Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Gallium arsenide based semiconductor laser design and growth by metal-organic chemical vapor deposition
(USC Thesis Other)
Gallium arsenide based semiconductor laser design and growth by metal-organic chemical vapor deposition
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
GALLIUM ARSENIDE BASED SEMICONDUCTOR LASER DESIGN AND GROWTH BY METAL-ORGANIC CHEMICAL VAPOR DEPOSITION by Yuanming Deng 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 2006 Copyright 2006 Yuanming Deng ii Dedication To my wife Jun Hu and to my son Botao Alex Deng iii Acknowledgements First I would like to give my greatest thanks to my thesis advisor, Professor P. D. Dapkus. He provided me the opportunity to work in the compound semiconductor lab, guided me with his knowledge and experience, and encouraged me to face the challenges of the research. He is a tutor, a mentor, a colleague and a friend. What I have learned from him will benefit me forever. I would like to give special thanks to Prof. John O’Brien, who provided the best academic training possible for anyone wants to work in the optoelectronic world. I also want to show my appreciation to all the other members of my dissertation committee and guidance committee for the Ph. D. qualifying exam, namely Professor Edward Goo, Professor William Steier, and Professor Eun Sok Kim. My fellow colleagues and friends in the CSL made my life at USC a memorable experience. We collaborated on research projects, shared our passions and frustrations in our research and in our life. I am grateful that I had the chance to be one of the CSL members. I would like to thank Sang-Wan Ryu and Won-Jin Choi, who introduced me to the wonderful world of MOCVD; Zhijian Wei and Ryan Stevenson worked with me together on the VCSEL project; Lisandra Pataro collaborated with me on the SHEDs project. I would like to thank Dawei Ren, Stephen Farrell and Lawrence iv Stewart for sharing their valuable time to proof-read my thesis. I am also thankful to all the other CSL members: Kostadin Djordjev, Xingang Zhang, Ruijuan Li, San Jun Choi, Thiruvikraman Sadagopan, Wei Zhou, Seung-June Choi, Qi Yang, and Peng Zhen. I would also like to give my thanks to Loring Smith. Our lab would not run smoothly without her assistants on various business issues. Last but not least, my wife Jun Hu gave me unconditional support over the lengthy period of my Ph. D. work. She gave me the best gift I can imagine: my son Botao Alex Deng. The happiness they brought to me helped me through even the most difficult times. v Table of Contents Dedication ................................................................................................................. ii Acknowledgements.................................................................................................. iii List of Tables ......................................................................................................... viii List of Figures .......................................................................................................... ix Abstract ................................................................................................................ xviii Chapter 1 : Introduction .............................................................................................1 References..................................................................................................3 Chapter 2 : Vertical Cavity Surface Emitting Laser Design and Growth..................5 2.1 Introduction of VCSEL........................................................................7 2.2 VCSEL Design.....................................................................................9 2.2.1 Active Region Design ...................................................................9 2.2.2 DBR Design I: Transmission Matrix Method.............................13 2.2.3 DBR Design II: DBR Resistivity ................................................19 2.2.4 Oxide Aperture Design ...............................................................26 2.3 MOCVD Growth of VCSEL..............................................................28 2.3.1 Growth Rate Calibration .............................................................28 2.3.2 QW Calibration ...........................................................................36 2.3.3 Doping Calibration......................................................................37 2.4 VCSEL Devices Performance............................................................38 2.4.1 Wafer Bonded 850nm Bottom-Emitting VCSEL .......................38 2.4.2 850 nm Top-Emitting VCSEL ....................................................43 2.5 Summary ............................................................................................45 References................................................................................................46 Chapter 3 : Super High Efficiency Laser Diodes.....................................................49 3.1 Excess Voltage Drop in Edge Emitting Lasers..................................50 3.1.1 Sources of Losses........................................................................51 3.1.2 Simulation Tool...........................................................................53 3.1.3 Voltage Defects...........................................................................54 3.2 Reduction of Excess Voltage in Symmetric Structures .....................61 3.3 Asymmetric Heterojunction Laser Diodes.........................................68 3.4 Transverse Junction Stripes ...............................................................74 3.4.1 TJS Simulation............................................................................75 3.4.2 Preliminary Experimental Results...............................................95 vi 3.5 Summary ..........................................................................................115 References..............................................................................................115 Chapter 4 : InGaAs Multi-QW Grown On Twist Bonded Compliant Substrate ...117 4.1 Introduction of Compliant Substrate................................................118 4.1.1 Ideal Compliant Substrate.........................................................118 4.1.2 Compliant Substrate Realized ...................................................120 4.1.3 Compliant Substrate Made by Twisted Wafer Bonding ...........121 4.2 Experiments .....................................................................................123 4.2.1 Compliant Substrate Preparation and MOCVD Growth...........123 4.2.2 Compliant Substrate Based on GaAs ........................................125 4.2.3 Compliant Substrate Based on Ion Implanted GaAs.................136 4.2.4 Compliant Substrate Based on InGaAs.....................................143 4.2.5 Transmission Electron Microscopy Results..............................148 4.3 Discussion and Summary.................................................................151 4.3.1 Compliant Substrate Experiment Uniformity and Reproducibility...................................................................................151 4.3.2 Effect of Compliant Substrate...................................................154 4.3.3 Future Works.............................................................................155 References..............................................................................................156 Chapter 5 : Adaptive Designed Asymmetric Electro-Absorption Modulators ......159 5.1 Introduction of Adaptive designed EAM.........................................159 5.1.1 Quantum Confined Stark Effect................................................159 5.1.2 Adaptive Design........................................................................161 5.1.3 Adaptive Designed Asymmetric EAM .....................................162 5.2 EAM Wafer Structure Design Details .............................................165 5.2.1 Active Region Design ...............................................................165 5.2.2 Effect of Background Doping in a Separate Confinement Layer ..................................................................................................168 5.3 Experiments .....................................................................................170 5.3.1 Broad Area Laser Performance.................................................170 5.3.2 EAM Performance ....................................................................171 5.4 Summary ..........................................................................................175 References..............................................................................................175 Chapter 6 : Doping in GaAs MOCVD...................................................................177 6.1 Introduction......................................................................................177 6.2 Doping in MOCVD grown GaAs/AlGaAs ......................................178 6.3 Doping Measurement Technologies ................................................181 6.4 GaAs-AlGaAs Si 2 H 6 and DEZn doping...........................................184 6.4.1 Routine Doping Calibration (Si, Zn).........................................184 6.4.2 Ultra-High P Type Doped GaAs by DEZn ...............................187 6.4.3 Zn and Si Doping at Low V/III Ratio .......................................192 vii 6.5 Carbon Doping Using CBr 4 .............................................................194 6.6 Low Background Doping Al 0.3 GaAs ...............................................207 6.7 Doping Dependency of Crystal Orientation.....................................209 6.8 Summary ..........................................................................................210 References..............................................................................................211 Bibliography...........................................................................................................214 Appendix A: LASTIP Codes for AlGaAs EEL Structure A..................................220 A1 Layer Definition File: AlGa6030.layer............................................220 A2 Simulation Condition Control: AlGa6030.sol..................................221 A3 Material Macro File: yuanming.mac................................................221 Appendix B: LASTIP Codes for TJS-GaAs Structure...........................................234 B1 Layer Definition File: deepgaas.layer ..............................................234 B2 Simulation Condition Control: deepgaas.sol....................................235 Appendix C: MATLAB Code for Transmission Matrix Method ..........................237 C1 VCSEL Reflectivity as A Function of Wavelength .........................237 C2 Electrical Field Distribution in VCSEL ...........................................238 Appendix D: Process Follower of GaAs Based Broad Area Laser........................241 viii List of Tables Table 2-1 Parameters Used in VCSEL Optimization ................................................11 Table 2-2 Confinement Factor as a Functions of QW numbers.................................11 Table 2-3 DBR Layer Thickness and Doping Used In the Simulation......................21 Table 2-4 Simulated Voltage Drops in DBR layers...................................................24 Table 2-5 Epi Structure of 850 nm Top Emitting VCSEL (#5610)...........................35 Table 2-6 GaAs 845nm QW Calibration Sample Epi Structure ................................37 Table 3-1 InGaAs/AlGaAs 940nm EEL Epi Structure “A”.......................................55 Table 3-2 Four Different InGaAs/AlGaAs 940nm EEL Epi Structures ....................61 Table 3-3 Two AHL Structures .................................................................................70 Table 3-4 Epi Structure of TJS-Channel structure middle column............................88 Table 3-5 Selective Area Growth Results with Varying V/III ratio and SiNx Width............................................................................................................107 Table 5-1 Epi Structure of Step Quantum Well EAM .............................................165 ix List of Figures Figure 2-1 Free Space Optical Interconnect Concept for Massive Parallel Communications between Adjacent Computer Boards. ..................................6 Figure 2-2 Schematic of Bi-Directional Board-to-Board Communications via FSOI. ................................................................................................................6 Figure 2-3 Four Basic VCSEL Structures: (a) Etched Air-Post; (b) Ion Implanted; (c) Buried Heterostructure; and (d) selectively oxidized .................................8 Figure 2-4 Driving Current and Wall Plug Efficiency of 850nm VCSEL as Functions of Mirror Reflectivity, Output Power at 1 mW.............................12 Figure 2-5 (a) Single Two-Port Network. (b) Two Networks Cascaded Together....15 Figure 2-6 Transmission Matrices for (a) an interface and (b) a dielectric block......15 Figure 2-7 Reflectivity of VCSEL Calculated by Transmission Matrix Method ......17 Figure 2-8 Field Distribution in VCSEL Calculated by Transmission Matrix Method ...........................................................................................................17 Figure 2-9 DBR Reflectivity as a Function of DBR Pairs.........................................18 Figure 2-10 Schematic Drawing of the Structure for DBR Resistance Simulation...19 Figure 2-11 Simulated Voltage-Current Curves of various VCSEL Structures ........21 Figure 2-12 Band Diagram and Quasi-Fermi Level Distribution of “10Grad” Structure .........................................................................................................22 Figure 2-13 Simulated Voltage Distribution in P-DBRs ...........................................23 Figure 2-14 Simulated Voltage Distribution in N-DBRs...........................................23 Figure 2-15 Influence of Grading on DBR Reflectivity ............................................25 Figure 2-16 Resonant Wavelengths of VCSELs without Oxide Aperture and with Oxide Aperture at Optical Field Node and Antinode.....................................27 Figure 2-17 Reflectivity Measurement Setup ............................................................29 Figure 2-18 Structures of Growth Rate Calibration Samples ....................................30 x Figure 2-19 Reflectivity of Al 0.95 GaAs/Al 0.15 GaAs Growth Rate Calibration Sample............................................................................................................31 Figure 2-20 Reflectivity of Grading Growth Rate Calibration Sample .....................32 Figure 2-21 Simulated Optical Filed Distribution near VCSEL Active Region........34 Figure 2-22 Reflectivity of VCSEL structure ............................................................36 Figure 2-23 PL Data of a Typical GaAs QW Sample................................................37 Figure 2-24 Schematic Drawing of Bottom Emitting VCSEL structure, not to scale................................................................................................................39 Figure 2-25 Bottom Emitting VCSEL LI Curves and Far Field Picture....................39 Figure 2-26 Bottom Emitting VCSEL External Efficiency, Threshold Current, and Series Resistance as Functions of Aperture Size ....................................40 Figure 2-27 Bottom Emitting VCSEL Array Performance, Original Design............41 Figure 2-28 Bottom Emitting VCSEL Array Performance, Improved Design..........42 Figure 2-29 Bottom Emitting VCSEL Array Performance, Improved Oxide Technology.....................................................................................................43 Figure 2-30 Schematic Drawing of Top Emitting VCSEL Structure, not to scale....44 Figure 2-31 SEM image of Top Emitting VCSEL, SiNx Isolation ...........................44 Figure 2-32 LI, IV Data and Lasing Spectrum for a 2μm-Aperture Top-Emitting VCSEL ...........................................................................................................44 Figure 2-33 Small Signal Responses for a 6µm-Aperture Top-Emitting VCSEL, Implant Isolated..............................................................................................45 Figure 3-1 Laser System Power Conversion Efficiency............................................50 Figure 3-2 Simulated VI and LI Curves for Structure “A”........................................56 Figure 3-3 Simulated Wall Plug Efficiency for Structure “A” ..................................56 Figure 3-4 Loss Distribution in Structure “A”...........................................................57 Figure 3-5 Band structures & Quasi-Fermi Level distributions.................................59 xi Figure 3-6 Excess Voltage Drop Distributions at Each Layer and Interface.............60 Figure 3-7 Simulated LIV Curves and Efficiency Curves of Structure “A” and Structure “B”..................................................................................................62 Figure 3-8 Voltage Drop Distributions of Structure “A” and Structure “B” .............62 Figure 3-9 Simulated LIV Curves and Efficiency Curves of Structure “B” and Structure “C”..................................................................................................63 Figure 3-10 Voltage Drop Distributions of Structure “B” and Structure “C” ...........64 Figure 3-11 Simulated Carrier Concentration in Structure “B” and “C”...................65 Figure 3-12 Simulated LIV Curves and Efficiency Curves of Structures “B”, “C” and “D” ..........................................................................................................66 Figure 3-13 Excess Voltage Drop Distributions for 940 EEL Epi Structures. ..........67 Figure 3-14 Electron and Hole Current in Structure “C” (left) and “D” (right) ........67 Figure 3-15 Carrier Concentration Distributions in Structure “C” (left) and “D” (right) .............................................................................................................68 Figure 3-16 Band structure of AHL Structure “A”....................................................70 Figure 3-17 Simulated LIV Curves and Efficiency Curves of AHL Structure “A” ..71 Figure 3-18 Excess Voltage Drop Distribution in EEL Structure “C” ......................72 Figure 3-19 Simulated LIV Curves and Efficiency Curve of AHL Structure “B” ....73 Figure 3-20 Excess Voltage Drop Distributions for AHL Structure “B” ..................73 Figure 3-21 Loss Distribution in AHL Structure “B”................................................74 Figure 3-22 TJS Structure Schematic Drawing .........................................................74 Figure 3-23 Column Settings and Optical Field Calculation Window in TJS Simulation ......................................................................................................76 Figure 3-24 Simulated Optical field in TJS ...............................................................77 Figure 3-25 Current in X direction in the middle of TJS...........................................78 Figure 3-26 Simulated LIV Curves of TJS-GaAs Structure ......................................78 xii Figure 3-27 Band Structures of TJS-GaAs. Simulation Results are Drew along the Lines Mentioned in Figure 3-23...............................................................79 Figure 3-28 Carrier Concentration in TJS-GaAs Along Line “D” ............................80 Figure 3-29 Simulated Laser performances of TJS-Al 0.3 GaAs Structure and EEL “B” Structure..................................................................................................81 Figure 3-30 Carrier Concentration in TJS-Al 0.3 GaAs Along Line “D” .....................82 Figure 3-31 Band Structures of TJS-Al 0.3 GaAs. Simulation Results are Drew along the Lines Mentioned in Figure 3-23.....................................................83 Figure 3-32 Quasi-Fermi Level Separation Minus Phone Energy along Line “D” ...84 Figure 3-33 Comparison of Laser Performances of TJS-Al 0.3 GaAs Structure Simulated with Single Optical Mode and Multi Optical Modes...................86 Figure 3-34 Simulated LI Curves for Each Optical Mode in TJS-Al 0.3 GaAs Structure .........................................................................................................86 Figure 3-35 Comparison of Carrier Concentration in QW of TJS-Al 0.3 GaAs Structure Simulated with Single Optical Mode and Multi Optical Modes ....87 Figure 3-36 Simulated Laser performances of TJS-Channel Structure Comparing with TJS-Al 0.3 GaAs Structure and EEL “B” Structure..................................89 Figure 3-37 Simulated Current Density in X direction in the middle of TJS- Channel Structure...........................................................................................89 Figure 3-38 Band Structures of TJS-Channel (along Line D Mentioned in Figure 3-23) ...............................................................................................................90 Figure 3-39 Wall Plug Efficiencies of TJS Channel Structures with Varying Channel Thickness .........................................................................................91 Figure 3-40 Simulated LI and IV Curves of TJS-Channel Structures with Varying Channel Thickness .........................................................................................91 Figure 3-41 Simulated Laser performances of Asymmetric TJS-Channel Structure Compared with TJS-Al 0.3 GaAs Structure and EEL “B” Structure .........................................................................................................93 Figure 3-42 Band Structures of TJS-Channel Asymmetric Structure (along Line D) ...................................................................................................................93 xiii Figure 3-43 Simulated Laser Performances of TJS-Channel Asymmetric Structures with Varying TJS mesa width.......................................................94 Figure 3-44 Band Structures of TJS-Channel Asymmetric Structure, 2 µm stripe width. (along Line D).....................................................................................95 Figure 3-45 TJS Structure Schematic Drawing .........................................................96 Figure 3-46 Cross-section Profiles after H 3 PO 4 :H 2 O 2 :H 2 O (1:9:1) etching ..............98 Figure 3-47 Cross-section Profiles after H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) etching ............99 Figure 3-48 Cross-section Profile after H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) etching............100 Figure 3-49 Cross-Section Profile after Remove SiNx Overhang...........................101 Figure 3-50 Photoluminescence Data from QWs Grown on Etched GaAs Wafer..102 Figure 3-52 Cross-section Profiles after TJS Al 0.3 GaAs Regrowth, V/III ~ 10,......108 Figure 3-53 Cross-section Profiles after TJS Al 0.3 GaAs Regrowth, V/III ~ 10 Stripes along <1 -1 0> direction...................................................................109 Figure 3-54 Mesa Side Wall Images after TJS Al 0.3 GaAs Regrowth, V/III ~ 10....110 Figure 3-55 SEM Cross Section Images of TJS Al 0.3 GaAs Regrowth, V/III ~ 5 ....110 Figure 3-56 Unsuccessful Second Step Etching ......................................................111 Figure 3-57 SEM Image of a Completed TJS Device .............................................112 Figure 3-58 SEM Image of TJS Mesa, After Device Processing ............................113 Figure 3-59 IV Curves of TJS Devices with Different QW width ..........................114 Figure 4-1 Several realizations of the compliant substrate ......................................120 Figure 4-2 Process of Making Compliant Substrate by Wafer Bonding with Twist121 Figure 4-3 InGaAs Multi-Quantum Well Epilayer Structure ..................................124 Figure 4-4 AFM Image of InGaAs MQWs Grown on GaAs Epi-Ready Substrate.126 Figure 4-5 AFM Image of InGaAs MQWs Grown on 20nm GaAs Compliant Substrate.......................................................................................................127 xiv Figure 4-6 AFM Image of InGaAs MQWs Grown on 10nm GaAs Compliant Substrate.......................................................................................................128 Figure 4-7 Schematic Drawing of X-Ray Diffraction Measurement Setup.............129 Figure 4-8 XRD Ω-2θ Scan of {026} plane of InGaAs MQWs on GaAs substrate 130 Figure 4-9 Ф scan of {026} XRD Peaks, 20nm GaAs Twist-Bonded Compliant Substrate.......................................................................................................130 Figure 4-10 Ф scan of {026} XRD Peaks, 10nm GaAs Twist-Bonded Compliant Substrate.......................................................................................................131 Figure 4-11 (004) Ω-2θ Scans of InGaAs QW grown on GaAs Based Compliant Substrates .....................................................................................................133 Figure 4-12 XRD Experiment and Simulation of InGaAs MQWs on GaAs...........134 Figure 4-13 Photoluminescence of InGaAs MQWs Grown on GaAs Based Compliant Substrate.....................................................................................135 Figure 4-14 Simulated Ion Range for H + 70 KeV 7 Degree....................................137 Figure 4-15 Simulated Collision Events for H + 70 KeV 7 Degree..........................138 Figure 4-16 Simulated Ion Range for As + 180 KeV 7 Degree ................................138 Figure 4-17 Simulated Collision Events for As + 180 KeV 7 Degree ......................139 Figure 4-18 XRD of Ion Implanted GaAs (Dosage=1E12 cm -3 ) .............................140 Figure 4-19 XRD of Ion Implanted GaAs (Dosage=1E13 cm -3 ) .............................141 Figure 4-20 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs....................................................142 Figure 4-21 XRD of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs .....................................................................................143 Figure 4-22 AFM Image of Compliant Substrate Based on InGaAs.......................144 Figure 4-23 AFM Image of InGaAs MQWs Grown on 10 nm GaAs Compliant Substrate Based on InGaAs..........................................................................145 xv Figure 4-24 AFM Image of InGaAs MQWs Grown on 20nm GaAs Compliant Substrate Based on InGaAs..........................................................................146 Figure 4-25 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on InGaAs ........................................................................147 Figure 4-26 XRD of InGaAs MQWs Grown on Compliant Substrates Based on InGaAs .........................................................................................................148 Figure 4-27 TEM Images of InGaAs MQWs Grown on Two Compliant Substrates .....................................................................................................149 Figure 4-28 TEM Image of Compliant Layer from 20nm GaAs Twist Bonded on GaAs.............................................................................................................149 Figure 4-29 TEM Image of Compliant Layer from 10nm GaAs Twist Bonded on Ion Implanted GaAs .....................................................................................150 Figure 4-30 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs....................................................152 Figure 4-31 InGaAs Grown on 10nm Twist Bonded GaAs Compliant Substrate, Insufficient Surface Rinsing before Growth ................................................153 Figure 5-1 Absorption Spectrum of a Rectangular Quantum Well under Electric Fields............................................................................................................160 Figure 5-2 Broken-Symmetry Double Quantum Well Obtained from Numerical ..162 Figure 5-3 Broken-Symmetry Step Quantum Well Obtained for Rapid Decreasing of Absorption with Increasing Electric Filed............................163 Figure 5-4 Absorption Modulation of Optimized Step Well Structure at 851 nm...164 Figure 5-5 Optical Field Distribution in the Asymmetric QW EAM Device..........166 Figure 5-6 Band Structure of Step Well Structure at Zero Bias ..............................167 Figure 5-7 Simulated Electric Filed in the EAM Active Region under Different Bias...............................................................................................................167 Figure 5-8 Simulated Electric Field with Different Reverse Bias, Background Doping Concentration 1E16 cm -3 ................................................................168 xvi Figure 5-9 Simulated Electric Field with Different Reverse Bias, Background Doping Concentration 1E15 cm -3 ................................................................169 Figure 5-10 Performances of Broad Area Lasers Made from Asymmetric QW EAM Wafer..................................................................................................170 Figure 5-11 SEM Images of GaAs EAM Devices...................................................171 Figure 5-12 Single Mode Lens Fiber Coupled EAM Device ..................................172 Figure 5-13 Measured Step Well EAM Output Power at Various Wavelengths.....173 Figure 5-14 Step Well EAM Absorption as Functions of Wavelength ...................173 Figure 5-15 Normal Square Well EAM Absorption as Functions of Wavelength ..174 Figure 6-1 Hall Effect of N-Type Semiconductor ...................................................181 Figure 6-2 Si Doped N GaAs AlGaAs Doping Calibration.....................................185 Figure 6-3 Zn doped P GaAs AlGaAs Doping Calibration .....................................186 Figure 6-4 GaAs Zn P+ Doping Concentration as a Function of Temperature .......188 Figure 6-5 Susceptor Temperature as a Function of Time after Shut Down RF Power ...........................................................................................................190 Figure 6-6 Circular test pattern for determination of specific contact resistance, from Ref. [, 14] ............................................................................................190 Figure 6-7 Transmission Line Measurement and Fitted Results .............................191 Figure 6-8 Si Doped N Doping Calibration at Low V/III Ratio ..............................192 Figure 6-9 Zn Doped P Doping Calibration at Low V/III Ratio..............................193 Figure 6-10 P GaAs/Al 0.15 GaAs Doping Concentration as Function of Ratio of CBr 4 to Group III Mole Fraction at 730 and 630 ºC ....................................195 Figure 6-11 Growth Temperature Dependence of P Doping Concentration in CBr 4 Doped GaAs........................................................................................196 Figure 6-12 Growth Temperature Dependence of P Doping Concentration in CBr 4 Doped Al 0.15 GaAs ...............................................................................197 xvii Figure 6-13 V/III Ratio Dependence of P Doping Concentration in GaAs Doped with CBr 4 , 730 ºC Growth Temperature ......................................................198 Figure 6-14 V/III Ratio Dependence of P Doping Concentration in GaAs Doped with CBr 4 , 630 ºC Growth Temperature ......................................................198 Figure 6-15 P Doping Concentration of CBr 4 Doped Al 0.3 GaAs at Low V/III ratio199 Figure 6-16 Growth Rate of GaAs and AlGaAs as a Function of CBr 4 flow ..........200 Figure 6-17 Lattice Constant Changes as a Function of C Doping Concentration and as a Function of Al Composition...........................................................203 Figure 6-18 XRD Ω-2θ Scans of GaAs N+ Wafer and Carbon Doped GaAs Grown on GaAs N+ Wafers, the P Doping concentrations are 1E18 and 1E19 cm -3 .....................................................................................................204 Figure 6-19 XRD Ω-2θ Scan of Carbon Doped GaAs on N+ GaAs Substrate, P Doping concentration is 8E19 cm -3 ..............................................................205 Figure 6-20 XRD Ω-2θ Scan of Carbon Doped Al 0.15 GaAs on N+ GaAs Substrate, P Doping concentration is 1E20 cm -3 ..........................................206 Figure 6-21 Background Doping Concentration of AlGaAs as a Function of V/III Ratio, Temperature and Aluminum Composition........................................208 xviii Abstract This dissertation presents research projects with the common theme: novel GaAs based device structures grown by metal organic chemical vapor deposition (MOCVD). Vertical-cavity surface-emitting laser (VCSEL) arrays for application to free space optical interconnections were manufactured. The design of VCSELs with oxide apertures is discussed. Low-threshold, uniform 20x20 arrays with a 3dB frequency of up to 5 GHz were demonstrated with MOCVD grown epi structures. Novel designs for high efficiency, high-power edge emitting lasers are discussed. Excess voltage drop is identified as an important cause limiting the efficiency. First, an asymmetric quantum well with a small conduction band offset in the n side and small valence band offset in the p side is proposed. Second, a transverse junction structure which has the carriers injected from the sides of the QW is also proposed. Simulations of both structures predict 80% overall efficiency. Preliminary selective area growth tests for the transverse junction structure were preformed. Compliant substrates are evaluated to extend the use of GaAs based materials to longer wavelengths. Thick InGaAs layers with InGaAs multiple quantum wells were grown on compliant substrates which were made by wafer bonding a thin GaAs layer on a thick GaAs substrate. A material quality improvement was observed compared with the same epi structure grown directly on GaAs substrates, but the improvement is too small for real device applications, so the limitations of compliant xix substrates are discussed. Preliminary research findings on adaptive designed asymmetric electro- absorption modulators are presented. Control of intentional and background doping in the MOCVD growth of AlGaAs is discussed. The doping behaviors of Zn and Si in AlGaAs under typical growth conditions and at low V/III ratios were investigated. Carbon doping of AlGaAs by CBr 4 was realized. The AlGaAs growth rate and lattice constant are influenced by the CBr 4 doping. MOCVD grown Al 0.3 GaAs with low background doping of 1~2E15 cm -3 was achieved by using TMAl, TEGa and Arsine as sources. 1 Chapter 1 : Introduction The operation of the semiconductor laser was first reported by four groups in late 1962 [1, 2, 3, 4]. Since then semiconductor lasers have been developed from simple pn junctions, double heterojunctions, and strained QWs. These devices have found applications from laser printing to medical applications. Depending on the specific application, semiconductor lasers can be designed to have a low threshold, high efficiency, high power, narrow linewidth, high speed, or long, short, or tunable wavelength. [5] Metal organic chemical vapor deposition (MOCVD) growth technology was invented by Manasevit in 1968 [6]. The successful fabrication of heterostructure devices, such as double heterostructure lasers in 1977 [7] and quantum well lasers in 1979 [8] by Dupuis and Dapkus, clearly demonstrated MOCVD as a dominant epitaxial growth technology. Today, MOCVD technology has been widely used for commercial production operations, particularly for the production of devices requiring large areas, such as LEDs, photocathodes and solar cells. The main advantages are the ability to provide large-scale production, high growth rates, and a high yield of suitable material. [9] The MOCVD system used in this work is a combination of a gas handling system made by Thomas Swan Scientific Equipment Ltd. and a low pressure (76 Torr) single 2-inch wafer vertical reactor designed by Won-Jin Choi [10]. A metal- organic precursor and a hydride gas are used as the source of a column III and column V element, respectively, and hydrogen is used as the main carrier gas. 2 This dissertation presents several research projects surrounding GaAs based semiconductor lasers. Most of these projects involve research work in device design, all of them require MOCVD growth. All growth related issues are discussed with the projects except for doping issues which are summarized in the last chapter. Vertical-cavity surface-emitting laser (VCSEL) arrays are a competitive candidate for application in free space optical interconnections. The device design and MOCVD growth of VCSELs for such a purpose is discussed in chapter 2. Low- threshold, uniform 20x20 arrays with a 3dB frequency of up to 5 GHz were demonstrated with MOCVD grown epi structures. Novel designs for high efficiency, high-power edge emitting lasers are discussed in chapter 3. Excess voltage drop is identified as the main cause for low efficiency. First, an asymmetric quantum well with a small conduction band offset in the n side and small valence band offset in the p side is proposed. A transverse junction structure which has the carriers injected from the sides of the QW is also proposed. Simulations of both structures predict 80% overall efficiency. The preliminary selective area growth results for the transverse junction structure are also presented. In Chapter 4, compliant substrates are evaluated to extend the use of GaAs based lasers to longer wavelengths. Thick InGaAs layers with InGaAs multiple quantum wells were grown on compliant substrates which were made by wafer bonding a thin GaAs layer on a thick GaAs substrate. Material quality improvements and limitations of compliant substrates are discussed. 3 Chapter 5 presents the preliminary research findings on adaptive designed asymmetric electro-absorption modulators. AlGaAs doping issues in MOCVD growth are summarized in chapter 6. Topics covered in this chapter include: the doping behaviors of Zn and Si at normal growth conditions and at low V/III ratios, Carbon doping of AlGaAs by CBr 4 , the influence of CBr 4 on the growth rate and lattice constant of AlGaAs, and MOCVD grown low background doping Al 0.3 GaAs. References 1 R. N. Hall, G. E. Fenner, J. D. Kingsley, T. J. Soltys, and r. O. Carlson, ”Cohernt Light Emission from GaAs Junctions,” Phys. Rev. Lett., Vol. 9, pp. 366-368, 1962. 2 M. I. Nathan, W. P. Dumke, G. Burns, F. H. Dill, Jr., and G. Lasher, “Stimulated emission of radiation from GaAs p-n junctions,” Appl. Phys. Lett., Vol. 1, pp. 62-64, 1962. 3 N. Holonyak, Jr. and S. F. Bevacqua, “Coherent (Visible) Light Emission from Ga(As x P 1-x ) Junctions,” Appl. Phys. Lett., Vol. 1, pp. 82-83, 1962. 4 T. M. Quist, R. H. Rediker, R. J. Keyes, W. E. Krag, B. Lax, A. L. McWhorter, and H. J. Zeiger, “Semiconducter maser of GaAs,” Appl. Phys. Lett., vol. 1 pp. 91- 92, 1962. 5 Donald A. Neamen, “Semiconductor Physics and Devices, Basic Principles,” McGraw-Hill Companies, Inc., 2003. 6 H. M. Manasevit, “Single-Crystal Gallium Arsenice on Insulating Substrates,” Appl. Phys. Lett., Vol. 12, No. 4, pp. 156-159, 1968. 7 R. D. Dupuis and P. D. Dapkus, “Room Temperature Operation of Ga 1- x Al x As/GaAs Double Heterostructure Lasers Grown by Metalorganic Chemical Vapor Deposition,” Appl. Phys. Lett., Vol. 31, No. 7, pp. 466-468, 1977. 8 R. D. Dupuis, P. D. Dapkus, R. Chin, N. Holonyak, Jr., and W. W. Kirchoefer, “Continuous 300 K Laser Operation of Single Quantum-Well AlGaAs 4 Heterostructure Diodes Grown by Metalorganic Chemical Vapor Deposition,” Appl. Phys. Lett., Vol. 34, No. 4, pp. 265-267, 1979. 9 Gerald Stringfellow, “Organometallic Vapor-Phase Epitaxy,” second edition, Academic Press, 1999. 10 Won-Jin Choi, “Design and Fabrication of Novel InGaAs and AlGaInP Quantum Well Lasers by Metalorganic Chemical Vapro Deposition,” Ph.D. thesis dissertation, 1999. 5 Chapter 2 : Vertical Cavity Surface Emitting Laser Design and Growth Vertical Cavity Surface Emitting Lasers (VCSELs) have the unique property of emitting light vertically from the top surface instead of laterally from the cleaved edge as edge emitting lasers does. This distinctive feature enables VCSELs with significant advantages. VCSELs have superior performance at low power regime. They have the potentials of low threshold, high efficiency and high speed. Their circularly shaped, low-numerical-aperture output beams make it easy for coupling to fibers or free-space optics. Their vertical emission nature also makes large two- dimensional array readily available. Besides these performance advantages, VCSELs also have attractive manufacturability features: the possibility of the low manufacture cost due to the elimination of fabrication steps such as wafer lapping, cleaving, facet coating and diode bonding; the possibility of being able to fabricate and test lasers on a wafer scale and perform non-intrusive testing; the possibility of formation of large laser arrays with no change in the fabrication procedure. These advantages have made VCSEL a superior candidate for short-haul data communications. Meanwhile, VCSELs are also considered for many other applications, such as printing, optical switch. [1, 2, 3] VCSELs discussed in this chapter are targeted for the applications in free space optical interconnect (FSOI). FSOI is an alternative to replace electrical wires for massively parallel chip-to-chip or board-to-board communications. This 6 application requires low threshold, high wall plug efficiency, single longitudinal mode, good beam profile, and high aggravated speed. The concept of FSOI is shown in Figure 2-1, from Ref. [4]. One Schematic of bi-directional board to board communications via FSOI is shown in Figure 2-2, also from Ref. [4]. VCSEL array and detector array are mounted on and driven by Si integrated circuits; microlens array, diffractive optical element (DOE) array and Fourier lens are used to direct optical beams. Process Units VCSEL Array Detector array Board-to-Board Optical Link Process Units VCSEL Array Detector array Board-to-Board Optical Link Figure 2-1 Free Space Optical Interconnect Concept for Massive Parallel Communications between Adjacent Computer Boards. VCSEL Array DOE Array Fourier Lens Microlens Array Detector Array Si IC VCSEL Array DOE Array Fourier Lens Microlens Array Detector Array Si IC Figure 2-2 Schematic of Bi-Directional Board-to-Board Communications via FSOI. 7 The goal of the research was to realize high performance uniform 20 by 20 850nm VCSEL arrays. This chapter is organized as follows: First, basic concept of VCSEL will be introduced, followed by detailed discussions on the design of VCSEL. The next section will focus on issues related to the VCSEL epi growth. Finally, performances of VCSELs will be presented. 