@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Applied Science, Faculty of"@en, "Electrical and Computer Engineering, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Vafaei, Raha"@en ; dcterms:issued "2010-10-06T14:49:54Z"@en, "2010"@en ; vivo:relatedDegree "Master of Applied Science - MASc"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """This thesis investigates the fabrication of 1550 nm emitting InP semiconductor racetrack resonator lasers (SRLs) via wet etching techniques. The method of choice for SRL fabrication is reported to be via dry etching. Dry etching is a complex, time consuming and expensive process which leaves relatively rougher surfaces and sidewalls compared to wet etching techniques. In this thesis, coupling, racetrack resonators, and edge emitter laser theory were studied for the SRL design. Then, a fabrication process for the SRLs was developed via wet etching techniques. The light emitting diode (LED) characteristics of the fabricated devices were observed and successfully measured. The spectrum of the device was also measured with optical spectrum analyzer (OSA) and resonances were observed. However, lasing was not observed. The cleaving process is a major limiting step in the fabrication and it is being improved. In parallel to the wet etching fabrication at UBC, dry etching (the common method for SRL fabrication) is being performed at the Centre de Recherche en Nanofabrication et Nanocaractérisation (CNR2) at the Université de Sherbrooke."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/28960?expand=metadata"@en ; skos:note "Fabrication And Characterization of InP Semiconductor Ring Lasers by Raha Vafaei B.Sc., The University of British Columbia, 2008 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Applied Science in The Faculty of Graduate Studies (Electrical and Computer Engineering) The University Of British Columbia (Vancouver) October, 2010 c￿ Raha Vafaei 2010 Abstract This thesis investigates the fabrication of 1550 nm emitting InP semicon- ductor racetrack resonator lasers (SRLs) via wet etching techniques. The method of choice for SRL fabrication is reported to be via dry etching. Dry etching is a complex, time consuming and expensive process which leaves rel- atively rougher surfaces and sidewalls compared to wet etching techniques. In this thesis, coupling, racetrack resonators, and edge emitter laser theory were studied for the SRL design. Then, a fabrication process for the SRLs was developed via wet etching techniques. The light emitting diode (LED) characteristics of the fabricated devices were observed and successfully mea- sured. The spectrum of the device was also measured with optical spectrum analyzer (OSA) and resonances were observed. However, lasing was not observed. The cleaving process is a major lim- iting step in the fabrication and it is being improved. In parallel to the wet etching fabrication at UBC, dry etching (the common method for SRL fab- rication) is being performed at the Centre de Recherche en Nanofabrication et Nanocaractérisation (CNR2) at the Université de Sherbrooke. ii Preface I am one of the co-authors in the Conference and Optics Letter titled, “Ring Resonator Reflector with a Waveguide Crossing” and “Design and Charac- terization of Microring Reflectors with a Waveguide Crossing” respectively. More specifically, I did the roundtrip loss value measurements and calcu- lations, rebuilt and improved the SOI setup used for the measurements, performed some of the temperature measurements, wrote the curve fitting code which was modified by Wei Shi to fit his design and helped with the design procedure used to determine device parameters such as effective in- dex, group index and coupling coefficients. The manuscript was written by Wei Shi. I am one of the co-authors in the paper published in the Journal of Light- waves titled, “Temperature Effects On Silicon-On-Insulator (SOI) Racetrack Resonators: a Coupled Analytic and 2D Finite Difference Approach”. More specifically I did some of the design and analysis of the racetrack resonator, designed the SOI racetrack resonators that were fabricated and used to ver- ify the design procedure and to study the temperature effects. I also did the quality factor analysis. The manuscript was written by Dr. Nicolas Rouger. I am one of the co-authors in the SPIE paper titled, “Simulation of a 1550 nm InGaAsP-InP transistor laser”, I researched the required p-type iii Preface contact composition for obtaining ohmic contact behaviour with the lowest reported resistance on the InGaAsP base contact layer. The manuscript was written by Wei Shi. I am one of the co-authors in the poster titled, “SOI Nanophotonic De- vices Analysis and Fabrication” presented at the Pacific Centre for Advanced Materials and Microstructures (PCAMM). More specifically I wrote the sec- tion on modelling and analysis of InP lasers, and provided the quality factor plot and the SOI racetrack resonator design. The list of publications include: 1. Wei Shi, Raha Vafaei, Miguel Ángel Guillén Torres, Nicolas A. F. Jaeger, Lukas Chrostowski, “Design and Characterization of Microring Reflectors with a Waveguide Crossing”, Optics Letters, vol. 35, issue 17, pp. 2901-2903, 09/2010. 2. Wei Shi, Raha Vafaei, Miguel Ángel Guillén Torres, Nicolas A. F. Jaeger, Lukas Chrostowski, “Ring Resonator Reflector with a Waveg- uide Crossing”, 2010 International Conference on Optical MEMS and Nanophotonics, 09/08/2010. 3. Nicolas Rouger, Lukas Chrostowski, Raha Vafaei, “Temperature Ef- fects On Silicon-On-Insulator (SOI) Racetrack Resonators: a Coupled Analytic and 2D Finite Difference Approach”, Journal of Lightwave Technology, vol. 28, issue 9, pp. 1380–1391, 05/2010. 4. Wei Shi, Zigang Duan, Raha Vafaei, Nicolas Rouger, Behnam Faraji, Lukas Chrostowski, “Simulation of a 1550 nm InGaAsP-InP transistor iv Preface laser”, Photonics and OptoElectronics Meetings, Proc. SPIE, vol. 7516, Wuhan, China, pp. 75160P-75160P-7, 08/2009. 5. Miguel Ángel Guillén Torres, Nicolas Rouger, Raha Vafaei, Shahrooz M. Amin, Robi Boeck, Behnam Faraji, Brendan Francis, Alina Kulpa, Juan Mario Michaan, Lukas Chrostowski, Nicolas Jaeger, Dan Dep- tuck, “SOI Nanophotonic Devices Analysis and Fabrication”, Pacific Centre for Advanced Materials and Microstructures (PCAMM) An- nual Meeting, 29/11/2008. v Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . xi Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Operating Principle . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.5 Novel Contributions . . . . . . . . . . . . . . . . . . . . . . . 12 1.6 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . 12 2 InP Racetrack Laser Design . . . . . . . . . . . . . . . . . . . 15 2.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.2 Mode Calculation . . . . . . . . . . . . . . . . . . . . . . . . 15 2.3 Racetrack Laser Coupler Design . . . . . . . . . . . . . . . . 18 2.4 Racetrack Laser Resonator Design . . . . . . . . . . . . . . . 29 2.5 Double Bus Passive Resonator . . . . . . . . . . . . . . . . . 37 3 Fabrication Via Wet Etching . . . . . . . . . . . . . . . . . . . 41 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.2 InP Material Structure . . . . . . . . . . . . . . . . . . . . . 45 3.3 InP Edge Emitter Racetrack Resonator Laser . . . . . . . . . 49 vi Table of Contents 3.3.1 Mask . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.3.2 Sample Preparation . . . . . . . . . . . . . . . . . . . 53 3.3.3 Photolithography . . . . . . . . . . . . . . . . . . . . 54 3.3.4 Wet Etching . . . . . . . . . . . . . . . . . . . . . . . 63 3.3.5 Planarization . . . . . . . . . . . . . . . . . . . . . . . 75 3.3.6 Metalization . . . . . . . . . . . . . . . . . . . . . . . 77 3.3.7 Cleaving . . . . . . . . . . . . . . . . . . . . . . . . . 81 3.3.8 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 82 3.3.9 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 88 3.4 InP Edge Emitter Ridge Waveguide Laser . . . . . . . . . . . 90 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 4 Fabrication Via Dry Etching . . . . . . . . . . . . . . . . . . . 92 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 5.1 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 Appendix A Clewin Matlab Script Used for the Mask Layout . . . . . . 109 vii List of Tables 2.1 Refractive Index and Lattice Constant Data . . . . . . . . . 16 2.2 Power Coupling for Rectangular Waveguides W = 2 µm H = 2 µm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.3 Power Coupling for Rectangular Waveguides W = 3 µm H = 2 µm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.4 Power Coupling for Trapezoidal Waveguide Wbottom = 3 µm, Wtop = 1 µm . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.5 Predicted Performance for Trapezoidal Waveguide Wbottom = 3, µm Wtop = 1 µm, R = 300 µm and a 200 nm Thick Polyimide Planarization, Biased @ 100 mA . . . . . . . . . . 28 2.6 Predicted Performance for Trapezoidal Waveguide Wbottom = 3 µm Wtop = 1µm R = 50µm and a 200nm Thick Polyimide Planarization, Biased @ 100mA . . . . . . . . . . . . . . . . . 29 2.7 Simulation Parameters. Data taken from Lecture Notes of Dr.Lukas Chrostowski and Ref [1] . . . . . . . . . . . . . . . . 31 3.1 Overview of The Fabrication Steps . . . . . . . . . . . . . . . 43 3.2 R2 - 1550 nm FP-LD Epi-wafer Structure[2] . . . . . . . . . . 46 3.3 Solvent and Properties (Data from [3]) . . . . . . . . . . . . . 55 3.4 Etchants Tried . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.5 Wet Etch Monitor Table for Sample : R2 epi-wafer, ID: D04 Date: Jul19/2010 . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.6 Summary of Ohmic Contact Data . . . . . . . . . . . . . . . . 77 4.1 Etch Depth Analysis, W = 2 µm . . . . . . . . . . . . . . . . 98 4.2 Etch Depth Analysis, W = 3 µm . . . . . . . . . . . . . . . . 99 4.3 Predicted Performance for Waveguide W = 2 µm R = 300 µm 99 viii List of Figures 1.1 Ring Laser Structure by Sorel and Group. . . . . . . . . . . . . 4 1.2 Ring Laser Spectrum by Sorel and Group. . . . . . . . . . . . 5 1.3 Micro-square Resonators. . . . . . . . . . . . . . . . . . . . . . 6 1.4 Microring Amplifier at University of Maryland. . . . . . . . . . 7 1.5 SEM of Dry Etched Test Coupler. . . . . . . . . . . . . . . . . 8 1.6 SEM of SRL By S. Park, S. S. Kim, L. Wang, and S. T. Ho. . . 9 1.7 Spectra of SRL By S. Park, S. S. Kim, L. Wang, and S. T. Ho. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.8 Hybrid Silicon Evanescent Device. . . . . . . . . . . . . . . . . 10 1.9 SRL with Tuneable Coupler by G.Mezosi S.Furst and M.Sorel. . . 11 2.1 Refractive Index Amplitude Profile. . . . . . . . . . . . . . . . 17 2.2 First and Second Order Mode Distributions. . . . . . . . . . . 18 2.3 Even and Odd Modes . . . . . . . . . . . . . . . . . . . . . . . 20 2.4 Crossover Length vs. Separation . . . . . . . . . . . . . . . . . 22 2.5 Power Coupling vs. Waveguide Spacing . . . . . . . . . . . . . 23 2.6 Mode Distribution in Trapezoidal Waveguide. . . . . . . . . . . 26 2.7 Transverse Component of Photons in Laser. . . . . . . . . . . . 26 2.8 Refractive Index Distribution in the Coupler .. . . . . . . . . . 27 2.9 neff and ng vs. λ .. . . . . . . . . . . . . . . . . . . . . . . . . 30 2.10 Threshold Current vs. κ .. . . . . . . . . . . . . . . . . . . . . 33 2.11 Output Power vs. κ At 3 Bias Conditions.. . . . . . . . . . . . 33 2.12 Output Power vs. κ for a Varying Ring Radius with Low Losses.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.13 Output Power vs. κ for a Varying Ring Radius with High Losses.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.14 Output Power vs. Current with High Losses.. . . . . . . . . . . 36 2.15 SOI Resonator.. . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.16 SOI Measurement Setup.. . . . . . . . . . . . . . . . . . . . . . 39 2.17 SOI Racetrack Resonator Drop Port Response.. . . . . . . . . 40 3.1 Fabrication Flow Diagram.. . . . . . . . . . . . . . . . . . . . . 43 ix List of Figures 3.2 The Epi-wafer PL Data .. . . . . . . . . . . . . . . . . . . . . . 48 3.3 Mask Layout.. . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.4 Layer 1(a) Mask Layout.. . . . . . . . . . . . . . . . . . . . . . 51 3.5 Layer 1(b) Mask Layout.. . . . . . . . . . . . . . . . . . . . . . 51 3.6 Layer 2 Mask Layout.. . . . . . . . . . . . . . . . . . . . . . . . 51 3.7 Theoretical Resolution Versus Mask to Photoresist Gap. . . . . 59 3.8 1µm Resolution Lithography. . . . . . . . . . . . . . . . . . . . 60 3.9 Rough Edge Mask Defects. . . . . . . . . . . . . . . . . . . . . 61 3.10 Photoresist Impurity Defect. . . . . . . . . . . . . . . . . . . . 62 3.11 Stubborn Photoresist Residue. . . . . . . . . . . . . . . . . . . 62 3.12 3H2SO4 : H2O2 : H2O System. . . . . . . . . . . . . . . . . . . 66 3.13 H3PO4/HCl System. . . . . . . . . . . . . . . . . . . . . . . . . 66 3.14 HCl/ CH3COOH/H2O2 System. . . . . . . . . . . . . . . . . . 67 3.15 HBr/H2O2/H2O/HCl System. . . . . . . . . . . . . . . . . . . 67 3.16 Etch Profile with HBr/H2O2/H2O/HCl System. . . . . . . . . 68 3.17 Spherical Lens Formation with HBr/H2O2/H2O/HCl System. . . 69 3.18 Side Wall Wet Etch Profile. . . . . . . . . . . . . . . . . . . . . 72 3.19 PR Lifting Defect. . . . . . . . . . . . . . . . . . . . . . . . . . 73 3.20 Bubble Formation Defect. . . . . . . . . . . . . . . . . . . . . . 74 3.21 Planarization Flow Diagram. . . . . . . . . . . . . . . . . . . . 78 3.22 IV Characteristic of the N-type Contact . . . . . . . . . . . . . 80 3.23 IV Characteristic of the P-type Contact . . . . . . . . . . . . . 80 3.24 Cleaved Region Optical Image. . . . . . . . . . . . . . . . . . . 83 3.25 Final Fabricated Device. . . . . . . . . . . . . . . . . . . . . . 83 3.26 Edge Emitter Setup. . . . . . . . . . . . . . . . . . . . . . . . . 84 3.27 LED Characteristic. . . . . . . . . . . . . . . . . . . . . . . . . 85 3.28 LED Characteristic. . . . . . . . . . . . . . . . . . . . . . . . . 86 4.1 Mask Designed for Sherbrooke . . . . . . . . . . . . . . . . . . 94 4.2 Cartoon Epi-wafer Structure and a Dry Etched SRL Side View.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 4.3 SEM Side View After Dry Etching . . . . . . . . . . . . . . . . 96 4.4 Measured ARDE of Dry Etched InP Sample . . . . . . . . . . 97 4.5 Predicted Mode Distribution in Coupler Region for Wet Etched Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 4.6 Predicted Mode Distribution in Coupler Region for Dry Etched Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 x List of Abbreviations • Semiconductor Racetrack Laser (SRL) • Ridge Waveguide (RWG) • Photoresist (PR) • Infrastructure of Nanosctructures and Femtosciences (INRS) • Centre de recherche en nanofabrication et nanocaracterisation (CNR2) • Counter Clockwise (CCW) • Clockwise (CW) • Optical Injection Locking (OIL) • Vertical Cavity Surface Emitting Laser (VCSEL) • Finite Difference Time Domain (FDTD) • Quantum Well (QW) • Photoluminescence (PL) • Ultraviolet (UV) • Scanning Electron Microscope (SEM) xi List of Abbreviations • Plasma Enhanced Chemical Vapor Deposition (PECVD) • Electron Cyclotron Resonance (ECR) • Reactive Ion Etching (RIE) • Aspect Ration Dependent Etching (ARDE) xii Acknowledgements I would like to acknowledge the support, wisdom, leadership and friendship of my supervisor Dr. Lukas Chrostowski. In addition, I would like to ac- knowledge the support of all of my lab-mates, specially Wei Shi, Dr. Mark Greenberg, Dr. Nicolas Roger, Behnam Faraji and Miguel Ángel Guillén Torres for their daily discussions and brain storming sessions. I am very grateful to have had the chance to work along side them and to have had the opportunity to learn about optics in such a dynamic group. Furthermore, I would like to acknowledge CMC Microsystems Inc. (Jes- sica Zhang, Dan Deptuck) for supporting the Silicon-On-Insulator racetrack resonator and InP semiconductor ring laser fabrication, and the Centre de Recherche en Nanofabrication et Nanocaractérisation (CNR2) at the Uni- versité de Sherbrooke (Vincent Aimee, Marie-Josée Gour, Jean-Francois Be- dard, Etienne Grondin) for fabricating the semiconductor ring lasers via dry etching. I thank the BC Innovation Council for funding this research. Also, I would like to thank the committee (Nicolas A. F. Jaeger, Alireza Nojeh, S. Tang) for taking the time to read my thesis and provide me with their feedback. I am very thankful to my family for all their love. My dad Masoud Vafaei, mom Sara Montazemi, sister Mina S. Vafaei and brother Rod Vafaei xiii Acknowledgements have been continually supportive throughout my life and eduction. It is their unconditional love that has always been my biggest motivation for growth and learning. I would also like to thank all of my great friends, for being there for me and supporting me. xiv Chapter 1 Introduction 1.1 Motivation The continual advancement in communications has placed a growing demand on the innovation of optical communication systems to proceed to deliver integrated photonics and systems that are low cost, small sized and high bandwidth[4]. The region near 1550 nm emission is of interest, since optical fibers are available with losses as small as 0.15dB/km at this wavelength, making it a good candidate for long-distance optical communications [5]. Ac- tive ring resonators are desirable for optical monolithic integrated systems; they have been explored for applications such as optical amplifiers, optical flip flops (memory), optical digital processing (digital logic and switching), filters, and lasers. These active resonators can solve many drawbacks that passive resonators have been experiencing such as 1) lack of tune-ability 2) high insertion losses and 3) high optical switching powers [6]. One exciting area of interest is to use these active resonators as ring lasers for monolithic optically injection-locked laser systems, without the use of an isolator [4]. The modulation bandwidth (in GHz) of a directly modulated laser re- veals the communication systems maximum achievable data rate (Gb/s). The best modulation bandwidths achieved by edge emitting lasers are up 1 Chapter 1. Introduction to 40GHz [4]. Optical injection locking (OIL) can significantly improve the performance of directly modulated lasers. Using SRLs in OIL allows the removal of optical isolators by relying on the lasers’ output asymmetry. Moreover, SRL cavity do not have any mirrors, gratings or cleaved facets; the rings’ physical structure (i.e. width, height, radius) determines the op- tical modes. It is these inherent qualities of SRL that make it an exciting candidate for future optical interface technologies such as backplane and inter-chip communication [4]. 1.2 Operating Principle SRLs are made of two main components: 1) a straight bus waveguide and a 2) bent waveguide forming a closed loop. The region where the bus waveg- uide is close enough to the ring waveguide such that the fields in the two waveguides can interact is called the coupler region. The lasing action oc- curs in the bent waveguide resonator structure. Power is then coupled out of the resonator by the bus waveguide (power coupling factor affects the device performance and is a function of the coupler region design). To couple the power out of the bus waveguide, either a cleaved facet or a grating coupler is required. Laser systems require three major components: 1) a pump (optical or electrical) to provide carriers required for light emission 2) an optical feed- back to provide oscillation and 3) a gain region to provide amplification. When current is applied to the ring/racetrack structure, lasing recombi- nation takes place in the higher refractive index, active quantum well (QW) 2 Chapter 1. Introduction region where spontaneous emission, stimulated emission and absorption are all happening simultaneously. When the laser is biased below threshold, only spontaneous emission can be observed. Laser threshold is reached once the device is biased at high enough currents such that the gain equals the cavity losses; the laser turns on at threshold. Increasing the bias current above threshold results in higher output power as long as the device is not overheating [5]. The wave-guiding layers provide orthogonal optical confine- ment. For a shallow etch design (where the waveguide structure is defined above the QWs), lateral confinement is defined by the air/wave-guiding layer refractive index contrast. Laser oscillation in most laser systems such as in a Fabry-Perot Etalon laser, is achieved by the constructive interference of two counter propagating waves travelling between the two Fabry-Perot mirrors and forming a stand- ing wave in the cavity. In the ring laser, the light is confined to circulate around the ring. However, two counter propagating waves can travel around the ring structure. One of the two counter propagating waves can be sup- pressed by experiencing higher losses and thus a unidirectional output can be obtained from the ring laser. 1.3 Literature Review The preferred laser fabrication method used in industry for regular ridge waveguide (RWG) edge emitters and vertical Cavity surface emitting lasers (VCSELS) is wet etching or a combination of dry and wet etching. Wet etching techniques are preferred due to their lower processing costs, lower 3 Chapter 1. Introduction processing time and lower optical losses [7–10]. To the best of our knowledge, there has been no reports of SRLs fabricated via wet etching techniques. Figure 1.1: Geometry of the device showing the contact layout; IR, IW1, IW2 indicate the current biases applied to the ring and to the two output wave- guide contacts, respectively. The ring radius is 600 µm and the output waveguides are 800 µm long. Figure from Ref [11]. There are a few groups making SRLs in InP based materials, all with dry etching. In 2002, M. Sorel and P. J. R. Laybourn at the University of Glasgow and G. Giuliani and S. Donati at the University of Pavia published results on unidirectional bistability in large-diameter semiconductor ring lasers fabricated by CH4H2 reactive ion etching (RIE). Their design is shown in Figure 1.1 and had the following design parameters: ring radius of 600 µm, total length of the coupled waveguide of 1.6 mm, ring to output waveguide spacing of 1 µm, waveguide width of 2 µm [11]. They reported continuous wave measurements at room temperature exhibiting a threshold current of 125 mA. Figure 1.2 shows the optical spectrum of their device when biased 4 Chapter 1. Introduction Figure 1.2: Showing a switching extinction ratio larger than 30 dB. The resolution of the optical spectrum analyzer is 0.1 nm. Figure from Ref [11]. at 175 mA. L. Bach, J. P. Reithmaier, A. Forchel, J. L. Gentner, and L. Gold- stein fabricated single-mode lasers by coupled micro-square resonators in GaInAsP-InP epitaxy; their stucture is shown in Figure 1.3. The devices were fabricated using Ar-Cl2 electron cyclotron resonance (ECR)-RIE pro- cess and had width 4 µm, height 6 µm, radius 10-30 µm and threshold currents as low as 48 mA [12]. K. Amarnath, R. Grover, S. Kanakaraju and P. T. Ho from university of Maryland, published results on 0.8-1 µm wide, 20 µm radius, 5-10 µm long coupler length, deep etched, 1550 nm emitting microring lasers in 2005. Their design is shown in Figure 1.4. They were able to achieve threshold values as low as 12-20 mA [13]. 5 Chapter 1. Introduction Figure 1.3: Schematic illustration of the device geometry with deeply etched RWG section and a strongly MMI-coupled pair of ring resonators with dif- ferent diameters D1 and D2. The rings have square-like geometry with 45 C facets of width e. The resonators are also covered by BCB like the left side of the RWG. Figure from Ref [12]. 6 Chapter 1. Introduction Figure 1.4: Schematic cross section of microring amplifier shoing the layer structure. Inset shows an SEM picture of a 20 µm radius microring device. The bus-waveguide (not visible) is shown as a dashed line. Figure from Ref [13]. 7 Chapter 1. Introduction Figure 1.5: A cross sectional view in the coupling region of the waveguides etched down to the InAlAs layer and subsequently covered with 200 nm PECVD SiO2 (after removing the etching mask) for electrical passivation. The effect of small gaps on the etching: the coupler is not etched to the InAlAs layer. Figure from Ref [14] S. Furst, M. Sorel, A. Scire, G. Giuliani and S. Yu Pavia published details on their high selectivity (between the upper cladding InP and the core layer AlInAs) RIE process in 2006. They achieved very smooth surfaces and a complete etch depth in the gap between the waveguide couplers [14]. Their dry etching recipe allowed precise control over the coupling (gaps as small as 0.5 µm) as well as reduction in losses resulting in a lower threshold current of 34 mA for a 150 µm ring radius for a structure emitting at 1300nm. They predicted theoretical values of 60 mA and 20 mA for rings of radii 100 µm and 170 µm correspondingly for an epitaxy emitting at 1550 nm [14]. Figure 1.5 shows the etching challenge of achieving a full etch depth in the gap region which was then solved by over etching with a selective recipe. S. Park, S. S. Kim, L. Wang, and S. T. Ho published results on high- 8 Chapter 1. Introduction density plasma etching of InP based SRLs using Cl2-N2 based gas mixtures. The addition of N2 had an excellent effect on the InP reaction chemistry; it helped reduce the surface roughness and lateral etching. They fabricated SRLs with diameters 10 µm and 20 µm, gaps as low as 0.2 µm, waveguide width 0.8 µm, and threshold currents as low as 1.1 mA. In order to clear the 0.2 µm gap, the devices were over etched to achieve a 3 µm etch depth in the gap region and a 5 µm etch depth outside the gap region [15]. Figure 1.6 shows an SEM image of their fabricated SRL and Figure 1.7 shows the device spectrum. Figure 1.6: SEM image of a 0.8-µm-wide waveguide-coupled 20-µm-diameter ring resonator etched by ICP with 10/35/10 sccm Cl2 N2 Ar plasma at 200- W ICP power, 350-V dc bias, 2.3-mtorr pressure, 250 C temperature, and 400-nm/min etch rate. Figure from Ref [15]. A. W. Fang, R. Jones, H. Park, O. Cohen, O. Raday, M. J. Paniccia, and J. E. Bowers at the university of California reported a SRL integrated 9 Chapter 1. Introduction Figure 1.7: Spectra of scattered light from 20-µm-diameter microring laser and currentoptical output characteristics of a 20-µm-diameter microring laser (inset). Figure from Ref [15]. Figure 1.8: The hybrid silicon-evanescent device cross section structure. Figure from Ref [16]. 10 Chapter 1. Introduction with two photo-detectors on the hybrid AlGaInAs-silicon evanescent device platform. The SRL had a 200 µm radius, 0.5 µm gap, and 600 µm parallel coupler length. Running on continuous-wave at 1590 nm the laser had a threshold of 175mA, maximum total output power of 29 mW and a maxi- mum operating temperature of 60 C [16]. The device cross section is shown in Figure 1.8. Figure 1.9: SEM image of fabricated SRL. Figure from Ref [14]. In 2008, Sorel and group made dry etched SRLs with tuneable evanescent couplers. Their fabricated design is depicted in Figure 1.9. They reported a coupler tuning ratio from 90% down to 10-20% with current injection values as low as 10 mA [14]. 11 Chapter 1. Introduction 1.4 Objectives The main objective of this thesis is to design, fabricate, and characterize SRLs in InP based materials. A large part of the project was invested in the fabrication of the SRLs. We aimed to develop a fabrication process for the SRLs via wet etching techniques. 1.5 Novel Contributions The LED characteristics of the fabricated devices via wet etching were ob- served and successfully measured. The spectrum of the device was also measured with OSA and resonances were observed. The following were ac- complished in the fabrication process: 1. A selective multiple step wet etch technique was developed to achieve InP waveguides with smooth surfaces, in-plane isotropy (equal etch rates in the x and y cross section), orthogonal to in-plane etch rate ratio of 2. 2. A planarization technique was developed using polyimide without any masking or plasma exposure. 1.6 Thesis Organization SRL design was divided into 3 main sections: 1. Mode simulations were performed via Finite Difference Time Domain 12 Chapter 1. Introduction (FDTD) mode solver to determine the main propagating mode, con- finement factor, group index, crossover length and κ 2. Plots to summarize the dependency of output power and threshold current on cavity length, κ and propagation loss were obtained using Matlab. 3. Double bus passive resonators were designed in silicon on insulator (SOI), fabricated by Interuniversity Microelectronics Centre (IMEC) to further study the racetrack resonator, test the design process, and develop a parameter extraction method. The drop port output was successfully measured and the design parameters were extracted. The predicted design parameters matched the extracted parameters from the measurements, verifying the design method. After the design section the devices were fabricated at the Advanced Materi- als and Process Engineering Laboratory (AMPEL) Nanofabrication facility in UBC. The following steps were taken and will be discussed in more detail: 1. The required mask was designed via a script and purchased. A com- mercial InP epi-wafer was purchased to eliminate epi-wafer design chal- lenges. 2. A photolithography process was developed to successfully achieve a resolution better than 1 µm. 3. A selective multiple step wet etch technique was developed to achieve InP waveguides with smooth surfaces, in-plane isotropy (equal etch rates in the x and y cross section), orthogonal to in-plane etch rate ratio 13 Chapter 1. Introduction of 2. A new planarization technique was developed using polyimide without any masking or plasma exposure. 4. Ohmic metal contacts with reasonably low resistance ( roughly 1Ω ) were found for the n-type and p-type contact layers. 5. SRL and RWG samples were fabricated via wet etching 6. A suitable setup was made for the measurements of edge emitting lasers. Matlab codes were written (some based on previously available codes) to automate most of the measurement process. 7. Devices were manually cleaved and the LED characteristics and optical spectrum were successfully measured. 8. Dry etching of SRLs were started in collaboration with CNR2 at the Université de Sherbrooke. 14 Chapter 2 InP Racetrack Laser Design 2.1 Purpose The aim of this chapter is to study the laser theory and design the SRL physical parameters such as waveguide dimensions, coupling spacings and predict the threshold and output power. 2.2 Mode Calculation The design is based on a conventional InP laser epitaxy commercially avail- able from LandMark Optoelectronics Inc with details provided in Table 3.2. A simplified structure was used for FDTD mode simulations using Lumer- ical. Because the waveguiding layers are followed by relatively thicker InP layers, only the waveguiding layers (layers 4 to 7 in Table 3.2) were in- cluded in the simulated structure and the rest of the material was assumed to be that of bulk InP. The index values for the layers were calculated using the material information depicted in Table 2.1 and Equations 2.1 and 2.2, all provided by Crosslight (a commercial software used for laser/active optic simulations and design). For calculating the refractive index of the quantum well (QW) and barrier layers, initial guesses for the x and y compositions in 15 Chapter 2. InP Racetrack Laser Design In1−x−yAlyGaxAs, based on the following references [17] [18], were inputed into Equations 2.1 and 2.2 to achieve the material stress values of Table 3.2. The refractive index amplitude profile results are depicted in figure 2.1. The QWs and wave-guiding layers have higher refractive indices and lower band gaps compared to the InP n and p regions, thus, providing lateral optical confinement. nInAlGaAs = nInAs + x(nGaAs − nInAs) + y(nAlAs − nInAs) (2.1) aInAlGaAs = aInAs + x(aGaAs − aInAs) + y(aAlAs − aInAs) (2.2) Table 2.1: Refractive Index and Lattice Constant Data Material Refractive Index Lattice Constant InP 3.164 5.8688 GaAs 3.65 5.65325 InAs 3.892 5.6605 AlAs 2.92 6.0584 The mode simulations via the FDTD mode solver showed 4 different possible modes; however, due to their distribution the higher order modes will be subjected to higher losses (i.e. scattering and bending). Eventually the other modes will lose and only the first order mode will be significant to the lasing. The main mode distribution is shown in Figure 2.2 (a) and the 16 Chapter 2. InP Racetrack Laser Design Figure 2.1: The index amplitude profile for λ= 1550 nm simulated in FDTD. The QW refractive index = 3.5889 with In1−0.65−0.15Al0.15Ga0.65As. The barrier refractive index = 3.5130 with In1−0.12−0.36Al0.36Ga0.12As. Taking x = 0.03 and y=0.44 as provided in Table 3.2, the upper In0.53Al0.44Ga0.03As layer’s refractive index = 3.4254. second order mode distribution is shown in Figure 2.2 (b). The portion of the mode in the active region will experience gain. The field intensity data was imported into Matlab in order to solve for the mode confinement factor Γ, for rectangular waveguides of width W and height H and trapezoidal waveguide of top width WT and bottom width WB: • Rectangular waveguide: W = 2 µm, H= 2 µm ⇒ Γ = 8.9% • Rectangular waveguide: W = 3 µm, H= 2 µm ⇒ Γ = 8.8% • Trapezoidal waveguide: WT = 1 µm, WB = 3 µm, H = 2 µm ⇒ Γ = 8.82% The trapezoidal structure was considered to take into account the effect of wet etching during fabrication as it will be discussed later in the Fab- 17 Chapter 2. InP Racetrack Laser Design rication chapter. However, it seems that the 3 considered structures have very similar mode confinement factors. The gain experienced by the mode varies with 1/t, where t is the thickness of the active layer. The two main factors responsible for achieving low threshold in this semiconductor are: 1) confinement of injected carriers in the QW. and 2) confinement of optical mode in the QW. Figure 2.2: (a) First order mode distribution for λ = 1550 nm simulated in FDTD. It has an effective index of 3.276593. (b) Second order mode distribution for λ = 1550 nm simulated in FDTD. It has an effective index of 3.238451. This mode will not lase since it will experience higher losses (i.e. bending and scattering losses) relative to the first mode. 2.3 Racetrack Laser Coupler Design The purpose of this section is to design the coupler of the semiconductor racetrack laser (SRL) and derive the coupling relationship with respect to the coupler’s physical parameters such as waveguide width, height, length 18 Chapter 2. InP Racetrack Laser Design and spacing. The SRL output coupling loss (analogous to the mirror loss), is calcu- lated using FDTD mode solver in combination with the super-mode coupling theory. An exchange (coupling) of power between the modes of two waveguides occurs where the guided mode functions have a physical overlap region. Since the evanescent field has an exponential decay, there will always be some coupling between the guided modes, however at very large separation distances between the waveguides the suffers coupling significantly. The Power coupling between waveguides is dependent on many parameters such as the waveguide structure (waveguide dimensions, separation (s), and ma- terial), wavelength, and the distance (z) travelled by the electromagnetic wave along the waveguide [19]. The following assumptions were made for the coupling calculations: • The couplers were rectangular, identical and phased matched • The coupling was mainly due to the parallel region of the waveguides. Coupling in the bent region leading to the parallel section was ignored. • The couplers were lossless • The coupling was unidirectional • The waveguide was transverse electric (TE) guided • The bent effects on the guided mode were ignored 19 Chapter 2. InP Racetrack Laser Design Figure 2.3: The E intensity for the coupler with rectangular waveguides of width=2 µm, gap=0.5 µm, λ=1.55 µm. (a) Symmetric/Even supermode. (b) Anti-symmetric/Odd supermode. 20 Chapter 2. InP Racetrack Laser Design Using the Lumerical software integrated with FDTD mode solver pack- age, the effective indices neff1 and neff2 for the even and odd supermodes were extracted for waveguides of width W = 2 µm and W = 3 µm with varying coupler spacings. Figure 2.3 depicts an example of such simulation for the even and odd modes for a coupler with W = 2 µm and Gap = 0.5 µm. The rectangular waveguide structure used for the effective index simulations has the epi-wafer layer structure of a typical commercial InP epi-wafer which was purchased for our fabrication. This epi-wafer structure details are shown and discussed later in the InP Material Structure section of chapter 2. Let the two electromagnetic modes with power amplitudes Pa and Pb represent the power in the input coupler waveguide and the power in the coupled waveguide, respectively. The power coupling coefficient κ, equiva- lent to the square of the field coupling coefficient is defined by κ = Pb Pa (2.3) Once neff1 and neff2 are extracted, using the super-mode theory the crossover length (coupler length at which 100% coupling occurs) Lx, and κ were cal- culated as [20]: β1 = 2πneff1(λ) λ (2.4) β2 = 2πneff2(λ) λ (2.5) where β is the propagation constant in units of rad/m. The cross over length 21 Chapter 2. InP Racetrack Laser Design occurs at a distance travelled by the propagating fields where the two fields have a phase difference of π. Lx(λ, s) = λ 2(neff1(λ)− neff2(λ)) (2.6) Coupling per length = κ∗ = π Lx(λ, s) (2.7) Pb = Pa sin(κ ∗z)2 (2.8) The exponential dependancy of Lx on the waveguide separation is de- picted in Figure 2.4. 0.5 1 1.5 20 1 2 3 4 5 6 7 8 9 x 10 4 Lx vs. Gap, W = 3µm H = 2µm Gap (µm) L x (µm ) Figure 2.4: Lx as a function of separation between waveguides. This plot was obtained for a 3 µm wide rectangular structure. Figure 2.5 shows the exponential relationship between the coupled mode 22 Chapter 2. InP Racetrack Laser Design 0.5 1 1.5 210 −5 10−4 10−3 10−2 10−1 100 Power Coupling vs. Gap, H = 2 µm Gap (µm) κ L κ = 200 µm, W = 2 µm L κ = 250 µm, W = 2 µm L κ = 200 µm, W = 3 µm L κ = 250 µm, W = 3 µm Figure 2.5: The evolution of κ is shown as a function of coupler waveguide spacing for waveguides with coupler lengths 200 µm and 250 µm at 1550 nm wavelength. The case for a constant waveguide height of H = 2 µm with widths of W = 2 µm and W = 3 µm, are shown. The data was obtained using FDTD mode solver and the super mode theory. 23 Chapter 2. InP Racetrack Laser Design power and the waveguide spacing for waveguides of widths 2 µm and 3 µm, with a fixed height of 2 µm at 1550 nm wavelength. Tables 2.2 and 2.3 show the simulated coupling design parameters for the SRLs that were fabricated in this project. These coupling values are rather small, which result in low output power values in the µW ranges ( the detectors in the lab are very sensitive and can measure power down to nW ranges). To increase the power coupling (this is analogous to the increase in the effective index difference between neff1 and neff2), smaller gaps are required (>0.5 µm); however, this is challenging for the fabrication process. These lower coupling factors are advantageous for the initial devices since they result in a lower threshold. After the first devices are fabricated and their lasing characteristics are measured successfully, the mask could be re-designed and the fabrication process improved in order to optimize for power (while satisfying reasonable threshold conditions) by increasing the power coupling coefficient. Table 2.2: Power Coupling for Rectangular Waveguides W = 2 µm H = 2 µm Gap (µm) neff1 neff2 κa κb Lx (µm) 0.5 3.257165 3.255365 0.444 0.625 430 1 3.25653 3.256057 0.036 0.056 1638 1.5 3.256362 3.256239 0.00248 0.00388 6301 2 3.256485 3.256453 0.00017 0.00026 24218 24 Chapter 2. InP Racetrack Laser Design Table 2.3: Power Coupling for Rectangular Waveguides W = 3 µm H = 2 µm Gap (µm) neff1 neff2 κa κb Lx (µm) 0.5 3.264419 3.263803 0.061 0.094 1258 1 3.26417 3.264023 0.0036 0.0055 5272 1.5 3.264131 3.264095 0.00021 0.00033 21527 2 3.26417 3.264161 0.00001 0.00002 86111 As later will be discussed in the wet etching fabrication section of this thesis, the 3 µm wide waveguide devices have a better chance of being suc- cessfully fabricated. Moreover, the waveguides end up with a more trape- zoidal structure rather than a rectangular structure after the wet etching process. To take into account the wet etching effects, a trapezoidal struc- ture was made in the FDTD solver. The first order mode distribution for the trapezoidal structure is shown in Figure 2.6 and its corresponding refractive index profile and field distribution are depicted in Figure 2.7. The trapezoidal structure results in even slightly lower coupling values. The coupling could be increased via deposition of a thin dielectric film. This dielectric film deposition will also provide planarization as will be explained in the Fabrication chapter. Considering a roughly 200 nm thick layer of poly- imide with a refractive index of 1.7, trapezoidal structures were simulated in the FDTD mode solver to obtain the coupling information by consider- 25 Chapter 2. InP Racetrack Laser Design Figure 2.6: The first order mode distribution in the trapezoidal waveguide structure with H = 2 µm, Wbottom = 3 µm and Wtop = 1 µm is shown here. Figure 2.7: Aspects of the trapezoidal waveguide structure with H = 2 µm, Wbottom = 3 µm and Wtop = 1µm : (a) the refractive index profile; (b) the electric field profile for a mode traveling in the y-direction. 26 Chapter 2. InP Racetrack Laser Design ing the wet etching fabrication process. Figure 2.8 shows this waveguide structure for the case of a coupler with a 0.5 µm waveguide spacing. Figure 2.8: Trapezoidal waveguide structure with a uniform 200 nm thick polyimide layer. The refractive index distribution for a coupler with a 0.5 µm spacing is shown here. Table 2.4 gives a summary of the extracted coupling information and Table 2.5 gives a summary of the expected device performance for the largest possible cavity design with bend radius R = 300 µm when biased at a current of 100 mA ( this value is higher than the threshold current). Table 2.6 gives a summary of the expected device performance for the largest possible cavity design with bend radius R = 50 µm when biased at a current of 100 mA. In 27 Chapter 2. InP Racetrack Laser Design the Tables, the subscripts a and b refer to a coupler length Lκa = 200 µm and a coupler length Lκb = 250 µm, respectively. Table 2.4: Power Coupling for Trapezoidal Waveguide Wbottom = 3 µm, Wtop = 1 µm with 200 nm Thick Polyimide Gap (µm) neff1 neff2 κa κb Lx (µm) 0.5 3.264004 3.263390 0.061 0.094 1260 1 3.263763 3.263606 0.004 0.0063 4936 1.5 3.263763 3.263679 0.0012 0.0018 9226 2 3.263763 3.263753 0.00002 0.00003 77500 Table 2.5: Predicted Performance for Trapezoidal Waveguide Wbottom = 3, µmWtop = 1 µm, R = 300 µm and a 200 nm Thick Polyimide Planarization, Biased @ 100 mA Ith (mA) Pout µW Gap (µm) 1a 1b a b 0.5 89 85 365 430 1 81 84.5 25 30 1.5 80 84 7 9 2 80 84 1 2 One of the parameters for the SRL that can be measured is the free spectral range (FSR). Taking into consideration the effect of wavelength on the effective index the FSR is [19] : 28 Chapter 2. InP Racetrack Laser Design Table 2.6: Predicted Performance for Trapezoidal Waveguide Wbottom = 3 µm Wtop = 1µm R = 50µm and a 200nm Thick Polyimide Planarization, Biased @ 100mA Ith (mA) Pout µW Gap (µm) 1a 1b a b 0.5 26 30 4300 5500 1 25 29 300 400 1.5 25 29 140 100 2 25 29 15 19 FSR = ∆λ = λ2 ngd (2.9) where d is defined as the total cavity length of the racetrack resonator in Equation 2.10 [19], R is the radius of curvature and Lκ is the parallel length in the coupler region. d = 2πR+ 2Lκ (2.10) The parameter ng is called the group index and it takes into account the effective index dependence on the wavelength. For further information on FSR and ng please refer to the reference [19]. FDTD simulations show approximately a linear dependency of the effective index and the group index on wavelength. Figure 2.9 shows this result. 2.4 Racetrack Laser Resonator Design We are primarily interested in the LI characteristic of the SRL. First, the effect of coupling (κ) and resonator length (d) on the output power and 29 Chapter 2. InP Racetrack Laser Design 1.4 1.45 1.5 1.55 1.6 1.653.25 3.3 3.35 3.4 3.45 3.5 3.55 neff and ng vs. λ, W = 2µm H = 2µm lambda(µm) n g an d n eff ng neff Figure 2.9: Effective index and group index as a function of wavelength. The calculations for group index here take into account the waveguide dispersions and the bulk InP material dispersion. 30 Chapter 2. InP Racetrack Laser Design threshold current were investigated. Then, an estimated LI curve was pre- dicted for the designed SRLs. These analyses were all carried out using Matlab. Table 2.7 lists the parameter values used. Table 2.7: Simulation Parameters. Data taken from Lecture Notes of Dr.Lukas Chrostowski and Ref [1] Symbol Boiling Point (Description) Value or Range λ Wavelength 1550 nm B Gain slope (near transparency) 1.5×10−16cm2 ￿ Gain compression 2×10−7 ηi Quantum efficiency 0.9 Ntr Transparency carrier density 3.3×1018cm−3 τs Carrier lifetime 2 ns The equivalent for the mirror reflectivity R (as normally seen for a Fabry- Perot Etalon) for the racetrack resonator is (1-κ). Thus, the output coupling loss was calculated [5]: αm = ( 1 d )ln √ (1− κ) (2.11) The additional propagation losses were roughly chosen to be 5 cm−1. This is a pessimistic value based on taking a higher value compared to reported values such as 0.34 cm−1[21]. 31 Chapter 2. InP Racetrack Laser Design The threshold and output power for the laser can be expressed [5]. Ith = ng c(ΓBτp) +Ntr τsηi qV (2.12) Pout = ηi ακ α+ ακ hν q (I − Ith) (2.13) Where τp is the photon lifetime and α is the propagation loss here taken to be 5 cm−1. τp = ng c[α− αm] (2.14) Figure 2.10 shows the threshold current as a function of κ for fixed ring radii. τp increases for larger ring radii given a fixed κ, which in turn reduces the threshold Ith; however, at the same time the volume is increasing for larger ring radii, which will increase Ith. Overall the increase in volume overcomes the increase in τp resulting in a linear relationship between the ring radii and the threshold current Ith. Increasing κ for a fixed ring radius allows more photons to escape in the coupling region; in other words, it reduces the photon lifetime; for high κ, this reduction in τp is followed by an exponential increase in Ith. Increasing the input current results in an increase in the unsaturated gain and the intensity inside the laser; this increase in the bias effectively shifts the Pout vs. κ curve to higher values as depicted in figure 2.11. The peak of each curve corresponds to the optimum output coupling (κ). For higher bias points the output coupling needs to shift to slightly higher values 32 Chapter 2. InP Racetrack Laser Design Figure 2.10: Ith of the laser is plotted as a function of κ for an increasing ring radius and a coupler length of 200µm. Ith has an exponential dependency on κ and almost a linear dependency with the ring radius. Round trip losses are taken to be 5cm−1 0 0.2 0.4 0.6 0.8 10 10 20 30 40 50 κ Po ut( mW ) Pout vs K, Fixed Radius=150um I=80mA I=90mA I=100mA Loss = 5/cm Loss = 0.34/cm Figure 2.11: Pout of the laser is plotted as a function of κ at 3 different bias conditions for a device wtih a coupler length of 200 µm and radius of 150 µm. Round trip losses are taken to be 5cm−1 33 Chapter 2. InP Racetrack Laser Design in order to track the peak power. Moreover, Figure 2.11 shows the effect of increased losses from a low loss value of 0.34 cm−1 [21] to a higher value of 5 cm−1. Increase in losses could be due to poor fabrication, high bending losses and/or mode conversion losses amongst other possibilities. From these plots it is observed that for lower losses (excluding coupling losses), the peak of the power shifts to lower coupling values at the same time allowing a wide range of couplings to be used for delivering high values very close to the peak output power. Thus, reducing losses plays a major roll in the performance of the device. Figure 2.12 shows the relationship between the output power, coupling efficiency κ and ring radius R. The output power has a rollover behaviour with respect to κ for a fixed R. Increasing κ allows more photons to es- cape/couple out of the ring, resulting in a higher output power; however, too much coupling will prevent the intensity inside the ring from recover- ing, resulting in a net decrease of intensity inside the ring and thus a drop in the output power. The output power decreases for larger SRLs. Figure 2.13 shows that with higher losses, the output power is more sensitive to κ; moreover, higher coupling is required to output more power from the laser. Finally we can predict the LI curve using the laser rate Equations [5]. Figure 2.14 shows the LI curve for a SRL with the following specs: W = 3 µm, H = 2 µm, G = 1 µm, R = 150 µm, Lκ = 250 µm. 34 Chapter 2. InP Racetrack Laser Design Figure 2.12: Output power of the laser is plotted as a function of κ for a varying ring radius and a coupler length of 200 µm. Round trip losses are approximate to be 0.34 cm−1. Figure 2.13: Output power of the laser is plotted as a function of κ for a varying ring radius and a coupler length of 200 µm. A higher round trip loss is taken to be 5 cm−1 in the case of poor fabrication, bending losses, and mode conversion losses amongst others. 35 Chapter 2. InP Racetrack Laser Design 0 50 100 150 200 2500 0.1 0.2 0.3 0.4 0.5 0.6 Light−Output Current(mA) Po ut( mW ) Figure 2.14: Output power is plotted as a function of input current for a SRL with the following specs: W = 3 µm, H = 2 µm, G = 1 µm, R = 150 µm, Lκ = 250 µm. Round trip losses are approximate to be 5 cm−1. This device has a threshold of 49 mA and a slope efficiency of 2.5 µW/mA. 36 Chapter 2. InP Racetrack Laser Design 2.5 Double Bus Passive Resonator The purpose of this section is to further study the racetrack resonator struc- ture, test the design process and develop a parameter extraction method. Passive racetrack resonators were designed in silicon on insulator (SOI) technology and fabricated by Interuniversity Microelectronics Centre (IMEC). Racetrack resonators were chosen instead of ring resonators in order to en- hance coupling as well as allowing more control over the coupling.The waveg- uide height is fixed to 0.22 µm according to IMEC fabrication steps and the width of the waveguide is determined to be 0.4 - 0.6 µm to ensure single mode operation at TE polarization. The buried oxide box is 2 µm thick. Dosage sweeping during the photolithography step provided a variation in the de- signed waveguide widths and spacings. To facilitate light coupling, vertical fiber couplers are chosen, as there is no need for cleaving, and although aligning is still challenging they are relatively easier to align. Couplers used here should have a coupling efficiency of 30%, with a 3 dB bandwidth of 60 nm around 1550 nm as designed by IMEC. SOI racetrack resonators were designed with add and drop coupler waveguides [19]. An optical image of a fabricated design is shown in Figure 2.15. A 2D finite difference methode integrated with Matlab developed by the Photonics Research Lab at the University of Maryland is used to estimate neff1, neff2 from which the coupling was calculated using the super-mode approach. The equations [19] to predict the add and drop port responses are as follows: 37 Chapter 2. InP Racetrack Laser Design Figure 2.15: Depicted here is the optical image of a fabricated SOI device. Add : |σt1|2 = t 2 1 + t 2 2a 2 − 2t1t2acos(δ) 1 + t21t 2 2a 2 − 2t1t2acos(δ) (2.15) Drop : |σt2|2 = (1− |t1| 2)(1− |t2|2)a 1 + t21t 2 2a 2 − 2t1t2acos(δ) (2.16) ,where t1 and t2 are the straight through field coefficients, δ = ωd/c [rad] with c [m/sec] being the phase velocity and ω [rad/sec] the angular frequency [19]. High quality factor (Q) devices with varying parameters were designed. In the Double Bus design, the response at the drop port is similar to the response expected from the SRL output. Figure 2.17 shows the drop port measurement for a device with the following specifications: Lx = 391 µm, Ly = 75 µm, radius = 4 µm. Using both the Lumerical FDTD mode solver and the Maryland Matlab mode solver, coupling was calculated to compare the two methods: • κ from FDTD mode solver = 0.07 • κ from Matlab 2-D FDTD = 0.1 38 Chapter 2. InP Racetrack Laser Design A set up was originally built by Dr. Nicolas Rouger and Nadai Makan, which was then improved to include polarization maintaining lensed fibres, a wider fiber movement range and a 3D microscope movement. A picture of the setup is shown in Figure 2.16. Figure 2.16: SOI Measurement Setup [22]. κ, ng, and FSR were extracted by curve fitting as shown in Figure 2.17. The extracted values were: • κ from curve fitting = 0.14 • ng = 4.474622 • FSR = 0.56 nm The κ extracted from measurements is higher than both predictions. This is expected since in the coupling predictions it was assumed that all the power coupling is from the parallel length region of the couplers and the contributions from the bent waveguide were ignored. The measured FSR is 39 Chapter 2. InP Racetrack Laser Design matches 99% to the prediction by Equation 2.9. The device shows a Q of 47000 and a 22 dB extinction ratio. The loss values extracted from these measurements were used in publications by other students. This design approach has been used for several silicon-on-insulator projects, some of which have been published and presented at conferences [23] [24] [25]. 1548 1549 1550 1551 1552 1553−65 −60 −55 −50 −45 −40 −35 Wavelength [nm] Dr op Tr an sm iss ion [d Bm ] Drop Transmission vs. Wavelength Curve Fit Measurement Figure 2.17: Drop port response of a SOI racetrack resonator with Lx = 391 µm, Ly = 75 µm, R = 4 µm. The solid line is the measurement data and the dotted line is the curve fitted function from which the parameters κ, FSR and ng were extracted to be 0.14, 0.56 nm, and 4.474622, respectively. 40 Chapter 3 Fabrication Via Wet Etching 3.1 Introduction This chapter presents the experiments and results of the investigation of fabrication techniques via wet etching utilized for SRLs and ridge waveguide RWG edge emitting lasers in InP based material systems. The objective is to develop a new fabrication process via the wet etching approach for SRLs. To the best of our knowledge, there has been no publication on SRLs made via wet etching in InP based materials. The preferred laser fabrication method used in industry for regular RWG edge emitters and VCSELS is wet etching or a combination of dry and wet etching. Wet etching techniques are preferred due to their lower process- ing costs, lower processing time and lower optical losses [7–10]. In order to achieve completely anisotropic sidewall profiles and zero in-plane etching, plasma and dry etching techniques have been mainly utilized for the fabri- cation of SRLs; these processes are usually very expensive, more complex, can leave relatively rough surfaces and may even damage the sample. This project investigates the fabrication steps for InP semiconductor ring laser using simpler, less costly, wet etching techniques with the aim of achieving smoother sidewalls/surfaces in order to reduce the scattering from the rough 41 Chapter 3. Fabrication Via Wet Etching surface typically present from plasma etched ring lasers. This would improve the efficiency of the laser and reduce the threshold. The wet etch process requires some optimization of the process conditions in order to achieve: isotropic etch rate in the x and y cross section, flat surfaces, maximize the etch rate ratio of orthogonal etching to in-plane etching (for higher sidewall verticality) and reduced under cuts. There are 3 major sections in this chapter: InP Material Structure, InP Edge Emitter Racetrack Resonator Laser, InP Edge Emitter Ridge Waveg- uide Laser. In the first section the InP epi-wafer structure details are in- troduced; the same epi-wafer was used for the fabrication investigation of both the SRLs and RWGLs. The analysis of the SRLs fabricated via wet etching revealed a set of possible areas for improvements in the fabrication and design. To pin point the main challenge in the SRL fabrication a sim- pler design was made to fabricate RWG edge emitting lasers via wet etching. Finally we started processing dry etched SRLs in collaboration with the fab- rication facility at Sherbrooke. The mask designs and fabrication processes for each of the devices are described in this chapter. Figure 3.1 show a flow diagram for the fabrication steps required. Table 3.1 shows an overview of the material and equipments used for the fabrication processes. 42 Chapter 3. Fabrication Via Wet Etching Figure 3.1: Depicts the fabrication steps required to process an SRL. Table 3.1: Overview of The Fabrication Steps Process Materials Equipments Settings/Action Solvent Clean- ing Sample Hotplate Acetone Microscope Isopropanol Waste Beaker DI Water, Com- pressed N2 gas Continued on next page 43 Chapter 3. Fabrication Via Wet Etching Table 3.1 – continued from previous page Process Materials Equipments Settings/Action Wet Etch Sample, H3PO4:HBr:H2O, DI Water, Compressed N2 gas 2 Beakers tweezers Etch ∼150nm of u-InP Spinning Sample AZ P4110 Spinner 1min @ 3000rpm Soft Bake Sample Hotplate 7min @ 100 C Optical Pat- terning Sample 320nm Optics 0 Gap contact mode Mask Karl Suss Mask Aligner 1 min exposure Pattern Devel- oping Sample, Beakers AZ400k Developer 1 : 4 90 sec DI Water, Microscope Wet Etch Sample, Compressed N2 gas, DI Water, H2SO4:H2O2:H2O 4 Beakers Etch InP and In- GaAs/InGaAsP for a total etch depth of ∼2µm Continued on next page 44 Chapter 3. Fabrication Via Wet Etching Table 3.1 – continued from previous page Process Materials Equipments Settings/Action H3PO4:HBr:H2O Dektak Pro- filometer 3.2 InP Material Structure The basic layered structure is grown epitaxially on a crystalline InP sub- strate by metal organic vapour phase epitaxy (MOCVD) at LandMark Op- toelectronics Corp. in Taninan, Taiwan. The orientation of the substrate used is (100) +/- 0.1 deg. Table 3.2 shows the epi-wafer structure in detail. This design is a commercial product used for edge emitting lasers and we use this proven design as a starting point for our fabrication. The substrate is n+ Sulphur doped InP with layers grown on top as follows: a lower cladding including a 500 nm n-type InP layer and a 185 nm transition quaternary layer; a 100 nm waveguide core; an upper cladding including a 75 nm InAlGaAs transition quaternary layer, a 50 nm InAlAs etch stop layer, a total of 72 nm transition quaternary layers, a 1.8 µm P- type InP layer, 50 nm of InGaAsP transition quaternary layers and finally a p-type InGaAs contact layer. For 1550 nm emission: Eg = 1.24eV 1.55µm = 0.8eV (3.1) 45 Chapter 3. Fabrication Via Wet Etching Therefore, the bandgap of the QW is 0.8 eV. The bandgap energy is grad- ually increased as we move away from the QW with changes in the x and y composition ratios. Using InP as substrate a range of lattice-matched qua- ternaries extending from InP to the InGaAs ternary line are accommodated. Here the quaternary is specified by an x and y value, i.e., In1−xGaxAsyP1−y. Choosing x equal 0.46 y results in approximate lattice matching to InP [2]. This structure includes 6 InAlGaAs QW layers each 6 nm thick, alter- nating with 7 barriers each 9 nm thick (In mode calculations the barriers are not included), sandwiched between two regions of varying layers (for simplicity of mode calculations here considered to be just InP). LandMark provided some test data of this epi-wafer including a photo- luminescence (PL) measurement reporting a peak wavelength of 1512.5±5 nm. The PL was measured again at UBC by Ph.D student Wei Shi and the data are shown in Figure 3.2. Table 3.2: R2 - 1550 nm FP-LD Epi-wafer Structure[2] # Name Value Unit Thickness Accuracy 0 InP Substrate S-Doped 2− 8× 1018 cm−3 1 N-InP Buffer Layer 0.5 µm ±10% (Concentration) 1× 1018 cm−3 ±20% 2 N − In0.53AlxGa0.47−xAs 0.01 µm ±10% Continued on next page 46 Chapter 3. Fabrication Via Wet Etching Table 3.2 – continued from previous page # Name Value Unit Thickness Accuracy (x:0.31→0.44) 1× 1018 cm−3 ±20% (Concentration) 3 N − In0.52Al0.48As 0.1 µm ±10% (Concentration) 1× 1018 cm−3 ±20% 4 U − In0.53AlxGa0.47−xAs 0.075 µm ±10% (x:0.44→0.3) 6× InAlGaAs QW 5.5 ∼ 6/ nm (+1.0 ∼ +1.15% CS)/ 5 7× InAlGaAs Barrier 9 nm ±10% (−0.45 ∼ −0.55% TS) (1512.5) nm ±5% (λPL) 6 U − In0.53AlxGa0.47−xAs 0.075 µm ±10% (x:0.3→0.44) 7 U − In0.52Al0.48As 0.05 µm ±10% 8 P-InP Layer 0.05 µm ±10% (Concentration) 5 ∼ 8× 1017 cm−3 9 P-InGaAsP Layer 0.022 µm ±10% (Concentration) 5 ∼ 8× 1017 cm−3 10 P-InP Layer 1.8 µm ±10% (Concentration) 1→ 1.5× 1018 cm−3 Continued on next page 47 Chapter 3. Fabrication Via Wet Etching Table 3.2 – continued from previous page # Name Value Unit Thickness Accuracy 11 P-InGaAsP Layer 0.025 µm ±10% (Concentration) > 3× 1018 cm−3 12 P-InGaAsP Layer 0.025 µm ±10% (Concentration) > 3× 1018 cm−3 13 P-InGaAs Layer 0.15 µm ±10% (Concentration) > 1× 1019 cm−3 14 U-InP Layer 0.15 µm ±10% Figure 3.2: The epi-wafer PL data at room temperature is shown here. Courtesy of Wei Shi. 48 Chapter 3. Fabrication Via Wet Etching 3.3 InP Edge Emitter Racetrack Resonator Laser 3.3.1 Mask Overview Once a range of possible SRL design parameters was determined, the re- quired mask was designed using a Matlab program integrated with Clewin (a drawing software). Mask was done using a script to facilitate varying the parameters. The mask was purchased from the Infrastructure of Nanosctructures and Femtosciences (INRS) Centre in Quebec. At INRS two types of direct writing techniques were available: a laser writer, an e-beam writer. The e-beam mask was chosen because it provides the highest possible resolution and that is necessary to obtain the 0.5 µm gaps in our design. The mask patterns were defined in chromium on a 5” × 5” glass plate with a thickness of 0.09”. The mask was later on cut down into a 4” × 4” plate in-order to fit the Karl Suss mask aligner used at the UBC AMPEL Nanofabrication facility. Figure 3.3 shows the mask design. The fabrication required two photolithography steps thus two mask lay- ers: Layer 1 to form the waveguide ridge (both the coupler waveguide and the racetrack waveguide), Layer 2 to define the contact opening before met- allization. For Layer 1, two designs were printed on the mask. Design of Layer 1(a) as shown in Figure 3.4 is simple and appropriate for wet etching techniques; however, the exposed area in this design is very large for dry 49 Chapter 3. Fabrication Via Wet Etching Figure 3.3: The SRL mask layout. This is the final version of the mask submitted. The metal layer is inverted. 50 Chapter 3. Fabrication Via Wet Etching Figure 3.4: Layer 1(a) shows the etch mask layout used for wet etching. Figure 3.5: Layer 1(b) shows the etch mask layout. A trench design is applied for the case of dry etching. Figure 3.6: Layer 2 shows the top contact mask layout. This is before inversion, thus the coloured region would be the contact area. 51 Chapter 3. Fabrication Via Wet Etching etching techniques and causes reduction in the uniformity throughout the wafer. In order to reduce the exposed area and increase the reliability of uniformity throughout the processed sample specially when dry etching, a second design Layer 1(b) using trenches was drawn as shown in Figure 3.5. Figure 3.6 shows the layer 2 design before inversion. SRL Design Variation Due to the approximations in modelling (for example ignoring coupling due to the bend region, estimating the overall losses, etc.), potential unexpected imperfections and/or limitations in the fabrication process, and also for the purpose of device characterization, the SRLs were designed with varying pa- rameters. The coupling is a function of the effective index, coupling spacing and coupler length. In order to study the coupling behaviour and its effects on quality factor, output power, resonance frequency, bandwidth and sen- sitivity, different coupling lengths each with several varying coupling gaps were designed. Over 80 different designs were included on the mask with a combination of the following parameters: • Bent Region Radius: 150 µm, 160 µm, 180 µm, 200 µm, 250 µm, 300 µm • Coupling Gap: 0.5 µm,1 µm, 1.5 µm, 2 µm, 2.5 µm • Coupler Lengths: 200 µm, 250 µm • Waveguide Width: 2 µm, 3 µm 52 Chapter 3. Fabrication Via Wet Etching Every device included a label to identify its parameters. The labels have a format of LxxxRxxWxGxxx. As an example a label reading L200R150W2G1, identifies a device with coupler length 200 µm, bent radius of 150 µm, waveg- uide width of 2 µm and coupling gap of 1 µm. Alignment Marks There are two important alignment steps required with the current mask design. First, when using Layer1 of the mask, the mask patterns should be aligned to the edge of the sample in such a way so that the waveguides run parallel to an edge of the sample. This alignment is necessary for the cleaving step required to achieve the output facet of the laser. If the patterns are misaligned with the edge of the sample, when the sample is cleaved along its crystallographic orientation, many devices can be damaged and thus the fabrication yield is reduced. The second critical alignment is between Layer 1 and Layer 2 of the mask. This alignment has a large tolerance, roughly ±10 µm since the metal pattern is very large. Large cross like alignment markers were placed throughout the mask to assist with coarse alignment. For finer alignment, vernier scales were designed with 6 µm to 6.2 µm spacings to improve the alignment accuracy up to ±2 µm. 3.3.2 Sample Preparation InP is very brittle; compared to Si it is a few times easier to break or shatter InP. The cleaving process was done manually. A tip of a diamond cutter was used to apply pressure to a point on the InP wafer. Once enough pressure was applied, the sample broke along its crystallographic orientation. The 2” 53 Chapter 3. Fabrication Via Wet Etching epi-wafer was cleaved into small pieces of roughly 1 cm by 1 cm. 3.3.3 Photolithography Cleaning InP based fabrication is costly; high yields are demanded for making prod- ucts more affordable. Cleanliness is a critical factor in achieving reliable, reproducible results with a high yield in all micro-fabrication processes in- cluding semiconductor laser fabrication. The cleaning process refers to the removal of any undesirable material from the sample surface prior to any further processing. Examples of such undesirable materials include chemical residues from previous steps, debris from cleaving or particles due to poor environmental control. Insufficient cleanliness leads to degradation of the process quality or sample integrity (because it can crack during alignment). Photoresist (PR) spin-coating and patterning, wet or dry etching, polymer application and oxidation for passivation and planarization, and metal ad- hesion are all affected by the initial cleanliness conditions of the sample surface. The environment cleanliness is defined by the size and number of particles present in the fabrication environment. The UBC yellow room fa- cility is a class 1000, which means the maximum concentration of airborne particles larger than 0.5 µm in diameter is controlled to be less than 1000 particles/ft3. Moreover, the cleanroom atmosphere is both temperature and humidity controlled [26]. Organic solvents were used to remove any organic material such as PR and oil, grease or wax residues from the InP surface. Acetone was used 54 Chapter 3. Fabrication Via Wet Etching to rinse the sample surface; unfortunately it may also leave some residue behind. Isopropanol or methanol was used to rinse away any residues left behind after the acetone rinse, then the sample was rinsed by de-ionized water and dried with a compressed nitrogen gas. Some properties of the solvents used are shown in Table 3.3. Table 3.3: Solvent and Properties (Data from [3]) Solvent Boiling Point (◦C) Flash Point(◦C) Water Solubility Safety Acetone 56.3 -16 100% Flammable Propanol-2 82.3 22 100% Flammable (Isopropyl alcohol) Methanol 64.7 12 100% Flammable The solvents used were disposed of in the appropriate non-halogenated waste containers [27]. After rinsing the sample surface and drying with the nitrogen gun, the sample was heated at 100◦C on a hotplate for 2 minutes to ensure that the surface was perfectly dried before the PR spinning step. After drying the sample on the hotplate, the sample was left for 2 minutes to cool before PR spinning. The sample surface was then observed under the microscope to ensure that it was sufficiently clean. The solvent cleaning procedure was 55 Chapter 3. Fabrication Via Wet Etching repeated as many times as necessary until the surface was acceptably clean. Photoresist Spin-coating The next step in the process of photolithography involves PR, a photosen- sitive organic material that is applied to sample surfaces a thin film. It is selectively subjected to UV light that creates exposed areas. Photolithog- raphy in short is a patterning technique used in semiconductor processing. There are two general types of PR, negative and positive. When positive resist is exposed to UV light, it gains enough energy to change its chemi- cal structure. This chemical change makes the resist sensitive to chemical etchants and more soluble in the developer. Therefore, any areas where the positive resist is exposed to UV light are being washed away in the devel- oper, exposing the underlying InP layer. Negative PR works in the opposite way. Exposure to UV polymerizes the PR and therefore the exposed areas become resistant to chemical etchants and do not dissolve easily in the de- veloper. In the negative resist, it is the unexposed portions that are removed during the developing process. In our fabrication a positive PR was used. A thin layer of roughly 1.4 µm thick PR was applied uniformly on the die surface via spin coating. The final thickness of the PR film is determined by the material of the stage surface, the viscosity of the PR and the speed at which the wafer spins. Given the spinning speed, spinning longer than a certain amount of time will not change the final thickness of the PR. The thickness was determined by a profilometer. After making sure that the sample was completely clean, dry and cool, the wafer was placed onto a spinner. A pipette was used to dispense 2-3 PR 56 Chapter 3. Fabrication Via Wet Etching drops onto the sample which covered the entire surface. The PR was spun at 3000 rpm for 1 min. A successful spin coat should result in a uniform layer of PR. If the layer was non-uniform or there were dots, lines or areas that were not covered by PR, the process was started from the beginning. Soft Bake The amount of water present in the PR film determines how soluble the UV exposed area becomes in the developing step. Optical exposure causes a chemical reaction in the PR film to form a ketone. If water is present, the ketone will react to form a product (indene carboxylic acid)which is highly soluble in the developer. In the absence of water, the degraded components form cross-linkages with other molecules and solubility decreases. Soft bak- ing removes some of the water and organics in the PR film which is critical to the PR exposure and developing steps. Some baking is required to harden the film, reduces its thickness and improves its adhesion. If the PR is over baked, it loses its sensitivity to light and thus degrades its solubility in the basic developer. However, if it is under-baked and there is still a significant amount of organics remained, PR resistance in the developer would degrade resulting in overdevelopment [28] and the sample’s PR can stick to the mask during the subsequent contact lithography step. Recipe for soft baking is given in the steps below[29]: • Adjust the hotplate temperature to 100 C • Place sample on the hotplate for 7 mins 57 Chapter 3. Fabrication Via Wet Etching • Remove sample from hotplate and keep away from UV light sources • Allow 3 minutes for the surface to cool before exposure Optical Patterning The AZ P4110 PR absorption spectrum is matched to the emission spectrum of mercury. When the PR is exposed to Hg emission, it gains energy which causes a chemical reaction necessary for the developing process. Using the appropriate photon energy and H2O concentration in the resist after soft bake, the PR development rate increases. For the positive PR the trans- parent patterns on the mask are the parts that will be exposed and later removed in the developing process [28]. The Karl Suss mask aligner used at the UBC fabrication facility is a manual system with 320 nm optics, which focuses UV light produced by a mercury lamp via a system of lenses onto the mask. This machine is limited to a maximum wafer diameter of 3” and a maximum mask plate size of 4” by 4”. The distance between the mask and the wafer, the intensity of the UV light, and the exposure time could all be adjusted to achieve the desired results [30]. The lateral resolution is limited by the emission wavelength and the gap between the mask and the sample. The lateral resolution d of an image transferred from the mask pattern to the PR has a relationship with the wavelength λ and the distance between the mask and PR g as shown in Equation 3.2 [28]. Figure 3.7 shows the relationship of Equation 3.2. 58 Chapter 3. Fabrication Via Wet Etching d = (λg) 1 2 (3.2) Figure 3.7: Theoretical maximum resolution that can be achieved as a func- tion of the distance between the mask and the PR surface is shown here.[28] Since the mask has a high resolution demand on the lithography (small- est coupling gap is 0.5 µm), it is very critical to reduce the distance between the mask and the PR surface sufficiently (g < 0.5 µm). Profilometer data revealed that the PR height close to the edges of the sample could be as much as roughly 1 µm to 1.5 µm higher than the uniform PR height ev- erywhere else on the sample. This height difference becomes an issue for patterns requiring lateral resolutions below 2 µm. To resolve this issue, be- fore transferring the mask patterns onto the PR surface a pre-exposure step was applied to remove the PR from the edges and corners of the sample. The sample was exposed in contact mode (the gap between the wafer and the mask was manually closed), for 1min. 59 Chapter 3. Fabrication Via Wet Etching Pattern Developing A low concentration developer solution was prepared with 1 part AZ400k developer and 4 parts DI water[29]. Developing refers to the process of agitating the sample in the developer solution for the appropriate amount of time needed for the exposed PR to dissolve in the solution. Developing for 1 min was found sufficient. Thanks to the corner and edge bead removal step as shown in Figure 3.8 a resolution of 1um was achieved. Figure 3.8: Shows a 1µm resolution achieved by optical lithography using the manual Karl Suss mask aligner and applying the edge beading compensation technique. Yeild Ananlysis With no unexpected issues in the above steps, a 100 percent yield was achieved on devices with 1µm resolution and higher. Since a near perfect 60 Chapter 3. Fabrication Via Wet Etching lithography step is possible under expected conditions, if any defects were observed at this point in the fabrication the sample was cleaned and the process was repeated once the occurred defects were investigated. Here are some factors that may contribute to defects and lower the yield: Figure 3.9: Defect revealed after developing due to rough edges present on the mask. Some of these mask damages were present originally due to the mask fabrication quality and some damages developed over time through usage. Particles in the PR due to contaminated substrate(insufficient cleaning or inherent defects), insufficient cleanroom conditions, contaminated PR con- tainer or pipette (PR dries very quickly and even reusing the same pipette could introduce some dry chunks), expired PR, mask contamination(by par- ticles or previous PR), bubbles introduced when dispensing the PR droplets onto the surface, edge bead. These problems will cause a g > 0. Also over or under baking and exposure would effect the developing results. Figures 3.10, 3.9 and 3.11 show optical images of some different defects. 61 Chapter 3. Fabrication Via Wet Etching Figure 3.10: Defect revealed after developing due to presence of impurities in the PR. This problem was resolved by using fresh PR and also ensuring that the sample surface was cooled before PR spinning. Figure 3.11: PR residue left over after cleaning including the following ap- proaches: over-flooding the PR with the UV and developing, two-solvent process at room temperature, two-solvent process at close to boiling tem- perature. This is most likely due to expired PR and/or over-baking the PR. A PR stripper solution can usually remove the residues. 62 Chapter 3. Fabrication Via Wet Etching 3.3.4 Wet Etching Wet etching refers to the process of removing layers from the epi-wafer sample through chemical reactions that occur at the surface of the material when exposed to liquid etchants. Once the sample has been patterned during the photolithography process, some areas are protected by PR from the etchant while the unprotected areas are exposed to the etchant and could be removed. Wet etching is used here to form mesa structures for device isolation, to define the waveguide structures, and to remove the protective u-InP layer. The wet etch requirements for the SRL fabrication are: • High selectivity: An acceptable selectivity allows the etchant to per- fectly remove the desired layer without damaging the underlying layers or completely removing the mask layer. In other words, the etch rate for the layer to be removed must be higher than that of the mask and the underlying layers. Selective etchants are needed to control the etch depth. • In-plane isotropy: Perfect isotropic in-plane etching results in equal erosion of the material in the x and y cross-sections. This is important for achieving the same waveguide dimensions everywhere on the SRL including the bend regions. Typical InP wet etchants are anisotropic. • Orthogonal to in-plane anisotropy: Ideally, a perfectly straight side- wall profile is desired with no undercuts, however, that is impossible with wet etching techniques. It is desired to maximize the etch rate 63 Chapter 3. Fabrication Via Wet Etching ratio of orthogonal to in-plane etching. Suitable Etchants The approach was to adjust the etchant concentrations in already reported InP etchant recipes to find the most suitable etchant recipe. A suitable recipe should achieve: • The highest possible in-plane isotropy • Flat bottom • Smooth surfaces • Maximize the sidewall verticality • Minimize the amount of mask undercut There are various wet etching recipes previously reported by other re- searchers including defect revealing etchants, material selective etchants or etchants for applications that were not highly sensitive to in-plane isotropy as reported in the review paper of Ref. [31] and Ref. [32].These recipes were tested on dummy InP wafer with the help of Dr. Mark Greenberg in order to determine the most suitable recipe for our application. Table 3.4 shows a summary of a number of approaches taken to achieve the wet etching requirement; the table lists the different etchant systems that were tried and the observations made during fabrication testing. 64 Chapter 3. Fabrication Via Wet Etching Table 3.4: Etchants Tried Etchant System Etched Material Selective Observations 1 H2SO4/ H2O2/ H2O InP No Large undercut 2 H3PO4/HCl InP No in-plane anisotropy 3 HCl/ CH3COOH/H2O2 InP No In-plane isotropy achieved but very slanted sidewall profile 4 HBr/H2O2/H2O/HCl InP No in-plane isotropy achieved but non-flat bottom profile 5 H3PO4/HBr/ H2 InP Yes, <4 in-plane isotropy achieved, 6 H2SO4/H2O2/H2O InGaAsP and In- GaAs Yes, <4 smooth surfaces Figures 3.12 to 3.15 show optical images of test samples etched with the etchants of approach 1 to 4 listed in Table 3.4. It was initially decided that Approach 4 listed in Table 3.4 , the HBr/H2O2/H2O/HCl system, was the most suitable for SRL fabrication. However, after the first batch of devices were fabricated, no light was observed from the cavity. Using the Dektak profilometer, the etching profile revealed that there is a 65 Chapter 3. Fabrication Via Wet Etching Figure 3.12: 3H2SO4 : H2O2 : H2O system is highly corrosive (etch rate mainly proportional to Sulphuric acid concentration) and results in large amount of undercut. Not suitable for our masking patterns. Figure 3.13: H3PO4/HCl systems are not suitable for our masking patterns. (a) High concentrations of phosphoric acid seems to result in holes and nonplanar bottoms. (b)Reducing the phosphoric acid concentration seems to result in smoother surfaces. 66 Chapter 3. Fabrication Via Wet Etching Figure 3.14: Etching results with HCl/ CH3COOH/H2O2 systems (a)High HCl and acetic acid concentrations enhance the in-plane anisotropic etch- ing behaviour. (b)Reducing the HCl concentration improved the in plane isotropy. (c) Reducing the acetic acid concentration to the HCl and hy- drogen peroxide concentration by a factor of 2 resulted in perfect in-plane isotropy; however, the sidewall profile was very slanted. Figure 3.15: Etching results with HBr/H2O2/H2O/HCl systems. (a) Reveals very non-flat bottom and sidewall profiles with high undercuts.(b)Diluting the recipe reduced the etch rate and the undercut behaviour; however, the bottom and sidewall profile is still very non flat. (c)The HBr concentration was reduced to help improve the flatness issue and then the recipe was diluted 12 times. This recipe was still inappropriate for our masking patterns due to the bottom and sidewall non-flatness. 67 Chapter 3. Fabrication Via Wet Etching higher etch rate near the waveguide edges, as shown in Figure 3.16. This non uniform etch depth is due to a higher concentration of Br2 molecules near the edges compared to far away from the edges. Br2 molecules diffuse to the mask corners since they can’t react with the substrate in the masked regions; moreover, due to their low mobility they have a higher probability of consumption near the mask corner rather than far away from the corners as shown in Figure 3.17 [33]. The HBr/H2O2/H2O/HCl systems are com- monly used for spherical lens formation and are not suitable for the SRL fabrication. Figure 3.16: Etching profile with HBr/H2O2/H2O/HCl system shows the Br2-based diffusion limited etching effect observed from the higher etch depth near the mask boundary relative to far away from it. 68 Chapter 3. Fabrication Via Wet Etching Figure 3.17: A schematic illustration of the semiconductor microlens fabri- cation process using Br2-based diffusion-limited etching [33]. The most suitable selective (higher etch rate of InP to InGaAs and In- GaAsP) recipe for wet etching InP was observed to be approach 5 of Table 3.4, H3PO4 : HBr : H2O solution. It resulted in a vertical etch rate that was twice as fast as the in-plane etch rate. Similar to approach 4, approach 5 system also uses HBr; however, it does not have the non-uniform etch depth problem of approach 4 since there are no H2O2 molecules oxidizing HBr molecules and thus no Br2 molecules are introduced into the solution. The epi-wafer includes some InP layers alternating with InGaAs/InGaAsP layers. For removing the InGaAs and InGaAsP layers selectively and stop- ping at the InP layer, approach 6 of table 3.4 was used [34] [35]. This ap- proach resulted in very smooth surfaces as observed with optical microscopy and profilometer measurements. The wet etching process for the epi-wafer of Table 3.2 required switching between approach 5 and approach 6 solutions for a total of 5 steps in order 69 Chapter 3. Fabrication Via Wet Etching to selectively remove the 3 InP layers (including the protective layer on top) and the 2 InGaAs/InGaAsP layers and stop at the InAlGaAs layer above the QWs. This switching makes use of etch stop layers to allow more control over the etch depth. Etch depth measurements with the Dektak profilometer after each etch step was carried out to further monitor the etching process. Table 3.5 shows an example of the data collected to monitor the wet etching process. As seen in table 3.5 the typical total etch time was roughly 5 minutes + 40 seconds for the desired etch depth of ∼ 2 µm. However, it was observed that for every 2 µm etched in the vertical direction, there was a 1 µm undercut from each side of the mask. This places a limit on the minimum width of our structures for the desired etch depth, as well as resulting in relatively low steep sidewall profiles (ideally straight sidewalls are desired: slope of infinity). The side profile in the SEM image of a wet etched InP sample in Figure 3.18 shows this 1:2 ratio. Figure 3.19 shows the device condition as the wet etching time is increased for both 2 µm wide waveguides and 3 µm wide waveguides. Table 3.5: Wet Etch Monitor Table for Sample : R2 epi- wafer, ID: D04 Date: Jul19/2010 Dektak Setting [µm/sample] Etch Time [min] PR+Metal Height [µm] Etchant Step’s Etch Rate [µm/min] 0.128 0 1.582 Continued on next page 70 Chapter 3. Fabrication Via Wet Etching Table 3.5 – continued from previous page Dektak Setting [µm/sample] Etch Time PR+Metal Height [µm] Etchant Step’s Etch Rate [µm/min] 0.128 2.5 1.763 H2SO4:H2O2:H2O 0.076 0.128 2.5 3.502 H3PO4:HBr:H2O 0.698 0.128 0.5 3.53 H2SO4:H2O2:H2O 0.03 0.128 0.117 3.593 H3PO4:HBr:H2O 0.53 Overall Etch Rate = Total Etch DepthTotal Etch Time = 2.01 5.62 = 0.36µm/min. H2SO4:H2O2:H2O and H3PO4:HBr:H2O Etch System Results With no unexpected issues in the above etching steps, roughly a 30% yield was achieved; devices with a 3 µm waveguide width and coupling gaps of 1 µm and higher survive. Here are some factors that may contribute to defects and lower the yield further: • The PR lifting from the the 2 µm wide devices introduces PR strips that are free to move around the sample and sit anywhere on the surface for a random amount of time during the etching process. This causes additional undesired masking effects and could be reduced by avoiding drying the sample with a nitrogen gun (or via applying very low pressure) in between etching steps. 71 Chapter 3. Fabrication Via Wet Etching Figure 3.18: SEM image showing the side wall profile of InP after wet etch- ing. 72 Chapter 3. Fabrication Via Wet Etching Figure 3.19: Shows the PR lifting effect due to increasing under cuts as the sample is etched for a longer time. (a)Shows the start of PR lifting at time 1 for waveguides of width 2µm. (b) Shows complete PR lifting at time 2 > time 1for waveguides of width 2µm. (c)Shows no PR lifting problems at time 1 for waveguides of width 3µm. (d) Shows no PR lifting problems at time 2> time 1for waveguides of width 3µm. 73 Chapter 3. Fabrication Via Wet Etching Figure 3.20: Shows defects due to formation of bubbles during wet etching. Also the PR lifting effect is observed in the 2µm wide devices. 74 Chapter 3. Fabrication Via Wet Etching • Bubbles forming when using some etchants containing hydrogen per- oxide may cause non uniform etching and introduce undesired masking effects as seen in Figure 3.20. This problem can be reduced by agita- tion. • HBr oxidation occurs in a solution that has been used and kept for a relatively longer time. This oxidation results in non uniform etching observed due to Br2-based diffusion-limited etching. A change in the color of this solution from clear to more reddish is an indicator of this problem. A fresh solution of H3PO4:HBr:H2O should always be used. 3.3.5 Planarization The aim here is to develop a planarization technique. A successful planariza- tion step must satisfy the following requirements: • Provide device isolation: The dielectric film should prevent un-intended electrical connections. Thus the film must provide coverage without any cracks and sufficiently low porosity. • Provide contact support: The dielectric film should minimize the height mismatches throughout the sample to increase contact stability. At first, SiNx and SiO2 dielectric film growth followed by an etch back step was considered for the planarization and passivation process. SiO2 film growth and etch processes were successfully developed using the PECVD machine at the UBC AMPEL Nanofabrication facility. However, this ap- proach required an additional mask layer with an alignment tolerance better 75 Chapter 3. Fabrication Via Wet Etching than 1 µm and it was due to this fabrication limitation that this technique was not applied. An alternative to the dielectric film solution is the use of polyimide. Polyimide is a photosensitive dielectric. It has been reported that UV sen- sitive polyimide can be lithographically masked as a dielectric film for pla- narization usage. However, an additional mask layer would be required, also aligning this mask layer to the small coupler gap would be very challeng- ing and not reliable with the current mask aligner available. Instead of the traditional masking of polyimide, a new ’etch-back’ process in a developer without UV irradiation was developed. Figure 3.25 shows an image of a sample planarized with Polyimide. The process steps are the following: Polyimide HD8220: • Spinning: (1) 60 sec, 500 rpm, (2) 60 sec, 5000 rpm • Softbake: 3 min at 120C on a hotplate • Developing: ∼ 5min in 10% TMA developer for the ’etch back’ pro- cess, in a few steps, checked under microscope until the top of the waveguides were exposed as monitored by the colour change • Hardbaking; 30 min at 320 C at a hotplate Polyimide spinning process allows a non-uniform film deposition in such a way that a higher thickness of polyimide is deposited in the trenches compared to the thickness deposited on top of the waveguides. Spinning polyimide allows sidewall coverage and partial filling of gaps. This ’filling 76 Chapter 3. Fabrication Via Wet Etching effect’ reduces the height difference throughout the sample, although it does not result in a perfectly flat surface. Figure 3.21 shows a flow diagram of the planarization step. 3.3.6 Metalization Ohmic contacts are needed to allow the flow of electrical current into and out of the laser. A good contact is one that contributes very little to resistance, it is stable and has a linear I-V characteristic. The contact compositions chosen for the n-type InP and p-type InGaAs layers for our laser were chosen based on previously reported results by other groups[36] [37][38]. A summary of the ohmic contact data used for n-type InP and p-type InGaAs are shown in Table 3.6. Table 3.6: Summary of Ohmic Contact Data Metalization Anneal Doping (cm−3) ρc(Ωcm2) N-InP Ni(30nm), Au/Ge(60nm), Au (80nm) 10min @ 400C in 5%H2 95%N2 2− 8× 1018 2 × 10−7 [36] P-InGaAs Pd(10nm), Ti(35nm), Pd(35nm), Au(80nm) 10min @ 400C in 5%H2 95%N2 1× 1019 7.7 × 10−6 [37] [38] 77 Chapter 3. Fabrication Via Wet Etching Figure 3.21: A flow diagram showing the polyimide planarization step. Once the polyimide was spinned onto the sample, an etch back process in the developer solution exposed the top of the waveguides. The next step was to evaporate the contacts. 78 Chapter 3. Fabrication Via Wet Etching The metal contacts were applied to the device by the e-beam evaporator available at the UBC cleanroom facility. A crystal thickness monitor was used to control the film thickness during the evaporation process [39]. The contacts were then annealed for 10 min at 400 C in 5%H2 95%N2. The contact quality was tested by measuring the IV characteristics of two separated contact regions. The contact geometry was a large circle of roughly 600 µm2 area and the current path was as follows: Probe 1 → metal/substrate(doped InP) interface layer → substrate(doped InP) → Probe 2 Thus, the measured resistance was a total resistance RT including re- sistance from the wirings Rw, probes Rp, contacts Rc and the InP material RInP . The N-type back side contact quality was tested by probing two sepa- rated metal strips and measuring the IV characteristic as shown in Figure 3.22. A total resistance of RT = 1.7 Ω was obtained. The P-type top contact was tested in the same way after removal of the U-InP protective layer; the IV characteristic is shown in Figure 3.23. A total resistance of RT = 1.5 Ω was obtained. The achieved contact quality was acceptable since it satisfied the follow- ing conditions: • The contacts exhibited reasonably low resistances, RT < Rdiode, where Rdiode = 37 Ω is shown in Figure 3.27. • The contact characteristics were ohmic 79 Chapter 3. Fabrication Via Wet Etching Figure 3.22: The IV characteristic of the N-type ohmic contact evaporated on the InP contact layer is shown here. A resistance of roughly 1.7 Ω. Figure 3.23: The IV characteristic of the p-type ohmic contact evaporated on the InGaAs contact layer is shown here. A resistance of roughly 1.5 Ω. 80 Chapter 3. Fabrication Via Wet Etching 3.3.7 Cleaving Since these devices are edge emitters, the samples need to be cleaved in-order to provide an output port for measurements. The cleaving requirements for our mask and design are: • Provide a smooth and flat facet free of any damages such as scratches or cracks • Separate devices, which means a cleaving pitch of about 200 µm The cleaving process was done manually. Similar to when the wafer was cleaved into small die pieces during the sample preparation step, a tip of a diamond cutter was used to brake the sample along its crystallographic orientation line. Placing a damper such as clean wipes underneath the sample when cleaving helps the process. Looking at the side wall of the cleaved samples as shown in the opti- cal images of Figure 3.24 and cleaved facets in the SEM image of Figure 3.18, many defects and damages are observed as the result of the cleaving technique and process. The device yield after the final cleaving step is very low for the following reasons: • Cleaving location is not well controlled and results in destruction of devices, • Coupler waveguides are damaged and/or terminated before the cleaved output edge 81 Chapter 3. Fabrication Via Wet Etching • Due to large cleaving pitch and limited location control, a number of resonators maybe left coupled to the same coupler waveguide. Testing devices away from the cleaved facet allows more losses due to defects (such as other devices coupled to the same output waveguide or dam- ages to the output waveguide) along the path to the output. • The cleaved facet has a poor quality as observed in Figure 3.18 and Figure 3.24 • A good cleaved output facet would significantly improve the amount of light coupled out as well as easing the task of fiber alignment when doing measurements. A good facet is specially critical for the RWG edge emitter lasers since it forms the cavity as further discussed in Section 5.4. Suggestions and current efforts to improve and to replace this manual cleaving technique are explained in the final analysis section. 3.3.8 Results Figure 3.25 shows an optical image of the fabricated devices. A suitable setup was made with the help of Wei Shi, in order to measure the diode and laser characteristics of t he edge emitting devices. A picture of the setup is shown in Figure 3.26. After cleaving, the samples were fixed using a carbon tape onto a 4 cm × 4 cm × 2 mm copper sheet which was placed on top of a 3-axis stage. The temperature of the copper sheet was controlled by a temperature controller through a Peltier cooler. The bus top 82 Chapter 3. Fabrication Via Wet Etching Figure 3.24: Optical images showing the side view of the wafer after cleaving. These samples were not thinned. Figure 3.25: The final fabricated SRLs are displayed before cleaving. The waveguides, top contacts and the polyimide regions are all shown in this Figure. 83 Chapter 3. Fabrication Via Wet Etching contact was kept at a fixed low bias while the ring contact current was swept. A Matlab program (LIVdownloader.m) was used to retrieve the data. Figure 3.26: The set up used for measuring edge emitter lasers. The IV and LI measurements of Figures 3.27(a) and 3.27(b) show that a good IV characteristic as well as light emission was successfully achieved. The device’s optical spectrum behaviour for varied bias conditions was measured using an optical spectrum analyzer (OSA), an example of such 84 Chapter 3. Fabrication Via Wet Etching Figure 3.27: (a)IV Characteristic of a fabricated SRL with a diode resistance Rdiode = 37Ω(b)LI characteristic of a fabricated SRL. 85 Chapter 3. Fabrication Via Wet Etching 1500 1520 1540 1560 1580 1600−80 −78 −76 −74 −72 −70 −68 −66 Spectrum of InP SRL Below Threshold Wavelength (nm) Lig ht Ou tpu t (d Bm ) R=500mA R=800mA R=1000mA R=1200mA R=1500mA Figure 3.28: (a)Optical spectrum of a fabricated SRL biased at 500 mA, 800 mA, 1000 mA, 1200 mA, and 1500 mA. The sample was cooled to roughly 5 degrees. 86 Chapter 3. Fabrication Via Wet Etching measurement is shown in Figure 3.28. Lasing, however was not observed. Going to higher currents showed that the device overheated and the power rolled off before lasing could occur. The voltage at 500mA was 6 V. Thus the power consumed by the device was: Pconsumed = IV = (1500mA)(6V ) = 9W (3.3) The output power as measured by the OSA was: Pout = IV = 10 −65dBm/10 = 0.3nW (3.4) Device area: Area = (3µm)(2π200µm+2× 250) = (3µm)(1756.6µm) ≈ 5300µm2 (3.5) The lasing wavelength shifts with a varying bias current. The lasing wave- length is dependent on the length (d) of the cavity and the index of refraction (n). λ = ngd0 m (m = 1, 2, 3, 4...) (3.6) The gain-peak shows a shift of ≈ 20 nm. This shift is due to the high temperature effects. 87 Chapter 3. Fabrication Via Wet Etching 3.3.9 Analysis The spectrum measurements of Figure 3.28 show the gain spectrum, some resonances and the effect of increasing the bias current on the emission spec- trum below threshold. Below threshold, increases in bias current contribute mostly to spontaneous emission, as also shown in the LI curve of Figure 3.27(b); there is almost no stimulated emission. An increase in the pump- ing rate directly contributes to an increase in the population inversion and the amplitude of the gain spectrum increases, but since there is no lasing, the emission spectrum is wide. Below threshold, the cavity losses exceed the gain coefficient, which means none of the emissions in the spectrum can experience a large enough gain to start lasing. The losses mainly due to fabrication have resulted in a very large threshold. Some sources contributing to the cavity losses could be: • Defects along the ring waveguide such as dirt and scratches could cause too much scattering or completely interfere with the light propagation. • Polyimide residues leftover on some areas of the ring waveguide (before the contact evaporation step) would result in areas where there is no pumping and thus no inversion occurs. • Metal deposition on the waveguide sidewalls could result in increased propagation losses, specially if the metal is closer to the bottom of the waveguide where it can have a higher affect on the field distribution. • Losses due to bending, mode conversion losses 88 Chapter 3. Fabrication Via Wet Etching • Self-heating effect: the device’s resistance could lead to heating, which would reduce the optical gain. • Another possibility could be problems with the epitaxy. The PL and the contacts are reasonable, so there is no evidence for this hypothesis, but it has not been ruled out. The major limiting step in the fabrication process was the cleaving. Due to the very low yield after cleaving, many devices did not have the chance to be properly tested. This process is currently under improvement. The common industry practice is to thin the wafers from their typical roughly 1 mm thickness down to 100 µm - 200 µm thickness before cleaving. We have tested the thinning idea, by manually thinning a sample on a sand paper and much better cleaving results were observed. However, doing this manually takes several hours of sanding by hand and it is not a reliable method. This wafer thinning process is done with automatic equipment in industry. We are currently putting parts together to test a wafer thinning machine at UBC. Once the wafer is thinned, still a better cleaving tool is required. The diamond tips are too wide, a better choice are surgery knives. Surgery knives are not the common option for cleaving InP wafers, however after thinning not much force is required and the surgery knife can easily be used to cleave devices. 89 Chapter 3. Fabrication Via Wet Etching 3.4 InP Edge Emitter Ridge Waveguide Laser Demonstrating a RWG edge emitting laser in InP would be a step closer towards fabricating InP SRLs. The same epi-wafer used to fabricate the SRLs was used for the fabrication of the RWG lasers. A mask was designed using Clewin’s Matlab script; the mask patterns consisted of long straight rectangular waveguides with varying waveguide widths. There were 2 mask layers: Layer 1 for etching and Layer 2 for contact evaporation. The etch mask was drawn wider in order to compensate for the under-etch effect. The etch mask included the following variations: Waveguide width [µm]: 6, 8, 12, 17, 22, 27, 32, 37, 47 Waveguide contact [µm]: 2, 4, 8, 10, 15, 20, 25, 30, 40 The fabricated devices were tested and LED characteristics similar to the SRL results were obtained; however, lasing was not observed. This is again most likely due to the cleaving technique leaving a poor quality facet which is very critical for the RWG laser since the cleaved facets form the cavity mirrors. Moreover, the cleaving pitch is also currently limited to roughly 1 mm and higher. Thus the cavity lengths are large, providing a larger area for inversion and thus having higher threshold currents. 3.5 Conclusion A commercial 1550 nm epi-wafer was purchased. The required mask was designed and purchased. The following fabrication steps were successfully achieved: 90 Chapter 3. Fabrication Via Wet Etching • A semi-automated mask drawing script • A photolithography process with a resolution better than 1 µm thanks to edge bead removal • A selective wet etching recipe giving smooth surfaces, in-plane isotropy, and an orthogonal to in-plane etch rate ratio of 2 • A planarizing technique via polyimide spinning, which does not require any extra masking step • Reasonably low resistance, ohmic contacts for n and p contacts An appropriate setup was made for testing edge emitters. The diode char- acteristics of the device was successfully measured. Light emission was also measured. The spectrum measurements were performed and the gain spec- trum and some resonances were observed. However, the output power was very low, in the tenths of nWs, and lasing did not occur. The cleaving process is being improved in order to allow further testing. 91 Chapter 4 Fabrication Via Dry Etching Dry etching techiques used for SRL fabrication are well published and more mature, thus we wanted to make some SRLs via dry etching to test the designs. We originally planned to pursue dry and wet etching techniques in parallel at the UBC Nanofabrication facility. The most common dry etching techniques are reported to be high tem- perature Cl2 based or hydrocarbon-based plasma. Usually the hydrocarbon- based dry etching is preferred for InP laser fabrication. The main reason for which hydrocarbon-based plasma is preferred over Cl2-based electron cyclotron resonance (ECR) plasma is that Cl2 etching rate of InP is tem- perature dependent [40]. In low temperatures of 80deg etching rate is <0.05 µm/min (this is due to the nonvolatile InCl reaction products, which also result in grassy-roughened surfaces), for higher temp of 230deg the rate is saturated at more than 1 µm/min [40]. This higher temperature requires a hard mask ( SiO2/or SiN). To achieve smoother side-walls and bottom surfaces high ion voltage of 900 V is reported and Cl2 gas pressure of 7E-5 torr [40]. Meanwhile the CH4 plasma etching is done at room temperature and it has been reported that addition of O2 decreases the etch rate but it gives selectivity of 60 (higher etching rate of InP over InAlAs) [40]. This 92 Chapter 4. Fabrication Via Dry Etching selectivity is explained by the oxidation of InAlAs layer, forming a thin layer of Al2O3 which could also provide electrical passivation. Addition of O2 also improves the side-wall verticality because it partially removes the polymer, which forms during the InP etching[40]. The UBC AMPEL Nanofabrication facility has an ECRmachine in work- ing condition. This machine was not suitable for CH4 gas; however, it has Cl2 gas available but the machine was not able to provide the required high temperatures. Investigation was done to replace the heating chuck in order to supply higher temperatures but this turned out to be an overly expensive task. The UBC facility also has an reactive ion etcher (RIE) machine which could be used for the preferred CH4 etching; however, this machine is not in working conditions. The machine was being repaired for over 1.5 years but it was decided that it was not going to reach working conditions. Thus we started to look dry etching fabrication outside UBC. Dry etching fabrication of the InP SRL has started at CRN2 facility of University of Sherbrooke funded by CMC. The high resolution SRL mask was redesigned in order to take into account the dry etching approach and include a planarization step. The planarization step will be done via SiO2 deposition and etch back process instead of the previous polyimide spinning technique. Figure 4.1 shows the 3 different mask layers with layer 1 for dry etching process, layer 2 for planarization and layer 3 for metal evaporation. Figure 4.2 shows a cartoon like cross section of the epi-wafer and SRL via dry etching. They already have a recipe developed for InP etching using a plasma 93 Chapter 4. Fabrication Via Dry Etching Figure 4.1: The three different mask layers used to first define the trench for dry etching, next to make a contact opening for the oxide planarization pro- cess and finally to define the metalization region for the top p-type contacts. The n-type contact is evaporated on the backside. 94 Chapter 4. Fabrication Via Dry Etching Figure 4.2: Cartoon of the epi-wafer structure and the side view of a dry etched SRL. 95 Chapter 4. Fabrication Via Dry Etching etcher STS ICP III-V. They have completed the first dry etching tests and tuned their recipe for our heterostructure; the etch depth is monitored with a laser end point detection system. An aspect ratio dependent etching (ARDE) was observed as depicted in Figure 4.3, which shows an SEM im- age of the cross section for the latest sample that was dry etched with their system. Figure 4.3: SEM image revealing the etch profile and etch depth in the coupler region and the trench region. Courtesy of Sherbrooke. The ARDE was summarized in Figure 4.4. The ARDE can be translated in terms of the thickness between the bottom of the etched trench and the etch stop layer. 96 Chapter 4. Fabrication Via Dry Etching Figure 4.4: A plot of etching rate as a function of the trench size is shown here to capture the ARDE for the dry etching of the InP samples. Courtesy of Sherbrooke. A study was done to analyze the designs etch depth tolerance. Struc- tures were simulated in FDTD mode solver and the corresponding coupling behavior was then extracted for the etch depths predicted by the graph of Figure 4.4. The results show much stronger field interaction between the waveguides and thus much shorter crossover coupling lengths as depicted in the data of Table 4.1 and Table 4.2. In the Tables, the subscripts a and b refer to a coupler length Lκa = 200 µm and a coupler length Lκb = 250 µm respectively. The corresponding threshold current and output power values for the devices of Table 4.1 are predicted and presented in Table 4.3 for a device with a bend region of radius R = 300 µm. The etch rate dependence of the dry etching on the gap size results in much shorter crossover lengths. However these Lx values are now shorter than the current designed coupler 97 Chapter 4. Fabrication Via Dry Etching lengths which could make the coupling determination more challenging, but it will provide more devices at higher powers to measure and analyze. Com- paring the results of Table 4.2 to Table 2.3, the achieved Lx values are very different. These two Tables present results for 3 wide devices, the only differences are the etch depths and the waveguide shapes (rectangular vs. trapezoidal). The Lx value for the trapezoidal device with an etch depth of 2 µm is 86111 µm, meanwhile for the rectangular device of etch depth 1.64 µm it is 171.1 µm. This can be explained by comparing the effect of the etch depth on the mode distributions for the structures. Comparing the first order mode distribution for the trapezoidal structure as shown in Figure 4.5 to the dry etched waveguide of etch depth 1.64 µm as depicted in Figure 4.6, a much stronger field interaction is observed between the coupler waveguides. They are currently working to improve their recipe and perhaps to change into a selective recipe. Table 4.1: Etch Depth Analysis, W = 2 µm Gap (µm) Etch Depth (µm) neff1 neff2 κa κb Lx (µm) 0.5 1.1 3.268150 3.256780 0.99 0.249 68.16 0.8 1.26 3.268454 3.258455 0.625 0.88 77.51 1 1.36 3.268499 3.259402 0.27 0.99 85.19 Continued on next page 98 Chapter 4. Fabrication Via Dry Etching 1.5 1.54 3.267857 3.261085 0.149 0.082 114.4 2 1.64 3.266478 3.261949 0.931 0.561 171.1 Table 4.2: Etch Depth Analysis, W = 3 µm Gap (µm) Etch Depth (µm) neff1 neff2 κa κb Lx (µm) 0.5 1.1 3.270061 3.264313 0.5 0.052 134.8 0.8 1.26 3.270126 3.264932 0.74 0.238 149.2 1 1.36 3.270069 3.265295 0.873 0.437 162.3 1.5 1.54 3.269434 3.265993 0.97 0.97 225.2 2 1.64 3.268362 3.266283 0.557 0.755 372.8 Table 4.3: Predicted Performance for Waveguide W = 2 µm R = 300 µm Ith (mA) Pout (mW) Gap (µm) 1a 1b a b 0.5 80 58 9 3 0.8 60 69 8.5 10.5 1 56 83 4 7.9 1.5 55 57 2 1.1 2 71 61 11.8 7.2 99 Chapter 4. Fabrication Via Dry Etching Figure 4.5: The mode distribution is shown for a trapezoidal structure with uniform 2 µm etch depth everywhere. Figure 4.6: The mode distribution is shown for a structure with a 1.64 µm etch depth in the gap region and 2 µm etch depth on the outside walls of the waveguides. 