integration, a number of alignment and pro cess variation issues will need to be considered. In Chapter 1 (the introduction), our stated research objective was to develop a general technological approach for the implementation o f signal processing tasks. It should be emphasized that the technological approach that we have chosen is not lim ited in its applicability to signal processing tasks alone. In particular, with the suc cessful development of integrated optical wide-aperture, short focal length lenses, large area rib waveguide arrays with surface outcoupling gratings, and external ele- 349 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ment hybridization, other highly active areas of research such as optical communica tions and optical interconnection will benefit. For example, currently available methods for fiber-optic wavelength division multiplexing (WDM) and demultiplexing consist of compact integrated optical mod ules that employ edge-mounted reflection gratings to obtain the necessary dispersion characteristics along with input and output fibers [Yen, et a l, 1981; Suhara, et al., 1982]. Conventional integrated optical lenses are not used because the typically small refractive index differences lead to very weak wavelength dispersion. How ever, the use of highly dispersive materials for embedded refractive elements such as waveguide prisms with large index differences may be used to spread the different wavelength components of an input signal in angle. An embedded lens or lens array could then be used to capture this light and direct it into an array of rib waveguides to dissect the wavelength-divided information. This type of WDM device could be fab ricated completely by planar techniques and feature a large number of channels with very low crosstalk. Additionally, weakly-coupled rib waveguide surface outcoupling gratings could be used with a surface-mounted linear array to sample the information in individual channels. For optical chip-to-chip interconnects, a channel or rib waveguide array with surface outcoupling gratings could be incorporated in a multi-layer stack and used as an optical power bus to illuminate devices on individual layers of the stack. With the appropriate outcoupling grating design, the light could be outcoupled vertically and hence the power bus would operate as a compact beam splitter. For example, if the illuminated devices in the layer above the optical power bus are spatial light modula tors that operate in reflection, then the reflected optical signal would pass back through the waveguide and illuminate the underlying layer. This arrangement could effectively serve as an optical chip-to-chip interconnect. Other host waveguide materials such as S i02/silicon or polymer/glass are potentially suitable for a broad array of applications. There are, for example, well 350 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without permission. known reactive etch techniques for Si02 that result in well-defined vertical sidewalls suitable for the embedded lens structure. However, the design of these structures and the fabrication sequence would need to be altered to provide for a higher refractive index embedded material such as T i02 (n = 2.2 to 2.4). The development o f the individual components discussed in this thesis can also be taken to many different levels. As previously mentioned, a large aperture flat-field lens doublet has been designed and will eventually be fabricated once the fabrication process control for single lens elements has been sufficiently developed. In general, imaging systems with several interfaces have more degrees of freedom available for the correction of aberrations; therefore, it is extremely desirable to make the design and fabrication of multi-element embedded lens sets a commonplace practice. Previ ously, Rastani demonstrated the operation of a TIPE lens triplet [Rastani and Tanguay, (to be published)]. Others have recently considered four-element lens designs with acircular interfaces accounting for waveguide anisotropies under the assumption of refractive index differences of 0.1 or less [Svarvas, et al., 1993]. For the larger refrac tive index differences available with embedded lenses, it is expected that the design flexibility will be increased. As a result of the large variety of thin film materials available for use in the embedded lens and the ease of designing arbitrarily curved interfaces, refractive ele ments not even conceived o f in bulk form may be implemented with standard planar fabrication technology in waveguides. These possibilities may lead to designs that relax the fabrication tolerances needed for lenses to have aberration-free performance. In addition, lenses could be designed to be athermal or achromatic, or they could pur posely be designed to be sensitive to these variations so that active feedback to the laser source or processor cooling system can stabilize the processor’s operation. The rib waveguide array with surface outcoupling gratings is also a very versa tile device. The array design may be modified to accommodate a fan-in or fan-out of the channels for appropriate scale matching with other devices, or the entrance aper- 351 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. ture of the array may be modified to match the curved focal plane of an uncorrected lens. The gratings may be made either highly efficient or weakly coupling; they may be modulated in grating depth to provide uniform outcoupling and modulated in period to provide external focusing of the radiated modes; or the grating profile may be blazed to direct the light into a particular radiation mode. A number of different devices can be used external to the rib waveguide array to modify or control the outcoupled light. For instance, a microlens array with cylindri cal diffractive or refractive focusing elements may be aligned above the rib waveguide array to collimate the light coupled out from each rib, and thus prevent crosstalk that results from the transverse spread of the light. Methods to filter unwanted radiation modes that travel at steep angles may be realized with an angu- larly-sensitive thin film structure such as a stacked AR coating. The externally- hybridized device could be a high dynamic range photodetector, CCD array, or a spa tial light modulator. A host of other passive or active devices could be considered such as diffractive elements for coupling into bulk optical systems, high speed detec tors, holographic storage media, fiber bundles, and so forth. As a final word, there are a number of well-established and useful waveguide devices that already exist that have not been specifically addressed herein, such as the electrooptic modulator. Alternative architectures designed to make use of these devices have different and possibly advantageous device characteristics. For exam ple, in the case of the electrooptic modulator, modulation bandwidths rely on elec trode finger width fabrication limitations and addressing schemes, and likewise the time constants are determined by the design and bounded by the limitations of capac itive charging and discharging. This device could be used in place o f the acoustooptic modulator to overcome the small SAW time aperture in lithium niobate that results from the high surface acoustic velocity of 3488 m/s. 352 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8.3 References M. M. Minot and C. C. Lee, “A New Guided-Wave Lens Structure,” J. Lightwave Technol., 8(12), 1856-1865,(1990). K. Rastani, “Advanced Integrated Optical Signal Processing Components,” Ph.D. Thesis, University of Southern California, (1988). K. Rastani and A. R. Tanguay, Jr., “Large Aperture Negative Meniscus Singlet and Triplet Lenses with Positive Focal Lengths Developed on L iN b03,” (to be published). T. Suhara, J. Viljanen, and M. Leppihalme, “Integrated-Optic Wavelength Multi- and Demultiplexers Using a Chirped Grating and an Ion-Exchanged Waveguide,” Appl. Opt., 21(12), 2159, (1982). G. Svarvas, M. Barabas, R. Richter, and L. Jakab, “Design o f Multielement Acircular Waveguide Lens Systems in Anisotropic Media,” Opt. Eng., 32(10), 2510-2516, (1993). H. W. Yen, H. R. Friedrich, R. J. Morrison, and G. L. Tangonan, “Planar Rowland Spectrometer for Fiber-Optic Wavelength De-Multiplexing,” Opt. Lett., 6(12), 639, (1981). 353 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Adaptive stochastic resonance.