2.1 Introduction of VCSEL The distinctive feature of VCSELs is that laser reflectors are parallel to the epi layers. Since the cavity volume of VCSEL is small, high mirror reflectivity is required to compensate the low round trip gain. Distributed Bragg Reflector (DBR) is the only practical way to realize the high reflectivity. Other than DBRs, initial research also [9] utilized metal mirrors which can reach high reflectivity. However, metal mirror is also absorptive, which introduces extra loss in the laser. In the GaAs based VCSEL, due to large refractive index difference between GaAs and AlAs, it becomes practical to have a full VCSEL structure, consisting of 2 DBRs with active region sandwiched in between, grown in a single epi growth run. In the long wavelength VCSEL which has the InP based active region, it is not practical to have a epitaxially grown InP based DBR structure for the following two reasons. First, the index difference available in the InP based material is small. Second, the inherent optical losses, i.e. Auger recombination and inter-valence band absorption, are large. Two hybrid mirror technologies have been developed to accommodate InP based 8 active region. One is wafer bonding of InGaAsP-InP active region and GaAs-AlAs DBRs [5,6]. The other is using a dielectric DBR deposited after semiconductor crystal growth [7]. The ultimate threshold current can be realized by decreasing the active region volume and increasing the overlap between the optical field and active region. Reflectors above and below active region define vertical optical mode in VCSEL. Both the current and optical field need to be confined laterally. Figure 2-3 Four Basic VCSEL Structures: (a) Etched Air-Post; (b) Ion Implanted; (c) Buried Heterostructure; and (d) selectively oxidized The primary VCSEL technologies that have been proven to provide lateral optical and/or electrical confinement include: ion implantation, etched mesa, oxide aperture, and buried heterostructure. Schematic drawings of these four basic 9 structures are shown in Figure 2-3, from Ref. [8]. These methods are often being combined to optimize the VCSEL design. Among these methods, oxide aperture is the most promising VCSEL technology by far. The oxide aperture is achieved by the selective wet oxidization of a high Al composition AlGaAs layer into AlOx and leaving small region of un-oxidized AlGaAs as the current path. Optical index guide is also formed due to the index contrast between the AlOx and the AlGaAs of the aperture region. The first VCSEL was presented by Soda et al. in 1979 [9]. The first room- temperature continuous-wave device using GaAs material was demonstrated in 1989 [10]. The discovery of a stable oxide in the AlGaAs material system by Dallesasse et al. [11] has significant impact on the VCSEL technology. VCSELs with high power conversion efficiency [12], low threshold [13, 14] and large modulation bandwidth [15] have been demonstrated. This chapter is focused on VCSELs with following features: GaAs substrate; fully epi grown structure; 850 nm wavelength; selective oxidized AlAs/AlGaAs aperture; peak performance at 1 mW. 2.2 VCSEL Design 2.2.1 Active Region Design Threshold current for a QW laser can be written as: 0 / ) ( 2 2 g i tr th m i e qVBN I Γ + > < ≅ α α η (2.1) 10 where V is the volume of quantum well(s), B is the bimolecular recombination coefficient, tr N is the transparency carrier density, i η is the internal quantum efficiency, > < i α is the average internal loss, m α is the mirror loss, Γ is the confinement factor, and 0 g is the gain coefficient used in the simplified two- parameter gain formula: tr N N g g ln 0 = (2.2) where N is the carrier density. Above the threshold, the laser output power increases linearly as the driving current increases: ) ( th d o I I q h P − = υ η (2.3) where d η is the differential quantum efficiency and defined as: m i m i d α α α η η + > < = (2.4) The overall power conversion or “wall-plug” efficiency is an important parameter to judge the laser performance. It is defined as in P P 0 , where in P is the total power into the VCSEL, which can be modeled as: s d s in IV IV R I P + + = 2 (2.5) where s R is the series resistance, s V is a current-independent series voltage, and d V is the idea diode voltage, which is equal to the quasi-Fermi level separation in QW. [3, 20] 11 Using above equations, the device design is optimized to have the highest wall plug efficiency at 1 mW output power. The optimization mainly focuses on the number of quantum wells and the reflectivity of top mirror. An 850nm top emitting laser is considered. The QW used in the VCSEL consists of 8 nm GaAs with Al 0.2 GaAs as barriers. The barriers between QWs are 10nm if there are multiple QWs in the active region. The lateral size of the VCSEL is 5 µm by 5 µm. The parameters used in the following simulation are listed in Table 2-1. tr N 2.6E18 cm -3 0 g 2400 cm -1 i α 20 cm -1 B 0.8E-10 cm 3 /s V 2E-13 cm 3 per QW L 1.15E-4 cm i η 0.9 s R 250 Ω s V 0.4 V Table 2-1 Parameters Used in VCSEL Optimization 1 QW 2 QWs 3 QWs 4 QWs 5 QWs Γ 0.016 0.030 0.040 0.046 0.050 Table 2-2 Confinement Factor as a Functions of QW numbers 12 The confinement factors are calculated using transmission matrix method which will be discussed later in this chapter. Calculated confinement factors for QW(s) are shown in Table 2-2. The maximum of optical field is very well confined in the center of cavity, the overlap of the optical field and the QW decreases significantly as the QW is placed further away from the center. The mirror loss m α is defined as: ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ = 2 1 1 ln 1 R R L m α (2.6) where L is the effective cavity length of VCSEL, 1 R and 2 R are the reflectivity of top and bottom mirrors. The bottom mirror has the reflectivity of 1. 0.92 0.94 0.96 0.98 1.00 1E-3 0.01 0.1 1 10 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 Output Power 1mW 1 QW 3 QWs 5 QWs Current(A) Top Mirror Reflectivity Wall Plug Efficiency Figure 2-4 Driving Current and Wall Plug Efficiency of 850nm VCSEL as Functions of Mirror Reflectivity, Output Power at 1 mW The driving current and wall plug efficiency of the 850nm VCSEL at 1 mW laser output as functions of top mirror reflectivity are shown in Figure 2-4. The 13 active region designs of 1 QW (solid curves), 3 QWs (dashed curves), and 5 QWs (dotted curves) are presented. The black curves are for the driving current and the red curves are for the wall plug efficiency. It is clear that very high top mirror reflectivity of 98% to 99% is required to have an optimized performance. The 3 QW design has the best wall plug efficiency for the parameters used. In order to get the best laser performance, QW gain peak should coincide with the cavity resonant wavelength. Both the gain peak and the cavity mode shift to the longer wavelength as the device temperature changes. The gain peak shifts about 0.3nm/°C [16, 17] while the cavity resonant wavelength shifts about 0.1nm/°C [18, 17]. VCSEL normally has relatively large thermal impedance because the power is being dissipated in a compact active region and the heat must then spread out into the substrate. The best design is to intentionally align the QW gain peak with the cavity resonant wavelength at the device working temperature. Since the gain peak shifts faster than the cavity resonant wavelength, so the QW gain peak is located in the short wavelength side of cavity resonant at room temperature. The optimum gain peak location varies with the device working temperature which is strongly influenced by device geometries and working conditions. In our current design of 850 nm VCSEL, the gain peak of QW is set at 845 nm, this optimization was experimentally determined by our former group members [19]. 2.2.2 DBR Design I: Transmission Matrix Method A distributed Bragg reflector is a structure formed from multiple layers of 14 alternating materials with different refractive index. Each interface reflects partial of the optical wave. A high quality reflector is constructed at the wavelength of which all the reflections combine with constructive interference. Transmission matrix method, which will be discussed next, is a powerful tool to simulate the multilayer structures. Considering the electric field which has the form of following: ) ~ ( 0 ) , ( ˆ ) , , , ( z t j e y x U E e t z y x β ω − − = Ε v (2.7) We can define a normalized amplitudes, j A , which have a magnitude equal to the square root of the power flow and a phase equal to the electric field: z j j j e E A β η ~ 0 2 − = (2.8) where j j n / 377 Ω = η is the mode impedance. β ~ in the equation is the complex propagation constant which includes any loss or gain: i j β β β + = ~ (2.9) λ π β / 2 n = (2.10) α β − = g i (2.11) where the n is the effective refractive index, g is the gain and a is the loss. The effect of periodic gratings on light propagation can be easily represented by transmission matrices. The transmission matrix of a two-port network is defined as 15 ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ = ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ 2 2 22 21 12 11 1 1 B A T T T T B A (2.12) where forward-going waves are denoted as A i and backward-going waves are denoted as B i , referring to Figure 2-5 (a), from Ref [20]. A i and B i have the form defined in equation (2.2). T A 1 A 2 B 1 B 2 T A 1 A 2 B 1 B 2 T' A 1 ' A 2 ' B 1 ' B 2 ' (a) (b) T A 1 A 2 B 1 B 2 T A 1 A 2 B 1 B 2 T A 1 A 2 B 1 B 2 T' A 1 ' A 2 ' B 1 ' B 2 ' T A 1 A 2 B 1 B 2 T' A 1 ' A 2 ' B 1 ' B 2 ' (a) (b) Figure 2-5 (a) Single Two-Port Network. (b) Two Networks Cascaded Together. The transmission matrix of two two-port networks cascaded together is defined as ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ′ ′ ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ′ ′ ′ ′ ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ = ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ = ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ 2 2 22 21 12 11 22 21 12 11 2 2 22 21 12 11 1 1 B A T T T T T T T T B A T T T T B A (2.13) Complicated cascade networks can be easily described by multiplication of simple transmission matrices. 12 t 12 r 12 12 21 12 21 t t r r = − = 22 L 2 L 2 ~ β φ − = (b) (a) ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ 1 1 1 12 12 2 r r t ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ − φ φ j j e e 0 0 12 t 12 r 12 12 t 12 r 12 12 21 12 21 t t r r = − = 22 L 2 22 L 2 L 2 ~ β φ − = (b) (a) ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ 1 1 1 12 12 2 r r t ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ − φ φ j j e e 0 0 Figure 2-6 Transmission Matrices for (a) an interface and (b) a dielectric block 16 Two basic optical elements described by two ports transmission matrices are shown in Figure 2-6, from Ref. [20]. The diagram shown in the left is an interface of two different materials. The reflection and transmission can be calculated according to Fresnel’s equation: 2 1 2 1 21 12 n n n n r r + − = − = (2.14) 2 1 2 1 21 12 2 n n n n t t + = = (2.15) where 1 n and 2 n are the refractive indices of two materials. The diagram shown in the right is a dielectric block. There are propagation delays which are same for both forward and backward waves. Normally incident plane waves are considered in both cases. Any complex multilayer structures can be broken down to interfaces and bulk dielectrics. The transmission matrix of the whole structure is the simple multiplication of transmission matrices of all the basic elements, namely interfaces and bulk dielectrics. The final transmission matrix of a multilayer structure is simply a 2 by 2 complex matrix which has the form of equation (2.6). Set 1 1 = A and 0 2 = B , the amplitude reflection and transmission coefficients of the structure can be calculated according to the following equations: 11 1 2 1 T A A t = = (2.16) 11 21 1 1 T T A B r = = (2.17) 17 The field amplitude distribution in the structure can also be obtained by using transmission matrices after both A1 and B1 are known. Considering a 980nm cold cavity (no active region) which consists of two AlAs/GaAs DBRs of 10 periods and a GaAs λ cavity, the reflectivity is calculated by transmission matrix method. The result is shown in Figure 2-7. Electrical field distribution in the structure is also obtained from transmission matrix method. The result is shown in Figure 2-8. Figure 2-7 Reflectivity of VCSEL Calculated by Transmission Matrix Method Figure 2-8 Field Distribution in VCSEL Calculated by Transmission Matrix Method 18 Matlab programs were written to perform the above simulations. The Matlab source codes are listed in Appendix C. Figure 2-9 DBR Reflectivity as a Function of DBR Pairs As discussed earlier in this chapter, the VCSEL bottom mirror reflectivity needs to be close to 1 and the top mirror reflectivity needs to be close to 99%. The DBR pairs required for such a high reflectivity is calculated. The results for GaAs/AlAs DBR and Al 0.15 GaAs/Al 0.95 GaAs DBR are shown in Figure 2-9. In order to reach 99% reflectivity, 14 pairs of GaAs/AlAs DBR and 18 pairs of Al 0.15 GaAs/Al 0.95 GaAs DBR are required respectively. In order to reach 99.99% reflectivity, 25 pairs of GaAs/AlAs DBR and 31 pairs of Al 0.15 GaAs/Al 0.95 GaAs DBR are required respectively. The smaller the difference between the refractive indices of those two DBR dielectrics is, the more number of DBR pairs are required to reach a certain reflectivity. 19 Even though GaAs/AlAs DBR can provide the high reflectivity with less number of pairs, this combination is not practical for 850 nm VCSEL for two reasons. In order to avoid the optical loss in the DBR, the minimum aluminum composition should be 15%, which has large enough band gap to prevent absorption. At the same time, AlAs or Al 0.98 GaAs is used for oxide aperture. In order to avoid the DBR structure to be oxidized in the wet oxidizing process, the highest aluminum composition in DBR is set as 95%. There are also other considerations which will significantly change the DBR structure. They will be discussed in the following sections. 2.2.3 DBR Design II: DBR Resistivity When the metal contact is on the outside of DBR, carriers must travel through the DBR structure to reach the active region. Large resistance of DBR multilayers seriously hampers the device’s DC and RF characteristics. LASTIP simulation was used to analyze the resistance of DBR structure. Detailed introduction of LASTIP is presented in chapter 3. 5 pairs P-DBR 5 pairs N-DBR Active Region P Metal N Metal 5 pairs P-DBR 5 pairs N-DBR Active Region P Metal N Metal Figure 2-10 Schematic Drawing of the Structure for DBR Resistance Simulation 20 The schematic drawing of the structures used in the simulation is shown in Figure 2-10. This simulation is trying to obtain information of serial resistance of DBRs used in an 850 nm VCSEL. So the active region is a GaAs QW with Al 0.2 GaAs as barriers. The active region is one wavelength thick. The active region is sandwiched in two 5-pair Al 0.15 GaAs/Al 0.95 GaAs DBRs. The bottom DBR is n type and the top DBR is p type. P and n metal contacts are formed on the surfaces of the DBR. The simulated device dimension is 5 μm by 5 μm. Maximum 2 mA current is simulated flowing through the structure. 5 different DBR designs are simulated. The DBR dielectric material, thickness and doping information are summarized in Table 2-3. The “NoGrad” DBR has no grading between the high refractive index material Al 0.15 GaAs and low refractive index material Al 0.95 GaAs. The layers are quarter wavelength thick and all the layers are uniformly doped to 1E18 cm -3 . The “10Grad” DBR has a 10 nm grading on each interface. The thicknesses of Al 0.15 GaAs and Al 0.95 GaAs are reduced accordingly to maintain the periodicity of the DBR. This DBR is uniformly doped to 1E18 cm -3 . The “10GradHigh” DBR has the same layer structure as the “10Grad” DBR except the doping is 4E18 cm -3 in the grading layers in which the carriers flow from the Al 0.15 GaAs to Al 0.95 GaAs. According to the transmission matrix method simulation, the optical field is at its minimum on these interfaces (see Figure 2-8). So the increasing in the doping won’t increase the free carrier absorption. The “20Grad” DBR has similar structure as “10Grad” DBR but has 20 nm Grading. The “20GradHigh” DBR is similar to “10GradHigh” DBR but has 20 nm Grading. 21 Sample Material NoGrad 10Grad 10GradHigh 20Grad 20GradHigh Al 0.15 GaAs 59 nm 1E18 49 nm 1E18 49 nm 1E18 39 nm 1E18 39 nm 1E18 Grading 0 nm 10 nm 1E18 10 nm 4E18 20 nm 1E18 20 nm 4E18 Al 0.95 GaAs 70 nm 1E18 60 nm 1E18 60 nm 1E18 50 nm 1E18 50 nm 1E18 Grading 0 nm 10 nm 1E18 10 nm 1E18 20 nm 1E18 20 nm 1E18 Table 2-3 DBR Layer Thickness and Doping Used In the Simulation 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 NoGrad 10Grad 20Grad 10GradHigh 20GradHigh Current (mA) Voltage (V) Figure 2-11 Simulated Voltage-Current Curves of various VCSEL Structures Simulated IV curves are shown in Figure 2-11. It is clearly shown that both increasing the grading layer thickness and increasing the doping in grading layer reduce the device resistance. Before the detailed discussion of each layer’s contribution to the serial resistance, a problem of LASTIP simulation should be discussed first. The “NoGrad” structure has extremely high resistance which is overestimated by the LASTIP simulation. When there is no grading in the junction, tunneling is the main source of the current flow. LASTIP gives out unrealistically 22 high resistance for abrupt junction since it underestimates the tunneling current through the junction. When there is grading presented, LASTIP can simulate the junction behavior accurately. In order to correctly simulate the abrupt junction, an add-on software package regarding tunneling effect must be used combining with the basic LASTIP simulation [21]. The following text will focus on the effect of various grading in DBRs. -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Conduction Band Valence Band Electron Quasi-Fermi Level Hole Quasi-Fermi Level Energy/Potential (eV) Position( μm) Figure 2-12 Band Diagram and Quasi-Fermi Level Distribution of “10Grad” Structure The band structure and quasi-Fermi level distribution of “10Grad” structure with 2 mA driving current are shown in Figure 2-12. The black curve is the conduction band structure, the cyan curve is the valence band structure, the green curve is the electron quasi-Fermi level distribution, and the blue curve is the hole quasi-Fermi level distribution. The voltage drop distribution can be obtained by following the electron quasi-Fermi level in the n side of the device and following the hole quasi-Fermi level in the p side of the device. Detailed discussion about obtaining voltage drop distribution is presented in chapter 3. 23 0.35 0.40 0.45 0.50 -0.14 -0.12 -0.10 -0.08 -0.06 -0.04 10Grad 20Grad 10GradHigh 20GradHigh P DBR Voltage Distribution Al 0.15 GaAs Al 0.95 GaAs Voltage (V) Position ( μm) Figure 2-13 Simulated Voltage Distribution in P-DBRs The voltage distributions in the p DBRs are shown in Figure 2-13. Four curves corresponding to four graded DBR structures are presented. They are marked by different colors according to the legends in the graph. Vertical black lines mark the positions of interfaces. For all these curves, the zero voltage points are located in the QW at the 0 μm position which is not shown. -0.55 -0.50 -0.45 -0.40 -0.35 0.00 0.02 0.04 0.06 0.08 0.10 0.12 10Grad 20Grad 10GradHigh 20GradHigh N DBR Voltage Distribution Al 0.15 GaAs Al 0.95 GaAs Voltage (V) Position ( μm) Figure 2-14 Simulated Voltage Distribution in N-DBRs 24 Holes flow from right to left in this drawing. Similar drawings for n DBRs are shown in Figure 2-14. The zero voltage points are still located in the QW at the 0 μm position. Electrons flow from left to right in the drawing. Figure 2-13 and Figure 2-14 are plotted with the same scale for the convenience of comparison. Sample Material 10Grad 20Grad 10GradHigh 20GradHigh p-Al 0.15 GaAs 0.002 0.002 0.002 0.002 p-Grading (LH) 0.013 0.006 0.004 0.002 p-Al 0.95 GaAs 0.009 0.007 0.009 0.007 p-Grading (HL) 0.013 0.005 0.013 0.005 P Total 0.037 0.020 0.028 0.016 n-Al 0.15 GaAs 0.001 0.001 0.001 0.001 n-Grading (LH) 0.008 0.006 0.002 0.002 n-Al 0.95 GaAs 0.003 0.002 0.003 0.002 n-Grading (HL) 0.007 0.007 0.007 0.007 N Total 0.019 0.015 0.013 0.011 Table 2-4 Simulated Voltage Drops in DBR layers Voltage drops in all layers of simulated DBRs are summarized in Table 2-4. The Grading (LH) is the grading from Al 0.15 GaAs to Al 0.95 GaAs in the current flow direction. The Grading (HL) is from Al 0.95 GaAs to Al 0.15 GaAs in the current flow direction. Compared to n DBR, p DBR has higher voltage drop due to heavy effective mass and low mobility. Increasing the length of grading minimize the voltage drop, especially for p DBR. Increasing the doping in the grading also helps to lower the voltage drop. Total resistance of the device can be further lowered if all grading layers are all highly doped. However, there is a trade off between the DBR resistance and free carrier absorption. Other than grading layers, p-Al 0.95 GaAs also contributes to the high voltage drop due to high bulk resistance. 25 10 15 20 25 30 35 40 96.0 96.5 97.0 97.5 98.0 98.5 99.0 99.5 100.0 25 30 35 40 99.86 99.88 99.90 99.92 99.94 99.96 99.98 100.00 NoGrad 10Grad 20Grad DBR Rreflectivity DBR Pairs Figure 2-15 Influence of Grading on DBR Reflectivity Using the grading instead of the abrupt junction in DBR decreases the resistivity, but it also influences the reflectivity. The reflectivity of Al 0.15 GaAs- Al 0.95 GaAs DBRs with no grading, with 10 nm grading, and with 20 nm grading is shown in Figure 2-15. According to the simulation, in order to reach the reflectivity of 99%, the number of DBR pairs need to be 16, 17, and 18 respectively for these three DBRs. In order to reach the reflectivity of 99.99%, 31, 32 and 35 pairs of DBRs are required respectively. Although the DBR resistivity increases linearly with the number of DBR pairs, the reduction of the DBR resistivity due to a wide grading is much larger. Other than DBR reflectivity, there is also another trade off between the free carrier absorption and the DBR resistivity. Doped DBR introduces the free carrier absorption especially in the p DBR side. In order to minimize the free carrier loss, few pairs of p DBR close to the active region may intentionally be doped less. The DBR resistivity is sacrificed to lower the loss in this case. 26 2.2.4 Oxide Aperture Design Oxide apertures provide lateral carrier confinement in VCSELs. The oxide apertures are placed very close to the active region to prevent carriers’ lateral diffusion. The oxide apertures can be placed in both of the p and the n side of the active region or just placed in one side. The oxide apertures are usually located in the p side of the active region. It was proven in the experiments that VCSELs with only the n side oxide apertures had large threshold and low efficiency [22]. After passing the aperture, electrons has a long diffusion length due to their high mobility. The carrier distribution profile in the active region extended much wider than the optical mode profile. So that n side oxide aperture doesn’t provide a good carrier confinement. Aluminum oxide is a transparent material with the refractive index of ~1.6. The oxide apertures also confine the optical mode by changing the effective index. However, scattering loss is introduced because the lasing mode travels through most of the structure unguided. The effective index profile for VCSEL’s can be determined by local changes in the Fabry-Perot resonance frequency: 0 0 λ λ Δ ≈ Δ eff eff n n (2.18) where eff n is the effective index and 0 λ is the resonance wavelength. [23] 27 842 844 846 848 850 90 91 92 93 94 95 96 97 98 99 100 Cold Cavity Core Oxide Aperture at Antinode Oxide Aperture at Node Reflectivity (%) Wavelength(nm) Figure 2-16 Resonant Wavelengths of VCSELs without Oxide Aperture and with Oxide Aperture at Optical Field Node and Antinode Resonant wavelengths of VCSELs without the oxide aperture and with the oxide apertures at different locations are shown in Figure 2-16. The effective index of our VCSEL structure in the aperture opening is 3.5. According to equation 2.18 and Figure 2-16, the effective index change is 0.03 when the oxide aperture is located at the antinode of optical field. The effective index change is only 0.0006 when the oxide aperture is located at the optical field node. However, this small effective index change is enough to provide the index guiding and a stable mode control. It also minimizes the aperture scattering loss. [24, 25] It is also believed that a tapered oxide aperture can further decrease the scattering loss. [26] While considering the aperture layer material, oxides formed from AlGaAs layers are better than those from AlAs layers in terms of oxidation isotropy, mechanical stability and lack of strain. Oxide formed from Al 0.98 GaAs which has only 2% Ga already shows clearly superiority over AlAs. [27] 28 2.3 MOCVD Growth of VCSEL The challenge of growth includes highly complex layer structure, precision control of material composition, layer thickness, and doping. Especially the layer thickness, a 3% change in the layer thickness will change the resonant wavelength by 25.5nm in an 850nm VCSEL. A similar 3% change in the layer thickness in an 850nm QW, which has Al 0.2 GaAs as the barrier and GaAs as the QW, will only shift the PL peak wavelength by 1.2nm. In the following sections, the growth rate calibration sequence in the VCSEL growth will be introduced first. Then the issues regarding doping concentration, QW wavelength calibration and growth uniformity and reproducibility will be discussed. 2.3.1 Growth Rate Calibration VCSEL structure has stringent requirement for the correct layer thickness. Both mirror reflectivity and cavity length (lasing wavelength) are controlled by epitaxial growth in our current design. The key factor influences these characteristics is the optical thickness, i. e. ) ( λ n L , where L is the physical thickness of the layer and ) ( λ n is the refractive index at certain wavelength λ . Instead of trying to get the accurate measurements of both the physical thickness and the refractive index, optical thickness itself as a parameter is relative easy to be determined with the combination of reflectivity measurement and TFCalc simulation. 29 Fiber Coupler Wafer or Reference Mirror Optical Spectrum Analyzer Tungsten Lamp Multimode Fiber Fiber Coupler Wafer or Reference Mirror Optical Spectrum Analyzer Tungsten Lamp Multimode Fiber Fiber Coupler Wafer or Reference Mirror Optical Spectrum Analyzer Tungsten Lamp Multimode Fiber Figure 2-17 Reflectivity Measurement Setup A reflectivity measurement setup consists of a Tungsten lamp light source, a fiber coupler, and an optical spectrum analyzer, as shown in Figure 2-17. Light from the Tungsten lamp is coupled into the multimode optical fiber. The multimode optical fiber approaches the sample surface vertically. The reflected light is collected by the same fiber and measured by optical spectrum analyzer. A metal mirror which has a near unity reflectivity spectrum is measured as the reference. The reflectivity of the measured sample is calibrated with respect to the metal mirror. During the course of current research, a commercial multilayer thin film coatings design software, TFCalc [28] was used. This software can simulate the performance of the multiple layer films. It can also optimize the layer structure to get the best fit of simulation result to a designated target. All the growth rate calibration growths were done in our GaAs MOCVD system. The growth temperature was 730 ºC. TMAl, TMGa and Arsine were the sources for GaAs, AlAs and Al x Ga 1-x As in which x is smaller than 95%. Al 0.95 GaAs was grown with TEG which has a lower vapor pressure, in order to maintain a stable but small gallium source flow. 30 All the calibration runs were done on quarters of 2 inch GaAs (100) wafers. Most of the VCSEL devices were grown on full 2 inch GaAs (100) wafers. Growth time for each layer on full wafer was increased by 4% to compensate the growth rate difference between the quarter wafer and full wafer in our reactor. VCSELs have very complicated layer structures, in order to have correct material composition and layer thickness, growth rate calibration starts from the simple structure consisting of the high refractive index material and the low refractive index material; afterwards the grading, the active region and the oxide layer are included and calibrated respectively. Normally four structures were grown in the whole sequence of growth rate calibration. 0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5 1.0 -4 -2 0 2 4 6 8 10 12 14 16 18 20 0.0 0.5 1.0 Oxide Aluminum Composition Position (QWOT) Cavity Grading Al(0.95)GaAs/Al(0.15)GaAs Substrate --> Epitaxial Direction Figure 2-18 Structures of Growth Rate Calibration Samples Schematic drawings of four growth rate calibration samples are shown in 31 Figure 2-18. The x axis is the location in the unit of quarter wavelength optical thickness (QWOT). The GaAs substrate is shown at x<0. The epilayer structure starts from x=0 point. Only limited pairs of DBRs are shown in the graph. The y axis shows the aluminum composition. These four structures will be discussed next with the real growth calibration runs as examples. The first sample (run # 5605) consisted of two 13 pairs ¼ λ Al 0.95 GaAs- Al 0.15 GaAs DBR and a λ long Al 0.15 GaAs Cavity. After the growth, the reflectivity spectrum was measured and used as optimization target in the TFCalc simulation. 700 800 900 1000 0 20 40 60 80 100 Simulation Experiment Reflectivity (%) Wavelength (nm) Figure 2-19 Reflectivity of Al 0.95 GaAs/Al 0.15 GaAs Growth Rate Calibration Sample All Al 0.15 GaAs layers were defined as one group while all Al 0.95 GaAs layers were defined as another group in the TFCalc. Thicknesses of the layers in the same group were varied simultaneously to get the best fit between the simulation and experiment. Simulation showed that Al 0.15 GaAs growth rate was 1.0126 times of the expectation and Al 0.95 GaAs growth rate was 0.9861 times of the expectation. 32 Simulation result is shown in Figure 2-19. The growth rates obtained in this growth would be used in the following calibrations. During the TFCalc simulation, both optical thickness and material refractive indices need to be correct in order to match the simulation with the experiment. The positions of the stop band and transmission peak are determined by the optical thickness. The shapes of the stop band and satellite peaks are influenced by the refractive index differences of DBR materials. The second sample (run #5607) included 20nm linear grading layers. The purpose of this growth was to determine the optical thickness of the grading layer. Grading was realized by linearly varying the TMAl and TMGa MFCs during the growth. The middle of the grading should be located in the quarter wavelength point as shown in Figure 2-18. The thicknesses of the Al 0.15 GaAs and Al 0.95 GaAs were decreased accordingly to maintain the quarter wavelength periodicity. The DBR pairs are increased to 16 in order to compensate the reduction of DBR reflectivity. 700 800 900 1000 0 20 40 60 80 100 Simulation Experiment Reflectivity (%) Wavelength (nm) Figure 2-20 Reflectivity of Grading Growth Rate Calibration Sample 33 The measured reflectivity spectrum and simulated curve are shown in Figure 2-20. During the TFCalc simulation, Al 0.15 GaAs and Al 0.95 GaAs layer thicknesses were varied together as a group. Thickness of grading layers was varied as another group. The data fitting showed that the layer thicknesses of Al 0.15 GaAs and Al 0.95 GaAs were exactly same as expected and the grading layer thickness was 1.06 times of expectation. The grading layer is simulated in the TFCalc as a layer with the aluminum composition of 55% which is the average aluminum composition in the grading layer. According to the simulation, 20nm grading should correspond to 0.3042 QWOT. However, the experimental data from real growth indicated a layer thickness of 0.3246 QWOT. In the following growths, instead of trying to adjust the grading layer thickness, the thickness of Al 0.15 GaAs and Al 0.95 GaAs layers in DBR was adjusted to 0.6754 QWOT to maintain the DBR periodicity. The third sample (run #5608) was the cavity length calibration. Al 0.3 GaAs waveguide layers and 3 GaAs QWs were grown in the cavity instead of Al 0.15 GaAs used in the last two calibration growths. With the knowledge of Al 0.15 GaAs, Al 0.95 GaAs and grading layers’ thicknesses and growth rates, only variable in the new growth was the thickness of active region. In the simulation, the thickness of the active region was 1.01 times of the expectation. The last sample (run #5609) was the oxide layer calibration. 20nm AlAs was inserted in between the first pair of Al 0.95 GaAs/Al 0.15 GaAs after the active region. There were grading layers surrounding the AlAs oxide layer in order to form a tapered oxide. The middle of the AlAs layer was set at the quarter wavelength point 34 as shown in Figure 2-18 to ensure the lowest optical field. Simulated optical field distribution near the VCSEL active region is shown in Figure 2-21. The QWs are located in the peak of optical field while the oxide aperture is sitting in the node of optical field. Figure 2-21 Simulated Optical Filed Distribution near VCSEL Active Region The reflectivity of the top and the bottom DBRs were calculated with the layers thicknesses and the refractive indices determined by the calibration experiments. DBR pairs were adjusted to reach a reflectivity of 99.99% in the bottom DBR and a 99.4% in the top DBR. The designed epi structure of this VCSEL is shown in Table 2-5. 35 Layer No. Iterations Material Thickness (nm) Thickness (QWOT) Doping (cm -3 ) 38 Contact GaAs 9.49 0.1623 P 5E19 37 Al 0.15 GaAs 40.88 0.6754 P 1E18 36 Grading 21.20 0.3246 P 1E18 35 Al 0.95 GaAs 47.53 0.6754 P 1E18 34 Grading 21.20 0.3246 P 1E18 33 Al 0.15 GaAs 40.88 0.6754 P 1E18 32 Grading 21.20 0.3246 P 1E18 31 X 14 High P Doping DBR Al 0.95 GaAs 47.53 0.6754 P 1E18 30 Grading 21.20 0.3246 P 5E17 29 Al 0.15 GaAs 40.88 0.6754 P 5E17 28 Grading 21.20 0.3246 P 5E17 27 X 4 Low P Doping DBR Al 0.95 GaAs 47.53 0.6754 P 5E17 26 Grading 21.20 0.3246 P 5E17 25 Al 0.15 GaAs 23.30 0.3850 P 5E17 24 Grading 21.20 0.3246 P 5E17 23 AlAs 20.0 0.2811 P 5E17 22 Grading 10.0 0.1476 P 5E17 21 Al0.6GaAs 10.0 0.1522 P 5E17 20 First P DBR and Oxide Layer Al0.95GaAs 27.09 0.3849 P 5E17 19 Grading 15.25 0.2320 - 18 Al 0.3 GaAs 100.0 1.6002 - 17 GaAs 8.5 0.1454 - 16 Al 0.3 GaAs 7.0 0.1120 - 15 GaAs 8.5 0.1454 - 14 Al 0.3 GaAs 7.0 0.1120 - 13 GaAs 8.5 0.1454 - 12 Al 0.3 GaAs 100.0 1.6002 - 11 Active Region Grading 15.25 0.2320 - 10 Al 0.95 GaAs 47.53 0.6754 N 1E18 9 Grading 21.20 0.3246 N 1E18 8 Al 0.15 GaAs 40.88 0.6754 N 1E18 7 X 36 DBR Grading 21.20 0.3246 N 1E18 6 Al 0.95 GaAs 47.53 0.6754 N 1E18 5 Grading 21.20 0.3246 N 1E18 4 Al 0.15 GaAs 40.88 0.6754 N 1E18 3 GaAs 9.49 0.1623 N 1E18 2 Etch Stop Al 0.8 GaAs 204.56 3 N 1E18 1 Buffer GaAs 300 - N 1E18 Substrate GaAs N+ (100) Table 2-5 Epi Structure of 850 nm Top Emitting VCSEL (#5610) The real VCSEL structure (run #5610) was grown after all these calibration runs. The expected sample reflectivity and experimental data closely match with each other as shown in Figure 2-22. 36 700 800 900 1000 0 20 40 60 80 100 Simulation Experiment Reflectivity (%) Wavelength (nm) Figure 2-22 Reflectivity of VCSEL structure 2.3.2 QW Calibration A typical GaAs QW calibration sample structure is shown in Table 2-6. As discussed earlier, in order to match the QW gain peak and the cavity resonant wavelength at VCSEL’s working condition, the QW peak wavelength is set at 845nm. This calibration sample has the same active region as the VCSEL. AlAs and Al 0.8 GaAs are used for the purpose of carrier confinement. This is also a standard growth test of the MOCVD reactor. The PL intensity and FWHM are very sensitive to the condition of reactor. PL result from such a sample is shown in Figure 2-23. 37 Layer No. Material Thickness (nm) 12 GaAs 20 11 Al 0.8 GaAs 100 10 Al 0.3 GaAs 100 9 GaAs 8.5 8 Al 0.3 GaAs 7.0 7 GaAs 8.5 6 Al 0.3 GaAs 7.0 5 GaAs 8.5 4 Al 0.3 GaAs 100 3 Al 0.8 GaAs 500 2 AlAs 100 1 GaAs 300 Substrate GaAs N+ (100) Table 2-6 GaAs 845nm QW Calibration Sample Epi Structure 780 800 820 840 860 880 900 M558 Peak Wavelength=845.2nm FWHM=15.8nm Intensity (a.u.) Wavelength (nm) Figure 2-23 PL Data of a Typical GaAs QW Sample 2.3.3 Doping Calibration P type DBR was realized by Zn doping and n type DBR was realized by Si doping. Details of doping calibration are introduced in chapter 6. Also shown in that chapter was the technique of forming a 10nm thick highly doped GaAs in order to form a decent ohmic metal contact. 38 Zn doping has the problem of diffusion which makes modulate doping unrealistic. Carbon is a perfect alternative p doping source which has a low diffusion coefficient. Carbon doping using CBr4 in GaAs-AlAs material system grown by MOCVD was realized in our lab. Detailed research work is also discussed in chapter 6. All the VCSELs discussed in this chapter have Zn doped P DBR. C coped DBR was not realized before the end of this project. 2.4 VCSEL Devices Performance The VCSEL devices processing and characterization were done by my colleagues Zhijian Wei and Ryan Stevenson. Fabrication details of these VCSELs were discussed in their thesis [22, 4]. 2.4.1 Wafer Bonded 850nm Bottom-Emitting VCSEL For the purpose shown in Figure 2-2, wafer bonded bottom-emitting VCSEL arrays are required. The VCSEL epi structure was wafer bonded on to sapphire due to its good mechanical strength, low electrical conductivity, high thermal conductivity and low absorption coefficient at 850 nm. The original GaAs substrate was then removed by mechanical lapping and wet chemical etching. Standard VCSEL processing was used to form the bottom emitting VCSEL, the light was emitted from the substrate side. A schematic drawing of the wafer bonded 850 nm bottom-emitting VCSEL is shown in Figure 2-24. The epi-structure was gown with p side up, so the p-DBR became the bottom DBR after wafer bonding. The VCSEL array was designed to be flip-chip bonded onto Si driving circuitry. 39 N-DBR Active Region N Metal P Metal P-DBR Sapphire Oxide Aperture N-DBR Active Region N Metal P Metal P-DBR Sapphire Oxide Aperture Figure 2-24 Schematic Drawing of Bottom Emitting VCSEL structure, not to scale 01 2 3 45 6 78 9 10 0 1 2 3/13/2000 Wei USC 8x8 μm 2 10x10 μm 2 14x14 μm 2 Power (mW) Current (mA) Figure 2-25 Bottom Emitting VCSEL LI Curves and Far Field Picture Typical bottom emitting VCSEL LI curves with different aperture sizes are shown in the Figure 2-25. The far field image of the fundamental mode is also shown. The interference fringes in the far field picture and the ripples in the LI curves are due to that the sapphire substrate behaves as an external cavity which can cause the interference. For detailed discussion, please see Wei’s thesis [22]. 40 23 4 567 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 Threshold Current ( μA) Aperture Size ( μm) Series Resistance ( Ω) 234 567 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 External Efficiency Aperture Size ( μm) Figure 2-26 Bottom Emitting VCSEL External Efficiency, Threshold Current, and Series Resistance as Functions of Aperture Size The bottom emitting VCSEL external efficiency, threshold current, and series resistance as functions of oxide aperture size are shown in Figure 2-26. As discussed earlier in this chapter, the external efficiency doesn’t change with the aperture size since the oxide aperture is located at the minimum of optical field. The threshold current decreases as the aperture size decreases due to the minimized volume of active region. At the same time, the device resistance increases since the oxide aperture confines the current path. 41 Figure 2-27 Bottom Emitting VCSEL Array Performance, Original Design The performance of bottom emitting VCSEL array is shown in Figure 2-27. This array has the original epi structure design. All 400 devices are working in this array, but the uniformity of the device performance is not perfect. All data curves are shown in the left column. Distributions of the threshold current, external efficiency and serials resistance over the whole array are shown in the middle column. Their statistic results are shown in the right column. It is shown that the threshold current and series resistance spatial distribution have the similar pattern. This pattern was also observed as the etching non-uniformity during the mesa etching. The etching should stop at a low Al composition layer for better ohmic contacts. However the etching non-uniformity exposes different layers of p DBR for p contact formation. High contact resistance is expected when the p contact is formed on high Al 42 composition layer. In an improved design, A 7/4 λ thick Al 0.15 GaAs layer in the p DBR was used for the contact layer. • Ave. I th = 491 μA Std. = 116 μA •Ave. QE= 42.2 % Std. = 2.6 % Figure 2-28 Bottom Emitting VCSEL Array Performance, Improved Design The performance of bottom emitting VCSEL array is shown in Figure 2-28. Limited improvement of array uniformity is observed. Wei noticed later that the uniformity of oxide aperture is the key issue in archive the array performance uniformity. After improving the oxide aperture uniformity by revising the oxidation furnace, a much improved uniformity was achieved, as shown in Figure 2-29. 43 Figure 2-29 Bottom Emitting VCSEL Array Performance, Improved Oxide Technology A much lower series resistance was realized by minimize the mesa size and better control of doping in DBR. The wafer used in this experiment was obtained from Prowtech Inc. [22] 2.4.2 850 nm Top-Emitting VCSEL Top emitting 850 nm VCSELs were also fabricated. The schematic drawing of such a VCSEL is shown in Figure 2-30. SEM image of a top emitting VCSEL structure which has the SiNx as the isolation layer is shown in Figure 2-31. 44 P-DBR Active Region P Metal N Metal N-DBR Substrate Oxide Aperture P-DBR Active Region P Metal N Metal N-DBR Substrate Oxide Aperture Figure 2-30 Schematic Drawing of Top Emitting VCSEL Structure, not to scale Figure 2-31 SEM image of Top Emitting VCSEL, SiNx Isolation 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3 4 5 6 7 8 9 844 846 848 850 852 854 856 0.0 0.5 1.0 1.5 2.0 2.5 Top-Emitting VCSEL: LI and IV Data Voltage (V) Current (mA) 0.0 0.1 0.2 0.3 0.4 0.5 Output Power (mW) λ o =850.6nm Lasing Spectrum Figure 2-32 LI, IV Data and Lasing Spectrum for a 2 μm-Aperture Top-Emitting VCSEL 45 The DC characteristic of the top emitting VCSEL is similar to the bottom emitting VCSEL. Data of a typical top emitting VCSEL is shown in Figure 2-32. The high speed performance of SiNx isolated VCSEL is not good due to the large capacitance formed between the contact pad and semiconductor. In order to improve the high speed performance of VCSEL, polyimide planarization and ion implantation isolation techniques were developed in our lab and utilized in the VCSEL process. Small signal response of an ion implantation isolated top emitting VCSEL is shown in Figure 2-33. The measured device has a 6 μm oxide aperture and a 750 µA threshold current. The maximum 3dB bandwidth is 5.1 GHz. High serial resistance is the main obstacle to high speed in this device. [4] 12 34 56 78 -85 -80 -75 -70 -65 -60 3.2mA 1.1mA f 3dB =1.7-5.1 GHz -3dB VCSEL Frequency Response: Implant Isolated--w=6 μm 20*Log 10 |S 21 | (dB) Frequency (GHz) Figure 2-33 Small Signal Responses for a 6µm-Aperture Top-Emitting VCSEL, Implant Isolated 2.5 Summary Basic VCSEL design was discussed first in this chapter, which includes 46 active region design, DBR design and oxide aperture design etc. Detailed calibration processes for VCSEL growth were introduced. Main focus was on four growth rate calibration runs which calibrated the high/low index material, grading, cavity and oxide aperture respectively. Uniform 20 by 20 wafer-bonded bottom-emitting VCSEL arrays were realized. VCSELs with sub mA threshold and high external efficiency (40% to 60%) were routinely manufactured in our group. High speed top emitting VCSELs were also developed with the maximum 3dB bandwidth of 5.1 GHz. High series resistance is the major obstacle in our VCSEL research. References 1 Kenichi Iga, “Surface-Emitting Laser – Its Birth and Generation of New Optoelectronics Field,” IEEE J. Selected Topics in Quantum Electron., Vol. 6, No. 6, pp. 1201-1215, 2000. 2 Govind P. Agrawal, “Semiconductor Lasers: Past, Present, and Future,” Chapter 5, American Institute of Physics, 1995. 3 Carl Wilmsen, Henryk Temkin and Larry A. Coldren, “Versitcal-Cavity Surface- Emitting Lasers,” Cambridge University Press, 1999. 4 Ryan Stevenson, “High Performance Components of Free-Space Optical and Fiber-Optic Communications Systems,” Ph.D. thesis dissertation, 2005. 5 N. M. Margalit, D. I. Babic, K. Streubel, R. P. Mirin, R. L. Naone, J. E. Bowers, and E. L. Hu, “Submiliamp long wavelength vertical cavity lasers,” Electron. Lett., Vol.32, No. 18, pp. 1675-1677 1996. 6 V. Jayaraman, J. C. Geske, M. H. MacDougal, F. H. Peters, T. D. Lowes, and T. T. Char, “Uniform threshold current, coutinuous-wave, single mode 1300nm vertical cavity lasers from 0 to 70 °C,” Electron. Lett., Vol. 34, No. 14, pp. 1405-1407, 1998. 7 T. Baba, Y. Yogo, K. Suzuki, F. Koyama and K. Iga, “Near room temperature continuous wave lasing characteristics of GaInAsP/InP surface emitting laser,” Electron. Lett., Vol. 29, No. 10, pp. 913-914, 1993. 47 8 W. W. Chow, K. D. Choquette, M. H. Crawford, K. L. Lear, and G. R. Hadley, “Design, Fabrication, and Performance of Infrared and Visible Vertical-Cavity Surface-Emitting Lasers,” IEEE J. Quantum. Electron., Vol. 33, No. 10, pp.1810- 1824, 1997. 9 Haruhisa Soda, Ken-ichi Iga, Chiyuki Kitahara and Yasuharu Suematsu, "GaInAsP/InP Surface Emitting Injection Lasers,” Japanese Journal of Applied Physics, vol. 18, no. 12, p. 2329–2330, 1979. 10 F. Koyama, S. Kinoshita, and K. Iga, “Room-temperature continuous wave lasing characteristics of GaAs vertical cavity surface-emitting laser,” Appl. Phys. Lett., Vol. 55, No. 3, pp. 221-222, 1989. 11 J. M. Dallesasse, N. Holonyak, Jr., A. R. Sugg, T. A. Richard, and N. El-Zein, “Hydrolyzation oxidation of AlGaAs-AlAs-GaAs quantum well heterostrcutres and superlattices,” Appl. Phys. Lett., Vol. 57, No. 26, pp. 2844-2846, 1990. 12 R. Jaeger, M. Grabherr, C. Jung, R. Michalzik, G. Reiner, B. Weigl, and K. J. Ebeling, “57% wallplug efficiency oxide-confined 850 nm wavelength GaAs VCSEL’s,” Electron. Lett., Vol. 33, No. 4, pp. 330-331, 1997. 13 G. M. Yang, M. H. MacDougal, and P. D. Dapkus, “Ultralow threshold current vertical-cavity surface-emitting lasers obtained with selective oxidation,” Electron. Lett., Vol. 31, No. 11, pp. 886-888, 1995. 14 D. G. Deppe, D. L. Huffaker, J. Shin, and Q. Deng, “Very-low-threshold index- confined planar microcavity lasers,” IEEE Photon. Technol. Lett., Vol. 7, No. 9, pp. 965-967, 1995. 15 K. L. Lear, K. D. Mar, K. D. Choquette, S. P. Kilcoyne, R. P. Schneider, Jr., and K. M. Geib, “High-frequency modulation of oxide-confined vertical cavity surface emitting lasers,” Electron. Lett., Vol. 32, No. 5, pp. 457-458, 1996. 16 S. L. Chuang, “Physics of Optoelectronic Devices,” John Wiley and Sons, 1995. 17 R. S. Geels, B.J. Thibeault, S.W. Corzine, J.W. Scott, and L.A. Coldren, “Design and characterization of In 0.2 Ga 0.8 As MQW vertical-cavity surface-emitting lasers,” IEEE J. Quantum Electron., Vol. 29, No. 12, pp. 2977-2987, 1993. 18 J. J. Dudley, D. L. Crawford, and J. E. Bowers, “Temperature Dependence of the Properties of DBR Mirrors Used in Surface Normal Optoelectronic Devices,” IEEE Photon. Technol. Lett., Vol.4, No. 4, pp. 311-314, 1992. 19 Private conversation with Sang-Wan Ryu. 48 20 L. A. Coldren and S. W. Corzine, “Diode Lasers and Photonic Integrated Circuits,” John Wiley & Sons, Inc. 1995. 21 LASTIP manual, Crosslight software Inc. 22 Zhijian Wei, “Semiconductor Surface Emitting Laser, Laser Array and Polarization Control,” Ph.D. thesis dissertation, 2006. 23 G. Ronald Hadley, “Effective index model for vertical-cavity surface-emitting lasers,” Optics Letters, Vol. 20, No. 13, pp. 1483-1485, 1995. 24 A. E. Bond, P. D. Dapkus, and J. D. O’Brien, “Aperture placement effects in oxide-defined vertical-cavity surface-emitting lasers,” IEEE Photo. Technol. Lett. Vol. 10, pp.1362-1364, 1998. 25 A. E. Bond, P. D. Dapkus, and J. D. O’Brien, “Aperture dependent loss analysis in oxide-defined vertical-cavity surface-emitting lasers,” IEEE Photo. Technol. Lett., Vol. 11, No. 4 pp.397-399, 1999. 26 Eric R. Hegblom, Dubravko I. Babic, Brian J. Thibeault, and Larry A. Coldren, “Sacttering Losses from Dielectric Apertures in Vertical-Cavity Lasers,” J. Selected Topics in Quantum Electron. Vol. 3, No. 2, pp. 379-389, 1997. 27 Kent D. Choquette, K. M. Geib, H. C. Chui, B. E. Hammons, H. Q. Hou, T. J. Drummond, and Robert Hull, “Selective oxidation of buried AlGaAs versus AlAs layers,” Appl. Phys. Lett. Vol. 69, No. 10, pp. 1385-1387, 1996. 28 Software Spectra, Inc., Portland, Oregon. 49 Chapter 3 : Super High Efficiency Laser Diodes Pursuing high efficiency is always a goal of Engineering. In the early stage of semiconductor laser research, the main goal was to decrease the threshold current. The next stage was to increase the maximum power output. Recently, achieving super high power efficiency becomes the new research highlight. It is common nowadays for a diode laser to have an overall power conversion efficiency of 50%, this is very high compared with gas and solid-state lasers whose efficiencies are in the range of 1% ~10%. The main cause is that laser diodes can be directly pumped by an electrical current. [1] High efficiency is especially important in high power diode laser. Due to the nature of high power diode lasers, all the power which is not converted into light in the semiconductor laser generates heat, and then this amount of heat must be extracted from the device to maintain the laser performance. Every portion of power wasted in the laser diode requires more power to extract the generated heat. Any wasted power in the laser diode will significantly decrease the efficiency of the total laser system, which includes both the semiconductor laser and its cooler. If we assume that the cooling power is 1.5 times of the wasted power in the laser diode, we can calculate the relationship between the laser system power conversion efficiency and the laser diode power conversion efficiency (shown in Figure 3-1). According to this graph, if we can increase the laser diode power conversion efficiency from 40% to 80%, the system power conversion efficiency can increase from 21% to 62%. 50 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 System Power Conversion Efficiency Laser Diode Power Conversion Efficiency Figure 3-1 Laser System Power Conversion Efficiency Super High Efficiency Diode Sources (SHEDS) program initiated by the Defense Advanced Research Projects Agency (DARPA) is the most ambitious program in this research area. [2,3] In this program, 80% wall plug efficiency in the generation of light from stacks of semiconductor diode laser bars is set as the program goal. In order to increase the efficiency, we need to have a clear understanding of all the losses and try to eliminate them. We believe the excess voltage drop is a key issue causing the low power conversion efficiency and it is possible to significantly improve the efficiency by lowering the driving voltage. In this chapter, I will first analyze the losses in edge emitting lasers. Then I will focus on the excess voltage drop and discuss two generous ideas of our group in this project. 3.1 Excess Voltage Drop in Edge Emitting Lasers At the beginning of SHEDS program, commercial semiconductor edge emitting lasers have the wall plug efficiency or power conversion efficiency around 51 50%. [3] In order to get 80% efficiency, we must first sort out the sources of losses and tackle them down respectively. 3.1.1 Sources of Losses The semiconductor Edge Emitting Laser (EEL) is a device which converts DC current into light output. The efficiency of an EEL is defined as the following: in out P P Efficiency Plug Wall = _ _ (3.1) P out is the light output power in the lasing mode and P in is the input electrical power. ) ( ) ( 0 0 V V I I V I P P P waste out in Δ + × Δ + = × = + = (3.2) 0 0 V I P out × = (3.3) I V I V V I P waste Δ × Δ + Δ × + Δ × = 0 0 (3.4) In the above equations, I and V are current and voltage applied onto the device respectively. 0 I counts for all the carriers which are converted into emitting photons in the lasing mode. 0 V corresponds to the photon energy ν h in the equation below: e h V / 0 ν = (3.5) where e is the elementary charge 1.60×10 -19 coulombs. I will call V Δ and I Δ the excess voltage and the excess current respectively in the following text. In the equations above, the loss of the device is described with the device 52 characteristic parameters. For the convenience of analysis, we can also split the right hand terms of (3.1) as in the following equation: energy electron number electron energy photon number photon P P in out _ _ _ _ × × = (3.6) In this equation, photon_number and photon_energy refer to the number and energy of photons emitted from the EEL at the lasing mode. The electron_number is the number of electrons injected into the EEL. The electron_energy is the energy of electrons which can be written as the charge of the electrons e times the applied voltage V. Complicated physical models are available [1] to describe the details of semiconductor laser operation mechanism. I will briefly discuss the mechanism of semiconductor laser diodes, and introduce the sources of losses at every stage. First, electrons are injected into the device through the metal contacts. Most of these electrons will be injected into the active region. The key parameter in this step is the quantum efficiency. Some of the electrons are lost due to current leakage. The energy of the electrons also decreases in this step. The contact resistance, bulk resistance of semiconductor and the resistance due to heterojunction consume the energy of the electrons. Second, photons are generated in the active region from the injected electrons. Some electrons are lost due to nonradiative recombination. And not all the photons generated are at the lasing mode. Spontaneous emission generates photons into other modes. 53 Third, not all the photons generated in lasing mode in the active region can be emitted from the end facet. There are optical losses, including free carrier absorption, interface scattering, facet scattering, etc. As stated above, the operation mechanism of semiconductor lasers is very complicate. It is difficult to gain the full picture and optimize the design just from the measured parameters of real devices. Therefore the comprehensive simulation tool is needed to assist the optimization process. 3.1.2 Simulation Tool During the research of SHEDS program, a commercial simulation software, LASTIP, is intensively used. LASTIP, Laser Technology Integrated Program, a product of Crosslight Software Inc., is primarily designed for 2D simulation of EEL diodes. Given the structural and material properties, LASTIP solves the following four basic equations under continuous wave or transient conditions. First, Poisson’s equation describes the potential charge relations. Second, the time-dependant electron and hole current continuity equations govern the carrier flux-recombination relation. Third, the complex wave equation finds the optical field distribution in the transverse and lateral directions. Fourth, the time-dependent photon rate equation calculates the optical power with modal gain and recombination. The finite element method is used to solve the basic differential equations. Complex physical models are implemented in LASTIP to obtain accurate 54 simulation data. The physical models include, but are not limited to, the following: quantum well subbands are solved to compute the carrier concentrations and optical gain; k.p theory is used to treat strained quantum wells; the optical gain function is computed starting from material parameters; Fermi statistics are used for accurate computation of carrier concentrations; a non-linear gain suppression model is included; a large number of material models have been implemented. [4] Comprehensive built-in physical models provide the program users with a quantitative insight into various aspects of semiconductor laser. With the understanding of the physical background behind these models, which is essential for users to successfully control the simulation processes and to interpolate the results, the users can concentrate on device optimization and design while leaving all the numerical modeling work to the computers. 3.1.3 Voltage Defects A InGaAs/AlGaAs 940nm EEL epi structure is listed in Table 3-1. this structure will be referred as structure “A” in the following text. In this structure, Al 0.6 GaAs is used as the cladding layer and Al 0.3 GaAs is used as the waveguide layer. This structure will be used as a starting point for the following optimization. 55 Layer No. Material Thickness (µm) Doping (cm -3 ) 9 GaAs 0.3 P 1E18 8 Grading 0.1 P 1E18 7 Al 0.6 GaAs 0.8 P 1E18 6 Al 0.3 GaAs 0.65 undoped 5 In 0.13 GaAs 0.012 undoped 4 Al 0.3 GaAs 0.65 undoped 3 Al 0.6 GaAs 0.8 N 1E18 2 Grading 0.1 N 1E18 1 GaAs 1 N 1E18 0 GaAs N+ (100) substrate Table 3-1 InGaAs/AlGaAs 940nm EEL Epi Structure “A” The following conditions are used in this simulation. The EEL width is 5 µm and its length is 1 mm. The front mirror reflectivity is 10% and the back mirror reflectivity is 90%. The background scattering loss is set at 1 cm -1 . The driving current is varying from 0 to 150 mA. The current density is comparable to a commercial high power EEL which is about 100 ~150 µm wide and is driven by the current of 3 to 5 A. In the simulation, the light is forced to be distributed uniformly in the lateral direction to correctly mimic a broad area EEL. It is also considering several ideal conditions which are actually engineering challenges in the device manufacturing: perfect ohmic contacts on both sides; perfect facet on both ends; no catastrophic optical damage; uniform and stable temperature distribution; et al. 56 The LASTIP simulated results are shown in Figure 3-2 and Figure 3-3. 0 20406080 100 120 140 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 0 20 40 60 80 100 120 140 160 180 200 Voltage (V) Current (mA) Laser Power (mW) Figure 3-2 Simulated VI and LI Curves for Structure “A” 0 20 40 60 80 100 120 140 0.0 0.2 0.4 0.6 0.8 1.0 Wall Plug Efficiency Current(mA) Figure 3-3 Simulated Wall Plug Efficiency for Structure “A” The wall plug efficiency simulated in this structure reaches 64%. I will focus on analyzing the laser at 100 mA in the following simulations unless a different condition is stated. Notations used in this analysis are same as those used in equation 3.2, 3.3 and 3.4. When driving current (I) is 100 mA, the driving voltage (V) is 1.790 Volt. Laser power consumption (P in ) is 179 mW and the output power (P out ) is 115 mW. The Lasing wavelength is 0.939 µm, the photon energy at this wavelength is 57 1.321eV, so the V 0 is 1.321 Volt. The “useful” driving current (I 0 ) is 87mA, which is enough to generate 115mW light output if all these carriers are converted to photons in lasing mode and emitted through the facet. The threshold current is 5.5 mA. After the laser is turned-on, dL/dI = 1.2194 V. The efficiency of converting carriers to “useful” photons above threshold is 1.2194/1.321 = 92.3%. The laser power input is split according to equation 3.2, 3.3 and 3.4. As shown in the pie graph (Figure 3-4) below, excess voltage counts for about 3/4 of the wasted power. (I 0 + ΔI) ∗(V 0 + ΔV) 100% Power Input ΔV ∗I 0 22.8% ΔI ∗ΔV 3.4% ΔI ∗V 0 9.6% I 0 ∗V 0 64.2% Laser output Figure 3-4 Loss Distribution in Structure “A” In order to eliminate the large loss which comes from the excess voltage, it is necessary to carefully evaluate the voltage drop distributions in the whole device. Voltage or Voltage Drop is a term loosely used for the potential difference between two specified points in a circuit or device [5]. The accurate expression of potential difference for a charged particle between two points is the product of the voltage drop and the particle’s charge. Although we use the term “Voltage Drop” through out this chapter, it should be more precisely expressed as “potential change”. 58 However, it is more convenient to use the term “Voltage Drop” which is linked with the driving voltage, a real device parameter. The information of potential distribution can be obtained from the quasi- Fermi level distribution in semiconductor devices. The electron quasi-Fermi level is also called in thermodynamics the electro-chemical potential. This is the average energy per electron which would be subtracted from the total energy of the electron system if we removed an electron. Similar definition also applies to the hole quasi- Fermi level. The spatial distribution of quasi-Fermi level reflects the potential of carriers everywhere in the device. The gradient of quasi-Fermi level is the driving force of carriers. By following the spatial variation of quasi-Fermi level, the potential drop distribution can be calculated. The separation of electron and hole quasi-Fermi levels comes from the non- equilibrium carrier distribution caused by the applied bias. In the EEL laser, electrons are injected from the n side and holes are injected from the p side. They recombine in the QW. We need to follow the electron quasi-Fermi level in the n side of the device and follow the hole quasi-Fermi level in the p side of the device. The potential change in the QW is the electron and hole quasi-Fermi level separation in the QW. Other than the portion which is corresponding to the photon energy, all the other potential drops should be treated as wasted which we should try to eliminate. 59 0.01.0 2.03.0 4.0 0.0 0.5 1.0 1.5 2.0 P GaAs Cap QW Wave- Guide N GaAs Buffer Wave- Guide P type Cladding N type Cladding Conduction & Valence Band Structure Quasi-Fermi Level for Electrons Quasi_Fermi Level for Holes Potential (eV) Position ( μm) Figure 3-5 Band structures & Quasi-Fermi Level distributions Figure 3-5 shows the band structures and quasi-Fermi level distributions simulated at 100 mA current injection condition. The conduction and valence band structures are shown with solid lines. The electron quasi-Fermi level is shown with dashed line and the hole quasi-Fermi level is shown with dotted line. By following the quasi-Fermi level changes, we can calculate the voltage drop on each layer and interface. 60 11~3 33~4 44~5 55~6 6 6~7 77~9 9 0.00 0.05 0.10 0.15 0.20 Excess Voltage Drop (V) Layers & Interfaces Figure 3-6 Excess Voltage Drop Distributions at Each Layer and Interface The calculated excess voltage drops from each layer and interface are shown in Figure 3-6. The bars labeled by single number index show voltage drop in the bulk layers. The layer numbers are listed in Table 3-1. The layer 4, for example, corresponds to the voltage drop in the n side Al 0.3 GaAs waveguide layer. The bars labeled by double number index indicate voltage drop across the interfaces. The bar marked by 6~7, for example, corresponds to the voltage drop from Al 0.6 GaAs cladding layer to Al 0.3 GaAs waveguide layer in the p side. All the voltage drops are calculated from the quasi-Fermi level changes across the bulk layers or interfaces except for layer 5, which is the QW. In this layer, the voltage drop is calculated by subtract the voltage corresponding to the photon energy from the difference between electron and hole quasi-Fermi levels in QW. Indicated in Figure 3-6, most of the excess voltage comes from the following parts. First is the bulk resistance of waveguide layers. Second is the voltage drops 61 across the QW-waveguide interfaces. Third is the extra split of quasi-Fermi levels. 3.2 Reduction of Excess Voltage in Symmetric Structures Layer No. Structure “A” Structure “B” Structure “C” Structure “D” 9 GaAs GaAs GaAs GaAs 8 Grading Grading Grading Grading 7 Al 0.6 GaAs Al 0.4 GaAs Al 0.4 GaAs Al 0.4 GaAs 6 Al 0.3 GaAs Al 0.3 GaAs Al 0.2 GaAs GaAs 5 In 0.13 GaAs In 0.13 GaAs In 0.13 GaAs In 0.13 GaAs 4 Al 0.3 GaAs Al 0.3 GaAs Al 0.2 GaAs GaAs 3 Al 0.6 GaAs Al 0.4 GaAs Al 0.4 GaAs Al 0.4 GaAs 2 Grading Grading Grading Grading 1 GaAs GaAs GaAs GaAs Table 3-2 Four Different InGaAs/AlGaAs 940nm EEL Epi Structures Four similar InGaAs/AlGaAs EEL structures are listed in Table 3-2. Only the Al composition of cladding layers and waveguide layers were varied in these structures. LASTIP simulations were done for all four structures. The effects of these different materials on the excess voltage drop and laser efficiency will be discussed. Let’s first focus on structure “A” and structure “B”. They have different Al compositions in the cladding layer. However, they have nearly the same performances, as shown in Figure 3-7. It is impossible to tell the difference between them on these graphs. 62 0 50 100 150 0.0 0.5 1.0 1.5 2.0 0 50 100 150 200 Structure "A" Structure "B" Voltage (Volt) Current (mA) Laser Power (mW) 0 50 100 150 0.0 0.2 0.4 0.6 0.8 1.0 Structure "A" Structure "B" Power Conversion Efficiency Current (mA) Figure 3-7 Simulated LIV Curves and Efficiency Curves of Structure “A” and Structure “B” Structure “A” has Al 0.6 GaAs/Al 0.3 GaAs interfaces. The band discontinuity is larger than those of Al 0.4 GaAs/Al 0.3 GaAs interfaces in structure “B”. So the excess voltages on the waveguide/cladding interfaces should be larger for structure “A” than for structure “B”. However, as shown in Figure 3-8, the voltage drops on these interfaces are very small compared to other layers and interfaces. Small changes in these interfaces do not alter the whole laser performance very much. 11~3 3 3~4 44~5 55~6 6 6~7 77~9 9 0.00 0.05 0.10 0.15 0.20 Voltage Drop (V) Layers and Interfaces Structure "A" Structure "B" Figure 3-8 Voltage Drop Distributions of Structure “A” and Structure “B” 63 0 50 100 150 0.0 0.5 1.0 1.5 2.0 0 50 100 150 200 Structure "B" Structure "C" Voltage (Volt) Current (mA) Laser Power (mW) 0 50 100 150 0.0 0.2 0.4 0.6 0.8 1.0 Structure "B" Structure "C" Power Conversion Efficiency Current (mA) Although lowering the cladding layer aluminum composition doesn’t change much the laser performance according to the simulation, there is a great practical advantage by doing it. Low aluminum composition material is relatively easy to be grown and processed. It is also less prone to oxidization. These features are beneficial to devices with higher yield and longer lifetime. For structures “B” and “C”, they have the same cladding layers but different waveguide layers. Aluminum composition is decreased from 30% in structure “B” to 20% in structure “C”. The simulation results are shown in Figure 3-9. They have identical LI curves but a significant difference in the IV characteristics. The structure “C” has a much lower driving voltage compared to structure “B”. This low voltage also reflects to the much higher power conversion efficiency. At 100 mA driving current, the efficiency is increased from 64.4% to 74.0%. The driving voltage of structure “B” is 1.788 V and that of structure “C” is 1.551 V. Figure 3-9 Simulated LIV Curves and Efficiency Curves of Structure “B” and Structure “C” Both structures have the laser photon energy 1.320eV. So the total excess voltage of structure “B” is 0.468 V and structure “C” is 0.231V. Voltage drop distributions of both structures are shown in Figure 3-10. 64 1 1~3 3 3~4 4 4~5 5 5~6 6 6~7 7 7~9 9 0.00 0.05 0.10 0.15 0.20 Voltage Drop (V) Layers and Interfaces Structure "B" Structure "C" Figure 3-10 Voltage Drop Distributions of Structure “B” and Structure “C” Both structures have the same cap layers, buffer layers and cladding layers. Layers and interfaces related to these layers, 1, 3, 7, 9, 1~3 and 7~9, have the same voltage drop. Band discontinuities from cladding layers to waveguide layers of structure “B” are smaller than those of structure “C”, as shown in the graph, indexes 3~4 and 6~7. For both waveguide layers, layer 4 and layer 6, in structure “C”, there is much less voltage drop than those in structure “B”. The voltage drops on these layers are due to the bulk resistance of the semiconductor. The resistivity of semiconductor is a function of mobility and carrier concentration [6]: () p n e p n μ μ ρ + = 1 (3.7) The electron and hole mobilities of Al 0.3 GaAs are 1606 cm 2 V -1 s -1 and 274 cm 2 V -1 s -1 respectively, which are smaller than those in Al 0.2 GaAs (4053 cm 2 V -1 s -1 and 312 cm 2 V -1 s -1 ). The waveguide layer carrier concentrations in the structure “B” 65 are also lower than those in structure “C”. 1.5 2.0 2.5 3.0 3.5 14 15 16 17 18 19 N Cladding Waveguide Waveguide QW Structure "B" Structure "C" Electron Concentration (log cm -3 ) Position ( μm) 1.5 2.0 2.5 3.0 3.5 14 15 16 17 18 19 P Cladding Waveguide Waveguide QW Hole Concentration (log cm -3 ) Position ( μm) Figure 3-11 Simulated Carrier Concentration in Structure “B” and “C” The small band discontinuity from barrier to QW in structure “C” also contributes to small voltage drops in interfaces 4~5 and 5~6. Both QWs need 35 meV extra quasi-Fermi level splitting to achieve population inversion. Although lowering the waveguide layer band gap is an efficient way to decrease the excess voltage drop, continuously decreasing the aluminum composition would bring new problems as indicated in structure “D”. In this structure, the cladding layers are GaAs which has very small barrier-QW band discontinuity and high mobility. The simulation result of structure “D” is shown in Figure 3-12 together with structure “B” and “C” results. As we expected, the excess voltage drops are even smaller for it. But the threshold is much higher and the slope efficiency is low. Combining all these factors, device structure “D” has low power conversion efficiency. 66 0 50 100 150 0.0 0.5 1.0 1.5 2.0 0 50 100 150 200 Structure "B" Structure "C" Structure "D" Voltage (Volt) Current (mA) Laser Power (mW) 050 100 150 0.0 0.2 0.4 0.6 0.8 1.0 Structure "B" Structure "C" Structure "D" Power Conversion Efficiency Current (mA) Figure 3-12 Simulated LIV Curves and Efficiency Curves of Structures “B”, “C” and “D” Let’s still first analyze the excess voltage drop. The total excess voltage in structure “D” is only 0.158 V. Excess voltage drop distributions for structures “B”, “C” and “D” are compared side by side in Figure 3-13. In structure “D”, the voltage drops at the waveguide-QW interfaces and in the waveguide layers are further decreased. The quasi-Fermi level splitting in the QW is nearly the same as in the structures “B” and “C”. The voltage drops at top cladding layers, buffer layer and their interfaces are also the same. The voltage drops at waveguide-cladding interfaces are actually increased in structure “D”. This is due to extremely high carrier concentration in the waveguide layers. This issue will be discussed with more details later. 67 1 1~3 3 3~4 4 4~5 5 5~6 6 6~7 7 7~9 9 0.00 0.05 0.10 0.15 0.20 Voltage Drops Layers & Interfaces 940 EEL Epi Structures "B" "C" "D" Figure 3-13 Excess Voltage Drop Distributions for 940 EEL Epi Structures. 012 34 0 500 1000 1500 2000 Electron Current Hole Current Current Density (A/m 2 ) Position ( μm) 01234 0 500 1000 1500 2000 Electron Current Hole Current Current Density (A/cm 2 ) Position ( μm) Figure 3-14 Electron and Hole Current in Structure “C” (left) and “D” (right) Electron and hole current in structures “C” and “D” are shown in Figure 3-14. As in structure “C”, most of the electrons and holes injected into the device flow into the QW and get recombined. Only 1 A/cm -2 electron current and 3 A/cm -2 hole current are lost before they reach the QW. The total current density is 2000 A/cm -2 . On the other hand, in structure “D”, more electrons and holes are lost outside the QW. 156 A/cm -2 electron current and 171 A/cm -2 hole current are lost on their 68 way to the QW. This means that 16.3% of the carriers injected into the device do not recombine in the QW. Most of these losses are in the waveguide layers. 01 23 4 0 5 10 15 20 Electrons Holes Carrier Concentration (log cm -3 ) Position ( μm) 01 23 4 0 5 10 15 20 Electrons Holes Carrier Concentration (log cm -3 ) Position ( μm) Figure 3-15 Carrier Concentration Distributions in Structure “C” (left) and “D” (right) Carrier concentration distributions in structure “C” and structure “D” are shown in Figure 3-15. Obviously, the significant difference appears in the waveguide layers of these two structures. In structure “C”, the carrier concentration in the waveguide layers is in the range of 1E16 cm -3 , while it is in the range of 4E17 cm -3 in structure “D” in which QW-barrier band offset is very small, and lots of carriers can pass though the QW without being trapped. Therefore, in order to maintain sufficiently high carrier concentration in the QW to reach the population inversion, much more carriers have to be injected into the structure. 3.3 Asymmetric Heterojunction Laser Diodes As shown in the last section, with lower the band offsets, the excess voltage can be continuously reduced. But less band offsets also cause lacking of confinement for the carriers, since the carriers have more chance of passing through the QW without generating photons. In order to fully utilize the advantage of lower band 69 offsets and avoid resulting shortcoming, an Asymmetric Heterojuntion Laser (AHL) design is proposed. In AHL, the materials used for each side of the QW are different from each other. Both sides will be optimized for the specific type of carriers being injected into QW. Namely in the n (p) side of the QW, the Barrier material should have a small (large) conduction band offset and large (small) valence band offset. In the AlGaAs/GaAs interface, 60% of the band discontinuity is in the conduction band. As a result, more interface voltage drop exists for the electron injection (n material side) than the hole injection (p material side) (Figure 3-10). At the same time, AlGaAs/GaAs interface provides a good confinement for the electron. Same principle applies to the n material side and another material combination which has larger portion of band discontinuity in the valence band is desired. InGaAsP/InGaAs is found out to be a good candidate for this purpose due to the 65% of band discontinuity in the valence band. Two AHL structures which will be discussed are listed in Table 3-3. They all have InGaAs 0.4 P as the waveguide layer and InGaP as the cladding layer in the n side. The AHL structure “A” (listed in Table 3-3) is using Al 0.2 GaAs as the waveguide layer and Al 0.4 GaAs as the cladding layer in the p side. 70 Layer No. Thickness (µm) Doping (cm -3 ) Structure “A” Structure “B” 9 0.3 P 1E18 GaAs GaAs 8 0.1 P 1E18 Grading Grading 7 0.8 P 1E18 Al 0.4 GaAs Al 0.4 GaAs 6 0.65 undoped Al 0.2 GaAs Al 0.08 GaAs 5 0.012 undoped In 0.13 GaAs In 0.13 GaAs 4 0.65 undoped InGaAs 0.4 P InGaAs 0.4 P 3 0.8 N 1E18 InGaP InGaP 2 0.1 N 1E18 Grading Grading 1 1 N 1E18 GaAs GaAs Table 3-3 Two AHL Structures 1.8 2.7 3.6 -0.7 0.0 0.7 QW Wave- Guide P type Cladding Wave- Guide N type Cladding P Side N Side Conduction Band Valence Band eV μm Figure 3-16 Band structure of AHL Structure “A” The intrinsic condition band structure for the AHL structure “A” is shown in Figure 3-16 and n side material is plotted in the left of the graph. The conduction 71 band in the n side has much smaller band offset and a larger portion of the band offset is located in the valence band side. In the EEL structure “C” simulated in the last section, both p and n sides have Al 0.2 GaAs as waveguide layers and Al 0.4 GaAs as cladding layers. The comparison of the simulation results (LIV curves and Efficiency curves) between EEL structure “C” and AHL structure “A” are shown in Figure 3-17. 050 100 150 0.0 0.5 1.0 1.5 2.0 0 50 100 150 200 AHL "A" EEL "C" Voltage (Volt) Current (mA) Laser Power (mW) 050 100 150 0.0 0.2 0.4 0.6 0.8 1.0 AHL "A" EEL "C" Power Conversion Efficiency Current (mA) Figure 3-17 Simulated LIV Curves and Efficiency Curves of AHL Structure “A” and EEL Structure “C” Both structures have the same slope efficiency, 92.3%. AHL structure “A” has a threshold current of 6.8mA, which is slightly larger than that of EEL structure “C” (5.8mA). At 100mA current, the driving voltage of AHL structure “A” is 1.504V. The driving voltage of EEL structure “C” is 1.551V. The wallplug efficiency of AHL structure “A” is 75.3%, which is 1.3% higher than that of EEL structure “C”. 72 1 1~3 3 3~4 4 4~5 5 5~6 6 6~7 7 7~9 9 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 Voltage Drop (V) Layers and Interfaces EEL Structure "C" AHL Structure "A" Figure 3-18 Excess Voltage Drop Distribution in EEL Structure “C” and AHL Structure “A” The main effect of the AHL structure, namely the InGaAsP-InGaP waveguide-cladding layer, is to minimize the band offset of conduction band in the n side of the devices. This effect is clearly shown in Figure 3-18, which illustrates the excess voltage drop distribution on all layers and interfaces. Significant decrease of voltage drops could be observed on all interfaces in the n side. The largest voltage drops happen in the bulk layers of both p and n side waveguides and in the extra quasi-Fermi level separation of the QW. So further increasing the mobility of waveguide material, especially in the p side, will result in extra improvement of excess voltage drop. As discussed in the previous section, the excess voltage drops decrease when the Al composition is decreased in the waveguide layer. However the slope efficiency will be affected if the band offsets are too low and the carriers can pass through the QW without generating any photon. Using the InGaAs 0.4 P-InGaP in the 73 n side of the device, the aluminum composition is optimized to get the highest wall plug efficiency. Based on the simulations, 8% is found out to be the optimized Al composition in the p side waveguide layer. That corresponds to the AHL structure “B” shown in Table 3-3. The simulation results are shown in Figure 3-19. 050 100 150 0.0 0.5 1.0 1.5 2.0 0 50 100 150 200 Voltage (Volt) Current (mA) Laser Power (mW) 0 50 100 150 0.0 0.2 0.4 0.6 0.8 1.0 Wall Plug Efficiency Current(mA) Figure 3-19 Simulated LIV Curves and Efficiency Curve of AHL Structure “B” The IV curve is nearly flat after turn on. The LI curve is still the same. The wallplug efficiency is 80.3% when the driving current is 100 mA. The excess voltage drop distribution is shown in Figure 3-20. 1 1~3 3 3~4 4 4~5 5 5~6 6 6~7 7 7~9 9 0.00 0.01 0.02 0.03 0.04 0.05 Voltage Drop (V) Layers and Interfaces Figure 3-20 Excess Voltage Drop Distributions for AHL Structure “B” 74 ΔV ∗I 0 5.94% ΔI ∗ΔV 0.95% ΔI ∗V 0 12.8% I 0 ∗V 0 80.3% Laser output (I 0 + ΔI) ∗(V 0 + ΔV) 100% Power Input Figure 3-21 Loss Distribution in AHL Structure “B” Similar to the analysis shown in Figure 3-4, the laser input power distribution is calculated according to equation 3.2, 3.3 and 3.4. As indicated by Figure 3-21, the loss caused by extra voltage only counts for 6.9% of the total power consumption. 3.4 Transverse Junction Stripes Quantum Well GaAs Substrate Cladding Layer Cladding Layer Waveguide Layer Waveguide Layer Carrier Injection Layer Carrier Injection Layer GaAs Cap + - + + - - Quantum Well GaAs Substrate Cladding Layer Cladding Layer Waveguide Layer Waveguide Layer Carrier Injection Layer Carrier Injection Layer GaAs Cap + + - - + + + + - - - - Figure 3-22 TJS Structure Schematic Drawing A novel EEL geometry, Transverse Junction Stripes (TJS), is proposed to solve the problem of excess voltage. A schematic drawing of TJS structure is shown in Figure 3-22. In the middle column of Figure 3-22 is a normal EEL structure: QW, 75 waveguide layer and cladding layers. But all these layers are undoped. QW is still the carrier confinement structure in the vertical direction and waveguide layers and cladding layers define the optical field. The main difference is that the carriers are not passing all the way through the vertical direction but injected laterally from the two sides of the QW. In the design of TJS, carriers are injected from two sides of the Laser mesa in order to avoid the junctions and bulk materials above and below the QW. Lateral injection in the semiconductor laser is not a brand new idea. The idea was investigated in the early stage of semiconductor lasers research as an alternative of vertical injection laser to achieve low threshold [7]. It has also been considered as an effective method to realize high-density integration of optoelectronic devices [8]. There are few literatures regarding theoretical simulation on this kind of device [9]. And it was not considered as a method to eliminate the excess voltage. 3.4.1 TJS Simulation Intensive simulations are preformed on TJS devices. There are several significant differences in the LASTIP simulations of TJS structures from those of EEL simulations. First, in order to describe the laser structure in the LASTIP, the whole structure is split into 3 columns, as shown in Figure 3-23. There is only undoped epi structure in the middle column. The carrier injecting layers are two columns on the sides. In a broad area EEL, the width of the laser is much longer than the light wavelength, so the effect of the boundary can be ignored, therefore a 76 uniformly distributed light in the lateral direction is considered in those simulations. But in the TJS structure, the effects of lateral injection and smaller laser width, which will be discussed in more details later, require a real 2-D simulation. Comparing with only 2 mesh points laterally in the EEL simulation, much more mesh points must be used. For the same reason, the boundary condition for the lasing mode is also changed. As shown in Figure 3-23, optical mode simulation window is confined only in the middle column and a zero intensity boundary condition is used. Quantum Well GaAs Substrate Cladding Layer Cladding Layer Waveguide Layer Waveguide Layer Carrier Injection Layer Carrier Injection Layer GaAs Cap Column I Column II Column III Optical Field Calculation Window Line A Line B Line C Line D Quantum Well GaAs Substrate Cladding Layer Cladding Layer Waveguide Layer Waveguide Layer Carrier Injection Layer Carrier Injection Layer GaAs Cap Column I Column II Column III Optical Field Calculation Window Line A Line B Line C Line D Figure 3-23 Column Settings and Optical Field Calculation Window in TJS Simulation In all the simulations regarding TJS structures in this chapter, the middle column has the same layer structure as that of structure “B” listed in Table 3-2. The materials of carrier injection layers are varied. In the initial simulation, GaAs is used as the carrier injection layer material. The layer in column I is n type doped and the layer in column III is p type doped. In this simulation, the middle column is 5 um 77 wide and two current injection columns are 2 um wide. Confining the optical field calculation window helps LASTIP to find the correct optical mode in the structure. Figure 3-24 shows the optical field intensity on the laser cross-section. In most of the simulations discussed in this section, only the first order optical mode is considered. In the later part of this section, evidences will be presented showing that considering only the first order optical mode gives out nearly identical simulation result as in the simulation where multiple optical modes are involved. Figure 3-24 Simulated Optical field in TJS The most unique characteristic of the TJS structure is that the carriers are injected from the sidewalls of the active region. The first task is to know whether all the carriers can be injected into the QW instead of leaking through waveguide or cladding layers. Figure 3-25 shows the lateral current density in the middle of TJS- GaAs. The curve is showing the simulated results along line B in Figure 3-23. 78 01 23 4 Current in X direction (a.u.) Position ( μm) Figure 3-25 Current in X direction in the middle of TJS By comparing the integral below the curve over the whole structure and just over QW, it is proved that clearly most of the lateral current, 99.28% of them, goes through the QW. 0 20406080 100 120 140 0.0 0.5 1.0 1.5 2.0 0 20 40 60 80 100 120 140 160 180 200 Voltage (Volt) Current (mA) Laser Power (mW) Figure 3-26 Simulated LIV Curves of TJS-GaAs Structure The simulated LIV curves of the TJS-GaAs structure are shown in Figure 3-26. The graph is using same scale as all the simulations shown in early part of this chapter. The turn on voltage is very low, close to the band gap of GaAs. But the resistance is high comparing to that of EEL structure “D” discussed earlier in this chapter. The only characteristic becoming even worse is the extremely low laser 79 power. The structure barely starts lasing at the current range shown. At 100 mA driving current, the driving voltage is 1.5556 V. Lasing wavelength is 0.9381 µm, which corresponds to photo energy of 1.3171 eV. The total voltage defect is 0.2383 V, which is not as low as the 940 EEL structure “D” which has the GaAs as waveguide layer. Figure 3-27 shows the band structure and quasi-Fermi level distribution in the structure. 0.0 3.0 6.0 9.0 0.0 0.5 1.0 1.5 01 23 4 0.0 0.5 1.0 1.5 01 23 4 -0.5 0.0 0.5 1.0 1.5 2.0 01 23 4 0.0 0.5 1.0 1.5 Line D Energy (eV) Conduction and Valence Bands Electron Quasi Fermi Level Hole Quansi Fermi Level Line B Line A Line C Position ( μm) Figure 3-27 Band Structures of TJS-GaAs. Simulation Results are Drew along the Lines Mentioned in Figure 3-23 Since this is more complex 2-D problem, it couldn’t be shown in a simple graph. Figure 3-27 illustrates band structures alone few important lines in the structure. The first graph is along line D, which is in the middle of the QW. The left 2 µm is the n type carrier injection layer, there is no voltage drop could be observed 80 from the electron quasi-Fermi level. There is also no voltage drop observed from the hole quasi-Fermi level in the p type carrier injection layer side. There are large voltage drops along the QW, which are shown in the graph as the large slopes in the electron and hole quasi-Fermi levels from 2 µm to 7 µm. Line A is in the middle of n carrier injection layer, column I, along the vertical direction. Line C is in middle of p carrier injection layer, column III, along the vertical direction. The metal contact is at 3.212 µm position and the QW is at 2.562 µm position. There is no big voltage drop could be observed from the top contact to the QW. Because it is obvious that most of the carriers are flowing through the QW in the middle column. So the graph along line D shows that nearly all the voltage drop is in the QW itself. 02 468 15.8 16.0 16.2 16.4 16.6 16.8 17.0 17.2 17.4 17.6 17.8 18.0 18.2 18.4 18.6 Electron Hole Carrier Concentration (log cm -3 ) Position ( μm) Figure 3-28 Carrier Concentration in TJS-GaAs Along Line “D” The carrier concentration in the QW and adjacent carrier injection layers is shown in Figure 3-28. There is carrier concentration gradient in the QW. The electrons have a higher mobility than the holes. Due to charge neutrality and 81 ambipolar transport properties, the carrier concentration is higher near the p side in the QW. Similar as structure “D” in EEL simulation, the QW now is sandwiched in GaAs laterally, and QW is along the current flow direction. So there is barrier high enough to prevent the carriers from flowing out of the QW. This is exactly what happens here, especially for electrons. The electron concentration in the p carrier injection layer side is very high. Lots of electrons pass right through the QW without recombining with holes and generating lights. As we discussed above, there is still a need for a high barrier in the lateral direction to confine the carriers. As a consequence, Al 0.3 GaAs is introduced in the next TJS simulation. This structure will be called TJS-Al 0.3 GaAs in the following discussion. The simulated LI and IV curves are shown in Figure 3-29. The simulated results of EEL structure “B” are also shown for comparison. -2.0 -1.5 -1.0 -0.5 0.0 -20 0 20 40 60 80 100 120 140 160 TJS-Al 0.3 GaAs Injection Layer 1D-Reference Current (mA) Voltage (Volt.) 0 2040 6080 100 120 140 0 25 50 75 100 125 150 175 200 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Wallplug Efficiency Output Power (mW) Current (mA) Figure 3-29 Simulated Laser performances of TJS-Al 0.3 GaAs Structure and EEL “B” Structure The turn on voltage of TJS-Al 0.3 GaAs structure is smaller than that of EEL structure “B”, in which Al 0.3 GaAs is also used as the waveguide layer. However, the resistance in the TJS structure is higher than 1D EEL structure. They have nearly the 82 same threshold current density, but the TJS has low slope efficiency. Comparing with TJS-GaAs structure, significant improvement could be observed from the TJS- Al 0.3 GaAs structure. This is mainly due to the fact that carriers can be successfully confined inside the QW itself. The carrier concentration along line “D” is shown in Figure 3-30. The electron concentration in the p injection layer is in the range of 10 14 cm -3 ~ 10 15 cm -3 instead of 10 17 cm -3 ~ 10 18 cm -3 in the TJS-GaAs structure. 02 46 8 13 14 15 16 17 18 19 Electrons Holes Carrier Concentration (log cm -3 ) Position ( μm) Figure 3-30 Carrier Concentration in TJS-Al 0.3 GaAs Along Line “D” 83 01 2 3 4 0.0 0.5 1.0 1.5 2.0 012 3 4 0.0 0.5 1.0 1.5 2.0 01 2 3 4 0.0 0.5 1.0 1.5 2.0 03 69 0.0 0.5 1.0 1.5 2.0 Line D Line D Line D Line D Electron Quasi Fermi Level Hole Quansi Fermi Level Conduction and Valence Bands Energy (eV) Line B Line A Line C Position ( μm) Figure 3-31 Band Structures of TJS-Al 0.3 GaAs. Simulation Results are Drew along the Lines Mentioned in Figure 3-23 The band structure and quasi-Fermi level distribution in TJS-Al 0.3 GaAs are shown in Figure 3-31. Data analyses are still presented along four lines mentioned in Figure 3-23. Similar as of TJS-GaAs structure, there is small voltage drop in the carrier injection layer according to band structure graph along line A and graph along line C. In the graph along line D, voltage drops from the carrier injection layers to the QW could be calculated. Meanwhile huge voltage drop exists in the QW itself. At 100 mA driving current, the total voltage applied onto the device is 1.806V; the photo energy is 1.322eV, which means that there is 0.484V extra voltage drop. Further analysis shows that the voltage drop from n side contact to QW is 0.144V and from p side contact to QW is 0.104V, then the QW counts for the rest, which is 84 0.236 V. Comparing with the EEL structure “B”, TJS-Al 0.3 GaAs structure avoids the voltage drop in the cladding layer but picks up more voltage drop from the lateral transportation in the QW. In the traditional vertical injection EEL structure, the quasi-Fermi level separation in the QW is only the source for achieving population inversion. But in the TJS structure, extra separation are needed to build up high enough carrier concentration gradient, especially in the both ends near the carrier injection layers, to guarantee the driving force for carriers flowing through QW. This extra quasi-Fermi level separation, the amount of energy larger than the photon energy, along line “D” is shown in Figure 3-32. 0246 8 0 50 100 150 200 250 Quasi-Fermi level separation - Photon Energy meV Position ( μm) Figure 3-32 Quasi-Fermi Level Separation Minus Phone Energy along Line “D” In the QW, from 2µm to 7µm position, the curve is an unevenly reverse bell shape. The minimum of separation energy is 22 meV, which is located about 1/3 of stripe width away from the n type carrier injection layer. Then the number rises up to 47 meV adjacent to n type carrier injection layer and to 166 meV adjacent to p side. 85 The shape of the curve resembles the carrier concentration distribution in QW shown in Figure 3-30. The slope efficiency of TJS-Al 0.3 GaAs structure is calculated as 72.4%, which is much lower than that of EEL structure B, 92.2%. In most part of the QW, the carrier concentration is significantly higher than what’s needed for population inversion. The high carrier concentration results in high non-radiative recombination and low L-I slope efficiency. As discussed above, both large excess voltage drop and large loss are caused by the lateral transportation of carriers in the QW since all carriers are forced to go through a very thin slab structure, i.e. QW. Another reason for the low L-I slope efficiency is the leakage current through other layers of TJS mesa. More detailed calculation shows that 99.28% of the current flow through the QW in the case of TJS-GaAs structure. However, there is only 95.39% of the current flow through the QW in the TJS-Al 0.3 GaAs structure, large resistance in the QW forces the current to take other paths, i. e. 1.14% leaks through the Al 0.3 GaAs layers, 0.04% through the Al 0.4 GaAs layers and 3.43% through the top GaAs cap layer and bottom GaAs buffer layer. In the above simulation, only the fundamental optical mode is considered. However, for this cross-section geometry, higher order modes could exist in the structure. Next, the effect of multiple optical modes is discussed. 86 0 25 50 75 100 125 150 0 25 50 75 100 125 150 175 Multi-Mode Single-Mode Output Power (mW) Current (mA) 0.0 0.5 1.0 1.5 2.0 0 20 40 60 80 100 120 140 160 Multi-Mode Single-Mode Current (mA) Voltage (Volt) Figure 3-33 Comparison of Laser Performances of TJS-Al 0.3 GaAs Structure Simulated with Single Optical Mode and Multi Optical Modes The simulated data considering multiple optical modes and considering only the fundamental optical mode are shown in Figure 3-33. They have nearly the identical IV curve. The multi-mode result has higher slope efficiency. 0 25 50 75 100 125 150 0 10 20 30 40 50 60 70 5 4 3 1 2 Output Power (mW) Current (mA) Figure 3-34 Simulated LI Curves for Each Optical Mode in TJS-Al 0.3 GaAs Structure As for multi-mode simulation, there are 5 possible optical modes in the current geometry. The output power for each optical mode is shown in Figure 3-34. They are marked by their mode index. Higher order modes have higher output power compared with lower order mode. The carrier concentrations in QW simulated under these two conditions are shown in Figure 3-35. 87 024 68 17.0 17.5 18.0 18.5 19.0 Multi-Mode Single-Mode Carrier Concentration (log cm -3 ) Position ( μm) Figure 3-35 Comparison of Carrier Concentration in QW of TJS-Al 0.3 GaAs Structure Simulated with Single Optical Mode and Multi Optical Modes Wave intensity of high order optical modes spread more uniformly across the active mesa comparing with the first order modes, so the carriers near the edge of the mesa can get more chance to recombine through stimulated emission. This effect lowers the carrier concentration near the edge of the mesa. It also helps decreasing non-radiative recombination and increasing the overall slope efficiency. Although the multiple optical modes simulation demonstrates different result than the simulation considering only fundamental optical mode, the difference is very small. Besides simulation calculation takes much longer time when considering multiple optical modes. Therefore, only the fundamental optical mode will be taken into account in the following simulations. As shown in the previous discuss, in order to improve TJS laser performance, either the carrier lateral transportation needs to be optimized or the lateral distance between two carrier injection layers has to be decreased. Both approaches will be discussed. 88 For the purpose of improving the lateral transportation of carriers, an extra path other than QW is proposed. Two thin layers of GaAs are added between the QW and AlGaAs waveguide layers. GaAs-InGaAs interface has smaller band offset comparing with AlGaAs-InGaAs interface, so not all the carriers will flow into QW. There will be significant amount of carriers moving in the GaAs layer first then drop into QW and recombine. This structure is referred as the TJS-Channel structure. The epi structure of this proposal before regrowth is shown in Table 3-4. Layer No. Material Thickness (µm) Doping (cm -3 ) 11 GaAs 0.3 P 1E18 10 Grading 0.1 P 1E18 9 Al 0.4 GaAs 0.8 P 1E18 8 Al 0.3 GaAs 0.65 Undoped 7 GaAs 0.05 Undoped 6 In 0.13 GaAs 0.012 Undoped 5 GaAs 0.05 Undoped 4 Al 0.3 GaAs 0.65 Undoped 3 Al 0.4 GaAs 0.8 N 1E18 2 Grading 0.1 N 1E18 1 GaAs 1 N 1E18 0 GaAs N+ (100) substrate Table 3-4 Epi Structure of TJS-Channel structure middle column 89 The simulated LI and EI curves are shown in Figure 3-36. The simulation results of TJS-Al0.3GaAs and EEL structure “B” are also shown for comparison. Significant improvement in the driving voltage can be observed. The TJS-Channel structure has a low turn on voltage and low resistance. The slope efficiency is better than that of the TJS-Al 0.3 GaAs structure but still worse than EEL structure “B”. 0.00.20.40.60.81.0 1.21.41.61.82.0 0 20 40 60 80 100 120 140 160 TJS-Channel EEL Structure "B" TJS-Al 0.3 GaAs Current (mA) Voltage (Volt.) 0 20 40 60 80 100 120 140 0 25 50 75 100 125 150 175 200 0.0 0.2 0.4 0.6 0.8 1.0 Wallpulg Efficiency TJS-Channel EEL Structure "B" TJS-Al 0.3 GaAs Output power (mW) Current (mA) Figure 3-36 Simulated Laser performances of TJS-Channel Structure Comparing with TJS- Al 0.3 GaAs Structure and EEL “B” Structure 2.4 2.5 2.6 2.7 GaAs Channel GaAs Channel QW TJS-Channel TJS-Al 0.3 GaAs Current Density (a.u.) Position ( μm) Figure 3-37 Simulated Current Density in X direction in the middle of TJS-Channel Structure Channel structure indeed provides extra paths for the carriers. The x-direction current density distribution is shown in Figure 3-37. Comparing with the TJS- 90 Al 0.3 GaAs structure, shown as dashed line in the graph, the TJS-Channel structure, shown as solid line in the graph, has much smaller current density in the QW itself. According to simulation, 58.33% of carriers flow through the QW, 39.79% flow through the channels and 1.34% leak through the GaAs cap and buffer layers. 0.02.04.06.08.0 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Excessive Voltage 1.694-1.312=0.382eV Voltage Drop in QW 0.382-0.129-0.088=0.166eV 0.088V 0.129V Conduction and Valence Bands Electron Quasi-Fermi Level Hole Quasi-Fermi Level eV μm Figure 3-38 Band Structures of TJS-Channel (along Line D Mentioned in Figure 3-23) Figure 3-38 illustrates the band structure drew along Line D in the middle of QW in TJS-Channel structure. At 100 mA, the excess voltage drop in this structure is 0.382 V. It is about 0.1 V less than that of TJS-Al 0.3 GaAs structure. For this 0.382 V of voltage drop, 0.129 V is wasted on the n type contacting layer and QW interface, 0.088 V is on p type contacting layer and QW interface. The QW counts for 0.166V of extra voltage drop. This is 30% less than the TJS-Al 0.3 GaAs structure. As discussed above, introducing channels between the QW and wave guide layers can significantly decrease the excess voltage. The thickness of these channels also plays an important role in the TJS Laser performance. The effect of channel thickness on the wall plug efficiency is shown in Figure 3-39. 91 0 25 50 75 100 125 150 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Channel Thickness 0 nm 50 nm 100 nm 200 nm 400 nm Wall Plug Efficiency Current (mA) Figure 3-39 Wall Plug Efficiencies of TJS Channel Structures with Varying Channel Thickness At 100 mA driving current, introducing 50nm channels largely increases the wall plug efficiency. As the channel thickness increase to 100 nm and 200 nm, the wall plug efficiencies also increases to about 65%. However, when the channels continue to become thicker, the wall plug efficiency starts to decrease. 0 25 50 75 100 125 150 0 25 50 75 100 125 150 175 Channel Thickness 0 nm 50 nm 100 nm 200 nm 400 nm Laser Output (mW) Current (mA) 0.00.5 1.01.52.0 0 20 40 60 80 100 120 140 Channel Thickness 0 nm 50 nm 100 nm 200 nm 400 nm Current (mA) Voltage (Volt) Figure 3-40 Simulated LI and IV Curves of TJS-Channel Structures with Varying Channel Thickness In order to better understand the changes of wall plug efficiency with various thicknesses of channels, the LI and IV curves are simulated and shown in Figure 3-40. When increasing the channel layers thickness, the resistance drops correspondingly due to larger conduction area. At the same time, the LI slope 92 increases and so is the threshold current. This is caused by the fact that much larger volume of GaAs must be populated to high enough carrier density to ensure the QW population inversion. Therefore, the thickness of QW influences the performance of TJS Lasers in several different and interconnecting ways. In the real device optimization, a design of experiment method should serve best. As discussed earlier about the excess voltage in the TJS-Channel structure, only 0.166 V out of total 0.382 V is wasted in the QW itself. All the rest is consumed by the carrier injection layer and QW interface, especially in the n side. The idea of asymmetric QW used earlier in this chapter can also be applied to the selection of carrier injection layers. We can use a n type material which has small band offset in the conduction band but large band offset in the valence band. So the n type carriers can be injected into QW easily but the holes will still experience a large barrier which prevents the holes from leaking into n carrier injection layer. In the next simulation, an asymmetric TJS-Channel structure is studied. The most of the structure is same as the TJS-Channel structure except the n type carrier injection layer. InGaAs 0.4 P is used in there instead of Al 0.3 GaAs. The simulation results are shown in Figure 3-41. The asymmetric carrier injection layer doesn’t change the LI characteristic but decreases the driving voltage. The wall plug efficiency at 100 mA increases to 67.7%. 93 0.00.20.40.60.81.01.21.4 1.6 1.8 0 25 50 75 100 125 150 TJS-Channel Asymmetric TJS-Channel Current (mA) Voltage (Volt) 0 25 50 75 100 125 150 0 25 50 75 100 125 150 175 0 25 50 75 100 125 150 0.0 0.2 0.4 0.6 0.8 1.0 Output Laser Power (mW) Current (mA) TJS-Channel Asymmetric TJS-Channel Wall Plug Efficiency Figure 3-41 Simulated Laser performances of Asymmetric TJS-Channel Structure Compared with TJS-Al 0.3 GaAs Structure and EEL “B” Structure 0.0 2.0 4.0 6.0 8.0 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0.090V 0.016V Excessive Voltage 1.581-1.312=0.269V Voltage Drop in QW 0.269-0.016-0.090=0.163V Conduction and Valence Bands Electron Quasi-Fermi Level Hole Quasi-Fermi Level eV Position μm Figure 3-42 Band Structures of TJS-Channel Asymmetric Structure (along Line D) As shown in Figure 3-42, introducing the InGaAs 0.4 P as the n type carrier injection layer significantly decreases the voltage drop at the n side. Compared with TJS-Channel Structure in Figure 3-38, voltage drop at the n side goes down from 0.129 V to 0.016 V. And this material change doesn’t change the performance in the rest of the structure. After introducing the channel and the asymmetric injection layers, the excess voltage drop in the TJS structure decreases from 0.484 V to 0.269 V. Out of this 94 0.269V, 0.163 V is still wasted on the lateral transportation in QW. Therefore, decreasing the TJS stripe width needs to be reduced. Simulations on various TJS stripe width are shown in Figure 3-43. In this Figure, the unit for the current is mA/µm. Because the devices have different widths, the correct way of comparing their currents is normalizing the currents over their widths. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0 5 10 15 20 25 30 2 μm Stripe 3 μm Stripe 5 μm Stripe Current Density (mA/ μm) Voltage(Volt) 0 5 10 15 20 25 30 0 5 10 15 20 25 30 35 40 0.0 0.2 0.4 0.6 0.8 1.0 2 μm Stripe 3 μm Stripe 5 μm Stripe Light Output (mW/ μm) Current Density (mA/ μm) Figure 3-43 Simulated Laser Performances of TJS-Channel Asymmetric Structures with Varying TJS mesa width As expected, narrow stripe width helps reducing the driving voltage and increasing the wall plug efficiency. When the stripe width narrows down to 2 µm, the wall plug efficiency increases to 78.7% at 20 mA/µm driving current. Figure 3-44 shows band structure and Quasi-Fermi levels along the middle of QW. The TJS structure with a 2 µm stripe width is studied in the graph, the current is 40mA (20 mA/µm normalized current density), which is comparable to 100mA when the TJS stripe width is 5 µm. As indicated in this graph, the excess voltage drop in the structure is only 0.124V. The voltage drop in the QW is 0.064V, which is proportional to that of 0.164V in the TJS structure with 5 µm stripe width. Both the voltage drops at the n and p side injection layers decreases a little. 95 0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 0.3 0.6 0.9 1.2 1.5 1.8 0.055V 0.005V Excessive Voltage 1.436-1.312=0.124eV Voltage Drop in QW 0.124-0.005-0.055=0.064eV Conduction and Valence Bands Electron Quasi-Fermi Level Hole Quasi-Fermi Level eV μm Figure 3-44 Band Structures of TJS-Channel Asymmetric Structure, 2 µm stripe width. (along Line D) When the active mesa width is 2 µm, only two optical modes exist. The second order mode is dominant. But the simulation results are very similar to the result from considering only one mode. The wall plug efficiency is 79.1% at 20 mA/µm driving current. 3.4.2 Preliminary Experimental Results An optimized TJS structure will require narrow TJS stripes and two different materials as p and n type carrier injection material. It is a big challenge to build such a device. The fabrication involves multiple steps of etching and regrowth. Due to unexpected termination of the project, we didn’t succeed in making a working TJS device, but lots of preliminary experiments were completed. Fabrication process key issues in the process and the experiment results will be discussed in this section. 96 1. Active Layers Epi on Substrate 1. Active Layers Epi on Substrate 2. Mesa Etching 2. Mesa Etching 3. Regrowth, N Type Injection Layer 3. Regrowth, N Type Injection Layer 4. Mesa Etching 4. Mesa Etching 5. Regrowth, P Type Injection Layer 5. Regrowth, P Type Injection Layer Figure 3-45 TJS Structure Schematic Drawing A schematic drawing of the TJS process is shown in Figure 3-45. The first step is MOCVD growth of TJS active layers. The epi structure is shown in Table 3-4. The growth itself is a simple routine MOCVD planar growth. The second step is the mesa etching. The interface of carrier injection layer and QW must be exposed. The third step is n type carrier injection layer regrowth. The regrowth should only happen in the trench exposed by the etching. During the regrowth process, the mesa top will be covered by dielectrics and should be free of any material deposition. The fourth step is another mesa etching. New interface between injection layer and QW is exposed. The fifth step is another regrowth to fill the trenches with p type carrier injection material. There will be more following steps to form metal contacts and cleave wafer before a TJS device is finally tested. Obviously, etching and regrowth are two most challenging steps in the whole process. There is always a choice needed to be made between dry etching and wet etching when etching is required in the semiconductor process. Both methods have 97 their advantage and disadvantage due to their different etching mechanism. The wet etching normally has less physical damage to the surface, but it is has the problem of undercut beneath the mask or anisotropic etching rate for certain crystal planes. Dry etching is good at forming straight side wall and precise control of following masks, but it forms defects on the etched surface [10]. According to simulations discussed earlier, mesa width needs to be only few microns for the TJS to have decent wall plug efficiency. And the mesa height would be larger than 3.5 µm in order to fully etch through the epi layer. This narrow mesa width requires that etching should closely follow the mask pattern. The dry etching is more suitable for this kind of precise control but leave large amount of surface defects on the exposed interface. This interface will be crucial because it is adjacent to QW and carriers will be injected through it to reach the QW. The non-radiative recombination due to the damaged interface will cause too much loss for the Laser. An extra wet-etching step might be needed to remove the damaged surface. At this early stage of research, we are just testing the fundamental ideas of TJS, therefore only wet etching is utilized for these testing devices and also these first trials of TJS devices will be far from each other and the width of the devices will be varied on same wafer. This approach will require less precise control during wet etching. However, a relative vertical sidewall is still required since the mesa (QW width) needs to be very narrow. Wet etching is carried out using a solution of H 3 PO 4 :H 2 O 2 :H 2 O. There is no heat generation when mixing the solution. This solution has an advantage over the 98 H 2 SO 4 :H 2 O 2 :H 2 O solution which is another popular etchant for GaAs/AlGaAs material system. For H 3 PO 4 :H 2 O 2 :H 2 O, the etching profile, etching speed and surface smoothness can be controlled by the mixture ratio and temperature. [11, 12] The H 3 PO 4 :H 2 O 2 :H 2 O solution is being agitated constantly during etching. It is very important to avoid the non-uniformity of the solutions so as to guarantee the stable etching speed and desired etching profile. The solution is kept at room temperature. In the etching and regrowth test, a double stripe photo mask is used. The distance between these two stripes is 15 µm. The stripe width varies from 10 to 50 µm. The pattern is repeated on the mask with 250 µm spacing. The pattern is transferred onto the SiNx film deposited on the wafer. SiNx serves as the dielectric mask for wet etching and regrowth. Stripes along <1 1 0> direction Stripes along <1 -1 0> direction Figure 3-46 Cross-section Profiles after H 3 PO 4 :H 2 O 2 :H 2 O (1:9:1) etching The cross-section profiles after H 3 PO 4 :H 2 O 2 :H 2 O (1:9:1) 40 seconds etching are shown in Figure 3-46. These are SEM pictures taken on the facets cleaved after wet etching. The graph shown in the left is the stripes along <1 1 0> direction. The 99 graph shown in the right is from the stripes along <1 -1 0> direction. In both cases the trench depth is slightly more than 3 µm and the undercut is about 2.5 µm, significantly different etching profiles are shown in these two perpendicular directions. This is what should be expected from a reactive limited wet etching [10]. The etching process picks up the (111) plane which forms two different kinds of slopes with the stripes along <1 1 0> and <1 -1 0> directions. Also etch rate is quite high, the surface smoothness after the etching is not very satisfying. Stripes along <1 1 0> direction Stripes along <1 -1 0> direction Figure 3-47 Cross-section Profiles after H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) etching The cross-section profiles after H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) etching are shown in Figure 3-47. The graph shown in the left is the stripes along <1 1 0> direction. The etching time is 12.5 minutes. The trench depth is 2.7 µm and the undercut of SiNx is 2.7 µm too. The graph shown in the right is the stripes along <1 -1 0> direction. The etching time is 14 minutes. The trench depth is 2.9 µm and the undercut is also 2.9 µm. Although these two set of stripes are perpendicular with each other, they share the same etching profile, which is the characteristics of diffusion limited, isotropic etching [10]. 100 Figure 3-48 Cross-section Profile after H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) etching A magnified SEM image (Figure 3-48) is taken. The area shown is marked as a square in Figure 3-47. The SiNx overhang is cracked during cleaving. The surface on the bottom of trench is very smooth. There are small vertical ripples on the side wall. The reason for these ripples is unknown. It may indicate the etching still has slightly crystallographic preference. There is a clear horizontal kink on the side wall which is close to the position of QW. The edge of the QW is within couple of tenths of a micron horizontally from the top edge of the mesa. It is nearly vertical sidewalls above the QW. If we want to etch through the whole Epitaxial layers, then the etch depth will be more than 3 µm. As observed from above SEM images, the SiNx overhang width is same as the etching depth due to the property of diffusion limited isotropic etching. A 3 µm SiNx overhang will make it more difficult to deposit thick injection layer material on whole side wall. So the SiNx overhang needs to be removed at the 101 same time as the SiNx on the mesa top should remain to act as dielectric mask in the following selective area growth. Another step of wet etching is utilized to remove the SiNx overhang. The initial SiNx film thickness increases to 200nm. After the first step mesa etching, the sample is treated by buffered oxide etchant (BOE) 10:1 for 3 minutes. The SiNx overhang is removed and there is still about 50nm SiNx on mesa top. This is because the SiNx overhang is attacked from both top and bottom sides but the SiNx on top of the mesa is only etched from top side. Figure 3-49 Cross-Section Profile after Remove SiNx Overhang The cross-section profile after removing SiNx overhang is shown in Figure 3-49. This is the image of stripes along <1 -1 0> direction after H 3 PO 4 :H 2 O 2 :H 2 O (1:9:1) 40 second etching. The image before the SiNx removal is at the right side of Figure 3-46. There is no SiNx overhang films observed. The edge of the mesa top is very clean. There is still SiNx film on the mesa top which could be observed with the optical microscope. 102 780 800 820 840 860 880 900 Epi-ready wafer chemical etched wafer PL Intensity (a. u.) Wavelength ( μm) Figure 3-50 Photoluminescence Data from QWs Grown on Etched GaAs Wafer In order to check the surface quality after all these wet etching processes, a GaAs-AlGaAs QW structure is grown on both epi-ready GaAs wafer and etched GaAs Wafer simultaneously. The etched GaAs wafer is etched by H 3 PO 4 :H 2 O 2 :H 2 O (3:1:25) for 15 minutes, followed by DI rinsing for 1 minute, BOE 10:1 for 3 minutes and DI rinsing for another 1 hour. The measured PL data from both samples is shown in Figure 3-50. There is no difference in these two curves. This proves that GaAs surface after all these chemical process is comparable to epi-ready GaAs wafer. The next step after chemical etching is MOCVD regrowth. Under the ideal condition, the regrowth is expected to fully cover the trench exposed by etching and form a planar surface. There should be no material deposition on dielectric mask. The selectivity of growth strongly depends on the growth condition. Low growth rate and low V/III ratio [13,14] are desired to have better selectivity. 103 According to the simulation, regrowth materials for n and p side carrier injection layers are different, namely Al 0.3 GaAs is for p side and InGaAs 0.4 P is for n side. At the initial stage of the research, regrowth experiments are focused on Al 0.3 GaAs for the following reasons. First, it should be much easier to introduce the InGaAs 0.4 P selective growth later since the InGaAsP quaternary selective growth is easier to realize compared with AlGaAs. Second, Al 0.3 GaAs is a good carrier confinement material for both electrons and holes although causing high voltage drop in the n side. The electron confinement of InGaAs 0.4 P is very weak. If we use InGaAs 0.4 P for both sides, large amount of electrons in the QW will flow into p side of injection layer which will cause extremely low slope efficiency. So Al 0.3 GaAs is selected for both carrier injection layers at this initial stage to prove the idea of TJS. Samples are loaded into MOCVD chamber right after all the chemical etching processes and blowing dry with Nitrogen. Material deposition starts 10 minutes after the samples are heated to the growth temperature, 730 ºC. The growth starts with a 10 nm GaAs, then nominal 1.2µm thick Al 0.3 GaAs is deposited. Finally another 10 nm GaAs is deposited to prevent the oxidation of the sample surface. This GaAs layer is also necessary for forming good ohmic metal contacts [15]. As claimed earlier, if we want to etch through the whole Epitaxial layers in the wet etching, the etch depth will be more than 3 µm. If we want to fully planarize the trenches, the regrowth thickness should be in the same range.1.2 µm Al 0.3 GaAs instead of 3 µm or more is chosen for the following reasons. A thick Al 0.3 GaAs requires long time for the MOCVD growth. A 3 µm layer increase the total growth 104 time to about 2 hours. It also increases the difficulty of maintain the selectivity of the growth. Besides we don’t have to fully planarize the sample surface, we just need to make sure the TJS sidewall is fully covered with carrier injecting material. The Al 0.3 GaAs growth rate is 6 Ǻ per second. This is about 1/2 ~ 1/3 of normal growth rate. The GaAs growth is also very low, 4.2 Ǻ per second. Three different V/III ratios, 47, 20 and 10, are used in three growths. When the V/III ratio is 47, there is nearly no growth area selectivity. There is thick material deposition on SiNx masks regardless of the stripe width. A SEM image is shown in Figure 84. As shown in this image, there is uniform thick polycrystal deposition on the SiNx masks. The material grown in open area is very smooth. There are bumps in the edges of SiNx stripes. And the side walls are rough. After decreasing the V/III ratio, there is much better growth selectivity. Table 3-5 summarizes the microscope images of the selective area growth results. Growths Figure 84 SEM Image of Wafer Grown with V/III ~ 47 SiNx SiNx 105 were performed with V/III ratio 20 and 10. The mask pattern width is varied from 50 µm to 10 µm. The width listed for each row is the original photo mask stripe width. The actual SiNx width during the regrowth is about 6~7 micron less than the photo mask width due to the undercut during wet etching. Every small grid shown on the rulers corresponds to 1 µm. The areas covered by the SiNx are marked by lines with arrows in the graph. The arrows point to the edges of the SiNx while the line itself sits on the SiNx covered area. 106 Mask Width V/III ~20 V/III~10 50 µm 40 µm 35 µm 30 µm Table 3-5 to be continued 107 Mask Width V/III ~20 V/III~10 25 µm 20 µm 15 µm 10 µm Table 3-5 Selective Area Growth Results with Varying V/III ratio and SiNx Width 108 Comparing with the growth done with high V/III ratio, improved selectivity could be clearly observed with low V/III ratio. When the V/III is about 20, there is only material deposition on the SiNx when the stripe is wide. The narrower the stripe width, the less material is deposited on the SiNx. When the V/III is about 10, there is no serious material deposition on the SiNx even when the strip width is more than 40 µm. However, there are still a few localized deposits on the SiNx mask which is not clearly related to the mask width. These deposited dots might be caused by that there are pinholes in the SiNx film or the SiNx film is extremely rough in those local areas. There is no solution found to solve this problem yet. Figure 3-52 Cross-section Profiles after TJS Al 0.3 GaAs Regrowth, V/III ~ 10, Stripes along <1 1 0> direction The cross-section images shown in Figure 3-52 are SEM images after the selective area growth. The image in the right is with high magnification from the box in the left side. The original mask width in these images is 35 µm. The stripe is along < 1 1 0 > direction. The mesa formed by wet etching could be seen from the image. The QW and remaining SiNx are visible and marked in the right image. There is no material deposit on top of the SiNx except at the very edge. These bumps on the edge SiNx QW 109 do not start from the SiNx film but from the grown side wall. The whole sidewall surface exposed by etching is well covered by regrowth material even though the grown Al 0.3 GaAs is only 1.2 µm which is much less than the 3 µm high mesa. One feature need to be noticed is the brightness contrast shown clearly in the bottom right corner of the right hand side image. This dark layer covers all the bottom of trenches but not the side wall. The contrast in the SEM image is due to either material composition difference or doping profile variation. Either case could result from the growth difference on different orientation planes. The shape of the dark layer changes when the TJS orientation is changed. When the stripe is along <1 1 0> direction, as shown in Figure 3-52, the dark layer stops near the bottom edge of etched mesa. But when the stripe is along <1 -1 0> direction, as shown in Figure 3-53, the dark layer extends over the side wall. Figure 3-53 Cross-section Profiles after TJS Al 0.3 GaAs Regrowth, V/III ~ 10 Stripes along <1 -1 0> direction Figure 3-54 shows the mesa side walls after regrowth for both strip orientations. Both of them have rough side walls shown in the images. The rough 110 side walls may cause larger scattering loss in the device. It is not clear whether this roughness is from the sidewall roughness formed during wet etching or is from the orientation dependent growth mechanism. Stripes along <1 -1 0> direction Stripes along <1 1 0> direction Figure 3-54 Mesa Side Wall Images after TJS Al 0.3 GaAs Regrowth, V/III ~ 10 The regrowth test, where the V/III is about 5, is also conducted. Two SEM images are shown in Figure 3-55. Decreasing the V/III ratio from 10 to 5 does not change the regrowth profile much. The bumps at the edges of stripes remain the same. The regrowth layer on the side wall is still rough and even thinner than that grown with V/III ratio ~ 10. Stripes along <1 -1 0> direction Stripes along <1 1 0> direction Figure 3-55 SEM Cross Section Images of TJS Al 0.3 GaAs Regrowth, V/III ~ 5 111 After the first regrowth, second step wet etching and regrowth should be continued to form the other carrier injection layer. There are some more technical challenges for these steps. First, the remaining SiNx from the first etching and regrowth needs to be removed. According to the SEM image, the remaining SiNx is less than 100 nm thick. It takes 3 min before the regrowth to remove the SiNx overhang by using BOE 10:1. However, it takes about 30min to remove the remaining SiNx on mesa top after regrowth. The regrowth process must serve as an annealing process for SiNx and makes it harder to remove. The next challenge comes from the second step wet etching. A new SiNx film is deposited and photolithograph is used to form the mask for the second step etching. The SiNx mask covers part of the mesa and protects the adjacent n injection layer. The etching will expose a fresh side wall for the p injection layer regrowth. However, the surface left after the first etching and regrowth is not flat. There are 3 µm height differences from the highest point to the lowest point. This makes the photolithograph very difficult. An example of unsuccessful second step etching is shown in Figure 3-56. Figure 3-56 Unsuccessful Second Step Etching 112 In this image, the mesa formed in the first etching is from point A to point B. After the first regrowth, two high ridges are formed in there. The SiNx mask for the second etching should extend from C to D. After the etching, the only TJS mesa remaining should be from C to B. From B to D, n contacting layer grown in the first step regrowth is preserved for future metal contact formation. But in this unsuccessful case, the TJS mesa is attacked from point B because the SiNx continuity is broken at point B. This is the highest point on the wafer after the first step regrowth. The photoresist used for cover up the SiNx is very thin and bleaches during the lithograph step to define the SiNx mask. Extra thick photoresist is the key to a successful lithograph and etching. Following the second step etching and regrowth, n and p metal are formed on the corresponding injection material. The device process is done after rapid thermal annealing and cleaving. Figure 3-57 SEM Image of a Completed TJS Device TJS Mesa 113 A completed TJS device SEM image is shown in Figure 3-57. The position of TJS mesa is marked in the graph. The white layers on top of the device are metal contacts. The left side metal contact is the n metal on n injection layer formed in the first regrowth. The right side is p metal on p injection layer formed in the second regrowth. Figure 3-58 SEM Image of TJS Mesa, After Device Processing Figure 3-58 zooms in on the TJS mesa of a completed TJS device. The device shown in this image has a much narrower stripe width than that shown in Figure 3-57. The QW width is about 3 µm. There is a ~ 1 µm by 1 µm large bump on the right side of the mesa top. This is the same as the bumps we noticed in the regrowth test (Figure 3-55). There is a much larger bump on the left side of the mesa top. It is located at the edge formed in the first etching and regrowth. After the first regrowth, there is a small bump which is supposed to be covered by the SiNx and no material should be deposited in the second step regrowth. However, there is still significant QW 114 material deposition in the second regrowth. In this case, where the width of the QW is 3 µm, the distance between two bumps on top of the mesa is about 1 µm. These bumps make it very difficult to make the mesa width small. -2 024 0 2 4 6 QW Width 6 μm 6 μm 5 μm 5 μm 4 μm 4 μm Current (mA) Bias (V) Figure 3-59 IV Curves of TJS Devices with Different QW width The measured IV curves are shown in Figure 3-59. Devices with QW width from 6 µm to 4 µm (measured by SEM) are shown in the graph. All of them have very high resistance (~K Ω). There is no correlation between the QW width and the resistance. There is no any light emitted from the device. The measured resistance is much higher than what is simulated early in this chapter. There are several possible reasons. First, doping behavior on the different crystal planes is different. The doping calibration conducted on the plenary growth may not be applied to the mesa sidewalls. This can causes the carrier injection layer adjacent to the active mesa to become intrinsic or even wrong type of doping. Detailed discussion on the doping issue will be presented. Second, the regrowth interface on the side wall may have 115 lots of defects. More experiments are required to pinpoint the problem. 3.5 Summary Extensive simulations are performed in order to get super high wall plug efficiency semiconductor laser. The simulations demonstrate that excess driving voltage is the main source of loss in the current design of EEL. Most of the excess voltage is caused by the band discontinuity from waveguide layer to QW and the low mobility of waveguide layer. One idea of decreasing the excess voltage is using an asymmetric QW design. The other idea is using a TJS design where the channel structure, asymmetric carrier injection layers and narrow stripes are the key issues to achieve high efficiency. The simulations in both cases predict the wall plug efficiency can be increased to about 80%. Some preliminary experiments on etching and regrowth related to the TJS project are discussed. Much more works are needed in order to realize working devices. References 1 Larry A. Coldren, Scott W. Corzine “Diode Lasers and Photonic Integrated Circuits,” 1995, John Wiley & Sons, Inc. 2 Paula M. Powell, “Laser Diode Manufacturers Turn Up the Power,” Photonics Spectra, vol. 38, no. 2, pp. 82, 2004. 3 Breck Hitz “Military Project Aims to Improve Diode Laser Efficiency,” Photonics Spectra, vol. 39, no. 1, pp. 89, 2005. 116 4 http://www.crosslight.com 5 Valerie Illingworth, “The Penguin Dictionary of Physics,” 1990, Market House Books Ltd. 6 Donald A. Neamen, “Semiconductor Physics and Devices, Basic Principles,” 2003, McGraw-Hill Companies, Inc. 7 W. Susaki, T. Tanaka, H. Kan, and M. Ishii, “New structures of GaAlAs lateral- injection laser for low-threshold and single-mode operation,” IEEE J. Quantum Electron., vol. QE-13, pp. 587–591, 1977. 8 D. A. Suda and J. M. Xu, “Lateral current injection lasers—A new enabling technology for OEIC’s,” NATO Proc. Future Trends in Microelectronics. New York: Kluwer, 1996. 9 Edward H. Sargent and J. M. Xu, “Lateral Injection Lasers,” International Journal of High Speed Electronics and Systems, Vol. 9, No. 4, pp. 941-978, 1998. 10 Ralph Williams, “Modern GaAs Processing Methods,” 1990 Artech House, Inc. 11 Yoshifumi Mori and Naozo Watanabe, “A New Etching Solution System, H 3 PO 4 - H 2 O 2 -H 2 O, for GaAs and Its Kinetics,” J. Electrochem. Soc: Solid-State Science and Technology, Vol. 125, No. 9, pp. 1510, 1978. 12 C. Angulo Barrios et al, “GaAs/AlGaAs Buried Heterostructure Laser by Wet Etching and Semi-insulating GaInP:Fe Regrowth,” Electrochemical and Solid-State Letters, vol. 3, no. 9, pp. 439-441, 2000. 13 Won-Jin Choi, P. Daniel Dapkus, “Selective growth and regrowth of high Al content AlGaAs for use in BH lasers,” Journal of Crystal Growth Vol. 195 pp. 495- 502, 1998. 14 Seigo Ando and Takashi Fukui, “Facet Growth of AlGaAs on GaAs with Si0 2 Grattings by MOCVD and Applications to Quantum Well Wires,” Journal of Crystal Growth Vol. 98 pp. 646-652, 1989. 15 S. M. Sze, “Physics of Semiconductor Devices,” 1981, John Wiley & Sons, Inc. 117 Chapter 4 : InGaAs Multi-QW Grown On Twist Bonded Compliant Substrate By combining different materials on the pre-existing substrate, it is possible to realize versatile functionalities. Researchers are able to use state of art growth technologies such as Molecular Beam Epitaxy (MBE) and Metal-Organic Chemical Vapor Deposition (MOCVD) to grow materials coherently. Although lattice matched materials are easier to be grown, growth of lattice mismatched materials can lead to exciting new frontiers. If the misfit between the epilayer and substrate is relatively small, the first atomic layer deposited is fully strained and coherently match to the mismatched substrate. As the thickness of the epilayer increases, the strain energy becomes very large. At a certain thickness, which is nominally called the critical thickness, misfit dislocations start to form in the epilayer and degrade the material quality. [1,2] Normally, the dislocations form only at the interface and in the epilayer because the substrate is much thicker and hence more rigid. When the substrate is relatively thin compared to the epilayer, it becomes flexible and dislocations formation will occur in both the epilayer and substrate. If it is thin enough, dislocations will form only the substrate. This effect will leave a dislocation-free epilayer. The thin & flexible substrate which accommodates the misfit strain is called a Compliant Substrate. [3] The outline of this chapter is as following. First, the concept of coherent growth and critical thickness will be explained. Then the idea of compliant substrate 118 will be discussed. Following this, the experimental results on the compliant substrate made by twisted wafer bonding will be presented. The working mechanism of this type of compliant substrate will be discussed base on obtained results. 4.1 Introduction of Compliant Substrate 4.1.1 Ideal Compliant Substrate Let’s consider the following condition: two cubic materials, an epilayer of thickness epi h is on a substrate of thickness s h [3]. The misfit between epilayer and substrate is f . The interface is a common (001) plane. The condition of mechanical equilibrium at the interface is: 0 = + s s epi epi h h σ σ (4.1) with s s s s s s s K ε ε ν ν μ σ = − + = 1 ) 1 ( 2 epi epi epi epi epi epi epi K ε ε ν ν μ σ = − + = 1 ) 1 ( 2 and f s epi = − ε ε (4.2) where ε , σ and h are, strain, stress and thickness respectively. Subscript epi refers to the epilayer and subscript s refers to the substrate. μ and ν are the shear modulus 119 and the Poisson ratio respectively. According to these equations, the strain actually partition between epilayer and substrate. Combining two equations above, we have the strain on the epilayer as: epi epi s s s s epi h K h K h K f + = ε (4.3) In normal epitaxy, the thickness of substrate is much larger than the thickness of epilayer, i. e. epi s h h >> . The last equation simplifies to f epi = ε , which means all the strain is located in the epilayer. When the strain energy becomes very large, the misfit dislocations start to form only in the epilayer. However, when the thickness of the substrate is comparable to the thickness of the epilayer, the strain energy will exist in both layers. The elastic stored energy per unit area of the interface is given by: s s epi epi epi epi h s s epi epi eip h K h K h f K h K h K W + = + = 1 2 2 2 2 ε ε (4.4) When the epilayer and substrate have similar elastic constants ( s epi K K = ), the critical thickness, c h , before any strain relief is: s c c h h h 1 1 1 0 − = (4.5) where 0 c h is the critical thickness on a bulk substrate. According to this equation, when the substrate thickness is small, the critical thickness c h could be significantly larger than the case of bulk substrate. If 0 c s h h < , then an infinitely thick epilayer can 120 be grown on top of the thin substrate. 4.1.2 Compliant Substrate Realized As discussed above, an ideal compliant substrate is a very thin layer of material on which the misfit material could be grown without generating any defects in the epilayer. The thickness of the compliant layer should be on the order of a few nm. It should remain flat and planar. If supported, the layer should be free to glide on the supporting media. Figure 4-1 Several realizations of the compliant substrate Several experimentally realized compliant substrates are shown in Figure 4-1, from Ref. [3]. The compliant substrate was first realized with GaAs membranes supported at several points, as shown in (a). However, film bulging was observed in the experiment. [4] A free standing thin InGaAs membrane supported on a central 121 pedestal has also been demonstrated, as shown in (b). [5] The structures formed by these two methods are fragile, not planar. It is very difficult to realize them for device applications. Shown in (c) is the compliant substrate on a mechanical host substrate with a viscous layer. The viscous layer could be SiO 2 , Indium or Glass. At typical growth temperatures, these viscous layers become so soft that the compliant layer can slip on the host substrate. But it is not very efficient in release the strain. [6, 7, 8, 9] The most widely researched technique so far is a thin layer bonded to a mechanical host substrate of the same material but with a large twist angle between them, shown in (d). [10] In this thesis, focus will be on the compliant substrate made by this method. 4.1.3 Compliant Substrate Made by Twisted Wafer Bonding The process of making a compliant substrate by twisted wafer bonding technology is shown in Figure 4-2. (a) Etch stop layer and compliant layer growth on sacrificial substrate (b) Wafer Bonding on mechanical host substrate (c) Selective etching of sacrificial substrate and etch stop layer Sacrificial substrate Etch stop layer Compliant Layer Mechanical host substrate (a) Etch stop layer and compliant layer growth on sacrificial substrate (b) Wafer Bonding on mechanical host substrate (c) Selective etching of sacrificial substrate and etch stop layer Sacrificial substrate Etch stop layer Compliant Layer Mechanical host substrate Figure 4-2 Process of Making Compliant Substrate by Wafer Bonding with Twist 122 First, the etch stop and compliant layers are grown on the sacrificial substrate. The wafer is then flipped over and wafer bonded on to the mechanical host substrate. The wafers are not aligned with each other during bonding. There is a large angle between the cleaved edges of the wafers. Finally, the sacrificial substrate and the etch stop layer are removed to expose the compliant layer. There is another method which uses the smart-cut® procedure instead of the etch-stop layer [11]. Wafer bonding is a process in which two wafers adhere to each other without the application of any macroscopic gluing layer. This process can be done at room temperature without any outside force. Van der Waals attraction force pulls two wafers together. An annealing process which follows initial bonding increases bonding strength by altering the bonding mechanism. After the annealing process, hydrogen atoms form bonds with the atoms on both sides of the bonding interfaces. This provides the main attraction force for the wafer bonding. [12] When the twist-bonded compliant substrate was first realized, the idea behind it was same as the compliant layer with a viscous layer. As the wafer is bonded to a bulk crystal with a large deliberated angular misalignment between their <011> surfaces, it is believed that bonds between them are weaker than the covalence bonds in the wafer. The qualitative argument is that bonds are so distorted in a twist boundary that they can be rearranged during a glide with a small energy or even non- existent barrier. Therefore it closely approximates an ideal compliant substrate described earlier in this chapter. [3, 10] 123 4.2 Experiments 4.2.1 Compliant Substrate Preparation and MOCVD Growth The compliant layer is first grown by MOCVD on top of an epi-ready GaAs N+ (001) wafer. There is 100nm of Al 0.8 GaAs, which serves as a etch stop layer, grown under the compliant layer. In the work presented here, GaAs is the only compliant layer material. The thicknesses of compliant layers grown are 10nm or 20nm. The compliant layer wafer and host substrate (epi-ready GaAs N+ (001) Wafer) surfaces are first checked with an optical microscope to ensure that there are no particles contaminating them. If necessary, particles are removed by a swab with TCE. The surfaces are cleaned by Acetone and Methanol for 5 minutes each in an ultrasonic bath. This is followed by 10 minutes DI water rinsing. The surface oxide is then removed by NH 3 OH 1:5 water solution in ultrasound bath for 5 minutes. It is followed by another 10 minutes DI rinsing. Finally, the two wafers are putting together face to face in the DI water. These last three steps must be done without exposing the wafer surface to air. The wafers are then transferred into the bonding fixture and loaded into bonding chamber. The bonding chamber is first pumped down by a mechanical pump, and the temperature is increased to 95 ºC for 30 minutes to remove the water vapor. Then the chamber is filled with hydrogen. The bonding is done under a hydrogen environment at one atmosphere pressure. The temperature is gradually increased to 600 ºC in 5 124 minutes. The sample is maintained at the bonding temperature for 30 minutes before gradually cooled down to 150 ºC in 15 minutes. The size of compliant substrate prepared is limited by our bonding fixture. The maximum sample size which can be loaded is 2 cm by 2 cm square. The sacrificial substrate with compliant layer is twist bonded, so it is smaller than the size of the host substrate. The largest compliant substrate is only slightly larger than 1 cm by 1 cm square. After wafer bonding, the sacrificial substrate is first lapped down to 50~100 µm thick. Then it is spray etched by NH 3 OH:H2O2~1:50 until the Al 0.8 GaAs, which is shining red, is exposed. The Al 0.8 GaAs etch stop layer is removed by HCl:H 2 O~1:1. Then the sample is rinsed by DI water for 30 minutes before being loaded into the reactor. The whole etching process is done right before the growth to avoid deterioration of surface when exposed to air. In 0.2 GaAs 30nm In 0.4 GaAs 7nm × 5 In 0.2 GaAs 300nm Figure 4-3 InGaAs Multi-Quantum Well Epilayer Structure The sample temperature is increased to 600 ºC. The growth starts after 10 minutes. First, a 300nm InGaAs buffer layer is grown. Then 5 pairs of In 0.2 GaAs /In 0.4 GaAs QWs are grown, as shown in Figure 4-3. The lattice constant of In 0.2 GaAs is 1.4% higher than GaAs. The structure grown here is about 50 times thicker than the critical thickness of about 10 nm. The growth conditions are listed as follows: 125 reactor pressure is 76 Torr; growth temperature is 600 ºC; In 0.2 GaAs growth rate is 6Å/S; TMGa, TMIn and Arsine are used as sources; V/III ratio is 20; H2 carrier flow is 10 slm. Indium composition and InGaAs growth rate are determined by X-Ray diffraction. 4.2.2 Compliant Substrate Based on GaAs In this section, I will discuss the results of growths on compliant substrates which have the epi-ready GaAs as the mechanical host wafer. Four characterization methods are used: Atomic Force Microscopy (AFM), X-Ray Diffraction (XRD), low temperature photoluminescence (PL), and Transmission Electron Microscopy (TEM). InGaAs MQWs structure was grown in the same growth run on three substrates. One was 10nm GaAs twist-bonded on GaAs, another one was 20nm GaAs twist-bonded on GaAs, and the last one was a GaAs reference wafer. The GaAs compliant layers thicknesses stated in here and in the future text are the thicknesses of original compliant layers grown but not the thicknesses of compliant layer left after sample preparation processes. During the sample preparation stage of the wafer bonding process, the surface oxidize layer is removed once by wet etching. Then the newly formed surface oxidize layer is removed again before the growth by thermal etching. The thickness of the oxidized GaAs surface layer is about 3nm [13], so the actual compliant layer thickness after the final growth 126 is about 6 nm less than what it is nominally stated as. For example, in the case of the 20nm GaAs twist-Bonded on GaAs, the final compliant layer thickness is about 14nm. This will be proved by the TEM images. An AFM image of the reference sample is shown in Figure 4-4. The <110> directions of substrate The <110> directions of substrate Figure 4-4 AFM Image of InGaAs MQWs Grown on GaAs Epi-Ready Substrate (x and y scale in µm) The thickness of InGaAs grown is much thicker than the critical thickness calculated. Hence, many defects form during growth. In Figure 4-4, a deep pattern of cross hatches, which directions are along the <110> directions of the substrate, is clearly shown. The arrows marked on the graph are along the <110> directions of the substrate. The directions of the cross hatches follow the directions of misfit dislocations, which are the <110> directions. This phenomenon has been extensively discussed in the literature [14,15]. In the twist-bonded compliant substrate, the host substrate and compliant layer have different crystal orientation. If it closely approximates the ideal free 127 standing compliant substrate, then the misfits and cross hatches should be located along <110> directions of the compliant layer, since the epilayer literally should not be effected by the existence of host substrate. However, this is not always true from what observed in the experiments. The <110> directions of host substrate The <110> directions of compliant layer The <110> directions of host substrate The <110> directions of compliant layer Figure 4-5 AFM Image of InGaAs MQWs Grown on 20nm GaAs Compliant Substrate An AFM image of InGaAs MQWs grown on 20nm GaAs compliant substrate is shown in Figure 4-5. The compliant layer is 30 degree twisted from the host substrate. During the AFM measurement, the scanning direction is along one of the <110> directions of host substrate. The cross hatches are along the directions of <110> of compliant layer, which is 30 degree twisted from the host substrate. 128 The <110> directions of host substrate The <110> directions of compliant layer The <110> directions of host substrate The <110> directions of compliant layer Figure 4-6 AFM Image of InGaAs MQWs Grown on 10nm GaAs Compliant Substrate An AFM image of an InGaAs MQWs grown on 10nm GaAs compliant substrate is shown in Figure 4-6. The compliant layer is 45 degree twisted from the host substrate. During the AFM measurement, the scanning direction is along one of the <110> directions of compliant layer. The cross hatches are along the <110> directions of host substrate, which is 45 degree twisted from the compliant layer. Therefore, two different orientations of cross hatch patterns observed in the growths on these two compliant substrates. It is easy to imagine that when the compliant layer is very thick, and the epilayer interface is far away from the bonding interface, the compliant layer could behave just like a normal substrate. But, when the compliant layer is very thin, apparently, it is not like free standing film according to the AFM images. The effects of the bonding interface strongly interfere with the epilayer growth. Misfit dislocations are normally generated at the interface between two misfit materials and glide along {110} planes. However, when the compliant 129 layer is very thin, an unusual dislocation generation and propagation may happen. Then next step is to determine whether the whole epilayer changed their crystal orientation to the host substrate or only the defects are affected. An effective method of verifying crystal orientation is X-ray diffraction. The diffraction of X-ray is the result of interaction between the incident X-ray and the crystal. The diffraction will occur when the Bragg’s condition is fulfilled: θ λ sin 2d n = (4.6) where λ is the wavelength of the X-ray, d is the perpendicular distance between planes that form a family, θ is half the angle through which the incident beam is diffracted by a particular family of planes, and n is the order of the diffraction. [16] [001] [110] X-Ray Ф Ω 2θ [001] [110] X-Ray Ф Ω 2θ Figure 4-7 Schematic Drawing of X-Ray Diffraction Measurement Setup The X-ray diffraction of {026} planes of (001) wafer can only be observed when the incident plane of X-ray is along the {100} planes, which is 45 degrees off from the cleaved edge of GaAs. By observing the X-ray diffraction from the {026} planes, we can determine that whether the crystal orientation of the compliant layer is same as that holding substrate. 130 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1 10 100 1000 10000 Ω−2θ scan of {026} plane InGaAs on GaAs (001) Φ=45 0 Intensity (a.u.) Ω Figure 4-8 XRD Ω-2 θ Scan of {026} plane of InGaAs MQWs on GaAs substrate The {026} X-ray diffraction of the reference sample is shown in Figure 4-8. The signal can only be collected when the Ф is ± 45° or ± 135°, as mentioned previously. Peaks from both the GaAs substrate and InGaAs epilayer can be seen at the same Ф because they have the same crystal orientation. 0 10000 20000 30000 40000 50000 0 15 30456075 90 105 0 10 20 30 40 50 60 70 80 90 (026) InGaAs Phi Intensity (a. u.) (026) GaAs Figure 4-9 Ф scan of {026} XRD Peaks, 20nm GaAs Twist-Bonded Compliant Substrate 131 The {026} XRD results of the sample grown on a 20nm GaAs twist-bonded compliant substrate are shown in Figure 4-9. Two measurements need to be done to generate this graph. First, the Ω and 2 θ are set at the peak of GaAs {026} diffraction before scanning Ф (sample rotation). The solid curve shown in the graph is from this scan. Then, Ф is scanned again after Ω and 2 θ are set at the peak of InGaAs {026} diffraction. The dotted curve is obtained from this Ф scan. The peak of GaAs is on Ф=45°, which is one of the <100> directions of host substrate. The peak of InGaAs is on Ф=75°. The difference between the peak positions of these two materials is 30°, which is the twisting angle during wafer bonding. So it is clear that in this sample, the epilayer has the same orientation as the compliant layer and the defects are generated along the <110> directions of the compliant layer. 0 153045 607590 105 0 10 20 30 40 50 60 70 80 0 10000 20000 30000 40000 50000 (026) InGaAs Intensity (a.u.) Phi (026) GaAs Figure 4-10 Ф scan of {026} XRD Peaks, 10nm GaAs Twist-Bonded Compliant Substrate The {026} XRD results of the sample grown on the 10nm GaAs twist-bonded compliant substrate are shown in Figure 4-10. Similar to the experiments shown in Figure 4-9, the Ф scans of both the GaAs and InGaAs (026) XRD peaks are shown. 132 The position of the GaAs peak is still at Ф=45°. The position of the InGaAs peak is at Ф=45° too, although the compliant substrate is 45° twisted from the host substrate. This proves that epilayer has the same orientation as the host substrate and the defects are generated along the <110> directions of the host substrate too. In the literature, the best results on twist-bonded compliant substrate come from layers with thickness of less than 10nm [3,10,17,18]. However, one study claims that epilayer has the same crystal orientation as twist-bonded compliant layer [10] while the other one claims that epilayer has the same crystal orientation as the host wafer [18]. Our experimental results are consistent with the latter one. From the above experiments, it is proved that the twist bonded compliant substrate in our experiments is not even close to an ideal free standing film. The host substrate still plays an important roll in the strain relaxation through the compliant layer – host substrate interface. Besides determining the crystal orientation, XRD is also used to investigate the lattice relaxation of the material grown on compliant substrate. 133 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 InGaAs Peak 20nm GaAs on GaAs 30 o twist 10nm GaAs on GaAs 45 o twist GaAs N+ (001) X-Ray Intensity (a.u.) Log Scale Δ Ω Figure 4-11 (004) Ω-2 θ Scans of InGaAs QW grown on GaAs Based Compliant Substrates XRD (004) Ω-2 θ scans of InGaAs QW grown on GaAs Based Compliant Substrates are shown in Figure 4-11. There are three samples shown in the graph. A (004) Ω-2 θ scan provides the information of lattices constant in the growth direction. In scanning range used in this graph, XRD shows the peak of GaAs, at ∆Ω=0, and the peak of In 0.2 GaAs. There are little differences between these results. The vertical line in the graph marked peak position of InGaAs peaks for viewing convenience. XRD (004) Ω-2 θ scan measures the vertical lattice constant ⊥ a of epilayer through following equation: ) sin( ) sin( ΔΩ + Ω Ω = ⊥ sub sub sub a a (4.7) Where sub a is the lattice constant of substrate. In our case it is the lattice constant of GaAs, 5.6533 Ǻ. sub Ω is the Bragg angle of substrate. For GaAs at (004) condition in our XRD system, sub Ω is 33.04°. Δ Ω is the distance from the substrate 134 XRD peak to the epilayer peak. In the case of fully relaxed In 0.2 GaAs, ⊥ a is 5.7343 Ǻ. ΔΩ calculated is -0.525°. In the case of elastically strained layer, which means the lateral lattice constant of defect free epilayer is forced to that of the substrate: ν ν ν − − + = ⊥ 1 2 ) 1 ( sub relax a a a (4.8) Where ν is the Poisson ratio, which is 0.33 for In 0.2 GaAs. relax a is the fully relaxed lattice constant of epilayer. Calculating from the above two equations, the ΔΩ of elastically strained In 0.2 GaAs grown on GaAs is -1.025°. Equations and constants used in the above calculation are from reference [19]. Comparing the calculation here and the data shown in Figure 4-11, it is clear that In 0.2 GaAs shown is very close to fully relaxed. -2.0 -1.5 -1.0 -0.5 0.0 0.5 10 0 10 1 10 2 10 3 10 4 10 5 (004) Ω−2 θ Scan InGaAs MQW on GaAs Experiment Simulation (Strained) X-Ray Diffraction Intensity ΔΩ Figure 4-12 XRD Experiment and Simulation of InGaAs MQWs on GaAs The simulated XRD data of InGaAs MQWs grown on GaAs and the 135 experimental data are shown in Figure 4-12. In the simulation, an elastically strained condition is considered. There are several major differences between the simulated curve and experiment. First, the peak position of InGaAs Bulk layer in the experiment is much closer to the GaAs peak, which indicates that InGaAs has relaxed significantly. Second, the FWHM of the InGaAs bulk layer peak in the experiment is very large compared to the simulation. This reflects the fact that the lattices of InGaAs are not uniform. Third, no clear peaks corresponding to the MQWs could be clearly observed in the experiment. This is due to the bad quality of the interfaces of the MQWs. All these features are also observed with AFM. 1100 1120 1140 1160 1180 1200 1220 PL Intensity (a.u.) GaAs (001) N+ 10nm GaAs on GaAs 45 o twist 20nm GaAs on GaAs 36 o twist Wavelength (nm) Figure 4-13 Photoluminescence of InGaAs MQWs Grown on GaAs Based Compliant Substrate The quality of the MQWs grown on compliant substrate was also check by PL. Photoluminescence data was taken at liquid nitrogen temperature, 77K. The PL was excited by a helium neon laser. The laser power was 5 mW, the laser wavelength was 632.8 nm. The PL signal was measured by a 0.5 meter monochrometer 136 (SpectraPro 500i, Acton Research Corporation) and an ultra sensitive germanium detector (Edinburgh Instruments). Three typical PL results are shown in Figure 4-13. The solid curve is the PL of MQWs grown on a normal GaAs substrate which serves as a reference sample. The dashed curve is the PL of MQWs grown on one compliant substrate, 10 nm GaAs bonded on GaAs with a 45 degree twist. The PL peak intensity and peak position is comparable to the reference. The dotted curve is the PL of MQWs grown on another compliant substrate, 20 nm GaAs bonded on GaAs with a 36 degree twist. The PL peak position is close to the reference but the peak intensity is about half. 4.2.3 Compliant Substrate Based on Ion Implanted GaAs As discussed earlier in this chapter, when the substrate is thicker and more rigid, all defects are generated in the epilayer. Introducing the compliant substrate should help to release part of or all strains in the compliant layer. But the experiments indicate that host substrate still strongly influences the growth on the compliant layer. In the next section, a substrate which is no longer rigid and nearly perfect, but is damaged and full of defects is used as the host substrate in order to release part of the misfit strain. Ion implantation is an effective way to generate defects in the crystal. In our case, most defects should be generated close to the surface in order to have strong impact on the compliant layer – host substrate interface. Software simulation is used to decide the ion type, energy, and dosage applied. The simulation program used is 137 SRIM. This program simulates ion distribution and damage to the crystal during ion implantation. [20] Results of two simulations are shown in Figure 4-14, Figure 4-15, Figure 4-16 and Figure 4-17. The first two plots are simulations of hydrogen ions of energy 70 KeV are implanted into GaAs with incident angle 7 degrees from the surface normal. The other two plots are simulations of arsenic ions energy 180 KeV are implanted into GaAs with incident angle 7 degrees from the surface normal. The distribution of implanted ions and generated vacancies are shown. Figure 4-14 Simulated Ion Range for H + 70 KeV 7 Degree 138 Figure 4-15 Simulated Collision Events for H + 70 KeV 7 Degree Figure 4-16 Simulated Ion Range for As + 180 KeV 7 Degree 139 Figure 4-17 Simulated Collision Events for As + 180 KeV 7 Degree There are two advantages in using arsenic ions for implantation. First, most of the implanted arsenic ions and generated vacancies are closer to the surface compared to hydrogen ions even though hydrogen ions have less energy. Second, arsenic ions cause about 3 orders more damage (vacancies) than hydrogen ions. So, arsenic ion was selected for implantation. The ion energy was 180 KeV and the incident angle was 7 degrees from normal. The ion energy was optimized to get damage deep into the crystal while maintain even amount of damage from the surface to the inside. Three ion implantation dosages, 1E12 cm -2 , 1E13 cm -2 and 1E14 cm -2 , were used in experiments. Calculating from Figure 4-17, the vacancy densities near the surface are 2E20 cm -3 , 2E21 cm -3 and 2E22 cm -3 respectively. The original ion density in GaAs is about 5E22 cm -3 [20]. At the highest dosage we used, the vacancy density was close to the atom density of GaAs itself. When ion implantation dosage is1E14 cm -2 , wafer bonded compliant substrate can not be successfully prepared. Bonded layers peel off easily.Compliant 140 substrates can be successfully prepared on ion implanted GaAs when the dosages are 1E12 cm -2 and 1E13 cm -2 . X-Ray results of the host substrates before growth are shown in Figure 4-18 and Figure 4-19. Shown in these two graphs are the (004) Ω-2 θ scans of GaAs (001) substrates before ion implantation, after ion implantation and after compliant layers are bonded. It is clear that ion implantation significantly alters the host substrates. There is a shoulder or even peaks in the left side of the main GaAs peak, which corresponds to larger lattice constants. According to the simulation, most of these changes happen on the surface. The wafer bonding process is literally an annealing process for the host substrate itself. After the compliant layer is bonded and sacrificial substrate is removed, two samples which have different ion implantation dosages have similar XRD results. The samples of high dosage have stronger side peaks than that of low dosage samples. -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 10 0 10 1 10 2 10 3 10 4 10 5 XRD (004) Ω−2 θ Scan GaAs As + 180KeV 1E12 after bonding GaAs As + 180KeV 1E12 before bonding GaAs (100) N+ Substrate XRD Intensity (a.u.) Δ Ω Figure 4-18 XRD of Ion Implanted GaAs (Dosage=1E12 cm -3 ) 141 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 10 0 10 1 10 2 10 3 10 4 10 5 XRD (004) Ω−2θ Scan GaAs As + 180KeV 1E13 after bonding GaAs As + 180KeV 1E13 before bonding GaAs (100) N+ Substrate XRD Intensity (a.u.) Δ Ω Figure 4-19 XRD of Ion Implanted GaAs (Dosage=1E13 cm -3 ) Four compliant substrates will be discussed in this section. First one is 10nm GaAs bonded on ion implanted GaAs with a 58 degree twist. The ion implantation dosage is 1E12 cm -3 . Second one is 20nm GaAs bonded on ion implanted GaAs with a 23 degree twist. The ion implantation dosage is 1E12 cm -3 . Third one is 10nm GaAs bonded on ion implanted GaAs with a 24 degree twist. The ion implantation dosage is 1E13 cm -3 . And the last one is 20nm GaAs bonded on ion implanted GaAs with a 20 degree twist. The ion implantation dosage is 1E13 cm -3 . I will only identify these compliant substrates with their thickness and ion implantation dosage. The crystal orientations of the growths on these compliant layers have the same features as observed with the compliant substrates on normal GaAs substrates: the growth has the same orientation as the substrate when the compliant layer is 10 nm thick and it has the same orientation as the compliant layer when the compliant layer is 20 nm thick. 142 1050 1100 1150 1200 0 50 100 150 200 250 300 350 400 Reference 10nm 1E12 20nm 1E12 10nm 1E13 20nm 1E13 PL intensity (a.u.) Wavelengh (nm) Figure 4-20 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs The PL data of these samples are shown in Figure 4-20. The PL of the reference sample, which is InGaAs MQWs grown on normal GaAs substrate, is shown as the solid line. Samples grown on 10 nm compliant substrates are slightly better than samples grown on 20 nm compliant substrates. Samples grown on compliant substrates in which host substrates receive high dosage ion implantation are better than those with low dosage. The best sample is on 10 nm twist bonded compliant substrate based on As+ 180 eV 1E13 cm -3 ion implanted GaAs. The PL peak intensity is only slightly better than the reference. 143 -2 -1 0 1 10 0 10 1 10 2 10 3 10 4 10 5 (004) Ω−2θ Scan InGaAs MQW 20nm 1E13 10nm 1E13 20nm 1E12 10nm 1E12 Reference XRD Intensity ΔΩ Figure 4-21 XRD of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs The XRD data of these samples are shown in Figure 4-21. As introduced earlier, the highest peak in the right is the peak of GaAs substrate. The broad peak in the middle is the peak of In 0.2 GaAs. The ripples in the left are the peaks of the MQWs. Data from all these samples are very similar to each other. The biggest difference comes from the In 0.2 GaAs peak from the sample of 10nm 1E13 cm -3 , which also has the largest PL signal among all growths on compliant substrates based on ion implanted GaAs. It is shifted to the right hand side comparing with all the other In 0.2 GaAs peaks. The position of In 0.2 GaAs peak from the reference sample is marked by a vertical dashed line in the graph. This shift indicates that InGaAs is more relaxed in this sample. 4.2.4 Compliant Substrate Based on InGaAs 300 nm In 0.2 GaAs epilayer grown on GaAs substrate was also used as the 144 host substrate. The critical thickness of In 0.2 GaAs grown on GaAs is about 10 nm. On the surface of 300nm epilayer, In 0.2 GaAs is nearly fully relaxed. When 10 nm GaAs is twist-bonded onto GaAs, the bonded compliant layer is deformed according to the host GaAs. If the bonded compliant layer can be deformed to the relaxed In 0.2 GaAs, then the InGaAs epilayer grown on the compliant layer doesn’t have any strain. However, there are lots of defects in the host substrate. The ideal condition is that defects generated in the host InGaAs layer will not propagate into the compliant layer and the InGaAs layer grown on it. Figure 4-22 AFM Image of Compliant Substrate Based on InGaAs An AFM image of compliant substrate made by 20 nm GaAs bonded on 300nm InGaAs with a 34 degree twist is shown in Figure 4-22. An AFM image of compliant substrate made by 10 nm GaAs bonded on 300nm InGaAs with a 47 degree twist is the same as the AFM shown here. The AFM scans are along the (110) direction of host substrate. The scale marked in the graph is µm. The pattern shown 145 here is the same as the cross hatch pattern of thick InGaAs grown on GaAs, for example, Figure 4-4. But, the depth of cross hatch is smaller. The bonding process makes the surface smoother. Figure 4-23 AFM Image of InGaAs MQWs Grown on 10 nm GaAs Compliant Substrate Based on InGaAs An AFM image of InGaAs MQWs Grown on compliant substrate made by 10 nm GaAs bonded on 300nm InGaAs with a 47 degree twist is shown in Figure 4-23. The AFM scans are along the (010) direction of host substrate. The scale marked in the graph is µm. The deep cross hatch pattern is clearly shown in the graph. 146 Figure 4-24 AFM Image of InGaAs MQWs Grown on 20nm GaAs Compliant Substrate Based on InGaAs An AFM image of InGaAs MQWs Grown on a compliant substrate made by 20 nm GaAs bonded on 300nm InGaAs with a 34 degree twist is shown in Figure 4-24. The AFM scans are along the (110) direction of host substrate. The scale marked in the graph is µm. The cross hatch pattern shown in the graph is not very clear. Observing the two AFM images above, we can still conclude: when the compliant layer is 10 nm GaAs, the growth has the same orientation as the substrate; when the compliant layer is 20 nm GaAs, the growth has the same orientation as the compliant layer. 147 1100 1150 1200 1250 0 200 400 600 800 1000 PL Intensity (a.u.) GaAs (100) N+ B47 10nm GaAs on 300nm InGaAs 47 o twist B48 20nm GaAs on 300nm InGaAs 34 o twist 300nm InGaAs Wavelength (nm) Figure 4-25 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on InGaAs The PL data of InGaAs MQWs grown on a compliant substrate based on InGaAs is shown in Figure 4-25. The solid line is the InGaAs MQWs grown on normal GaAs substrate as reference. The dotted line is for InGaAs MQWs grown on the 20nm GaAs compliant substrate based on InGaAs with a 34 degree twist. The peak PL intensity of this sample is 70% of the reference. The dash dot line is for InGaAs MQWs grown on the 10nm GaAs compliant substrate based on InGaAs with a 47 degree twist. The peak PL intensity of this sample is 4 times higher than the reference. Also in the graph is the PL of InGaAs MQWs grown directly on the host substrate. The peak PL intensity of this sample is 2.5 times higher than the reference. 148 -2.0 -1.5 -1.0 -0.5 0.0 0.5 10 0 10 1 10 2 10 3 10 4 10 5 (004) Ω−2 θ Scan InGaAs MQW 20nm GaAs on 300nm InGaAs 34 o twist 10nm GaAs on 300nm InGaAs 47 o twist GaAs (100) N+ XRD Intensity (a.u.) ΔΩ Figure 4-26 XRD of InGaAs MQWs Grown on Compliant Substrates Based on InGaAs The XRD data of these samples are shown in Figure 4-26. Data for all of these samples are very similar to each other. 4.2.5 Transmission Electron Microscopy Results Limited TEM images of compliant substrate samples are available. But some important information is still obtained from them. Shown in Figure 4-27 are two images with small magnification. The host substrate, compliant layer and whole epilayer can be found from them. The graph in the left shows the growth on 20nm GaAs bonded on GaAs substrate with a 36 degree twist. The graph in the right shows the growth on 10 nm GaAs bonded on ion implanted GaAs with a 24 degree twist. In the left graph, more dislocations extend through whole epilayer. All the dislocations shown in the left graph are confined 149 above the compliant layer. There are no dislocations shown in the host substrate. There are less dislocations extending through epilayer shown in the right graph. There are defects in the host substrate due to ion implantation. There are also dislocations extending from the compliant layer to the host substrate because some of the strain energy is released into the host substrate. ( ) Bright Field TEM Substrate: 20nm GaAs bonded on GaAs with 36 degrees twist ( ) Bright Field TEM Substrate: 10nm GaAs bonded on ion implanted GaAs with 24 degrees twist 500 nm 0 2 2 004 ( ) Bright Field TEM Substrate: 20nm GaAs bonded on GaAs with 36 degrees twist ( ) Bright Field TEM Substrate: 10nm GaAs bonded on ion implanted GaAs with 24 degrees twist 500 nm 500 nm 0 2 2 004 Figure 4-27 TEM Images of InGaAs MQWs Grown on Two Compliant Substrates ( ) Bright Field TEM Substrate: 20nm GaAs bonded on GaAs with 36 degrees twist 20nm GaAs Holding Substrate GaAs Bonding layer ~14nm 0 2 2 InGaAs Epilayer Figure 4-28 TEM Image of Compliant Layer from 20nm GaAs Twist Bonded on GaAs 150 Shown in Figure 4-28 is a TEM image of the growth on 20nm GaAs bonded to GaAs with a 36 degree twist. This image has much larger magnification to show the compliant layer and interfaces. The two interfaces shown in the image are the interfaces of GaAs compliant layer. As discussed earlier, the real thickness of the compliant layer is about 6 nm thinner than the nominal compliant layer thickness due to the removal of natural oxide layers. The thickness shown in this image proves it. There are clear dislocations on the compliant layer - epilayer interface. There are dislocations extending from that interface to the compliant layer. There is also a dislocation extending from the compliant layer - epilayer interface to the compliant layer. ( ) beam Dark Field TEM Substrate: 10nm GaAs bonded on ion implanted GaAs with 24 degrees twist 20nm As ion implanted GaAs Substrate GaAs Compliant layer ~4nm InGaAs Epilayer 0 2 2 Figure 4-29 TEM Image of Compliant Layer from 10nm GaAs Twist Bonded on Ion Implanted GaAs Shown in Figure 4-29 is a TEM image of the growth on 10nm GaAs bonded on ion implanted GaAs with a 24 degree twist. This image has nearly the same magnification as the last TEM image. The GaAs compliant layer can be seen in the 151 image but the interfaces are blurry. There are several dislocations confined to the compliant layer. There are lots of defects in the host substrate due to ion implantation. There are not many TEM images beyond these shown here. More TEM should be done with different beam conditions to reveal the details of defects observed. The propagation and interaction between defects near the interfaces are the keys to understanding the role of compliant substrate. We also lack TEM images of compliant substrates in which InGaAs acts the host substrate. 4.3 Discussion and Summary In this section, some problems in the compliant substrate research will first be presented. Then our experimental results will be summarized and the working mechanism of compliant substrate will be discussed. Suggestions will be made on possible future research. 4.3.1 Compliant Substrate Experiment Uniformity and Reproducibility The process of fabricating twist bonded compliant substrate is very complicated. This complication causes the problems of non-uniformity and non- reproducibility in my compliant substrate research work. Shown in Figure 4-30 is the PL data of InGaAs MQWs grown on an ion implanted GaAs based compliant substrate. The data from these samples was shown in Figure 4-20 earlier. In Figure 4-30, PL data taken from multiple locations on these samples is shown. Each PL curve in the graph corresponds to a different location on 152 the wafer. The divergence of the data reflects the non-uniformity problem in the compliant substrate research. 1050 1100 1150 1200 1250 0 100 200 300 400 1050 1100 1150 1200 1250 0 100 200 300 400 10nm 1E12cm -3 PL Intensity (a.u.) 10nm 1E13cm -3 20nm 1E12cm -3 Wavelength (nm) 20nm 1E13cm -3 PL of InGaAs MQW Grown on Ion Implanted GaAs Based Complaint Substrate Figure 4-30 Photoluminescence of InGaAs MQWs Grown on Compliant Substrates Based on Ion Implanted GaAs Although there is large divergence in the data shown, the trend of compliant substrate effects is represented if we select only the best data from each sample. That was the approach for choosing the data shown in this thesis. Non-reproducibility was a problem which jeopardized credibility of experiments. The primary uncertainty came from the fabrication of compliant substrate. First, the wafer bonding process is very sensitive to the bonding surfaces. A particle on bonding interface often prevents a piece of material, as large as 4 mm by 4 mm, from being bonded effectively. Second, the condition of exposed 153 compliant substrate surface is critical to the result of regrowth. There was a certain period of time in the early stage of the research in which hazy surfaces resulted from the InGaAs grown on GaAs compliant substrates. An AFM image of such a surface is shown in Figure 4-31. The surface of the growth is extremely rough but there is no crosshatch pattern of dislocations which normally can be seen when growing thick InGaAs on GaAs. Later experiments prove that there are two causes of this roughness. First, the sample was not rinsed long enough under DI water after the HCl etching of the AlGaAs etch stop layer. Second, the sample was sitting in the air for too long after DI rinsing before growth. The combination of HCL residue and oxygen degraded the sample surface. Figure 4-31 InGaAs Grown on 10nm Twist Bonded GaAs Compliant Substrate, Insufficient Surface Rinsing before Growth One possible way to solve the problem of reproducibility is splitting a wafer bonded compliant substrate into lots of pieces and using them in different growths. Then these structures can be grown from nearly identical compliant substrates. 154 However, the size of the compliant substrate is limited by the bonding system we have. Most of the bonded wafer can only be split into two or four pieces at the most in which case each piece is still sufficient for characterizations. Acknowledging the possible problem of non-reproducibility, I am very careful on presenting the data. All of the experiments reported in this thesis were repeated at least once. Lots of experiment results were disregarded, such as varying growth temperature, growth rate, V/III ratio and InGaAs layer thickness etc. 4.3.2 Effect of Compliant Substrate Although we have the problems of non-uniformity and non-reproducibility, we can still draw important conclusions from our experiments. First, the crystal orientation of the epilayer is always same as the host wafer when the twist bonded compliant layer is 10 nm GaAs; it is always same as the compliant layer when the compliant layer is 20 nm GaAs. This was proved by asymmetric XRD and AFM. The compliant layer is still strongly influenced by the mechanical host substrates. It is unlikely that the layer approximates a free standing film on the host substrate. Strictly speaking, it is inappropriate to call it a “compliant” layer or substrate since it is more likely a “fixed” layer. Second, the InGaAs multi-QW samples grown on a 10 nm twist bonded GaAs compliant layer are always better than those grown on 20 nm GaAs compliant layer. PL intensity is higher and cross hatch pattern is clearer. The thin compliant layer does have a positive effect on the epilayer compared to a thick layer. However, 155 the result from thin GaAs twist-bonded on GaAs is not as good as what has been achieved in the literature [21]. We may still have problems with the compliant substrate preparing processes. Third, the compliant substrate based on normal GaAs and ion-implanted wafers have nearly identical results. Although the epilayer is strongly influenced by the host substrate, defects generated by ion implantation do not influence the epilayer. Fourth, the best InGaAs MQWs PL result was measured from samples grown on the 10nm GaAs twist bonded on 300 nm InGaAs. It is possible that the bonded layer is distorted to the lattices of InGaAs buried beneath it but blocked further propagation of dislocations. Unfortunately, we lack TEM images to support this claim. Other interpretations, besides free standing film, presented in the literature about the role of this layer in producing high-quality hetero-epitaxial films include gliding of large angel grain boundary, correlated misfit dislocation generation, and contaminant gettering. [3,22] None of them was proven to be the dominant effect by experiments. The working mechanism of twist bonded compliant substrates is still being debated. 4.3.3 Future Works Wafer bonded compliant substrates are just now becoming realizable. Although there is improvement in the material grown, it is far from any realistic 156 applications. Perfect compliant substrates have not been achieved. There is no comprehensive theory yet to explain this. In order to gain more understanding of the compliant substrate effect, a well controlled and repeatable fabrication process which can provide large and uniform compliant substrates must be realized. With consistent compliant substrate supply, different growth conditions can be explored in researching the working mechanism of these compliant substrates. Varying compliant layer thickness, twist angle, epilayer composition, strain and thickness can also help us to understand the strain relaxation process in compliant substrate. As for the characterization of compliant substrates, more TEM results are necessary to investigate the defect generation process of lattice mismatched materials. References 1 F. C. Frank and J. H. van der Merwe, “One-dimensional dislocations. ”Proc. R. Soc. London, Ser. A, Vol. 198 pp. 205-225, 1949. 2 J. H. Van der Merwe, “Crystal Interfaces. Part II. Finite Overgrowths” J. Appl. Phys., Vol. 34 No. 117 pp. 123, 1963. 3 A. Bourret, “Compliant substrates: a review on the concept, techniques and mechanisms” Applied Surface Science, Vol. 164, pp.3–14, 2000. 157 4 C. L. Chua, W. Y. Hsu, C. H. Lin, G. Christenson, and Y. H. Lo “Overcoming the pseudomorphic critical thickness limit using compliant substrates,” Appl. Phys. Lett. Vol. 64, No. 26, pp. 3640-3642, 1994. 5 A.M. Jones, J.L. Jewell, J.C. Mabon, E.E. Reuter, S.G. Bishop, S.D. Roh, J.J. Coleman, “Long-wavelength InGaAs quantum wells grown without strain-induced warping on InGaAs compliant membranes above a GaAs substrate,” Appl. Phys. Lett. Vol. 74, No. 7, pp. 1000-1002, 1999. 6 A. R. Powell, S. S. lyer, and F. K. LeGoues, “New approach to the growth of low dislocation relaxed SiGe material,” Appl. Phys. Lett. Vol. 64 No.14, pp. 1856-1858 1994. 7 Z. Yang, F. Guarin, I. W. Tao, and W. I. Wang and S. S. Iyer, “Approach to obtain high quality GaN on Si and SiC-on-silicon-on-insulator compliant substrate by molecular-beam epitaxy,” J. Vac. Sci. Technol. B, Vol. 13, No. 2, pp. 789-791, 1995. 8 D.M. Hansen, P.D. Moran, K.A. Dunn, S.E. Babcock, R.J. Matyi and T.F. Kuech, “Development of a glass-bonded compliant substrate,” Journal of Crystal Growth Vol. 195, pp. 144-150, 1998. 9 Carrie Carter-Coman, Robert Bicknell-Tassius, April S. Brown, and Nan Marie Jokerst, “Analysis of In 0.07 Ga 0.93 As layers on GaAs compliant substrates by double crystal x-ray diffraction,” Appl. Phys. Lett., Vol. 70, No. 13, pp. 1754-1756, 1997. 10 F.E. Ejeckam, M.L. Seaford, Y.H. Lo, H.Q. Hou, B.E. Hammons, “Dislocation- free InSb grown on GaAs compliant universal substrates,” Appl. Phys. Lett. Vol. 71, No. 6, pp. 776-778, 1997. 11 http://www.soitec.com/en/techno/t_2.htm 12 Q.-Y. Tong and U. Goesele, “Semiconductor Wafer Bonding: Science and Technology,” John Wiley & Sons, Inc. 1999. 13 Werner Kern, “Handbook of Semiconductor Wafer Cleaning Technology: Science, Technology, and Applications,” pp. 260, Noyes Publications, 1993. 14 K. H. Chang, R. Gibala, D. J. Srolovitz, P. K. Bhattacharya, and J. F. Mansfield, “Crosshatched surface morphology in strained III-V semiconductor films,” J. Appl. Phys. Vol. 67, No. 9, pp. 4093-4098, 1990. 15 Meeyoung Yoon, Bun Lee, Jong-Hyeob Baek, Hyo-Hoon Park, and El-Hang Lee, Jeong Yong Lee, “Evolution of the surface cross-hatch pattern in In x Ga 1-x As/GaAs 158 layers grown by metal-organic chemical vapor deposition,” Appl. Phys. Lett. Vol. 68, No. 1, pp. 16-18, 1996. 16 D. J. Dyaon, “X-Ray and Electron Diffraction Studies in Materials Science,” Maney Publishing, 2004. 17 Z. H. Zhu, et al, “Growth of InGaAs multi-quantum wells at 1.3 mm wavelength on GaAs compliant substrates,” Appl. Phys. Lett., Vol. 72, No. 20, pp. 2658, 1998. 18 G. Patriarche, C. Me´riadec, G. LeRoux, C. Deparis, I. Sagnes, J.-C. Harmand and F. Glas, “GaAs/GaAs twist-bonding for compliant substrates: interface structure and epitaxial growth,” Applied Surface Science, Vol. 164, pp. 15–21, 2000. 19 Pallab Bhattacharya, “Properties of Lattic-Matched and Strained InGaAs,” INSPEC, the institution of Electrical Engineers, 1993. 20 SRIM, The Stopping and Range of Ions in Matter, http://www.srim.org/ 21 Z. H. Zhu, R. Zhou, F. E. Ejeckam, Z. Zhang, J. Zhang, J. Greenberg, and Y. H. Lo, H. Q. Hou and B. E. Hammons, “Growth of InGaAs multi-quantum wells at 1.3 µm wavelength on GaAscompliant substrates,” Appl. Phys. Lett., Vol. 72, No. 20, pp.2598-2600, 1998. 22 T.Y. Tan and U. Gösele, “Twist wafer bonded “fixed-film” versus “compliant” substrates: correlated misfit dislocation generation and contaminant gettering,” Appl. Phys. A, Vol. 64, pp. 631–633, 1997. 159 Chapter 5 : Adaptive Designed Asymmetric Electro- Absorption Modulators Quantum Confined Stark Effect (QCSE) structures are widely used in Electro-Absorption Modulators (EAM) due to their large tuning range and great absorption strength. [1, 2] Simple square quantum wells in AlGaAs/GaAs or InP/InGaAsP material systems are normally used. Unlike the conventional ad hoc approaches in device design, adaptive design finds the broken-symmetry quantum well potential profile for a target optical response using numerical search with adaptive algorithms. Adaptive design fully employs the ability of modern crystal growth technique by varying the quantum well profiles on an atomic monolayer scale in the growth direction. Devices with advanced functionalities can be effectively developed by this method. Research work on realizing the devices optimized by adaptive design is discussed in this chapter. The outline of this chapter is as follows: First, adaptive designed EAM devices based on QCSE will be introduced. The next section will focus on device design issues. Finally, performances of adaptive designed EAMs will be presented. 5.1 Introduction of Adaptive designed EAM 5.1.1 Quantum Confined Stark Effect The quantum confined stark effect describes the excitonic optical absorption 160 in the semiconductor quantum well structures under electric fields. In bulk semiconductor the band edge absorption broadening that occurs when an electric field is applied is mainly due to the Franz-Keldysh effect. Considering the influence of the Coulomb effect, electron and hole pairs form excitons which have a Stark shift of the resonance to lower energies; However, the resonance is severely broadened due to the field ionization when the resonant shift is only ~10% of the binding energy. [2] However in quantum wells, a strong resonance with a large shift can be observed even when the electric field is 50 times larger than the classical ionization field, which is called the QCSE. [2] A simulated absorption spectrum of a 10nm thick rectangular quantum well under electric fields is shown in Figure 5-1 (From Ref [3]) Figure 5-1 Absorption Spectrum of a Rectangular Quantum Well under Electric Fields 161 The electron (solid curves) and hole (dashed curves) wave functions are shown with the quantum well band structures to the left. At finite electric field, F, applied in the x direction, the quantum well bands are tilted. The electrons and holes are pushed towards opposite sides of the QW resulting in a reduction in the exciton energy. Unlike in bulk semiconductors, the walls of the quantum well impede the electrons and holes from tunneling out of the well in rapid field ionization. Electron- hole interaction is still strong because the well is narrow compared to the three- dimensional exciton size (e.g. ~300 Ǻ), but the shifting of the wave functions reduces their spatial overlap so as to the exciton resonance. [2, 3] 5.1.2 Adaptive Design Historically laser diode and photodiode designs have incrementally built on previous designs and concepts; however, in adaptive design software solves the inverse problem by identifying the best configuration, which is typically a broken- symmetry spatial configuration, to achieve a desired target device function. [4] Adaptive design is possible by combining modern computer power, adaptive algorithms, and realistic physical models. The physical model used for this work was the QCSE. The search algorithm used was a genetic algorithm. Genetic algorithms were first used for simulating population and dynamics in the biological sciences, and they have also been used for finding the optima of arbitrary mathematical functions. [4] A similar concept, adaptive quantum design, was first used to find the best 162 broken-symmetry configurations of atoms and molecules to implement a target function with a specified quasiparticle density of states. [5] 5.1.3 Adaptive Designed Asymmetric EAM The first example of an adaptive designed asymmetric EAM considers following functionality: exciton absorption peak intensity is the same at F=0 kV/cm and F=70 kV/cm, but the peak position shifts in photon energy by at least 10 meV. This device switches the frequency of the quantum well exciton absorption without loss of its absorption strength. The solution space for the broken-symmetry structure was limited to double wells with variable depths and widths. The best solution found by the adaptive design is shown in Figure 5-2 (From Ref [3]). Figure 5-2 Broken-Symmetry Double Quantum Well Obtained from Numerical Optimization of Well Width and Depth Parameters. 163 The left graph shows the conduction band structure and the ground state wave functions of electrons (solid curves) and holes (dashed curves) at different fields. In the optimized double well design, the ground state wave function of the hole has two maxima whose relative weight is shifted from right to left as the electric field is increased. Meanwhile, the center of the electron wave function moves from left to right, having a maximum spatial overlap with the right peak of the hole wave function at F=0 kV/cm and with the left peak at F=70 kV/cm. The absorption spectrum as a function of electric field, shown on the right, has the desired strength and separation. The second example of an adaptive designed asymmetric EAM considers a target function that has the highest possible absorption strength at low applied field, and with a small applied electric field this absorption quickly decreases with negligible shift in frequency and low chirp. Figure 5-3 Broken-Symmetry Step Quantum Well Obtained for Rapid Decreasing of Absorption with Increasing Electric Filed 164 The optimized solution is shown in Figure 5-3. At low electric field strength, both the electron and the hole are localized within the subwell. When an electric field is applied in the reverse direction, holes ionize out of the subwell while the electron wave function remains localized within the subwell. The overlap between the two wave functions falls rapidly to zero. Figure 5-4 Absorption Modulation of Optimized Step Well Structure at 851 nm The simulated absorption modulation result of the above said step well is shown in Figure 5-4. A sharp decrease of absorption occurs as the electric field increases from 50 kV/cm to 55 kV/cm. This behavior is very attractive for building a low voltage, high speed modulator. The following sections in this chapter will focus on our efforts to realize this design in actual devices. All the simulations discussed in the above section were done by Jason Thalken. 165 5.2 EAM Wafer Structure Design Details 5.2.1 Active Region Design The adaptive design discussed in the previous section only provides the structure of the quantum wells; we also have to find appropriate waveguide and cladding layer designs to create a real modulator structure. After considering both electric and optical properties, a final design was decided as shown in Table 5-1. Material Thickness (nm) Doping (cm -3 ) GaAs 50 P >1E19 GaAs 200 P 1E18 Al 0.6 GaAs 800 P 1E18 Al 0.3 GaAs 191.5 - Al 0.3 GaAs 8.5 - Al 0.1 GaAs 14.42 - GaAs 7.07 X6 - Al 0.3 GaAs 200.0 - Grading 15.25 - Al 0.6 GaAs 800 N 1E18 GaAs 300 N 1E18 GaAs N+ (100) Table 5-1 Epi Structure of Step Quantum Well EAM The first consideration is that the transverse optical mode should be well confined. According to a simple finite difference simulation [6], 800 nm Al 0.6 GaAs cladding layers are adequate to prevent the optical field from leaking out. A simulated optical field distribution of the designed EAM structure is shown in Figure 5-5. The curve shown at the top is the refractive index profile and the curve shown at the bottom is the optical field distribution. 166 Figure 5-5 Optical Field Distribution in the Asymmetric QW EAM Device EAM devices modulate absorption properties by adjusting the electric field applied on the active region. The correct range of electric field must be obtainable with a reasonable applied bias. This field in the active region consists of the internal built-in field at zero bias and also the applied field from the external bias. These field strengths are mainly determined by the intrinsic waveguide layer thickness. In our current asymmetric QW design the strong absorption switches off at 50 kV/cm ~ 55 kV/cm. In this design the built-in field should be smaller than the switching field while having a switching bias voltage about 1~2 Volt. 0.2 μm Al 0.3 GaAs is the optimum design from the above criteria. 167 0.0 0.2 0.4 0.6 0.8 1.0 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Conduction Band Valence Band Fermi Level Band structure at Zero Bias Energy (eV) Position ( μm) Figure 5-6 Band Structure of Step Well Structure at Zero Bias 0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 80 100 120 Field under different reverse bias 0 V -0.5 V -1 V Field (kV/cm) Position ( μm) Figure 5-7 Simulated Electric Filed in the EAM Active Region under Different Bias The band structure of the current design at zero bias is shown in Figure 5-6. Only 0.2 μm Al 0.6 GaAs was considered in this and following field simulations. The Al 0.3 GaAs and QWs extend from 0.2 μm to 0.8 μm on the x axis of the graph. At zero bias, the Fermi level is flat in the whole structure. The bands in the intrinsic region are tilted due to the built-in potential. The simulated electric field in the EAM active region under different bias is shown in Figure 5-7. Reverse biases of 0V, 0.5V and 1V are shown in the graph. The electric field strength in the active region are 168 27.7 kV/cm, 36.3 kV/cm, and 44.8 kV/cm respectively. According to the simulation, the absorption of the device will switch between 1.3 V and 1.6 V reverse bias. 5.2.2 Effect of Background Doping in a Separate Confinement Layer A perfect intrinsic active region was considered in the prior simulation; however, there might be background doping in the active region in the real growth especially in the MOCVD with TMAl, TMGa, and Arsine as sources. The Al 0.3 GaAs may have a p type background doping level as high as 1E17 cm -3 . Details of the background doping of AlGaAs are discussed in chapter 6. Doping in the active region strongly influences the electric field distribution. For example, when the doping in Al 0.3 GaAs is 1E17 cm -3 , the built-in potential is not even enough to deplete the whole active region. An EAM device cannot work with this high amount of background doping, and even when the background doping is low enough to fully deplete the active region, it still influences the device performance. 0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 80 100 120 Field under different reverse bias 0 V -0.5 V -1 V Field (kV/cm) Position (μm) Figure 5-8 Simulated Electric Field with Different Reverse Bias, Background Doping Concentration 1E16 cm -3 169 The simulated electric field with different reverse biases for a device which has an active region background doping of 1E16 cm -3 is shown in Figure 5-8. Free carrier depletion of the p type active region leaves positive charges which alter the electric field. According to simulation, the electric field difference between the QWs is 8kV/cm, which corresponds to ~0.5V bias change. This electric field difference is even larger than the simulated 5kV/cm switching field; therefore, the sharp transition will be covered by the slow transition due to the field difference; and consequently, lower background doping is required to have the expected EAM performance. 0.0 0.2 0.4 0.6 0.8 1.0 0 20 40 60 80 100 120 Field under different reverse bias 0 V -0.5 V -1 V Field (kV/cm) Position ( μm) Figure 5-9 Simulated Electric Field with Different Reverse Bias, Background Doping Concentration 1E15 cm -3 The simulated electric field with different reverse biases for a device which has an active region background doping of 1E15 cm -3 is shown in Figure 5-9. A relatively uniform electric field is shown in the simulation; the electric field difference between QWs is less than 1kV/cm. According to the above simulations, a low background doping, in the vicinity of 1E15 cm -3 , is essential for the expected device performance. 170 5.3 Experiments The epi-structure was grown in our GaAs MOCVD reactor. A low background doping (1~2E15 cm -3 ) Al 0.3 GaAs was realized by using TMAl, TEGa and Arsine as sources. Details of the low background doping Al 0.3 GaAs are discussed in chapter 6. Other than the Al 0.3 GaAs, the whole structure was grown with TMAl, TMGa, and Arsine as sources at a growth temperature of 730 ºC. 5.3.1 Broad Area Laser Performance The wafer was first processed into broad area lasers to test the material qualities. The standard laser process in Appendix D was used. 0 200 400 600 800 1000 1.0 1.5 2.0 Quantum Efficiency 73% Loss 4 cm -1 1/ η Device Length ( μm) 0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 400 450 500 550 600 650 700 750 J tr =514A/cm 2 J tr per QW =86A/cm 2 J th (A/cm 2 ) 1/Device_Length Figure 5-10 Performances of Broad Area Lasers Made from Asymmetric QW EAM Wafer The performance of the broad area lasers is summarized in Figure 5-10. The graph on the left shows the reciprocal device external efficiencies as a function of device length, and the graph on the right shows the threshold current densities as a function of reciprocal device length. The scattered black dots are experimental data and the red lines are linear data fittings. The internal device parameters can be 171 extrapolated from these two graphs [6, 7]: the Quantum efficiency is 73%, internal loss is 4 cm -1 , and transparent current density is 86 A/cm 2 per well. These are typical numbers for the normal GaAs/AlGaAs QW broad area lasers. [7] 5.3.2 EAM Performance The GaAs step-well EAM was processed into a 250 μm long single mode ridge waveguide structure. The SEM images of the processed GaAs devices are shown in Figure 5-11. The processing was done at AdTech Photonics Inc., City of Industry, CA. Figure 5-11 SEM Images of GaAs EAM Devices Light from the GaAs tunable laser is coupled into the EAM waveguide with a single mode lensed fiber. The output of the EAM is collected by another single mode fiber and is measured by HP8153A lightwave multimeter. The image of the EAM device being measured is shown in Figure 5-12. A HP 4155 parameter analyzer is used to control the applied bias and to collect the device current and output light power data. 172 Au bond wire Light couple into waveguide Light couple out of waveguide EAM ridge waveguide Au bond wire Light couple into waveguide Light couple out of waveguide EAM ridge waveguide Figure 5-12 Single Mode Lens Fiber Coupled EAM Device The measured EAM output power as function of bias voltage is shown in Figure 5-13. The incident light was varied from 845 nm to 855 nm at 1 nm steps. The incident light power was 0.5 mW before coupling into the waveguide. Strong tuning of absorption as a function of bias voltage is shown in the graph; however, it is hard to notice the exciton peaks in the graph. It is easier to distinguish the exciton peaks and to compare with the simulation data if the absorption curves could be shown as functions of wavelength. The data in Figure 5-13 was converted accordingly to the graph in Figure 5-14. A better way to obtain the data curves shown in Figure 5-14 is to fix the bias voltage and scan the input light wavelength; however, the GaAs tunable laser does not have a constant power mode and the tuning can not be controlled by the automation program (LabView). These problems prevent us from making the direct measurement. 173 ER of USC step well EAM (L=250 μm) -65 -60 -55 -50 -45 -40 -35 -4 -3 -2 -1 0 Bias voltage Vea (volts) EAM output power (dBm) 845nm (0.5mW) 846nm (0.5mW) 847nm (0.5mW) 848nm (0.5mW) 849nm (0.5mW) 850nm (0.5mW) 851nm (0.5mW) 852nm (0.5mW) 853nm (0.5mW) 854nm (0.5mW) 855nm (0.5mW) Figure 5-13 Measured Step Well EAM Output Power at Various Wavelengths Low power (0.5 mW) USC step well absorption 0 50 100 150 200 250 845 850 855 Photon wavelength (nm) Absorption (1/cm) 0.5 V 0 V -0.5 V -1 V -1.5 V -2 V -2.5 V -3 V -3.5 V -4 V Figure 5-14 Step Well EAM Absorption as Functions of Wavelength The absorption from the continuum state, which red-shifts with the applied reverse bias, is shown clearly in Figure 5-14; however, there is no clear exciton peak shown. There is one kink at 851 nm for low bias, it disappears soon after the bias is 174 increased. There is also another kink at 854 nm which slowly disappears with the increasing bias. These two features are too weak and are covered by the continuum state absorption. It is impossible to observe the effect predicted by the adaptive design. A similar EAM wafer with 6 normal square quantum wells was also grown by MBE (Intelligent Epitaxy Technology, Inc., Richardson, TX). It was processed into single mode EAM devices using identical processing. The EAM device was also measured by fixing the input laser wavelength and scanning the reverse bias. After the data conversion, the absorption curves as functions of wavelength at different reverse biases are shown in Figure 5-14. GaAs 6 QWs EAM Absorption Input 3dBm 0 50 100 150 200 250 300 350 400 830 832 834 836 838 840 842 844 846 848 Wavelength (nm) Absorption (1/cm) 0 V -1 V -2 V -3 V -4 V -5 V -6 V -7V Figure 5-15 Normal Square Well EAM Absorption as Functions of Wavelength Clear exciton peaks shift from 840.5 nm to 843 nm as the reverse bias increases from 0V to 5V. The intensity of the resonance decreases during this shift. 175 The absorption from the continuum state shows a red-shift with increase bias. These effects are all predicted by the simulation; however, compared to the theoretical simulation, Figure 5-1, the exciton absorption intensity is very small. This simulation is consistent with the earlier experiments done on GaAs/AlGaAs quantum EAM devices. [8] Both the step well EAM and square well EAM have low exciton absorption intensities compared with the simulations. The possible reasons for the low intensity include: interface roughness, width non-uniformity and field inhomogeneity between quantum wells. One immediate experiment in the future works should simply using only one QW in the EAM instead of six. Although the total absorption decreases with the small absorption volume, but it eliminates all the potential problems related to the multiple quantum wells. 5.4 Summary The preliminary research work on adaptive designed asymmetric QW EAM was presented in this chapter. The basic concept of such devices was first introduced. The device design details were discussed with the focus on the optimum optical field and electric field distributions. The actual devices were measured but the simulated effect was not achieved. References 176 1 T. H. Wood, C. A. Burrus, D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, and W. Wiegmann, “High-speed optical modulation with GaAs/GaAlAs quantum wells in a p-i-n diode structrue,” Appl. Phys. Lett. Vol. 44, No. 1, pp. 16-19, 1984. 2 D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wiegmann, T. H. Wood, and C. A. Burrus, “Band-Edge Electroabsorption in Quantum Well Structures: The Quantum-confined Stark Effect,” Phys. Rev. Lett. Vol. 53, No. 22, pp. 2173-2177, 1984. 3 Jason Thalken, Weifei Li, Stephan Haas and A. F. J. Levi “Adaptive design of excitonic absorption in broken-symmetry quantum wells,” Appl. Phys. Lett., Vol. 85, No. 1, pp. 121-123, 2004. 4 Jason Thalken, Stephan Hass, and A. F. J. Levi, “Synthesis for semiconductor device design,” J. of Appl. Phys., Vol. 98, pp. 044508, 2005. 5 Jason Thalken, Yu Chen, A. F. J. Levi, and Stephan Haas “Adaptive quantum design of atomic clusters,” Phys. Rev. B, Vol. 69, pp. 195410, 2004. 6 L. A. Coldren and S. W. Corzine, “Diode Lasers and Photonic Integrated Circuits,” John Wiley & Sons, Inc. 1995. 7 Julian S. Osinski, “Analysis, Design, and Fabrication of Low Threshold Semiconductor Lasers Using Compressively Strained Quantum Wells,” Ph. D. Dissertation, University of Southern California, 1992. 8 David A. B. Miller, Joseph S. Weiner, and D. S. Chemla, “Electric-Field Dependence of Linear Optical Properties in Quantum Well Structures: Waveguide Electroabsorption and Sum Rules,” IEEE J. Quan. Elec. Vol. QE-22, No. 9, 1986. 177 Chapter 6 : Doping in GaAs MOCVD 6.1 Introduction Foreign atoms in semiconductor crystals are called impurities. Impurities can have significant impacts on the properties of semiconductor materials. On one hand, great efforts are devoted into preventing unwanted impurities incorporation in order to get more purified materials; On the other hand, designated impurities are added deliberately into purified semiconductors to change its conductivity. Introducing impurities into semiconductor in a controlled manner is called doping. These deliberately added impurities are called dopants. By using modern semiconductor technology, doping concentration can be controlled in a wide range from 1E14 cm -3 to 1E21 cm -3 . Within this range the material conductivity varies from the semi- insulating via semi-conductive to the semi-metallic. From a simple PN junction device to the state of art optoelectronic devices, precise control of doping concentration is one of the key factors to realize the functionality of devices. [1] Dopants can be incorporated into semiconductor during growth, by implantation or by diffusion. In this chapter, doping in GaAs/AlGaAs materials during MOCVD growth is the main topic. The outline of this chapter is as following. First, doping mechanism in MOCVD will be discussed. Two methods of doping concentration measurement will be introduced after that. Then the doping behaviors of two common types of dopants, Zn as p type dopant and Si as n type dopant, will be presented. Carbon doping by 178 using CBr 4 will be discussed as an alternative p doping method. The research work on achieving ultra low background doping level in AlGaAs MOCVD growth is also presented in the chapter. 6.2 Doping in MOCVD grown GaAs/AlGaAs Impurities incorporated into a semiconductor lattice predominantly occupy substitutional lattice sites. For compound semiconductors like GaAs, the lattice site can be either a cation or an anion site. Due to the vapor phase nature of MOCVD, only gaseous doping compounds can be used in the growth process. Such gaseous doping compounds are called doping precursors. Incorporation of doping atoms is same as the growth of other atoms in MOCVD. The processes includes but not limited to the following: transportation of precursors to the stagnant boundary layer near the substrate surface by gas phase diffusion; decomposition of precursors; gas phase reaction between precursors; adsorption and desorption on the surface, surface reaction, transportation of byproducts from the stagnant boundary layer, etc. Therefore, dopants incorporation strongly depend on the chemical properties of precursors, growth condition and surface condition etc. The doping precursors can be inorganic compounds, e.g. Disilane (Si 2 H 6 ) and CBr 4, or metal-organic compounds, e.g. Diethylzinc (DEZn, (C 2 H 5 ) 2 Zn) and Trimethylgalium (TMGa, (CH 3 ) 3 Ga). There are three basic requirements for the 179 doping precursors: (i) a sufficient high vapor pressure at room temperature, (ii) thermal decomposition of the precursor at the growth temperature, and (iii) no parasitic chemical reactions before and after thermal decomposition. [1,2] Before any other doping precursors are introduced, unintended carbon doping must be discussed mainly because carbon can originate from the metal-organic precursors used in the growth of the main material. Carbon is group IV atom. Carbon atoms occupy the position of group V atoms in MOCVD. The group III precursors containing alkyl groups, such as TMGa, Triethylgalium (TEGa, (C 2 H 5 ) 3 Ga), Trimethylaluminum (TMAl, (CH 3 ) 3 Al), and Triethylaluminum (TEAl, (C 2 H 5 ) 3 Al) etc., can behave as C-doping precursors simultaneously. They introduce intrinsic C doping from the methyl and ethyl radicals which are inevitable byproducts during the growth. [3] The carbon concentration of unintentionally doped AlGaAs grown using methyl based sources, TMGa and TMAl and AsH 3 , typically ranges from 10 16 to 10 18 cm -3 , depending on the aluminum composition. The higher the Al composition is, the higher the carbon incorporation rate is. The increased carbon concentration can be explained by the difference in bond strength between Al-C (Al-CH 3 65 kcal/mol) and Ga-C (Ga-CH 3 , 59 kcal/mol). [3] The carbon incorporation in AlGaAs decreases as the V/III ratio increases. It is believed that AsH x species transfer hydrogen to the surface methyl groups, providing a route for carbon removal through CH 4 formation. [3] There are significantly different C incorporation rates between using TMGa 180 and using TEGa as Ga precursor. The dissociation of an ethyl group in TEGa happens according to the following reaction: 3 2 5 2 3 5 2 ) ( ) ( CHCH GaH H C Ga H C + ↔ (6.1) This process of dissociation of the ethyl group from TEGa and transferring of a H atom to the Ga is known as the β-elimination process. This process effectively reduces the stability of the Ga-C bond. The β-elimination process doesn’t happen to TMGa. The stronger Ga-C bond in TMGa therefore leads to the incorporation of more C atoms simultaneously with the Ga atom. [4] Similar effect also happens to the aluminum precursors. AlAs grown with aluminum precursors containing ethyl radicals, e.g. TEAl, has significantly less C incorporation rate. Al precursors without Al-C bonds, such as Trimethylamine alane (AlH 3 :N(CH 3 ) 3 , TMAA), can also generate ultra-low C doping AlAs or AlGaAs. [2] There are few technical details needed to be clarified before the discussion of specific dopants. First, in the growth of V/III semiconductors, the V/III ratio normally is much larger than unity. The group III precursors determine the growth rate in most cases. When intentionally adding doping precursors, it is convenient to interpret the doping efficiency by using the ratio of mole fraction between doping precursors and group III precursors. However, it doesn’t mean that doping efficiency won’t change as the ratio of mole fraction changes. Second, when calculating the ratio of mole fraction involving TMAl, which is a dimer in the gas phase [1], each dimer is considered as two molecules. 181 6.3 Doping Measurement Technologies There are two main methods to measure the doping concentration, one is the Hall measurement, and the other one is the capacitance-voltage (CV) profiling. B x y z + + ++ + + ++ + L w d I B x y z + + ++ + + ++ + B x y z x y z + + ++ + + ++ + + + ++ + + ++ + L w d I Figure 6-1 Hall Effect of N-Type Semiconductor The basic physical principle of the Hall effect is the Lorentz force. Considering an n-type bar-shaped semiconductor as shown in Figure 6-1, constant current flow I is along x axis, magnetic field B is applied vertically along z axis. The electrons are main carriers in the n-type semiconductor, they move in the negative x direction with velocity v v . When an electron moves along a direction perpendicular to applied magnetic field, it experiences the Lorentz force in the form of following: B v e F l v v v × − = (6.2) where –e is the electron charge. Initially, the electrons are deflected to negative y axis due to Lorentz force until the electric field generated by excess electrons applies same force on electrons: l H H F E e F v v v = − = (6.3) 182 H E v is called Hall Field. According to (6.3), Hall carrier concentration can be calculated as the following equation: d eV IB n H H 1 ⋅ = (6.4) where w E V H H = is the Hall voltage, w is the width of the bar and d is the thickness of the bar. Similar analysis also can be done on p-type semiconductor where the Hall voltage has an opposite sign from the n-type semiconductor. By combining Hall measurement and conductivity measurement, Hall mobility of carriers can be determined: H H en σ μ = (6.5) σ is the conductivity of semiconductor. The Hall carrier concentration and Hall mobility is different from the real carrier concentration and drift mobility by a Hall factor H r : H H n r n = (6.6) H H r μ μ = (6.7) H r is larger or equal to 1 depending on which scattering mechanism dominates the carrier transportation. The van der Pauw technique is the common technique of performing Hall and conductivity measurements. [1,5,6] CV profiling utilizes the surface charge depletion to measure the doping concentration spatial distribution. 183 The width of the surface depletion region can be obtained from Poisson’s equation and is given by ) ( 2 ) ( 2 V eN V V eN W B D bi D D − ≅ − = φ ε ε (6.8) where bi V is the built-in voltage, which is B bi V φ ≅ barrier height, for typical doping concentrations; V is the external voltage applied. When the applied voltage changes, the depletion width also changes: dV V eN dW B D D ) ( 2 2 1 − − = φ ε (6.9) Then the change in depletion charge per unit area is given by D D dW eN dQ = (6.10) The differential depletion capacitance per unit area is defined as dV dQ C / = (6.11) Substituting 6.9 and 6.10 into 6.11: D B D W V N e C ε φ ε = − = 2 2 1 (6.12) The CV-concentration is then defined as dC dV e C dV dC e N CV ε ε 3 1 2 2 = ⎟ ⎟ ⎠ ⎞ ⎜ ⎜ ⎝ ⎛ − = − − (6.13) CV measures compensated carrier concentration ND-NA, which is the concentration of electrically active impurities. The depth of doping profile could be obtained by adjusting reverse bias. [1] 184 However, the depth measurement of normal CV is constrained by the reverse breakdown voltage of the Schottky diode. The Electrochemical C-V Profiler (ECV) overcomes this limitation by utilizing an electrochemical etching system. Liquid chemical etchant is also served as contact to form the Schottky diodes, which significantly simplifies the sample preparation process. Due to these advantages, ECV profiler is the most effective method of determining the doping profile of semiconductor materials. [7] 6.4 GaAs-AlGaAs Si 2 H 6 and DEZn doping 6.4.1 Routine Doping Calibration (Si, Zn) Silicon is a group IV atom. Silicon atoms take the position of group III atoms in GaAs/AlGaAs and behave as donors. Disilane (Si 2 H 6 ) is a popular hydride for Si doping. It has a nearly unity doping efficiency in a large temperature range. Its incorporation efficiency is almost linearly dependent on precursor mole fraction. [8] Some typical Si doped n doping calibration curves are shown in Figure 6-2. The X axis is the ratio of Disilane to Group III precursor (TMGa+TMAl) mole fraction, the Y axis is the doping concentration measured by ECV profiler. Both axes are shown in log scale. The growth temperature was 730 ºC. The growth rates of GaAs, Al 0.15 GaAs, Al 0.40 GaAs and Al 0.60 GaAs were 14.7 Ǻ/s, 17.3 Ǻ/s, 11.2 Ǻ/s, and 17.4 Ǻ/s respectively. The V/III ratios during the growths of these samples were 22.9, 17.3, 30 and 20.7 respectively. 185 10 -5 10 -4 10 -3 1E17 1E18 1E19 GaAs V/III=22.9 M708 15% AlGaAs V/III=19.7 M748 40% AlGaAs V/III=30 M711 60% AlGaAs V/III=20.7 M756 Si doping Calibration Doping Concentration(cm -3 ) Ratio of Si 2 H 6 to Group III Precursor Mole Fraction Figure 6-2 Si Doped N GaAs AlGaAs Doping Calibration The difference of n doping among various Al compositions is dominantly originated from the carbon background doping. As discussed earlier in this chapter, the carbon doping concentration increases as the aluminum composition in AlGaAs increases. When the aluminum composition is high, more Si doping atoms are needed to compensate higher carbon doping. Zinc is a group II atom. Zinc atoms take the position of group III atoms in III- V semiconductors and behave as acceptors. Zinc is one of the most popular p-type doping impurities in MOCVD. It has been well studied and much is known about the interdependent of Zinc doping precursors and crystal growth parameters [1]. Dimethylzinc (DMZn, (CH 3 ) 2 Zn) and Diethylzinc (DEZn, (C 2 H 5 ) 2 Zn) are commonly used as Zn doping precursors. At the same growth condition, DEZn has smaller 186 activation energy so as to higher doping efficiency compared with DMZn. [9] DEZn is used in our MOCVD as the P type doping precursor. 10 -2 10 -1 10 0 1E17 1E18 1E19 7E17 4.7E17 GaAs V/III=22.9 M714 Al 0.15 GaAs V/III=17.34 M344 Al 0.40 GaAs V/III=30 M715 Al 0.60 GaAs V/III=20.7 M753 Zn doping Calibration. Doping Concentration(cm -3 ) Ratio of DEZn to Group III Precursor Mole Fraction Figure 6-3 Zn doped P GaAs AlGaAs Doping Calibration Some typical Zn doping calibration curves are shown in Figure 6-3. The X axis is the ratio of DEZn to Group III precursor (TMGa+TMAl) mole fraction, the Y axis is the doping concentration measured by ECV profiler. Both axes are shown in log scale. Two numbers marked on the y axis are the background doping level of Al 0.6 GaAs and Al 0.4 GaAs, both of which are P type. The growth temperature was 730 ºC. The growth rates of GaAs, Al 0.15 GaAs, Al 0.40 GaAs and Al 0.60 GaAs were 14.7 Ǻ/s, 17.3 Ǻ/s, 11.2 Ǻ/s, and 17.4 Ǻ/s respectively. The V/III ratios of these samples were 22.9, 17.3, 30 and 20.7 respectively. It is clearly shown in the graph that the higher the aluminum composition is, the higher the background doping level is. Also shown in the graph is that doping 187 efficiency drops as the aluminum composition increase. It was explained as the increasing in activation energy at higher aluminum composition. [10] 6.4.2 Ultra-High P Type Doped GaAs by DEZn Good ohmic contacts are essential for all optoelectronic devices. But the space charge region, which is depleted of mobile carriers, forms a potential barrier on the metal-semiconductor interface. For Si, Ge, GaAs and most of the other III-V compound semiconductors, the surface potential barrier height is approximately 2/3 of the band gap for n type material and approximately 1/3 of the band gap for p type material. In the GaAs-AlGaAs material system, the contact behavior of AlGaAs is very sensitive to the Al concentration and the difficulty of making good ohmic contacts increases as the Al mole fraction gets higher; therefore, a GaAs contact layer is preferred. Also, a good ohmic contact can be realized when the doping in the contact layer is high enough so the conduction becomes field-emission dominated. [11] A GaAs layer with N type doping of 1~3E18 cm -3 is sufficient to form an ohmic contact by using Ni/AuGe/Au metal contacts. On the other hand, in order to form good ohmic contact on p-GaAs with Ti/Pt/Au, the doping concentration should be higher than 5E19 cm -3 . [12] In the application of top emitting 850nm VCSEL, this contact layer must be thin since light passes through the top GaAs contact layer, which is absorbent at 850nm. A growth process is designed to grow a 100 Å thick highly doped p-GaAs contact layer. It is well known in the literature that Zn doping concentration is mainly 188 limited by Zn desorption from the sample surface. Extra high doping concentration can be realized by decreasing the growth temperature. [9] 1E18 1E19 1E20 750 700 650 600 1.00x10 -3 1.05x10 -3 1.10x10 -3 Temperature (C) Doping Concentration (cm -3 ) Activation Enenergy T<670C: 2.6eV T>670C: 1.6eV Reciprocal Temperature (1/K) Figure 6-4 GaAs Zn P+ Doping Concentration as a Function of Temperature GaAs Zn doping concentration vs. growth temperature is shown in Figure 6-4. The x axis is reciprocal temperature and y axis is hole doping concentration in log scale. The temperature in ºC is also marked in the graph for viewing convenience. The growth conditions for the data shown in the figure are: the GaAs growth rate was 14.67 Ǻ/s; the DEZn to TMGa mole fraction ratio was 0.12; and V/III ratio was 23. Three data points of 670, 640 and 610 ºC are set on a straight line while data points of 730, 700 and 670 are set on another straight line. Therefore, two different Zn activation energies can be calculated for two temperature ranges. When the temperature is lower than 670 ºC, the Zn activation energy is 2.6eV. When the temperature is higher than 670 ºC, the Zn activation energy is 1.6eV. Similar effect 189 was discussed in Ref. [9], where the activation energy is 3.3eV when the temperature is lower than 675 ºC and is 2.1eV when the temperature is higher than 675 ºC. According to this paper, hole concentrations of zinc doped GaAs films depends upon Arsine partial pressures. Therefore between 575 and 675°C, zinc incorporation changes with temperature and decomposed Arsine concentration. Above 675°C Arsine is fully decomposed and zinc incorporation is temperature dependent. According to Figure 6-4, doping concentration of Zn doped p-GaAs can reach 5E19 cm -3 when the growth temperature is low than 610 ºC. In our current GaAs MOCVD system, growth temperature is controlled manually. The growth control computer doesn’t have a direct control over the substrate temperature. And the graphite susceptor is heated by RF coil and cooled by gas flow, it has a very slow response to the variation of RF power. It normally takes more than 10 min to stabilize the substrate temperature after adjusting the set point. In order to simplify the growth control, other than waiting for the susceptor temperature becomes stable at a low temperature, the time dependence of the susceptor temperature after turning off the RF power was recorded, this recorded data is used to determine the start time of the contact layer growth. 190 0 50 100 150 200 550 600 650 700 750 Temperature readout( o C) Time after cooling down(second) Figure 6-5 Susceptor Temperature as a Function of Time after Shut Down RF Power The temperature of susceptor as a function of time after turning off the RF power is shown in Figure 6-5. The contact layer growth start time is calculated so that the contact layer is grown at 600 ºC. To check the contact resistance of the p-contact, a transmission line experiment was performed to determine the specific contact resistance of the Ti/Pt/Au-GaAs interface. A circular test pattern as shown in Figure 6-6 was used for the experiment. 650 μm d r 2 r 1 r o Figure 6-6 Circular test pattern for determination of specific contact resistance, from Ref. [13, 14] The above pattern of metal contact was defined on the contact surface by 191 photo lithograph, and the contact metals were evaporated at a starting pressure of 5 ×10 -7 Torr. The metal layers used for the p-contact were 300Å of titanium, 500Å of platinum, and 2000Å of gold. Following the liftoff, the wafer was annealed at 400°C for 30 seconds in a rapid thermal annealing (RTA) system under forming gas ambient environment. A constant current of 20mA was injected and the voltage drop across the various gap lengths, d, was measured for the different pattern sizes. The data was then fit to a transmission line model of the following form [14]: 1 111 11 ln 2 os T iR r VL rd r rd π ⎡ ⎤ ⎛⎞ ⎛ ⎞ Δ= + + ⎢ ⎥ ⎜⎟ ⎜ ⎟ −− ⎝⎠ ⎝ ⎠ ⎣ ⎦ (6.14) where i o is the injected current, ΔV is the voltage drop, R s is the sheet resistance, and L T is the transfer length. The specific contact resistance is calculated from c T s R L R = . In the test patterns used here, r 1 is 75 μm, and d varies from 6 to 28 μm. The voltage drop data is plotted below and fit to equation (6.14). 5 101520 2530 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 ΔV Transmission Line Fit Contact Resistance Measurement R s =273.16 Ω/sq Rc=3.73x10 -5 Ω-cm 2 Voltage (V) Separation Distance ( μm) Figure 6-7 Transmission Line Measurement and Fitted Results The results (Figure 6-7, from Ref. [13]) show a sheet resistance of 273.16 192 Ω/sq and a specific contact resistance of 3.73 ×10 -5 Ω-cm 2 , which is a number normally achieved from Ti/Pt/Au contacts to p-GaAs [15]. The post growth processing and measurement were performed by Ryan Stevenson. [13] 6.4.3 Zn and Si Doping at Low V/III Ratio In the growth of TJS structure in the SHEDs project, a low V/III ratio ~10 must be used to ensure that the deposition does not happen on top of the dielectric masks. However, varying the V/III ratio not only changes the growth profile but also changes doping behavior. Both Si and Zn impurities are taking the position of group III atoms, the incorporation of these impurities can be suppressed by a low V/III ratio. [1] This effect was clearly observed in the experiments. 10 -5 10 -4 10 -3 1E17 1E18 1E19 Al 0.15 GaAs V/III=19.7 Al 0.4 GaAs V/III=30 Al 0.3 GaAs V/III=10 N doping Calibration Doping Concentration(cm -3 ) Ratio of Si 2 H 6 to Group III precursor mole fraction Figure 6-8 Si Doped N Doping Calibration at Low V/III Ratio The Si doping calibration of Al 0.3 GaAs at low V/III ratio is shown in Figure 6-8. The Al 0.3 GaAs growth rate was 6 Ǻ per second. The V/III ratio was 10. This 193 growth condition was proved to result in no material deposition on the dielectric masks (see chapter 3). Also shown in the graph is the N doping calibrations of AlGaAs at higher V/III ratios as comparison. Compared to doping calibration data at high V/III ratio, the ratio of Si 2 H 6 to group III precursor mole fraction has to be increased by an order of magnitude in order to realize n type doping level of 1E18 cm -3 . 10 -2 10 -1 10 0 1E17 1E18 1E19 Al 0.15 GaAs V/III=17.34 Al 0.4 GaAs V/III=30 Al 0.3 GaAs V/III=10 P doping Calibration. Doping Concentration(cm -3 ) Ratio of DEZn to Group III precursor mole fraction Figure 6-9 Zn Doped P Doping Calibration at Low V/III Ratio A similar experiment for Zn doping calibration of Al 0.3 GaAs at low V/III ratio is shown in Figure 6-9. The result is very similar to that happens in the n type doping calibration. However, the DEZn source in current MOCVD system can not provide high enough flow for the p type doping to reach 1E18 cm -3 unless increasing the bubbler temperature form 10 ºC to 20 ºC. Zn doped material also suffers from potential problem of Zn diffusion, which will impose more serious problems in the TJS structure because there is no undoped buffer material between the Zn doped 194 AlGaAs and QW. In stead of Zn, carbon was selected to realize p doping in this growth condition. Details will be discussed in the next section. 6.5 Carbon Doping Using CBr 4 Carbon is a shallow acceptor in GaAs i.e. it has a low acceptor binding energy. Unlike Zn, Carbon nearly has no memory effect during the growth. Carbon is also an impurity with an exceptionally low diffusion constant. It is perfect for the devices with stringent doping requirement. [1] Instead of intrinsic carbon doping from TMGa, TMAl or other precursors containing alkyl radicals as discussed earlier in this chapter, extrinsic carbon doping precursors such as CCl 4 and CBr 4 provide easy controls in carbon doping without having to change growth conditions such as growth temperature or V/III ratio. The initial focus was mainly on CCl 4 . However, recent legislation implies that usage of CCl 4 will be severely restricted in the future due to its toxicity and effect on the environment. Since this legislation does not apply to CBr 4 , [16] CBr 4 is used as the carbon doping precursor in our GaAs reactor. 195 0.00 0.02 0.04 0.06 0.08 0.10 0.0 2.0x10 18 4.0x10 18 6.0x10 18 8.0x10 18 1.0x10 19 1.2x10 19 1.4x10 19 1.6x10 19 1.8x10 19 2.0x10 19 2.2x10 19 A=5.33E17 B=7.96E19 A=1.33E17 B=1.30E19 GaAs 1.5nm/s, T=630 o , V/III=14 GaAs 1.5nm/s, T=730 o , V/III=38 Al 0.15 GaAs 1.8nm/s, T=730 o , V/III=23 Y = A + B * X A=3.28E17 B=1.94E20 Doping Concentration (cm -3 ) Ratio of CBr4 to Group III Mole Fraction Figure 6-10 P GaAs/Al 0.15 GaAs Doping Concentration as Function of Ratio of CBr 4 to Group III Mole Fraction at 730 and 630 ºC Three sets of CBr 4 doped p type GaAs/AlGaAs calibration data are shown in Figure 6-10. In this graph, the x axis is the ratio of CBr 4 to group III precursor mole fraction; the y axis is the doping concentration measured by ECV. In all three datasets, p doping concentrations are linearly dependent on CBr 4 mole fraction. The coefficients of linear regression are shown next to the fitted lines. In AlGaAs material system doped with CBr 4 , p doping concentration strongly depends on Al composition and growth temperature according to this graph. P doping concentration of Al 0.15 GaAs (shown as diamond dots in the graph) is 6 time higher than that of GaAs (round dots) grown at same temperature, 730 ºC. Similar as the intrinsic carbon doping discussed earlier in this chapter, strong Al-C bonds greatly enhance the carbon incorporation. For CBr 4 doped GaAs, the doping concentration increases by an order when the growth temperature changes from 730 ºC to 630 ºC (square dots). Part of this increase is due to the V/III ratio change, it will be discussed later. 196 0.00100 0.00105 0.00110 0.00115 0.00120 0.00125 1E18 1E19 1E20 750 700 650 600 550 GaAs Growth Rate 1.5nm/s Ratio of CBr 4 to TMGa Mole Fraction 0.079 V/III ~25.8 Y=a*e bx a=2.2E9±3.4E9 b=20E3±1.3E3 Activation energy 1.7eV Doping Concentration (cm -3 ) Temperature 1/(K) Figure 6-11 Growth Temperature Dependence of P Doping Concentration in CBr 4 Doped GaAs The growth temperature dependence of p doping concentration in CBr 4 doped GaAs is shown in Figure 6-11. The x axis is reciprocal temperature and y axis is p doping concentration in log scale. The p type doping concentration strongly depends on the growth temperature. The highest doping shown in this graph is 8E19 cm -3 , which is high enough for forming a good ohmic contact layer. The activation energy obtained from this dataset is 1.7eV. It is slightly higher than 1.4eV from Ref [17]. The doping concentration saturate at 8E19 cm -3 when the growth temperature is lower than 550 ºC. When the temperature is higher than 680 ºC, the decrease of doping concentration becomes less rapid, this effect is similar to what happens in Zn doped GaAs. When the temperature is higher than 675 ºC, which is the Arsine cracking temperature, Arsine is fully decomposed and carbon incorporation is solely depends on temperature. One the other hand, when the temperature is lower than 675 197 ºC, carbon incorporation is affected by both temperature and decomposed Arsine concentration. 0.00100 0.00105 0.00110 0.00115 0.00120 0.00125 1E18 1E19 1E20 750 700 650 600 550 Al 0.15 GaAs Growth Rate 1.8nm/s Ratio of CBr 4 to TMGa Mole Fraction 0.065 V/III ~23 Y=a*e bx a=3.4E11±4.6E11 b=16E3±1.1E3 Activation energy 1.4eV Doping Concentration (cm -3 ) Temperature 1/(K) Temperature (C) Figure 6-12 Growth Temperature Dependence of P Doping Concentration in CBr 4 Doped Al 0.15 GaAs The growth temperature dependence of p doping concentration in CBr 4 doped Al 0.15 GaAs is shown in Figure 6-12. Compared with CBr 4 doped GaAs, Al 0.15 GaAs has a very similar temperature dependence behavior. But Al 0.15 GaAs has a saturated doping level of 1E20 cm -3 and the doping concentration is slightly less sensitive to temperature comparing with GaAs. The calculated activation energy is 1.4eV. P doping concentration of CBr 4 doped AlGaAs also depends on V/III ratio. Carbon takes the position of arsenic atom. It competes for the group V positions during the growth. The V/III ratio dependence of CBr 4 doped GaAs at 730 ºC is shown in Figure 6-13. The x axis is V/III ratio and y axis is doping concentration in unit of cm -3 . The p doping concentration increases from 1E18 cm -3 to 1.8E18 cm -3 198 when the V/III ratio is changed from 38 to 3. Further decreasing V/III ratio to 2 causes the doping concentration to start decreasing and wafer surface to become rough. V/III ratio of 2 is too low to grow high quality materials. 0 5 10 15 20 25 30 35 40 1.0x10 18 1.2x10 18 1.4x10 18 1.6x10 18 1.8x10 18 GaAs Growth Rate 1.5nm/s CBr 4 to TMGa Mole Fraction Ratio 0.079 Growth temperature 730 o C Doping Concentration (cm -3 ) V/III ratio Figure 6-13 V/III Ratio Dependence of P Doping Concentration in GaAs Doped with CBr 4 , 730 ºC Growth Temperature 0 5 10 15 20 25 30 35 40 6.0x10 18 8.0x10 18 1.0x10 19 1.2x10 19 1.4x10 19 1.6x10 19 1.8x10 19 2.0x10 19 2.2x10 19 GaAs Growth Rate 1.5nm/s CBr 4 to TMGa Mole Fraction Ratio 0.079 Growth temperature 630 o C Doping Concentration (cm -3 ) V/III Ratio Figure 6-14 V/III Ratio Dependence of P Doping Concentration in GaAs Doped with CBr 4 , 630 ºC Growth Temperature 199 The V/III ratio dependence of CBr 4 doped GaAs at 630 ºC is shown in Figure 6-14. The doping level increases from 7E18 cm -3 to 1.9E19 cm -3 as the V/III ratio decreases from 38 to 11. With growth temperature of 630 ºC, Carbon incorporation becomes more sensitive to V/III ratio compared with the similar experiments performed at 730 ºC. Hole concentration of CBr 4 doped AlGaAs increases when the V/III ratio decreases. This dependency is opposite to that in DEZn doped AlGaAs. This feature could be utilized in the TJS regrowth mentioned earlier in this chapter. Hole concentration in DEZn doped Al 0.3 GaAs is not high enough because the V/III ratio must be low to have good regrowth selectivity. It is not a problem in CBr 4 doped AlGaAs where low V/III ratio is desired to increase the doping concentration. 1E-3 0.01 0.1 1E17 1E18 1E19 Al 0.3 GaAs V/III=10 Doping Concentration(cm -3 ) Ratio of CBr 4 to Group III Precursors Mole Fraction Figure 6-15 P Doping Concentration of CBr 4 Doped Al 0.3 GaAs at Low V/III ratio P doping concentration of CBr 4 doped Al 0.3 GaAs grown at low V/III ratio is shown in Figure 6-15. The growth rate was 6 Ǻ per second. Doping concentration of 200 1E18 cm -3 can be easily realized by adding very small amount of CBr 4 . During the growth of CBr 4 doped AlGaAs, one of the byproducts is HBr, which etches AlGaAs. Therefore, CBr 4 doped AlGaAs may have a reduced growth rate and different Al composition compared with undoped AlGaAs grown at same growth condition. In the growth of VCSEL structure, the device performance is very sensitive to both layer thickness and composition. It is essential to have a good control over these two parameters. The dependence of GaAs and Al 0.15 GaAs growth rates on CBr 4 doping was investigated. Growth rate calibration sample structures, which contain 2 sets of DBRs and one λ cavity (for details, see chapter 2), were grown. The reflectivity spectrums were measured and TFCalc was used to determine the layers’ thicknesses. 0 20 40 60 80 100 120 140 160 180 10 12 14 16 18 20 22 GaAs, V/III=27 GaAs, V/III=38.4 Al 0.15 GaAs, V/III=23 T=730 0 Slope=-0.0138+/-0.0003 Slope=-0.0161+/-0.0004 Growth Rate(A/s) CBr4 Flow (SCCM) Figure 6-16 Growth Rate of GaAs and AlGaAs as a Function of CBr 4 flow The growth rate of GaAs and Al 0.15 GaAs as a function of CBr 4 flow is shown in Figure 6-16. The x axis is the CBr 4 flow in SCCM; the y axis is the growth rate. Three sets of data are shown in the graph. All growths were done at 730 ºC. V/III 201 ratios are labeled in the graph. The growth rate reduction linearly depends on CBr 4 flow. The slopes of the linear regression curves are marked in the graph. Al 0.15 GaAs curve has a steeper slope compared with CBr 4 doped GaAs, which indicates that the effect of CBr 4 etching is stronger for Al 0.15 GaAs than for GaAs. Two data sets for GaAs are shown in the graph and V/III ratios are 38 (marked as circle) and 27 (marked as triangle) respectively. There is nearly no difference between these two data sets except the data point at the largest CBr 4 flow. Combining the growth rate calibration data and the p doping concentration calibration data, it is possible to determine the growth rate change at various doping concentrations. For example, the growth rate of CBr 4 doped Al 0.15 GaAs with p doping level of 1E18 cm -3 grown at 730 ºC is about 1.5% slower than undoped Al 0.15 GaAs. At the same time, the growth rate of CBr 4 doped GaAs with p doping level of 1E18cm -3 grown at 730 ºC is about 15% slower than undoped GaAs. GaAs and Al 0.15 GaAs serve as the high index materials of DBR layers in 980nm and 850nm VCSELs. Growth rate of CBr4 doped AlGaAs with higher aluminum composition is not checked due to the fact that intrinsic carbon doping, which does not influence the growth rate, is enough to supply the donors required. It is also a big concern that CBr 4 doping might change the Al composition since the etching speed depends on the Al composition shown in the above experiment. In experiments for determining the carbon doped Al 0.15 GaAs growth rates, the aluminum composition used in TFCalc simulations was always 15%. There was 202 no strong evidence that aluminum composition needs to be changed for a better fitting of the experimental data. So at the growth conditions used in that experiment (see Figure 6-16), the aluminum composition changes are within a couple of percents. X ray diffraction is normally used to determine the compound composition by measuring the lattice constant. But carbon doping in AlGaAs also changes the lattice constant. The lattice constant of the hypothetical binary material GaC is 4.66 Ǻ which is estimated from the covalent radii of carbon (0.77 Ǻ) and Ga (1.25 Ǻ) [18]. The concentration of carbon on Ga and interstitial sites is assumed to be negligible. The molecular concentration in AlGaAs materials is 2.21E22 cm -3 . The Poisson ratio of AlGaAs material is 3.324 [19]. Assuming pseudomorphic growth on GaAs, the change of lattice constant in the growth direction as a function of C doping concentration is shown in Figure 6-17 (solid line). Also shown in the figure is the lattice constant change as a function of Al composition (dash line). 203 1E18 1E19 1E20 -0.020 -0.015 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.0 0.2 0.4 0.6 0.8 1.0 Δa (A) in Growth Direction C doping concentration (cm -3 ) Al composition Figure 6-17 Lattice Constant Changes as a Function of C Doping Concentration and as a Function of Al Composition When the carbon doping concentration is 1E18 cm -3 , the GaAs-C lattice constant in the growth direction is 8.8E-5 Ǻ larger than GaAs. This change, corresponding to 0.6% difference in the Al composition, is negligible in XRD measurement. But when the carbon doping concentration is 1E20 cm -3 , the change in lattice constant is same as AlGaAs with 60% Al composition. Accurate measurement of carbon doping concentration is required to estimate the Al composition. It is worthwhile mentioning that there is discrepancy in carbon doping concentration and p doping concentration due to the fact that not all of the carbon atoms incorporated are activated. According to ref [17], about 80% to 100% of carbon incorporated in AlGaAs is activated. 204 -0.04 -0.02 0.00 0.02 0.04 10 100 1000 10000 100000 GaAs n+ wafer GaAs C doped 1E18 GaAs C doped 1E19 Intensity (a.u.) Ω (degree) Figure 6-18 XRD Ω-2 θ Scans of GaAs N+ Wafer and Carbon Doped GaAs Grown on GaAs N+ Wafers, the P Doping concentrations are 1E18 and 1E19 cm -3 XRD Ω-2 θ scan of GaAs N+ wafer and two carbon doped GaAs on GaAs N+ wafers are shown in Figure 6-18. The carbon doped GaAs layers are 2 µm thick. The dotted line is the data of GaAs N+ epi ready wafer. The FWHM is 0.0044 degree. This scan is used as the reference. The lattice constant of Si 2E18 cm -3 doped GaAs is very close to GaAs because the covalent radii of Si (1.17 Ǻ) is similar that of Ga (1.25 Ǻ). The solid line is from the carbon doped GaAs with a p doping concentration of 1E18 cm -3 . The FWHM of this curve is 0.0045 degree. The dashed line is from the carbon doped GaAs with a p doping concentration of 1E19 cm -3 . The FWHM is 0.0083 degree. According to the simulation, the peak positions of XRD Ω- 2 θ scans of GaAs with carbon doping concentrations of 1E18 and 1E19 cm -3 are 5.8E-4 and a 5.8E-3 degree separated respectively from the peak of pure GaAs. Both of these separations are all too small to form a separate peak. Only FWHM is influenced. 205 -0.05 0.00 0.05 0.10 1 10 100 1000 10000 GaAs GaAs-C 0.05515 o XRD Intensity Ω Figure 6-19 XRD Ω-2 θ Scan of Carbon Doped GaAs on N+ GaAs Substrate, P Doping concentration is 8E19 cm -3 Another XRD Ω-2 θ scan data of carbon doped GaAs on N+ GaAs substrate is shown in Figure 6-19. The carbon doped GaAs is 2 µm thick. P doping concentration is 8E19 cm -3 measured by ECV. There is a clear GaAs-C XRD peak locating 0.05515 degree away from the GaAs substrate XRD peak. According to the simulation, this separation corresponds to a carbon doping level of 9.5E19 cm -3 , which is consistent with the ECV measurement if accuracy of ECV and the non- unity carbon activation ratio are considered. 206 -0.10 -0.05 0.00 0.05 0.10 1 10 100 1000 10000 GaAs Al 0.15 GaAs 0.06047 o Intensity (a.u.) Ω Figure 6-20 XRD Ω-2 θ Scan of Carbon Doped Al 0.15 GaAs on N+ GaAs Substrate, P Doping concentration is 1E20 cm -3 A XRD Ω-2 θ scan data of carbon doped Al 0.15 GaAs on N+ GaAs substrate is shown in Figure 6-20. The carbon doped Al 0.15 GaAs is 2 µm thick. P doping concentration is 1E20 cm -3 measured by ECV. The aluminum composition is determined from the undoped material. The aluminum composition is unknown in this experiment. Undoped Al 0.15 GaAs has a Ω-2 θ scan peak position of -0.0138 degree compared to GaAs. If aluminum composition is unchanged, the carbon doping concentration calculated from the position of the XRD peak is 1.2E20 cm -3 , which is consistent with the ECV measurement again. Summarizing from the experimental results and analysis above, it is difficult to determine the aluminum composition with XRD because accurately measured carbon concentration becomes hard to achieve when the hole concentration is high. Other analysis methods, such as SIMS (Secondary Ion Mass Spectrometry) should be considered. 207 6.6 Low Background Doping Al 0.3 GaAs As discussed in chapter 5, EAM project requires that the doping concentration of AlGaAs with aluminum composition 25%~30% is close to 1E15 cm -3 . The research efforts on reducing the background doping level of AlGaAs are summarized in this section. For MOCVD in which TMGa and Arsine are source materials, it is a common practice to adjust the V/III ratio to grow the optimized semi-insulating GaAs. At low V/III ratio, the GaAs is p type with high carbon concentration. As the V/III ratio is increased, the carbon doping is suppressed while at the same time the donor impurities, such as Si and Ge, incorporations are less dependent on the V/III ratio. The semi-insulating condition is reached when the donor and acceptor concentration compensate with each other. GaAs will become n type upon further increasing of V/III ratio. Similar trends were also observed in AlGaAs. The V/III ratio required to realize the p doping to n doping transition is increased due to the higher intrinsic carbon doping level in AlGaAs. [20] Background doping of AlGaAs was investigated by growing samples with different aluminum compositions, growth temperatures, and V/III ratios. Each of the background doping test samples consists of a 2 µm thick AlGaAs layer covered by 100nm GaAs. The samples were grown on semi-insulation GaAs (001) wafer. TMAl, Arsine, and TMGa or TEGa were used as growth precursors. The doping concentrations were measured by Hall measurement. The aluminum compositions were determined by X-ray diffraction. 208 30 40 50 60 70 80 90 1E15 1E16 1E17 Background Doping of AlGaAs Doping Concentration (cm -3 ) V/III ratio Al 0.3 GaAs 730 o C Al 0.3 GaAs 760 o C Al 0.3 GaAs 790 o C Al 0.25 GaAs 760 o C Al 0.25 GaAs 730 o C Al 0.3 GaAs 730 o C TEG Figure 6-21 Background Doping Concentration of AlGaAs as a Function of V/III Ratio, Temperature and Aluminum Composition The background doping concentrations of AlGaAs are summarized in Figure 6-21. Triangle symbols represent Al 0.3 GaAs samples grown with TMAl, TMGa, and Arsine. Square and diamond symbols represent Al 0.25 GaAs samples grown also with TMAl, TMGa and Arsine. The growth rates of these samples are ~7 Ǻ/s. The background doping concentrations of Al 0.3 GaAs samples grown with TMAl, TEGa and Arsine are shown as stars. The growth rates for these samples are ~7 Ǻ/s. Shown from the Figure 6-21, the doping concentration is lowered as the V/III ratio increases. The donor impurities are unable to fully compensate the acceptor impurities even with the highest V/III ratio which can be realized in our reactor. The doping concentration also increases for higher aluminum composition. The lowest doping concentrations of Al 0.3 GaAs and Al 0.25 GaAs grown with TMGa are about 3E16 cm -3 and 1E16 cm -3 respectively. The doping concentrations of those samples 209 grown with TMAl, TEGa and Arsine are much smaller than the samples grown with TMAl, TMGa, and Arsine at the similar growth condition. The highest doping concentration is near 1E16 cm -3 . Doping concentration of ~ 1E15 cm -3 could be realized by increasing V/III ratio to ~80. As introduced earlier in this chapter, AlGaAs grown with precursors containing ethyl radicals has less carbon incorporation. The idea case is using both TEAl and TEGa as the aluminum and gallium sources respectively. But there was only TEGa available in our MOCVD system. Hole doping concentrations in the samples grown with TMAl, TEGa and Arsine is reduced due to that the β- elimination process of TEGa in which the carbon incorporation is suppressed during growth. The ethyl radicals in TEGa not only help to break the Ga-C bonds, they must also help to break the Al-C bonds since the background doping of GaAs grown with TMGa at similar condition is already in the range of 1E15 cm -3 . The exchange of radicals between TEGa and TMAl [2] is likely the explanation. 6.7 Doping Dependency of Crystal Orientation MOCVD growth, especially the doping atoms incorporation, is sensitive to the exposed surface bonds, which are determined by growth surface crystal orientation [21]. This is essential in the TJS device growth where the growth of contacting material is on the mesa side wall instead of normal (001) surface. As varying the crystal orientation, Si doping concentration variation is believed to be negligible when using Si 2 H 6 as the doping precursor in AlGaAs 210 material system. [21] There were different results reported regarding the carbon doping efficiency on different crystal orientation. Generally the C doping concentration on the (111) plane is different than that of (001) plane. Either (111)A>(001)>(111)B or (111)B>(001)>(111)A are reported. [22, 23, 24] There is not a conclusive explanation for the carbon doping dependence on crystal orientation. A detailed study on this problem under the growth condition used in TJS growth is required. 6.8 Summary Research activities regarding doping in MOCVD grown AlGaAs were summarized in this chapter. Routine p and n type doping calibrations in AlGaAs by using precursors of Si 2 H 6 and DEZn were first introduced. Extremely high p doping GaAs was realized by Zn doping at low growth temperature. The limitations of Zn and Si doping of AlGaAs at low V/III ratio were investigated. Both the Si 2 H 6 and DEZn to group III mole ratios have to be increased by an order of magnitude in order to reach the similar doping concentration realized at normal growth condition. Extrinsic carbon doping in AlGaAs by using CBr 4 as precursor was explored. The carbon doping concentration strongly depends on growth temperature, V/III ratio, aluminum composition etc. CBr4 doping also has etching effect on the AlGaAs during growth. Growth rate reductions of GaAs and Al 0.15 GaAs were measured, which show that the etching effect depends on aluminum composition. CBr4 doping may also influence the aluminum composition. However, it is difficult to determine 211 the aluminum composition in the carbon doped AlGaAs by XRD because the doped carbon atoms also influence the lattice constant. Extremely low background doped Al 0.3 GaAs was achieved by using TEGa instead of TMGa. Finally, doping dependency on crystal orientation was briefly discussed. References 1 E. Fred Schubert, “Doping in III-V semiconductors,” Cambridge University Press, 1993. 2 Gerald Stringfellow, “Organometallic Vapor-Phase Epitaxy,” second edition, Academic Press, 1999. 3 T.F. Kuech , J.M. Redwing, “Carbon doping in metalorganic vapor phase epitaxy,” J. Crystal Growth, Vol. 145, pp. 382-389, 1994. 4 N. Putz, H. Heinecke, M. Heyen and P. Balk, “A Comparative Study of Ga(CH 3 ) 3 and Ga(C 2 H 5 ) 3 In the MOMBE of GaAs,” J. Crystal Growth, Vol. 74, pp. 292-300, 1986. 5 http://www.eeel.nist.gov/812/hall.html 6 Jasprit Singh, “Physics of Semiconductors and Their Heterostructures,” McGraw- Hill, Inc., 1993. 7 “PN4300PC ECV Profiler Operating Manual,” Accent Optical Technologies, 1994. 8 T.F. Kuech, E. Veuhoff and B.S. Meyerson, “Silicon Doping of GaAs and Al x Ga 1-x As Using Disilane in Metalorganic Chemical Vapor Deposition,” J. Crystal Growth, Vol. 68, No. 1, pp 48-53, 1984. 9 R.W. Glew, “Zinc Doping of MOCVD GaAs,” J. Crystal Growth, Vol. 68, No. 1, pp 44-47, 1984. 10 Jane J. Yang, William I. Simpson, and Lavada A. Moudy, “Electrical properties of Zn in Metalorganic chemical vapor deposition GaAlAs,” J. Appl. Phys., Vol. 53, No. 1, pp. 771-773, 1982. 212 11 V. L. Rideout, “A Review of the Theory and Technology for Ohmic Contacts to Group III-V Compound Semiconductors,” Solid-State Electronics, Vol. 18, pp. 541- 550, 1975. 12 A.G. Baca, F. Ren, J.C. Zolper, R.D. Briggs, S.J. Pearton, “A Survey of Ohmic Contacts to III-V Compound Semiconductors,” Thin Solid Films, Vol. 308–309, pp. 599–606, 1997. 13 Ryan Stevenson, “High Performance Components of Free-Space Optical and Fiber-Optic Communications Systems,” Ph.D. thesis dissertation, pp. 60, 1999. 14 G.S. Marlow and M.B Das, “The effects of contact size and non-zero metal resistance on the determination of specific contact resistance,” Solid-State Electronics, vol. 25, no. 2, pp. 91-94, 1982. 15 G. Stareev, “Formation of extremely low resistance Ti/Pt/Au ohmic contacts to p- GaAs,” Appl. Phys. Lett., Vol. 62, No. 22, pp. 2801-2803, 1993. 16 http://www.epichem.com/metalorganics/products/carb_brom.html 17 Kouta Tateno, Yoshitaka Kohama, Chikara Amano, “Carbon doping and etching effects of CBr 4 during Metalorganic chemical vapor deposition of GaAs and AlAs,” J. Crystal Growth, Vol. 172, No. 1-2, pp 5-12, 1997. 18 “Size of Atoms,” http://chemed.chem.purdue.edu/genchem/topicreview/bp/ch7/size.html 19 M. Krieger, H. Sigg, N. Herres, K. Bachem, and K. Köhler, “Elastic constants and Poisson ratio in the system AlAs–GaAs,” Appl. Phys. Lett . , Volume 66, No. 6, pp. 682-684, 1995. 20 T. Nakanisi, “The growth and characterization of high quality MOVPE GaAs and GaAlAs,” J. Crystal Growth, Vol. 68, No. 1, pp. 282-294, 1984. 21 Makoto Kondo, Chikashi Anayama, Naoko Okada, Hiroshi Sekiguchi, Kay Domen, and Toshiyuki Tanahashi, “Crystallographic orientation dependence of impurity incorporation into III-V compound semiconductors grown by metalorganic vapor phase epitaxy,” J. Appl. Phys., Vol. 76, No. 2, pp. 914-927, 1994. 22 Makoto Kondo, and Toshiyuki Tanahashi, “Dependence of carbon incorporation on crystallographic orientation during metalorganic vapor phase epitaxy of GaAs and A1GaAs,” J. Crystal Growth, Vol. 145, pp. 390-396, 1994. 23 Young K. Park, Chang-Sik Son, Seong-Il Kim, Yong Kim, Eun Kyu Kim, Suk-Ki Min and In-Hoon Choi, “Effect of atomic bond structure on crystallographic 213 orientation dependence of carbon doping in GaAs,” J. Appl. Phys., Vol. 83, No. 5, pp. 2519-2524, 1998. 24 T.F. Kuech and E. Veuhoff, “Mechanism of Carbon Incorporation in MOCVD GaAs,” J. Crystal Growth, Vol. 68, pp. 148-156, 1984. 214 Bibliography Govind P. Agrawal, “Semiconductor Lasers: Past, Present, and Future,” American Institute of Physics, 1995. Seigo Ando and Takashi Fukui, “Facet Growth of AlGaAs on GaAs with Si0 2 Grattings by MOCVD and Applications to Quantum Well Wires,” Journal of Crystal Growth Vol. 98 pp. 646- 652, 1989. T. Baba, Y. Yogo, K. Suzuki, F. Koyama and K. Iga, “Near room temperature continuous wave lasing characteristics of GaInAsP/InP surface emitting laser,” Electron. Lett., Vol. 29, No. 10, pp. 913-914, 1993. A.G. Baca, F. Ren, J.C. Zolper, R.D. Briggs, S.J. Pearton, “A Survey of Ohmic Contacts to III-V Compound Semiconductors,” Thin Solid Films, Vol. 308–309, pp. 599–606, 1997. C. Angulo Barrios et al, “GaAs/AlGaAs Buried Heterostructure Laser by Wet Etching and Semi- insulating GaInP:Fe Regrowth,” Electrochemical and Solid-State Letters, vol. 3, no. 9, pp. 439- 441, 2000. Pallab Bhattacharya, “Properties of Lattic-Matched and Strained InGaAs,” INSPEC, the institution of Electrical Engineers, 1993. A. E. Bond, P. D. Dapkus, and J. D. O’Brien, “Aperture placement effects in oxide-defined vertical-cavity surface-emitting lasers,” IEEE Photo. Technol. Lett. Vol. 10, pp.1362-1364, 1998. A. E. Bond, P. D. Dapkus, and J. D. O’Brien, “Aperture dependent loss analysis in oxide-defined vertical-cavity surface-emitting lasers,” IEEE Photo. Technol. Lett., Vol. 11, No. 4 pp.397-399, 1999. Bourret, “Compliant substrates: a review on the concept, techniques and mechanisms” Applied Surface Science, Vol. 164, pp.3–14, 2000. Carrie Carter-Coman, Robert Bicknell-Tassius, April S. Brown, and Nan Marie Jokerst, “Analysis of In 0.07 Ga 0.93 As layers on GaAs compliant substrates by double crystal x-ray diffraction,” Appl. Phys. Lett., Vol. 70, No. 13, pp. 1754-1756, 1997. K. H. Chang, R. Gibala, D. J. Srolovitz, P. K. Bhattacharya, and J. F. Mansfield, “Crosshatched surface morphology in strained III-V semiconductor films,” J. Appl. Phys. Vol. 67, No. 9, pp. 4093-4098, 1990. Won-Jin Choi, P. Daniel Dapkus, “Selective growth and regrowth of high Al content AlGaAs for use in BH lasers,” Journal of Crystal Growth Vol. 195 pp. 495-502, 1998. Won-Jin Choi, “Design and Fabrication of Novel InGaAs and AlGaInP Quantum Well Lasers by Metalorganic Chemical Vapro Deposition,” Ph.D. thesis dissertation, 1999. Kent D. Choquette, K. M. Geib, H. C. Chui, B. E. Hammons, H. Q. Hou, T. J. Drummond, and Robert Hull, “Selective oxidation of buried AlGaAs versus AlAs layers,” Appl. Phys. Lett. Vol. 69, No. 10, pp. 1385-1387, 1996. 215 W. W. Chow, K. D. Choquette, M. H. Crawford, K. L. Lear, and G. R. Hadley, “Design, Fabrication, and Performance of Infrared and Visible Vertical-Cavity Surface-Emitting Lasers,” IEEE J. Quantum. Electron., Vol. 33, No. 10, pp.1810-1824, 1997. A. L. Chua, W. Y. Hsu, C. H. Lin, G. Christenson, and Y. H. Lo “Overcoming the pseudomorphic critical thickness limit using compliant substrates,” Appl. Phys. Lett. Vol. 64, No. 26, pp. 3640-3642, 1994. S. L. Chuang, “Physics of Optoelectronic Devices,” John Wiley and Sons, 1995. L. A. Coldren and S. W. Corzine, “Diode Lasers and Photonic Integrated Circuits,” John Wiley & Sons, Inc. 1995. J. M. Dallesasse, N. Holonyak, Jr., A. R. Sugg, T. A. Richard, and N. El-Zein, “Hydrolyzation oxidation of AlGaAs-AlAs-GaAs quantum well heterostrcutres and superlattices,” Appl. Phys. Lett., Vol. 57, No. 26, pp. 2844-2846, 1990. A. G. Deppe, D. L. Huffaker, J. Shin, and Q. Deng, “Very-low-threshold index-confined planar microcavity lasers,” IEEE Photon. Technol. Lett., Vol. 7, No. 9, pp. 965-967, 1995. J. J. Dudley, D. L. Crawford, and J. E. Bowers, “Temperature Dependence of the Properties of DBR Mirrors Used in Surface Normal Optoelectronic Devices,” IEEE Photon. Technol. Lett., Vol.4, No. 4, pp. 311-314, 1992. A. J. Dyaon, “X-Ray and Electron Diffraction Studies in Materials Science,” Maney Publishing, 2004. R. D. Dupuis and P. D. Dapkus, “Room Temperature Operation of Ga 1-x Al x As/GaAs Double Heterostructure Lasers Grown by Metalorganic Chemical Vapor Deposition,” Appl. Phys. Lett., Vol. 31, No. 7, pp. 466-468, 1977. R. D. Dupuis, P. D. Dapkus, R. Chin, N. Holonyak, Jr., and W. W. Kirchoefer, “Continuous 300 K Laser Operation of Single Quantum-Well AlGaAs Heterostructure Diodes Grown by Metalorganic Chemical Vapor Deposition,” Appl. Phys. Lett., Vol. 34, No. 4, pp. 265-267, 1979. F.E. Ejeckam, M.L. Seaford, Y.H. Lo, H.Q. Hou, B.E. Hammons, “Dislocation-free InSb grown on GaAs compliant universal substrates,” Appl. Phys. Lett. Vol. 71, No. 6, pp. 776-778, 1997. A. C. Frank and J. H. van der Merwe, “One-dimensional dislocations. ”Proc. R. Soc. London, Ser. A, Vol. 198 pp. 205-225, 1949. R. S. Geels, B.J. Thibeault, S.W. Corzine, J.W. Scott, and L.A. Coldren, “Design and characterization of In 0.2 Ga 0.8 As MQW vertical-cavity surface-emitting lasers,” IEEE J. Quantum Electron., Vol. 29, No. 12, pp. 2977-2987, 1993. R.W. Glew, “Zinc Doping of MOCVD GaAs,” J. Crystal Growth, Vol. 68, No. 1, pp 44-47, 1984. A. Ronald Hadley, “Effective index model for vertical-cavity surface-emitting lasers,” Optics Letters, Vol. 20, No. 13, pp. 1483-1485, 1995. R. N. Hall, G. E. Fenner, J. D. Kingsley, T. J. Soltys, and r. O. Carlson, ”Cohernt Light Emission from GaAs Junctions,” Phys. Rev. Lett., Vol. 9, pp. 366-368, 1962. 216 D.M. Hansen, P.D. Moran, K.A. Dunn, S.E. Babcock, R.J. Matyi and T.F. Kuech, “Development of a glass-bonded compliant substrate,” Journal of Crystal Growth Vol. 195, pp. 144-150, 1998. Eric R. Hegblom, Dubravko I. Babic, Brian J. Thibeault, and Larry A. Coldren, “Sacttering Losses from Dielectric Apertures in Vertical-Cavity Lasers,” J. Selected Topics in Quantum Electron. Vol. 3, No. 2, pp. 379-389, 1997. Breck Hitz “Military Project Aims to Improve Diode Laser Efficiency,” Photonics Spectra, vol. 39, no. 1, pp. 89, 2005. N. Holonyak, Jr. and S. F. Bevacqua, “Coherent (Visible) Light Emission from Ga(As x P 1-x ) Junctions,” Appl. Phys. Lett., Vol. 1, pp. 82-83, 1962. Kenichi Iga, “Surface-Emitting Laser – Its Birth and Generation of New Optoelectronics Field,” IEEE J. Selected Topics in Quantum Electron., Vol. 6, No. 6, pp. 1201-1215, 2000. Valerie Illingworth, “The Penguin Dictionary of Physics,” 1990, Market House Books Ltd. R. Jaeger, M. Grabherr, C. Jung, R. Michalzik, G. Reiner, B. Weigl, and K. J. Ebeling, “57% wallplug efficiency oxide-confined 850 nm wavelength GaAs VCSEL’s,” Electron. Lett., Vol. 33, No. 4, pp. 330-331, 1997. V. Jayaraman, J. C. Geske, M. H. MacDougal, F. H. Peters, T. D. Lowes, and T. T. Char, “Uniform threshold current, coutinuous-wave, single mode 1300nm vertical cavity lasers from 0 to 70 °C,” Electron. Lett., Vol. 34, No. 14, pp. 1405-1407, 1998. A. M. Jones, J.L. Jewell, J.C. Mabon, E.E. Reuter, S.G. Bishop, S.D. Roh, J.J. Coleman, “Long- wavelength InGaAs quantum wells grown without strain-induced warping on InGaAs compliant membranes above a GaAs substrate,” Appl. Phys. Lett. Vol. 74, No. 7, pp. 1000-1002, 1999. Werner Kern, “Handbook of Semiconductor Wafer Cleaning Technology: Science, Technology, and Applications,” pp. 260, Noyes Publications, 1993. Makoto Kondo, Chikashi Anayama, Naoko Okada, Hiroshi Sekiguchi, Kay Domen, and Toshiyuki Tanahashi, “Crystallographic orientation dependence of impurity incorporation into III-V compound semiconductors grown by metalorganic vapor phase epitaxy,” J. Appl. Phys., Vol. 76, No. 2, pp. 914-927, 1994. Makoto Kondo, and Toshiyuki Tanahashi, “Dependence of carbon incorporation on crystallographic orientation during metalorganic vapor phase epitaxy of GaAs and A1GaAs,” J. Crystal Growth, Vol. 145, pp. 390-396, 1994. A. Koyama, S. Kinoshita, and K. Iga, “Room-temperature continuous wave lasing characteristics of GaAs vertical cavity surface-emitting laser,” Appl. Phys. Lett., Vol. 55, No. 3, pp. 221-222, 1989. M. Krieger, H. Sigg, N. Herres, K. Bachem, and K. Köhler, “Elastic constants and Poisson ratio in the system AlAs–GaAs,” Appl. Phys. Lett . , Volume 66, No. 6, pp. 682-684, 1995. T.F. Kuech, E. Veuhoff and B.S. Meyerson, “Silicon Doping of GaAs and Al x Ga 1-x As Using Disilane in Metalorganic Chemical Vapor Deposition,” J. Crystal Growth, Vol. 68, No. 1, pp 48- 53, 1984. 217 T.F. Kuech and E. Veuhoff, “Mechanism of Carbon Incorporation in MOCVD GaAs,” J. Crystal Growth, Vol. 68, pp. 148-156, 1984. T.F. Kuech , J.M. Redwing, “Carbon doping in metalorganic vapor phase epitaxy,” J. Crystal Growth, Vol. 145, pp. 382-389, 1994. K. L. Lear, K. D. Mar, K. D. Choquette, S. P. Kilcoyne, R. P. Schneider, Jr., and K. M. Geib, “High-frequency modulation of oxide-confined vertical cavity surface emitting lasers,” Electron. Lett., Vol. 32, No. 5, pp. 457-458, 1996. H. M. Manasevit, “Single-Crystal Gallium Arsenice on Insulating Substrates,” Appl. Phys. Lett., Vol. 12, No. 4, pp. 156-159, 1968. N. M. Margalit, D. I. Babic, K. Streubel, R. P. Mirin, R. L. Naone, J. E. Bowers, and E. L. Hu, “Submiliamp long wavelength vertical cavity lasers,” Electron. Lett., Vol.32, No. 18, pp. 1675- 1677, 1996. G.S. Marlow and M.B Das, “The effects of contact size and non-zero metal resistance on the determination of specific contact resistance,” Solid-State Electronics, vol. 25, no. 2, pp. 91-94, 1982. D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wiegmann, T. H. Wood, and C. A. Burrus, “Band-Edge Electroabsorption in Quantum Well Structures: The Quantum-confined Stark Effect,” Phys. Rev. Lett. Vol. 53, No. 22, pp. 2173-2177, 1984. David A. B. Miller, Joseph S. Weiner, and D. S. Chemla, “Electric-Field Dependence of Linear Optical Properties in Quantum Well Structures: Waveguide Electroabsorption and Sum Rules,” IEEE J. Quan. Elec. Vol. QE-22, No. 9, 1986. Yoshifumi Mori and Naozo Watanabe, “A New Etching Solution System, H 3 PO 4 -H 2 O 2 -H 2 O, for GaAs and Its Kinetics,” J. Electrochem. Soc: Solid-State Science and Technology, Vol. 125, No. 9, pp. 1510, 1978. T. Nakanisi, “The growth and characterization of high quality MOVPE GaAs and GaAlAs,” J. Crystal Growth, Vol. 68, No. 1, pp. 282-294, 1984. M. I. Nathan, W. P. Dumke, G. Burns, F. H. Dill, Jr., and G. Lasher, “Stimulated emission of radiation from GaAs p-n junctions,” Appl. Phys. Lett., Vol. 1, pp. 62-64, 1962. Donald A. Neamen, “Semiconductor Physics and Devices, Basic Principles,” McGraw-Hill Companies, Inc., 2003. Julian S. Osinski, “Analysis, Design, and Fabrication of Low Threshold Semiconductor Lasers Using Compressively Strained Quantum Wells,” Ph. D. Dissertation, University of Southern California, 1992. Young K. Park, Chang-Sik Son, Seong-Il Kim, Yong Kim, Eun Kyu Kim, Suk-Ki Min and In- Hoon Choi, “Effect of atomic bond structure on crystallographic orientation dependence of carbon doping in GaAs,” J. Appl. Phys., Vol. 83, No. 5, pp. 2519-2524, 1998. A. Patriarche, C. Me´riadec, G. LeRoux, C. Deparis, I. Sagnes, J.-C. Harmand and F. Glas, “GaAs/GaAs twist-bonding for compliant substrates: interface structure and epitaxial growth,” Applied Surface Science, Vol. 164, pp. 15–21, 2000. 218 A. R. Powell, S. S. lyer, and F. K. LeGoues, “New approach to the growth of low dislocation relaxed SiGe material,” Appl. Phys. Lett. Vol. 64 No.14, pp. 1856-1858, 1994. Paula M. Powell, “Laser Diode Manufacturers Turn Up the Power,” Photonics Spectra, vol. 38, no. 2, pp. 82, 2004. N. Putz, H. Heinecke, M. Heyen and P. Balk, “A Comparative Study of Ga(CH 3 ) 3 and Ga(C 2 H 5 ) 3 In the MOMBE of GaAs,” J. Crystal Growth, Vol. 74, pp. 292-300, 1986. T. M. Quist, R. H. Rediker, R. J. Keyes, W. E. Krag, B. Lax, A. L. McWhorter, and H. J. Zeiger, “Semiconducter maser of GaAs,” Appl. Phys. Lett., vol. 1 pp. 91-92, 1962. V. L. Rideout, “A Review of the Theory and Technology for Ohmic Contacts to Group III-V Compound Semiconductors,” Solid-State Electronics, Vol. 18, pp. 541-550, 1975. Edward H. Sargent and J. M. Xu, “Lateral Injection Lasers,” International Journal of High Speed Electronics and Systems, Vol. 9, No. 4, pp. 941-978, 1998. E. Fred Schubert, “Doping in III-V semiconductors,” Cambridge University Press, 1993. Jasprit Singh, “Physics of Semiconductors and Their Heterostructures,” McGraw-Hill, Inc., 1993. Haruhisa Soda, Ken-ichi Iga, Chiyuki Kitahara and Yasuharu Suematsu, "GaInAsP/InP Surface Emitting Injection Lasers,” Japanese Journal of Applied Physics, vol. 18, no. 12, p. 2329–2330, 1979. G. Stareev, “Formation of extremely low resistance Ti/Pt/Au ohmic contacts to p-GaAs,” Appl. Phys. Lett., Vol. 62, No. 22, pp. 2801-2803, 1993. Ryan Stevenson, “High Performance Components of Free-Space Optical and Fiber-Optic Communications Systems,” Ph.D. thesis dissertation, 2005. Gerald Stringfellow, “Organometallic Vapor-Phase Epitaxy,” second edition, Academic Press, 1999. D. A. Suda and J. M. Xu, “Lateral current injection lasers—A new enabling technology for OEIC’s,” NATO Proc. Future Trends in Microelectronics. New York: Kluwer, 1996. W. Susaki, T. Tanaka, H. Kan, and M. Ishii, “New structures of GaAlAs lateral-injection laser for low-threshold and single-mode operation,” IEEE J. Quantum Electron., vol. QE-13, pp. 587–591, 1977. S. M. Sze, “Physics of Semiconductor Devices,” 1981, John Wiley & Sons, Inc. Zhijian Wei, “Semiconductor Surface Emitting Laser, Laser Array and Polarization Control,” Ph.D. thesis dissertation, 2006. T.Y. Tan and U. Gösele, “Twist wafer bonded “fixed-film” versus “compliant” substrates: correlated misfit dislocation generation and contaminant gettering,” Appl. Phys. A, Vol. 64, pp. 631–633, 1997. Kouta Tateno, Yoshitaka Kohama, Chikara Amano, “Carbon doping and etching effects of CBr 4 during Metalorganic chemical vapor deposition of GaAs and AlAs,” ,” J. Crystal Growth, Vol. 172, No. 1-2, pp 5-12, 1997. 219 Jason Thalken, Weifei Li, Stephan Haas and A. F. J. Levi “Adaptive design of excitonic absorption in broken-symmetry quantum wells,” Appl. Phys. Lett., Vol. 85, No. 1, pp. 121-123, 2004. Jason Thalken, Yu Chen, A. F. J. Levi, and Stephan Haas “Adaptive quantum design of atomic clusters,” Phys. Rev. B, Vol. 69, pp. 195410, 2004. Jason Thalken, Stephan Hass, and A. F. J. Levi, “Synthesis for semiconductor device design,” J. of Appl. Phys., Vol. 98, pp. 044508, 2005. Q.-Y. Tong and U. Goesele, “Semiconductor Wafer Bonding: Science and Technology,” John Wiley & Sons, Inc. 1999. A. H. Van der Merwe, “Crystal Interfaces. Part II. Finite Overgrowths” J. Appl. Phys., Vol. 34 No. 117 pp. 123, 1963. Ralph Williams, “Modern GaAs Processing Methods,” 1990 Artech House, Inc. Carl Wilmsen, Henryk Temkin and Larry A. Coldren, “Versitcal-Cavity Surface-Emitting Lasers,” Cambridge University Press, 1999. T. H. Wood, C. A. Burrus, D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, and W. Wiegmann, “High-speed optical modulation with GaAs/GaAlAs quantum wells in a p-i-n diode structrue,” Appl. Phys. Lett. Vol. 44, No. 1, pp. 16-19, 1984. G. M. Yang, M. H. MacDougal, and P. D. Dapkus, “Ultralow threshold current vertical-cavity surface-emitting lasers obtained with selective oxidation,” Electron. Lett., Vol. 31, No. 11, pp. 886-888, 1995. Jane J. Yang, William I. Simpson, and Lavada A. Moudy, “Electrical properties of Zn in Metalorganic chemical vapor deposition GaAlAs,” ,” J. Appl. Phys., Vol. 53, No. 1, pp. 771-773, 1982. Z. Yang, F. Guarin, I. W. Tao, and W. I. Wang and S. S. Iyer, “Approach to obtain high quality GaN on Si and SiC-on-silicon-on-insulator compliant substrate by molecular-beam epitaxy,” J. Vac. Sci. Technol. B, Vol. 13, No. 2, pp. 789-791, 1995. Meeyoung Yoon, Bun Lee, Jong-Hyeob Baek, Hyo-Hoon Park, and El-Hang Lee, Jeong Yong Lee, “Evolution of the surface cross-hatch pattern in In x Ga 1-x As/GaAs layers grown by metal- organic chemical vapor deposition,” Appl. Phys. Lett. Vol. 68, No. 1, pp. 16-18, 1996. Z. H. Zhu, et al, “Growth of InGaAs multi-quantum wells at 1.3 mm wavelength on GaAs compliant substrates,” Appl. Phys. Lett., Vol. 72, No. 20, pp. 2658, 1998. Z. H. Zhu, R. Zhou, F. E. Ejeckam, Z. Zhang, J. Zhang, J. Greenberg, and Y. H. Lo, H. Q. Hou and B. E. Hammons, “Growth of InGaAs multi-quantum wells at 1.3 µm wavelength on GaAscompliant substrates,” Appl. Phys. Lett., Vol. 72, No. 20, pp.2598-2600, 1998. 220 Appendix A: LASTIP Codes for AlGaAs EEL Structure A A1 Layer Definition File: AlGa6030.layer begin_layer $ column column_num=1 w= 0.500000E+01 mesh_num=2 r=1. bottom_contact column_num=1 from=0 to= 0.500000E+01 && contact_num=1 contact_type=ohmic $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer d= 0.100000E+01 n= 60 && n_doping1= 0.1000E+25 && r= 0.8000E+00 $ layer_mater macro_name=algaas column_num=1 grade_var=1 grade_from=0 && grade_to=0.6 var_symbol1=x var_symbol2=xlam var2=0.94 layer d= 0.01 n=10 n_doping1= 0.1000E+25 shift_center= 0.2630E+00 && r= -0.1200E+01 $ layer_mater macro_name=algaas column_num=1 var_symbol1=x var1=0.6 && var_symbol2=xlam var2=0.94 layer d= 0.8 n=126 shift_center= 0.263 r= -0.1200E+01 n_doping1=1.e+24 $ layer_mater macro_name=algaas column_num=1 var_symbol1=x var1=0.3 && var_symbol2=xlam var2=0.94 layer d= 0.65 n= 102 r= -0.1200E+01 shift_center=-0.2137 $ layer_mater macro_name=ingaas column_num=1 active_macro=InGaAs/AlGaAs && var_symbol1=x var1=0.13 avar1=0.13 avar2=0.3 avar3=300. layer d= 0.012 n=10 shift_center= 0.2137E+00 r= -0.1200E+01 $ layer_mater macro_name=algaas column_num=1 var_symbol1=x var1=0.3 && var_symbol2=xlam var2=0.94 layer d= 0.65 n=102 shift_center= 0.2137 r= -0.1200E+01 $ layer_mater macro_name=algaas column_num=1 var_symbol1=x var1=0.6 && var_symbol2=xlam var2=0.94 layer d= 0.8 n= 128 p_doping1= 1.e+24 shift_center= -0.2506 && r= -0.1200E+01 $ layer_mater macro_name=algaas column_num=1 grade_var=1 grade_from=0.6 && grade_to=0 var_symbol1=x var_symbol2=xlam var2=0.94 layer d= 0.01 n= 10 p_doping1= 5.e+24 r= -1.2 shift_center=-0.000737 $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer d=0.3 n=30 r=1.2 p_doping1=1.e+25 221 $ top_contact column_num=1 from=0 to= 0.500000E+01 && contact_num=2 contact_type=ohmic $ end_layer A2 Simulation Condition Control: AlGa6030.sol $file:alga6030.sol begin use_macrofile macro1=yuanming.mac load_mesh mesh_inf=alga6030.msh include file=alga6030.mater include file=alga6030.doping output sol_outf=alga6030.out newton_par damping_step=5. max_iter=100 print_flag=3 use_sor max_iter=3000 print_sor=noprint $multimode mode_num=1 pick_ymode=1 pick_xmode=1 && $ boundary_type1=[2 2 1 1] $direct_eigen init_wave && length= 1.0000E+03 backg_loss=100. && boundary_type=[2 2 1 1] init_wavel= 0.9400E+00 front_back=[0.1 0.9] && wavel_range=[ 0.9200E+00 0.9600E+00] equilibrium newton_par damping_step=1. print_flag=3 scan var=voltage_1 value_to= -0.1187E+01 print_step= 0.1187E+01 && init_step= 0.2374E+00 min_step=1.e-5 max_step=0.5 scan var=current_1 value_to= 1.5000E+02 print_step= 0.5000E+02 && init_step= 0.1000E+01 min_step=1.e-5 max_step= 0.1000E+02 end A3 Material Macro File: yuanming.mac $*****************MATERIAL MACRO********************************* $Yuanming selected several materials might be used in the TJS $structure from the original macro. Changes will be make to $these macros with reminders. $macros in the file now: $GaAs free-style bulk $AlGaAs free-style bulk $InGaAs/AlGaAs free-style active $InGaAs free-style bulk $************************************************************* $ macro : gaas $ [free-style] $ Bulk GaAs macro at 300K. $ Typical use: 222 $ load_macro name=gaas var1=#xlam mater=#m var_symbol1=xlam $ where xlam is wavelength in microns. $********************************************************* begin_macro gaas material type=semicond band_valleys=(1 1) && el_vel_model=n.gaas hole_vel_model=beta dielectric_constant value=13.1 electron_mass value=0.067 hole_mass value=0.642 band_gap value=1.424 affinity value=4.07 max_electron_mob value=0.85 min_electron_mob value=0. electron_ref_dens value=1.69d23 alpha_n value=0.436 max_hole_mob value=0.04 min_hole_mob value=0. hole_ref_dens value=2.75d23 alpha_p value=0.395 $added by yuanming to override above statement electron_mobility value=0.85 hole_mobility value=0.04 beta_n value=2. electron_sat_vel value=1.0d5 beta_p value=1. hole_sat_vel value=1.d5 norm_field value=4.e5 tau_energy value=1.d-13 lifetime_n value=1.e-7 lifetime_p value=1.e-7 radiative_recomb value=2.d-16 auger_n value=2.d-42 auger_p value=2.d-42 real_index variation=function function(xlam) xev=1.24/xlam; 4.177137-3.512283*xev+5.163323*xev**2 -3.242022*xev**3+0.7935377*xev**4 end_function $absorption value=0. absorption variation=function function(xlam) for 1.24/xlam<1.424 500. for else xev=1.24/xlam; eg=1.424; 223 alpha=-59938.21*(xev-eg)+561976.8*(xev-eg)**2 -855408.5*(xev-eg)**3+430717.0*(xev-eg)**4; 500.+100.*alpha end_function $add by yuanming to override the absorption information above absorption value=100. thermal_kappa value=46. end_macro gaas $*********************************** $ macro algaas $ Bulk Al(x)Ga(1-x)As at 300K $ [free-style] $ Typical use: $ load_macro name=algaas var1=#x var2=#xlam mater=#m && $ var_symbol1=x var_symbol2=xlam $ where xlam is wavelength in microns. $*********************************** begin_macro algaas material type=semicond band_valleys=(1 1) && el_vel_model=n.gaas hole_vel_model=beta dielectric_constant variation=function function(x) 13.1 - 3 * x end_function electron_mass variation=function function(x) for 0.<x<0.45 0.067 + 0.083 * x for 0.45<x<1. 0.85 - 0.14 * x end_function hole_mass variation=function function(x) ( ( 0.087 + 0.063 * x ) ** (3 / 2) + ( 0.62 + 0.14 * x ) ** (3 / 2) ) ** (2 / 3) end_function band_gap variation=function function(x) for 0.<x<0.45 1.424 + 1.247 * x for 0.45<x<1. 1.9 + 0.125 * x + 0.143 *x * x end_function $ $ band gap as external function to be used by absorption bulk_xfunc1 variation=function function(x) for 0.<x<0.45 224 1.424 + 1.247 * x for 0.45<x<1. 1.9 + 0.125 * x + 0.143 *x * x end_function affinity variation=function function(x) for 0<x<0.45 4.07 - 0.748 * x $ 4.07 - 1.1*x for 0.45<x<1. 3.7964 - 0.14 * x end_function max_electron_mob variation=function function(x) for 0<x<0.45 0.85 * exp(-18.516 * x ** 2 ) for 0.45<x<1. 0.02 end_function min_electron_mob value=0. electron_ref_dens value=1.69d23 alpha_n value=0.436 max_hole_mob variation=function function(x) 0.04 - 0.048 * x + 0.02 * x * x end_function min_hole_mob value=0. hole_ref_dens value=2.75d23 alpha_p value=0.395 $added by yuanming to override above statement electron_mobility variation=function function(x) for 0<x<0.45 0.85 * exp(-18.516 * x ** 2 ) for 0.45<x<1. 0.02 end_function hole_mobility variation=function function(x) 0.04 - 0.048 * x + 0.02 * x * x end_function beta_n value=2. electron_sat_vel variation=function function(x) for 0<x<0.45 0.77e5 * (1 - 0.44 * x ) for 0.45<x<1. 225 8.e4 end_function beta_p value=1. hole_sat_vel value=1.d5 norm_field value=4.e5 tau_energy value=1.e-13 radiative_recomb value=1.d-16 auger_n value=1.5d-42 auger_p value=1.5d-42 lifetime_n value=1.e-7 lifetime_p value=1.e-7 real_index variation=function function(x,xlam) for 0<x<0.1 xev=1.24/xlam; n1=4.177137-3.512283*xev+5.163323*xev**2 -3.242022*xev**3+0.7935377*xev**4; n2=3.1715+0.8963*xev-1.4964*xev**2+1.0665*xev**3-0.2310*xev**4; n1+(n2-n1)*(x-0.)/0.1 for 0.1<x<0.2 xev=1.24/xlam; n1=3.1715+0.8963*xev-1.4964*xev**2+1.0665*xev**3-0.2310*xev**4; n2=2.3080+3.5046*xev-4.4066*xev**2+2.4046*xev**3-0.4481*xev**4; n1+(n2-n1)*(x-0.1)/0.1 for 0.2<x<0.3 xev=1.24/xlam; n1=2.3080+3.5046*xev-4.4066*xev**2+2.4046*xev**3-0.4481*xev**4; n2=2.4397+2.7106*xev-3.2388*xev**2+1.6831*xev**3-0.2919*xev**4; n1+(n2-n1)*(x-0.2)/0.1 for 0.3<x<0.4 xev=1.24/xlam; n1=2.4397+2.7106*xev-3.2388*xev**2+1.6831*xev**3-0.2919*xev**4; n2=3.3968-0.8732*xev+1.3142*xev**2-0.7823*xev**3+0.1881*xev**4; n1+(n2-n1)*(x-0.3)/0.1 for 0.4<x<0.5 xev=1.24/xlam; n1=3.3968-0.8732*xev+1.3142*xev**2-0.7823*xev**3+0.1881*xev**4; n2=4.1143 -3.5076*xev +4.5832*xev**2 -2.4916*xev**3 +0.5027*xev**4; n1+(n2-n1)*(x-0.4)/0.1 for 0.5<x<0.6 xev=1.24/xlam; n1=4.1143 -3.5076*xev +4.5832*xev**2 -2.4916*xev**3 +0.5027*xev**4; n2=3.2169 -0.5468*xev +0.8554*xev**2 -0.4952*xev**3 +0.1178*xev**4; n1+(n2-n1)*(x-0.5)/0.1 for 0.6<x<0.7 xev=1.24/xlam; n1=3.2169 -0.5468*xev +0.8554*xev**2 -0.4952*xev**3 +0.1178*xev**4; n2=3.3236 -1.1576*xev +1.6762*xev**2 -0.9628*xev**3 +0.2096*xev**4; n1+(n2-n1)*(x-0.6)/0.1 for 0.7<x<0.8 xev=1.24/xlam; 226 n1=3.3236 -1.1576*xev +1.6762*xev**2 -0.9628*xev**3 +0.2096*xev**4; n2=3.1106 -0.4848*xev +0.6741*xev**2 -0.3351*xev**3 +0.0700*xev**4; n1+(n2-n1)*(x-0.7)/0.1 for 0.8<x<0.9 xev=1.24/xlam; n1=3.1106 -0.4848*xev +0.6741*xev**2 -0.3351*xev**3 +0.0700*xev**4; n2=3.0202 -0.3050*xev +0.4331*xev**2 -0.2045*xev**3 +0.0447*xev**4; n1+(n2-n1)*(x-0.8)/0.1 for 0.9<x<1. xev=1.24/xlam; n1=3.0202 -0.3050*xev +0.4331*xev**2 -0.2045*xev**3 +0.0447*xev**4; n2=2.835238+0.1040565*xev-8.6519323e-2*xev**2 +8.2142524e-2*xev**3 -1.4580269e-2*xev**4; n1+(n2-n1)*(x-0.9)/0.1 end_function $absorption value=0. $ wavelength dependent absorption $ define backgroud aborption as bulk_xfunc2 in units 1/m bulk_xfunc2 variation=function function(x,xlam) 500. end_function absorption variation=function function(x,xlam) for 1.24/xlam<bulk_xfunc1 bulk_xfunc2 for 0<x<0.1 xev=1.24/xlam; eg=bulk_xfunc1; a1=-59938.21*(xev-eg)+561976.8*(xev-eg)**2 -855408.5*(xev-eg)**3+430717.0*(xev-eg)**4; a2=-18043.81*(xev-eg)+470037.7*(xev-eg)**2 -762741.0*(xev-eg)**3+409836.0*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.)/0.1; bulk_xfunc2+alpha*100. for 0.1<x<0.2 xev=1.24/xlam; eg=bulk_xfunc1; a1=-18043.81*(xev-eg)+470037.7*(xev-eg)**2 -762741.0*(xev-eg)**3+409836.0*(xev-eg)**4; a2=91721.91*(xev-eg)+107193.1*(xev-eg)**2 -399274.5*(xev-eg)**3+323630.5*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.1)/0.1; bulk_xfunc2+alpha*100. for 0.2<x<0.3 xev=1.24/xlam; eg=bulk_xfunc1; a1=91721.91*(xev-eg)+107193.1*(xev-eg)**2 -399274.5*(xev-eg)**3+323630.5*(xev-eg)**4; a2=103024.9*(xev-eg)+285712.4*(xev-eg)**2 227 -821371.1*(xev-eg)**3+592386.0*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.2)/0.1; bulk_xfunc2+alpha*100. for 0.3<x<0.4 xev=1.24/xlam; eg=bulk_xfunc1; a1=103024.9*(xev-eg)+285712.4*(xev-eg)**2 -821371.1*(xev-eg)**3+592386.0*(xev-eg)**4; a2=95788.52*(xev-eg)+13612.16*(xev-eg)**2 -176194.3*(xev-eg)**3+275328.6*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.3)/0.1; bulk_xfunc2+alpha*100. for 0.4<x<0.5 xev=1.24/xlam; eg=bulk_xfunc1; a1=95788.52*(xev-eg)+13612.16*(xev-eg)**2 -176194.3*(xev-eg)**3+275328.6*(xev-eg)**4; a2=90497.19*(xev-eg)-118715.9*(xev-eg)**2 +136604.8*(xev-eg)**3+105422.4*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.4)/0.1; bulk_xfunc2+alpha*100. for 0.5<x<0.6 xev=1.24/xlam; eg=bulk_xfunc1; a1=90497.19*(xev-eg)-118715.9*(xev-eg)**2 +136604.8*(xev-eg)**3+105422.4*(xev-eg)**4; a2=76853.11*(xev-eg)+161217.3*(xev-eg)**2 -360155.9*(xev-eg)**3+318543.7*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.5)/0.1; bulk_xfunc2+alpha*100. for 0.6<x<0.7 xev=1.24/xlam; eg=bulk_xfunc1; a1=76853.11*(xev-eg)+161217.3*(xev-eg)**2 -360155.9*(xev-eg)**3+318543.7*(xev-eg)**4; a2=32454.24*(xev-eg)+50188.12*(xev-eg)**2 -16573.62*(xev-eg)**3+106340.0*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.6)/0.1; bulk_xfunc2+alpha*100. for 0.7<x<0.8 xev=1.24/xlam; eg=bulk_xfunc1; a1=32454.24*(xev-eg)+50188.12*(xev-eg)**2 -16573.62*(xev-eg)**3+106340.0*(xev-eg)**4; a2=17639.67*(xev-eg)-31695.61*(xev-eg)**2 +116720.0*(xev-eg)**3+33719.42*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.7)/0.1; bulk_xfunc2+alpha*100. for 0.8<x<0.9 xev=1.24/xlam; eg=bulk_xfunc1; a1=17639.67*(xev-eg)-31695.61*(xev-eg)**2 228 +116720.0*(xev-eg)**3+33719.42*(xev-eg)**4; a2=6957.562*(xev-eg)+964.3477*(xev-eg)**2 -74366.28*(xev-eg)**3+101554.3*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.8)/0.1; bulk_xfunc2+alpha*100. for else xev=1.24/xlam; eg=bulk_xfunc1; a1=6957.562*(xev-eg)+964.3477*(xev-eg)**2 -74366.28*(xev-eg)**3+101554.3*(xev-eg)**4; a2=6957.562*(xev-eg)+964.3477*(xev-eg)**2 -74366.28*(xev-eg)**3+101554.3*(xev-eg)**4; alpha=a1+(a2-a1)*(x-0.9)/0.1; bulk_xfunc2+alpha*100. end_function $add by yuanming to override the absorption information above absorption value=100. thermal_kappa value=46. end_macro algaas $ *********************************************** $ acitve layer macro : In(xw)Ga(1-xw)As/Al(xb)Ga(1-xb)As $ [free-style] $ xw=Indium composition in well, $ xb=Al. comp. in barrier, temper=temperature $ Typical use: $ get_active_layer name=InGaAs/AlGaAs mater=#m && $ var1=#xw var2=#xb var_symbol1=xw var_symbol2=xb $ ********************************************** $ begin_active_layer InGaAs/AlGaAs $ layer_type type=strained_well valley_gamma=1 valley_l=4 && valley_hh=1 valley_lh=1 $ eg0_well variation=function function(xw,temper) shift0=-5.5d-4*300.**2/(300.+225.) ; shift=-5.5d-4*temper**2/(temper+225.) ; $1.424-1.4895*xw+0.4205*xw*xw+shift-shift0 $ Ref: Appl. Phys. Lett. 58 (20) p. 2208 1.424-1.614*xw+0.55*xw*xw+shift-shift0 end_function $ eg0_bar variation=function function(xb,temper) for 0.<xb<0.45 shift0=-5.5d-4*300.**2/(300.+225.) ; shift=-5.5d-4*temper**2/(temper+225.) ; 1.424+1.247*xb+shift-shift0 229 for 0.45<xb<1. shift0=-5.5d-4*300.**2/(300.+225.) ; shift=-5.5d-4*temper**2/(temper+225.) ; 1.9+.125*xb+0.143*xb**2+shift-shift0 end_function $ lband_well value=0.28 $ lband_bar value=0.28 $ $ compressive strain is negative: $ strain_well variation=function function(xw,temper) a0ga=5.65325; a0in=6.0584; acomp=a0ga+xw*(a0in-a0ga); (a0ga-acomp)/acomp end_function $ strain_bar value=0. $ band_offset value=0.6 $ delta_so_well variation=function function(xw) 0.366-0.0451*xw+0.0691*xw*xw end_function $ delta_so_bar value=0.366 $ mass_gamma_well variation=function function(xw) 0.063-0.036*xw end_function $ mass_l_well value=0.56 $ mass_gamma_bar variation=function function(xb) 0.067+0.083*xb end_function $ mass_l_bar variation=function function(xb) 0.56+0.1*xb end_function $ gamma1_well variation=function function(xw) g1ga=6.9 ; g1in=19.7 ; 230 g1ga+xw*(g1in-g1ga) end_function $ gamma2_well variation=function function(xw) g2ga=2.2 ; g2in=8.4 ; g2ga+xw*(g2in-g2ga) end_function $ gamma3_well variation=function function(xw) g3ga=2.9 ; g3in=9.3 ; g3ga+xw*(g3in-g3ga) end_function $ a_well variation=function function(xw) dhga=-9.8 ; dhin=-5.0 ; dhga+xw*(dhin-dhga) end_function $ b_well variation=function function(xw) duga=-1.76 ; duin=-1.8 ; duga+xw*(duin-duga) end_function $ c11_well variation=function function(xw) c11ga=11.9 ; c11in=8.33 ; c11ga+xw*(c11in-c11ga) end_function $ c12_well variation=function function(xw) c12ga=5.38 ; c12in=4.53 ; c12ga+xw*(c12in-c12ga) end_function $ gamma1_bar variation=function function(xb) g1ga=6.9 ; g1al=3.45 ; g1ga*(1.-xb)+g1al*xb end_function $ 231 gamma2_bar variation=function function(xb) g2ga=2.2 ; g2al=0.68 ; g2ga*(1.-xb)+g2al*xb end_function $ gamma3_bar variation=function function(xb) g3ga=2.9 ; g3al=1.29 ; g3ga*(1.-xb)+g3al*xb end_function $ a_bar variation=function function(xb) dhga=-9.8 ; dhal=-9.8 ; dhga*(1.-xb)+dhal*xb end_function $ b_bar variation=function function(xb) duga=-1.76 ; dual=-1.76 ; duga*(1.-xb)+dual*xb end_function $ c11_bar variation=function function(xb) c11ga=11.9 ; c11al=12.02 ; c11ga*(1.-xb)+c11al*xb end_function $ c12_bar variation=function function(xb) c12ga=5.38 ; c12al=5.70 ; c12ga*(1.-xb)+c12al*xb end_function $ lattice_constant value=5.65325 $ end_active_layer InGaAs/AlGaAs $*********************************** $ macro in(x)ga(1-x)as $ Bulk In(x)Ga(1-x)As, lattice matched to GaAs. $ For InGaAs grown on InP, please see macro ingaasp $ [free-style] $ Typical use: 232 $ load_macro name=ingaas var1=#x mater=#m var_symbol1=x $*********************************** $ begin_macro ingaas material type=semicond band_valleys=(1 1) && el_vel_model=n.gaas hole_vel_model=beta dielectric_constant variation=function function(x) 13.1 + ( 15.15 - 13.1 )* x end_function electron_mass variation=function function(x) 0.067 - 0.036 * x end_function hole_mass variation=function function(x) ( 1/( 12.6 + 23.9* x )** (3 / 2) + 1/( 1.8 + 1.1* x )** (3 / 2) ) ** (2 / 3) end_function band_gap variation=function function(x) a0=5.65325+x*(6.05840-5.65325); dh=-9.8+x*(-5.+9.8); du=-1.76+x*(-1.8+1.76); c11=11.9+x*(8.33-11.9); c12=5.38+x*(4.53-5.38); egu=1.424-1.4895*x+0.4205*x*x; epsxx=(5.65325-a0)/a0; deh=2.*dh*(c11-c12)/c11*epsxx; des=1.*du*(c11+2.*c12)/c11*epsxx; egu+deh-des end_function affinity variation=function function(x) a0=5.65325+x*(6.05840-5.65325); dh=-9.8+x*(-5.+9.8); du=-1.76+x*(-1.8+1.76); c11=11.9+x*(8.33-11.9); c12=5.38+x*(4.53-5.38); egu=1.424-1.4895*x+0.4205*x*x; epsxx=(5.65325-a0)/a0; deh=2.*dh*(c11-c12)/c11*epsxx; des=1.*du*(c11+2.*c12)/c11*epsxx; eghh=egu+deh-des; bohh=1.424-eghh; 4.07+0.6*bohh end_function electron_mobility variation=function function(doping_n,doping_p,trap_1) mu_max=1.46; mu_min=0.; 233 ref_dens=2.1d23; alpha=0.4; total_doping=doping_n+doping_p+trap_1; mu_min+(mu_max-mu_min)/(1+(total_doping/ref_dens)**alpha) end_function hole_mobility variation=function function(doping_n,doping_p,trap_1) mu_max=0.22; mu_min=0.; ref_dens=9.6d20; alpha=0.4; total_doping=doping_n+doping_p+trap_1; mu_min+(mu_max-mu_min)/(1+(total_doping/ref_dens)**alpha) end_function $added by yuanming to override above statement electron_mobility value=1.46 hole_mobility value=0.22 beta_n value=2. electron_sat_vel value=2.1e5 beta_p value=1. hole_sat_vel value=1.d5 norm_field value=4.e5 tau_energy value=1.e-13 radiative_recomb value=2.d-16 auger_n value=2.d-42 auger_p value=2.d-42 real_index variation=function function(x) 3.65 + (3.892 - 3.65) * x end_function lifetime_n value=1.e-7 lifetime_p value=1.e-7 absorption value=0. $add by yuanming to override the absorption information above absorption value=100. thermal_kappa value=46. end_macro ingaas 234 Appendix B: LASTIP Codes for TJS-GaAs Structure B1 Layer Definition File: deepgaas.layer begin_layer $ column column_num=1 w=2. mesh_num=15 r=0.8 column column_num=2 w=5. mesh_num=30 r=-1.2 column column_num=3 w=2. mesh_num=15 r=1.2 $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=gaas column_num=2 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=gaas column_num=3 solve_wave=no var_symbol1=xlam && var1=0.94 layer d= 0.100000E+01 n= 20 r= 0.8000E+00 $ layer_mater macro_name=algaas column_num=1 grade_var=1 grade_from=0 && grade_to=0.4 solve_wave=no var_symbol1=x var_symbol2=xlam var2=0.94 layer_mater macro_name=algaas column_num=2 grade_var=1 grade_from=0 && grade_to=0.4 solve_wave=no var_symbol1=x var_symbol2=xlam var2=0.94 layer_mater macro_name=algaas column_num=3 grade_var=1 grade_from=0 && grade_to=0.4 solve_wave=no var_symbol1=x var_symbol2=xlam var2=0.94 layer d= 0.01 n=10 r= -0.1200E+01 $ layer_mater macro_name=algaas column_num=1 solve_wave=no var_symbol1=x && var1=0.4 var_symbol2=xlam var2=0.94 layer_mater macro_name=algaas column_num=2 var_symbol1=x var1=0.4 && var_symbol2=xlam var2=0.94 layer_mater macro_name=algaas column_num=3 solve_wave=no var_symbol1=x && var1=0.4 var_symbol2=xlam var2=0.94 layer d= 0.8 n=20 r= -0.1200E+01 $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=algaas column_num=2 var_symbol1=x var1=0.3 && var_symbol2=xlam var2=0.94 layer_mater macro_name=gaas column_num=3 solve_wave=no var_symbol1=xlam && var1=0.94 layer d= 0.65 n= 30 r= -0.1200E+01 n_doping1=1.e+24 p_doping3=1.e+24 $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=ingaas column_num=2 active_macro=InGaAs/AlGaAs && var_symbol1=x var1=0.13 avar_symbol1=xw avar1=0.13 avar_symbol2=xb && avar2=0.3 layer_mater macro_name=gaas column_num=3 solve_wave=no var_symbol1=xlam && var1=0.94 235 layer d= 0.012 n=10 r= -0.1200E+01 n_doping1=1.e+24 p_doping3=1.e+24 $ layer_mater macro_name=gaas column_num=1 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=algaas column_num=2 var_symbol1=x var1=0.3 && var_symbol2=xlam var2=0.94 layer_mater macro_name=gaas column_num=3 solve_wave=no var_symbol1=xlam && var1=0.94 layer d= 0.65 n= 30 r= -1.2 n_doping1=1.e+24 p_doping3=1.e+24 $ layer_mater macro_name=void column_num=1 solve_wave=no layer_mater macro_name=algaas column_num=2 var_symbol1=x var1=0.4 && var_symbol2=xlam var2=0.94 layer_mater macro_name=void column_num=3 solve_wave=no layer d=0.65 n=20 r=-1.2 $ layer_mater macro_name=void column_num=1 solve_wave=no layer_mater macro_name=algaas column_num=2 grade_var=1 grade_from=0.4 && grade_to=0 solve_wave=no var_symbol1=x var_symbol2=xlam var2=0.94 layer_mater macro_name=void column_num=3 solve_wave=no layer d=0.01 n=10 r=-1.2 $ layer_mater macro_name=void column_num=1 solve_wave=no layer_mater macro_name=gaas column_num=2 solve_wave=no var_symbol1=xlam && var1=0.94 layer_mater macro_name=void column_num=3 solve_wave=no layer d=0.3 n=20 r=1.2 top_contact column_num=1 from=0 to=2. contact_num=1 contact_type=ohmic top_contact column_num=3 from=0 to=2. contact_num=2 contact_type=ohmic $ end_layer B2 Simulation Condition Control: deepgaas.sol $file:deepgaas.sol begin use_macrofile macro1=yuanming.mac load_mesh mesh_inf=deepgaas.msh include file=deepgaas.mater include file=deepgaas.doping output sol_outf=deepgaas.out newton_par damping_step=5. max_iter=100 print_flag=3 $use_sor max_iter=3000 print_sor=noprint multimode mode_num=1 pick_ymode=1 pick_xmode=1 && boundary_type1=[1 1 1 1] direct_eigen init_wave && length= 1.0000E+03 backg_loss=100. && boundary_type=[1 1 1 1] init_wavel= 0.9400E+00 front_back=[0.1 0.9] && wavel_range=[ 0.9200E+00 0.9600E+00] equilibrium 236 newton_par damping_step=1. print_flag=3 scan var=voltage_1 value_to= -0.1187E+01 print_step= 0.1187E+01 && init_step= 0.2374E+00 min_step=1.e-5 max_step=0.5 scan var=current_1 value_to= 1.5000E+02 print_step= 0.5000E+02 && init_step= 0.1000E+01 min_step=1.e-5 max_step= 0.1000E+02 end 237 Appendix C: MATLAB Code for Transmission Matrix Method C1 VCSEL Reflectivity as A Function of Wavelength clear all; format long e; idbr=10; icavity=4*idbr; lamda=980e-9; nair=1; nhigh=3.663; nlow=2.957; ncavity=3.663; nsub=2.957; dbrhigh=lamda/4/nhigh; dbrlow=lamda/4/nlow; cavity=4*dbrhigh; topdbr=10; bottomdbr=10; %generate index matrix index(1,1)=nair; index(1,2)=0; n=1; for m=1:topdbr n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh+index(n-1,2); n=n+1; index(n,1)=nlow; index(n,2)=dbrlow+index(n-1,2); end n=n+1; index(n,1)=ncavity; index(n,2)=index(n-1,2)+cavity; for m=1:bottomdbr n=n+1; index(n,1)=nlow; index(n,2)=dbrlow+index(n-1,2); n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh+index(n-1,2); end n=n+1; index(n,1)=nsub; 238 index(n,2)=index(n-1,2); for ii=1:4001 lamda=780e-9+(1180e-9-780e-9)/4000*(ii-1); % calculate the transmission matrix matrix=eye(2); for m=1:n-1 d(1,1)=0.5*(1+index(m+1,1)/index(m,1)); d(1,2)=0.5*(1-index(m+1,1)/index(m,1)); d(2,1)=d(1,2); d(2,2)=d(1,1); p(1,1)=exp(+i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); p(1,2)=0; p(2,1)=0; p(2,2)=exp(-i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); matrix=matrix*d*p; end x(ii)=lamda; y(ii)=abs(matrix(2,1)/matrix(1,1))^2; end plot(x,y,'linewidth',1) hold on title('VCSEL Reflectivity','fonts',18) ylabel('Reflectivity','fonts',16) xlabel('Wavelength','fonts',16) hold off C2 Electrical Field Distribution in VCSEL clear all; format long e; idbr=10; icavity=4*idbr; lamda=980e-9; nair=1; nhigh=3.663; nlow=2.957; ncavity=3.663; nsub=2.957; dbrhigh=lamda/4/nhigh; dbrlow=lamda/4/nlow; cavity=4*dbrhigh; topdbr=10; bottomdbr=10; %generate index matrix index(1,1)=nair; index(1,2)=0; n=1; 239 for m=1:topdbr n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh+index(n-1,2); n=n+1; index(n,1)=nlow; index(n,2)=dbrlow+index(n-1,2); end n=n+1; index(n,1)=ncavity; index(n,2)=index(n-1,2)+cavity; for m=1:bottomdbr n=n+1; index(n,1)=nlow; index(n,2)=dbrlow+index(n-1,2); n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh+index(n-1,2); end n=n+1; index(n,1)=nsub; index(n,2)=index(n-1,2); % calculate the transmission matrix matrix=eye(2); for m=1:n-1 d(1,1)=0.5*(1+index(m+1,1)/index(m,1)); d(1,2)=0.5*(1-index(m+1,1)/index(m,1)); d(2,1)=d(1,2); d(2,2)=d(1,1); p(1,1)=exp(+i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); p(1,2)=0; p(2,1)=0; p(2,2)=exp(-i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); matrix=matrix*d*p; end field=[1;matrix(2,1)/matrix(1,1)] %generate index matrix index(1,1)=nair; index(1,2)=0; n=1; for m=1:topdbr for k=1:idbr n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh/idbr+index(n-1,2); end for k=1:idbr n=n+1; index(n,1)=nlow; index(n,2)=dbrlow/idbr+index(n-1,2); 240 end end for k=1:icavity n=n+1; index(n,1)=ncavity; index(n,2)=index(n-1,2)+cavity/icavity; end for m=1:bottomdbr for k=1:idbr n=n+1; index(n,1)=nlow; index(n,2)=dbrlow/idbr+index(n-1,2); end for k=1:idbr n=n+1; index(n,1)=nhigh; index(n,2)=dbrhigh/idbr+index(n-1,2); end end n=n+1; index(n,1)=nsub; index(n,2)=index(n-1,2); x(1)=index(1,2); y(1)=real(field(1)+field(2))*3; for m=1:n-1 d(1,1)=0.5*(1+index(m+1,1)/index(m,1)); d(1,2)=0.5*(1-index(m+1,1)/index(m,1)); d(2,1)=d(1,2); d(2,2)=d(1,1); p(1,1)=exp(+i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); p(1,2)=0; p(2,1)=0; p(2,2)=exp(-i*2*pi*index(m+1,1)/lamda*(index(m+1,2)-index(m,2))); field=inv(d*p)*field; x(m+1)=index(m+1,2); y(m+1)=real(field(1)+field(2))*3; end ne2=0; e2=0; for m=1:n ne2=ne2+index(m,1)*y(m)*y(m); e2=e2+y(m)*y(m); end neff=ne2/e2 fmax=max(y.*y) plot(x,y.*y/fmax*4,'linewidth',1) hold on drawindex title('Electrical Field','fonts',18) ylabel('Field(a.u.)','fonts',16) xlabel('Position','fonts',16) 241 Appendix D: Process Follower of GaAs Based Broad Area Laser Process name Process Description Process Action Process Parameter P-Metal Patterning IR Lithography Clean Ace, Meth, DI Rinse Bake 120C for 30sec. Spin AZ5214 @ 3Krpm 3 times Edge Wipe Swabs /w acetone Bake 120C for 30s. Expose P-Metal Mask; 70mJ Bake 120C for 1 min. Flood Exposure 270mJ Develop AZ400K 1:4, 25-30sec. DI Rinse 1min under running water Inspection Complete development PR Descum/RIE O2, 60W, 200mT, 30sec P-Metal Deposition Oxide removal Etch HCl:H2O(1:10) 1min, DI Rinse Metal Evaporation Ti Deposition 300A, 2A/sec Pt Deposition 500A, 3A/sec Au Deposition 1500A, 4A/sec Liftoff Ace /w swab Clean Ace, Meth, DI Rinse Mesa Forming Lithography Spin S1813 @ 3Krpm Edge Wipe Swabs /w acetone Bake 120C for 1min Expose 10 mm mask twice; 140mJ Develop MF321, 45sec. DI Rinse 1min under running water Inspection Complete development PR Descum/RIE O2, 60W, 200mT, 30sec Bake 120C for 20S Etching Etch H2SO4:H2O2:H2O(4:1:40) DI Rinse 1min under running water Remove PR Aceton, Methanol, DI Inspection Complete remove of PR PR Descum/RIE O2, 60W, 200mT, 1 min N-Contact Deposition Lapping Wax Mounting Check thickness Substrate Thinning Thin to 4.5 mils (~110 ٛ m) Clean TCE, ACE, Meth, DI Rinse Metal Evaporation AuGe Deposition 1000A, 4A/sec Ni Deposition 500A, 3A/sec Au Deposition 2000A, 5A/sec Anneal RTA 400C 1C/sec 30sec Cleaving Scribing Sticky tab mounting Tape to glass slide Scribing Use diamond-tipped scriber Removing Remove sample from tape Breaking Break with roller
Abstract (if available)
Abstract
This dissertation presents research projects with the common theme: novel GaAs based device structures grown by metal organic chemical vapor deposition (MOCVD).
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The growth and characterization of III-V semiconductor nanowire arrays by nanoscale selective area metalorganic chemical vapor deposition
PDF
Coupled multiple micro-resonators design and active semiconductor micro-resonator fabrication
PDF
Vapor phase deposition of polymers in the presence of low vapor pressure liquids
PDF
The selective area growth and coalescence of indium phosphide nanostripe arrays on silicon through MOCVD for NIR monolithic integration
PDF
Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells
PDF
Inorganic/organic hybrid materials and grafted coatings via vapor phase deposition processes
PDF
Controlled synthesis of large grain single crystal graphene by chemical vapor deposition
PDF
Modulation response and linewidth properties of microcavity lasers
PDF
Topological and non-Hermitian semiconductor lasers
PDF
Understanding structural changes of polymer films during vapor phase deposition
PDF
Fabrication of polymer films on liquid substrates via initiated chemical vapor deposition: controlling morphology and composition
PDF
Growth and field emission of multi-walled carbon nanotubes
PDF
Photodetector: devices for optical data communication
PDF
A study of semiconductor microresonators in chip-scale wavelength division multiplexing (CS-WDM) systems
PDF
Fabrication of silicon-based membranes via vapor-phase deposition and pyrolysis of organosilicon polymers
PDF
Simultaneous monomer deposition and polymerization at low substrate temperatures for the formation of porous polymer membranes
PDF
Chemical vapor deposition of graphene: synthesis, characterization, and applications
PDF
Continuum and pore netwok modeling of preparation of silicon-carbide membranes by chemical-vapor deposition and chemical-vapor infiltration
PDF
The patterning of polymer thin films on porous substrates via initiated chemical vapor deposition
PDF
Novel processing of liquid substrates via initiated chemical vapor depostion
Asset Metadata
Creator
Deng, Yuanming
(author)
Core Title
Gallium arsenide based semiconductor laser design and growth by metal-organic chemical vapor deposition
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Electrical Engineering
Publication Date
09/27/2006
Defense Date
08/08/2006
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
compliant substrate,gallium arsenide,high efficiency,metal-organic chemical vapor deposition,OAI-PMH Harvest,semiconductor laser
Language
English
Advisor
Dapkus, P. Daniel (
committee chair
), Goo, Edward K. (
committee member
), O'Brien, John (
committee member
)
Creator Email
yuanmind@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m42
Unique identifier
UC1108781
Identifier
etd-Deng-20060927 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-4368 (legacy record id),usctheses-m42 (legacy record id)
Legacy Identifier
etd-Deng-20060927.pdf
Dmrecord
4368
Document Type
Dissertation
Rights
Deng, Yuanming
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
compliant substrate
gallium arsenide
high efficiency
metal-organic chemical vapor deposition
semiconductor laser