100 Chapter 5 Conclusion This thesis discussed the design process and process development for the fabrication of SRLs based on a commercial InP epitaxy. Unlike the well developed dry etching approach taken by other groups, we developed a new selective multi step wet etching technique for the fabrication of SRLs. This wet etching approach is a much simpler, cheaper and faster process compared to the dry etching methods. Moreover, achieving smooth surfaces by dry etching is possible however it adds more complexity to the etching process where as wet etching inherently results in smooth surfaces. Using this wet etching approach SRLs with waveguide widths as small as 3 µm can be fabricated. This limitation on the width of the device is due to the 1 to 2 relationship of the in-plane etch rate to the orthogonal etch rate. Also a polyimide planarization technique was developed that didn’t re- quire any additional masking, UV exposure or plasma exposure. We were able to measure the spectrum and LED characteristics of the fabricated SRLs. 101 Chapter 5. Conclusion 5.1 Future Work Further design, fabrication and measurement work are required in future to improve the SRL design model and optimize for device performance. For a more accurate SRL modelling, a more detailed loss model should be incorporated into the simulations which considers: • Bending losses: This occurs in the bend waveguide region and is a function of the bend radius. • Mode conversion losses: This occurs when light travels from the straight waveguide region to the bend waveguide region as well as when light travels into the coupling region. Amongst other parameters, this loss is a function of the effective index difference between the regions along the light’s path of travel. • Scattering losses: This loss is due to interaction of the field at crystal imperfections and surface roughness. It is a function of the amount of field intensity at the waveguide walls as well as the waveguide surface roughness root mean square (rms) amongst other parameters. Moreover, the power coupling prediction accuracy can be increased by con- sidering power coupling contributions from the bend regions near the cou- pler region of the SRL device. Other considerations contributing to the SRL modelling accuracy are: • The refractive index changes due to varying bias conditions since the coupling and the losses are refractive index dependent 102 Chapter 5. Conclusion • The effect of bias conditions and surface roughness on the dominating resonant mode (CW/CCW/bistable) • Mode competition analysis and effects Developing a cleaving process via wafer thinning (thinning the wafers down to 100 µm to 200 µm thickness) that results in high quality facets and allows high cleaving accuracy (allowing device separation) would sig- nificantly increase the fabrication yield and reduce the thermal effects to the substrate resistance. Moreover, in order to optimize for output power, higher power coupling coefficients are required. To achieve a higher power coupling coefficient, the higher lithography resolution (better than 300nm) needs to be increased to reduce the coupling gap during fabrication. Measurement future work: • Pulsed Measurement, self heating could be reduced by this technique • Modification of the set up to include a high magnification in order to be able to see 2 µm and 3 µm wide waveguides. This would be helpful for both seeing the waveguide condition (for example damages) as well as seeing the facet, which will help both the alignment and the inspection of facet quality. • Modifying the experimental set up so that the edge emitting devices could be measured from either one of the cleaved facets. This would allow easier lasing detection, since the device may be operating in clockwise (CW) or counter clockwise (CCW) mode depending on the bias conditions. 103 Bibliography [1] Diode Lasers and Photonic Integrated Circuits. John Wiley and Sons INC., 1995. [2] B. Wen, “1550nm fp-ld epiwafer specifications,” LandMark, Tech. Rep., 2008. [3] P. J. T. (Ed.), “Solvent guide,” Burdick and Jackson Laboratories, Inc, Tech. Rep., 1980. [4] L.Chrostowski and W.Shi, “Monolithic injection-locked high-speed semiconductor ring lasers,” Lightwave Technology, vol. 26, 2008. [5] A. ariv and P. Y. eh, Photonics,OpticalElectronicsInModern Commu- nications, 6th ed. Oxford University Press, 2007. [6] R. K.Amarnath and R.Ho, “Electrically pumped InGaAsP-InP micror- ing optical amplifiers and lasers with surface passivation,” Photonics Technology Letters, vol. 17, no. 11, 2005. [7] Y. Fedoryshyn, P. Strasse, P. Ma, F. Robin, and H. Jäckel, “Optical waveguide structure for an all-optical switch based on intersubband transitions in InGaAs/AlAsSb quantum wells,” Optics Letters, vol. 32, no. 18, 2007. 104 Bibliography [8] T.-C. Peng, C.-C. Yang, Y.-H. Huang, M.-C. Wu, C.-L. Ho, and W.- J. Ho, “Self-terminated oxide polish technique for the waveguide ridge laser diode fabrication,” Journal of American Vacuum Society, 2005. [9] K. K. Masaki Yanagisawa, T. Kawasaki, R. Yamabi, S. Yaegassi, and H. Yano, “A robust all-wet-etching process for mesa formation of ingaas–inp hbt featuring high uniformity and high reproducibility,” IEEE Transactions On Electrom Devices, vol. 51, no. 8, 2004. [10] K. Shinoda, K. Nakahara, and H. Uchiyama, “InGaAlAs/InP ridge-waveguide lasers fabricated by highly selective dry etching in CH4/H2/02 plasma,” IEEE, 2003. [11] M. Sorel, P. J. R. Laybourn, G. Giuliani, and S. Donati, “Unidirectional bistability in semiconductor waveguide ring lasers,” Applied Physics Letters, vol. 80, no. 17, 2002. [12] J. P. R. L. Bach, Member, A. Forchel, J. L. Gentner, and L. Goldstein, “Wavelength stabilized single-mode lasers by coupled micro-square res- onators,” IEEE Photonics Technology Letters, vol. 15, no. 3, 2003. [13] K. Amarnath, R. Grover, Member, S. Kanakaraju, and P.-T. Ho, “Elec- trically pumped IGaAsP–InP microring optical amplifiers and lasers with surface passivation,” IEEE Photonics Technology Letters, vol. 17, no. 11, 2005. [14] S. F. rst, M. Sorel, A. Scire‘, G. Giuliani, and S. Yu, “Technological challenges for cw operation of small-radius semiconductor ring lasers,” Proc. of SPIE, vol. 6184, 2006. 105 Bibliography [15] S. Park, S.-S. Kim, L. Wang, and S.-T. Ho, “InGaAsP–InP nanoscale waveguide-coupled microring lasers with submilliampere threshold cur- rent using Cl–N-based high-density plasma etching,” IEEE Journal Of Quantum Electronics, vol. 41, no. 3, 2005. [16] A. W. Fang, R. Jones, H. Park, O. Cohen, O. Raday, M. J. Paniccia, and J. E. Bowers, “Integrated AlGaInAs-silicon evanescent racetrack laser and photodetector,” Optical Society of America, 2007. [17] O. Tadanaga, K. Tateno, H. Uenohara, T. Kagawa, and C. Amano, “An 850-nm InAlGaAs strained quantum-well vertical-cavity surface- emitting laser grown on GaAs (311)b substrate with high-polarization stability,” IEEE Photonics Technology Letters, vol. 12, no. 8, 2000. [18] S. F. rst, M. Sorel, A. Scire‘, G. Giuliani, and S. Yu, “Technological challenges for cw operation of small-radius semiconductor ring lasers,” Proc. of SPIE, vol. 6184, 2006. [19] R. Vafaei, “Silicon on insulator add-drop racetrack resonators,” sOI Paper Report. [20] A.Yariv and P.Yeh, Photonics, Optical Electronics In Modern Commu- nications, 6th ed. Oxford University Press, 2007. [21] R. et al, “Low loss GaAs/AlGaAs optical waveguides on InP sub- strates,” IEEE Photonics Technology Letters, vol. 2, 1990. [22] [Online]. Available: www.mina.ubc.ca/course nanophotonics2010 106 Bibliography [23] W. Shi, R. Vafaei, M. A. G. Torres, N. A. F. Jaeger, and L. Chrostowski, “Design and characterization of microring reflectors with a waveguide crossing,” Optics Letters, vol. 35, 2010. [24] ——, “Design and characterization of microring reflectors with a waveg- uide crossing,” in 2010 International Conference on Optical MEMS and Nanophotonics, 2010. [25] N. Rouger, L. Chrostowski, and R. Vafaei, “Temperature effects on silicon on insulator (soi) racetrack resonators: a coupled analytic and 2d finite difference approach,” Lightwave Technology, vol. 28, 2010. [26] [Online]. Available: www.nanofab.ubc.ca/content/about-facility [27] [Online]. Available: http://www.nanofab.ubc.ca/process view [28] [Online]. Available: www.microchemicals.eu/technical-information [29] [Online]. Available: www.nanofab.ubc.ca/process-photolithography [30] [Online]. Available: http://www.nanofab.ubc.ca/equipment view [31] A.R.Clawson, “Guide to references on iii-v semiconductor chemical etching,” Material Science and Engineering, 2001. [32] S. Adachi and H. Kawaguchi, “Chemical etching characteristics of (001)lnP,” J. Electrochem. Soc.: Solid-State Science And Technology, vol. 128, no. 6, 1981. [33] Y.-S. Kim, J. Kim, J.-S. Choe, Y.-G. Roh, H. Jeon, J. Kim, J.-S. Choe, Y.-G. Roh, H. Jeon, and J. C. Woo, “Semiconductor microlenses fabri- 107 Bibliography cated by one-step wet etching,” IEEE PHOTONICS TECHNOLOGY LETTERS, vol. 12, no. 5, 2000. [34] F. Fiedler, A. Schlachetzki, and G. Klein, “Material-selective etching of InP and an InGaAsP alloy,” Journal of Materials Science, vol. 17, 1982. [35] P. H.Holloway and G. E.McGuire, Eds., Handbook Of Compund Semi- conductors Growth, Processing, Characterization and Devices. Noyes Publications, 1995. [36] J. R. ABELSON, T. W. SIGMON, G. BAHIR, and J. L. MERZ, “Rapid thermal alloyed ohmic contact on inp,” Journal of Electronic Materials, vol. 16, no. 4, 1987. [37] E. F. Chor, W. K. Chong, and C. H. Heng, “Alternative Pd,Ti,Au con- tacts to Pt,Ti,Au contacts for In0.53Ga0.47As,” JOURNAL OF AP- PLIED PHYSICS, vol. 84, no. 5, 1998. [38] E. Chor, D.Zhang, H.Hong, W. Chong, and S. Ong, “Electrical char- acterization, metallurgical investigation, and thermal stability stud- ies of (pd, ti, au)-based ohmic contacts,” JOURNAL OF APPLIED PHYSICS, vol. 87, no. 5, 2000. [39] Evaporator. [Online]. Available: http://www.nanofab.ubc.ca/ equipment view [40] Wada and Hasegawa, Eds., InP-based Materials and Devices. Wiley- Interscience, 1999. 108 Appendix A Clewin Matlab Script Used for the Mask Layout %This script will create the etch mask layer % Due to Clewin file script size limitation, separate scripts had to be written and positioned %after each other to make the whole mask. % Each of the scripts titled Script 01a to Script 01f will make 3 rows of the mask. %All scripts are the same except for varying ring\\radius and gaps. %%%%%%%%%%%%%%%%%%%%%%%%% Script 01a %%%%%%%%%%%%%%%%%%%%%%%%% %Create an angle vector between -90deg to 90deg, ex:100 angles angle = linspace(-pi/2,pi/2,100); offset = 80; yshift = 0; yshiftNext = 0; xshift = 0; %Create a radius vector giving a radius for each angle: rInner = [300,250,200,180,160,150,100];%Radius for the inner circle lenR = length(rInner); width = [2,3];%rOuter-rInner; lenw = length(width); gapArray1 = [0.5,0.5,1,1,1.5,1.5,2,2,1,1,1.5,1.5,2,2]; gapArray2 = [1.5,1.5,2,2,2,1.5,1,1]; 109 Appendix A. Clewin Matlab Script Used for the Mask Layout CouplingL = 200; LofDye=6500; x1Shift_init=0; for iR = 1:3;%lenR iG=1; for iw = 1:lenw w = width(iw); if w==2 gapArray=gapArray1; lenG = length(gapArray); else gapArray = gapArray2; lenG = length(gapArray); end rOuter = width(iw)+rInner; %Radius for the outer circle for a given width %Calculate the nodes of a inner circle and an outer circle, nodesInnery = rInner(iR).*sin(angle)+yshift; nodesOutery = rOuter(iR).*sin(angle)+yshift; lenN = length(nodesInnery); Rnodes = []; Lnodes = []; LofRaceTrack = rOuter(iR)*2+CouplingL+offset; CurrentNumofRaceTracks= round(LofDye/(LofRaceTrack)); %Calculate the nodes of a half ring, %i.e. a matrix with 2 columns and 1000 rows: for iRx = 1:CurrentNumofRaceTracks/2+1 if iRx < CurrentNumofRaceTracks/2 && iR==2 && iRx==1 xshift = xshift+50; end gap = gapArray(iRx); 110 Appendix A. Clewin Matlab Script Used for the Mask Layout if gap == 0.5 && w==2 shiftUp = (gap - 0.5)+2*w; else shiftUp = (gap - 0.5)+2*w; end nodesInnerx = rInner(iR).*cos(angle)+xshift; nodesOuterx = rOuter(iR).*cos(angle)+xshift; for index = 1:lenN if (nodesOuterx(index)-nodesInnerx(index))>= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; 111 Appendix A. Clewin Matlab Script Used for the Mask Layout y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && iR == 2 && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); xshiftNext = 2*rOuter(iR)+CouplingL+offset-400; xshift = xshift+xshiftNext; else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end end xshift = 0; if iR= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; 114 Appendix A. Clewin Matlab Script Used for the Mask Layout Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && iR == 2 && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); xshiftNext = 2*rOuter(iR)+CouplingL+offset-250; xshift = xshift+xshiftNext; else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB 115 Appendix A. Clewin Matlab Script Used for the Mask Layout +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end end xshift = 0; if iR= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; 118 Appendix A. Clewin Matlab Script Used for the Mask Layout x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && iR == 2 && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); xshiftNext = 2*rOuter(iR)+CouplingL+offset-250; xshift = xshift+xshiftNext; else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end end xshift = 0; if iR= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; 121 Appendix A. Clewin Matlab Script Used for the Mask Layout Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && (iR == 2 || iR ==1 || iR==4) && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); if iR==1 AlignmentG=500 else AlignmentG=250; end xshiftNext = 2*rOuter(iR)+CouplingL+offset-AlignmentG; 122 Appendix A. Clewin Matlab Script Used for the Mask Layout xshift = xshift+xshiftNext; else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end end xshift = 0; if iR= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; 125 Appendix A. Clewin Matlab Script Used for the Mask Layout y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && iR==4 && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end 126 Appendix A. Clewin Matlab Script Used for the Mask Layout end xshift = 0; if iR= 0 nodesy(index) = nodesOutery(index)+shiftUp; Rnodesx(index) = nodesOuterx(index)+CouplingL/2; Rnodes = [Rnodesx(:), nodesy(:)]; Lnodesx(index) = nodesOuterx(index) - 2.*nodesOuterx(index) - CouplingL/2+2*xshift; Lnodes = [Lnodesx(:), nodesy(:)]; end lenNr = length(Rnodes(:,1)); %calculate nodes of coupling length x2couplingT = Rnodes(lenNr,1); x1couplingT = x2couplingT - CouplingL; y2couplingT = Rnodes(lenNr,2)+w/2; y1couplingT = y2couplingT - w; x2couplingB = Rnodes(1,1); x1couplingB = x2couplingB - CouplingL+ x1Shift_init; y2couplingB = Rnodes(1,2)+w/2; y1couplingB = y2couplingB - w; x1Line_init = -400; x1Line = x1Line_init + xshift; x2Line = x1Line_init+LofRaceTrack+10; %*CurrentNumofRaceTracks; y2Line = y1couplingB - gap; y1Line = y2Line- w; iG = iG+1; end if iRx > CurrentNumofRaceTracks/2 && iR == 1 && w==2 rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB-100,y2Line); rectangle(x2couplingB-100, y1Line-0.5,x2couplingB+rOuter(iR)+offset,y2Line+0.5); xshiftNext = 2*rOuter(iR)+CouplingL+offset-250; xshift = xshift+xshiftNext; 129 Appendix A. Clewin Matlab Script Used for the Mask Layout else wire(1,w,Rnodes); %Right half ring rectangle(x1couplingT,y1couplingT,x2couplingT,y2couplingT); rectangle(x1couplingB,y1couplingB,x2couplingB,y2couplingB); wire(1,w,Lnodes); %Left half ring %rectangle(x1Line,y1Line,x2Line,y2Line); rectangle(x1couplingB-rOuter(iR)-offset*0.5, y1Line,x2couplingB +rOuter(iR)+offset*0.5,y2Line); xshiftNext = 2*rOuter(iR)+CouplingL+offset; xshift = xshift+xshiftNext; end if iRx==1 && w==2 rectangle(-700, y1Line,x1couplingB-rOuter(iR)-offset*0.5,y2Line); elseif iRx>=CurrentNumofRaceTracks/2 && w==3 rectangle(x2couplingB+rOuter(iR)+offset*0.5, y1Line,7600,y2Line); end end end xshift = 0; if iR