PDF

Design, fabrication, and integration of a 3-D hybrid electronic/photonic smart camera

PDF

Adaptive signal processing with the gradually increasing filter length (GRINFIL) method

PDF

A 1.2 V micropower CMOS active pixel sensor

PDF

A monolithic multichannel receiver array for gigabit/s parallel synchronous optical digital data links

PDF

Adaptive soft-input soft-output algorithms for iterative detection

PDF

Blind equalization techniques using maximum likelihood criterion

PDF

3-designs from the Goethals codes over Z(4) via a new Kloosterman sum identity and as Assmus-Mattson-type theorem for linear codes over Z(4)

PDF

A modified Kalman filter analysis of semiconductor deep-level impurity spectra from simulated and experimental high-resolution Hall effect measurements

PDF

A measurement-based admission control algorithm for integrated services packet networks.

PDF

A new monolithic VCO

PDF

Classification of digital modulations using constellation analyses

PDF

Cellular satellite communication system interference analysis

PDF

A uniform framework for efficient data remapping

PDF

A CMOS Clock and Data Recovery circuit for Giga-bit/s serial data communications

PDF

Analysis of multiplicative noise in laser amplifiers and its effect on laser linewidth

PDF

Array processing algorithms for multipath fading and co-channel interference in wireless systems

PDF

High -speed CMOS continuous -time switched -current sigma -delta modulators

PDF

A process elaboration formalism for writing and analyzing programs

$Hybrid fractal/wavelet methods for image compression$

PDF

Hybrid fractal/wavelet methods for image compression

Asset Metadata

Creator DeMars, Scott David (author)

Core Title Advanced hybrid bulk/integrated optical signal processing modules

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag engineering, electronics and electrical,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Tanguay, Armand R. (committee chair), [illegible] (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-260117

Unique identifier UC11350027

Identifier 9733045.pdf (filename),usctheses-c17-260117 (legacy record id)

Legacy Identifier 9733045.pdf

Dmrecord 260117

Document Type Dissertation

Rights DeMars, Scott David

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA