|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200397Ka 4500
controlfield tag 001 001789995
007 cr bnu uuuuu
008 070612s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001479
Myers-Ward, Rachael L.
High growth rate SiC CVD via hot-wall epitaxy
h [electronic resource] /
by Rachael L. Myers-Ward.
[Tampa, Fla] :
b University of South Florida,
ABSTRACT: This dissertation research focused on the growth of 4H-SiC epitaxial layers in low-pressure horizontal hot-wall chemical vapor deposition (CVD) reactors. The goal of the research was to develop a growth process that maximized the growth rate and produced films of smooth morphology. The epitaxial growth of SiC was carried out in two different reactor sizes, a 75 mm reactor and a 200 mm reactor. The maximum repeatable growth rate achieved was 30-32 um/h in the 200 mm reactor using the standard chemistry of hydrogen-propane-silane (H2-C3H8-SiH4) at growth temperatures < 1600 C, which is the highest growth rate reported to date in a horizontal hot-wall reactor at these temperatures. This growth rate was achieved with a silane flow rate of 30 sccm. The process development and characterization of 4H-SiC epitaxial films grown using the standard chemistry are presented.There are many ways to increase the growth rate, such as changing the pressure, increasing the reactant flow rates, or increasing the temperature. The method of choice for this dissertation work was to first increase the reactant flow rates, i.e. silane flow rate, and then to alter the growth chemistry by using a growth additive. When the silane flow is increased, while maintaining a specific growth temperature, supersaturation of silicon may occur. When this happens, particulates may form and deposit onto the sample surface during growth which degrades the film morphology of the epitaxial layers. In order to overcome this severe limitation in the growth of SiC, hydrogen chloride (HCl) was added to the standard chemistry of H2-C3H8-SiH4 during growth when the SiH4 flow was increased beyond 30 sccm. With the addition of HCl, the Si supersaturation was suppressed and the growth rate was increased from ~32 um/h to ~ 49 um/h by increasing the silane precursor up to 45 sccm, while maintaining the Si/C ratio of the standard chemistry process. The addition of HCl to the standard chemistry for growth of SiC films was pioneering work that has since been duplicated by several research groups.
Dissertation (Ph.D.)--University of South Florida, 2006.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 214 pages.
Adviser: Stephen E. Saddow, Ph.D.
Chemical vapor deposition.
High growth speeds.
x Electrical Engineering
t USF Electronic Theses and Dissertations.
High Growth Rate SiC CVD via Hot-Wall Epitaxy by Rachael L. Myers-Ward A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Stephen E. Saddow, Ph.D. Andrew M. Hoff, Ph.D. John T. Wolan, Ph.D. Grisselle Centeno, Ph.D. Olle Kordina, Ph.D. Yevgeniy Shishkin, Ph.D. Date of Approval: January 13, 2006 Keywords: silicon carbide, chemical vapor deposition, epitaxial layers, high growth speeds, low-pressure Copyright 2006, Rachael L. Myers-Ward
i TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES iv ABSTRACT xiii PREFACE xv CHAPTER 1 INTRODUCTION 1 1.1 Silicon Carbide Overview 1 1.2 Motivation for Dissertation Research 5 1.3 SiC CVD Epitaxy Background 7 1.4 Organization of Thesis 14 1.4.1 Publications and Presentations 15 CHAPTER 2 HOT-WALL CHEMICAL VAPOR DEPOSITION 18 2.1 Chemical Vapor Deposition 19 2.2 CVD Thermodynamics 22 2.2.1 Non-linear Equation Method 23 2.2.2 Minimization of Gibbs Free Energy 26 2.2.3 Thermodynamic Simulations 27 2.3 Kinetics 31 2.3.1 Fluid Dynamics 32 2.3.2 Mass Transport Limited Versus Surface Reaction Limited CVD 35 2.4 Reactor Hardware 40 2.4.1 75 mm Hot-wall System 41 2.4.2 200 mm Hot-wall System 46 2.5 Computational Fluid Dynamic (CFD) Simulations of the 200 mm Reactor 51 2.6 Summary 55 CHAPTER 3 4H-SIC EPITAXIAL GROWTH EXPERIMENT 56 3.1 Characterization Techniques 56 3.2 75 mm Hot-Wall Reactor Experiments 57 3.3 200 mm Reactor Process Development 59
ii 3.3.1 Flat Ceiling Susceptor Design 61 3.3.2 Angled Ceiling Susceptor Design 65 3.4 200 mm Reactor Process 74 3.5 200 mm Reactor Growth Characterization 78 3.5.1 Growth Rate as a Function of Silane Flow 79 3.5.2 Growth Rate as a Function of Process Pressure 82 3.5.3 Analysis of Thick Films 83 3.5.4 Analysis of Doping Concentration 87 3.5.5 Growth Variation in Hot-Zone 90 3.6 Summary 93 CHAPTER 4 4H-SIC EPITAXIAL GR OWTH USING HCL AS A GROWTH ADDITIVE 95 4.1 Proposed HCl Mechanisms 96 4.2 HCl Manifold 101 4.3 HCl Growth Additive Expe riments: 75 mm Reactor 103 4.3.1 Experimental Results 103 4.3.2 Characterization (AFM, XRD, and LTPL) 112 4.4 HCl Growth Additive Expe riments: 200mm Reactor 120 4.4.1 Growth Rate as a Function of Silane Flow 121 4.4.2 Characterization 126 4.5 Summary 130 CHAPTER 5 SUMMARY AND FUTURE WORKS 132 5.1 Summary 132 5.2 Recent and Future Work 135 5.2.1 Chemistry 135 5.2.2 Temperature Uniformity 140 5.2.3 CVD Simulations 142 REFERENCES 152 APPENDICES 159 Appendix A NASA-Lewis Chemical Equilibrium Computations and Applications (CEA) 160 Appendix B Characterization Techniques 175 Appendix C Operating Procedure for 4H-SIC Epitaxial Growth in the USF 200 mm Reactor 199 Appendix D HCl Purge Panel Operating Procedure 204 Appendix E Thickness Determination via Cross-Section SEM and FTIR Analysis 207 ABOUT THE AUTHOR End of Page
iii LIST OF TABLES Table 1.1 Properties of commonly used SiC polytypes compared with Si. 2 Table 3.1 4H-SiC baseline process in the 75 mm LP horizontal cold-wall reactor. 58 Table 3.2 Experimental results of five samples distributed across a 50 mm wafer area. 91 Table 4.1 200 mm CVD reactor film pr operties versus growth rate. 129 Table E.1 Epi layer thickness as de termined by FTIR and cross-section SEM analysis. 208 Table E.2 HCl growth additive epi layer thickness summary from Chapter 4. 211
iv LIST OF FIGURES Figure 1.1 Stacking sequences of the three most common SiC polytypes. 4 Figure 1.2 Schematic drawing of a horiz ontal hot-wall CVD reactor growth zone showing gas inlet from the left and exit on the right. 9 Figure 1.3 Schematic drawing of a ver tical cold-wall CVD reactor: (1) double-walled chamber, (2) gas diffuser, (3) SiCwafer, (4) substrate holder, (5) RF coil, and (6) graphite susceptor. 10 Figure 1.4 Schematic drawing of a ver tical hot-wall, or chimney, reactor showing gas inlet from the bottom and exhausts out of the top. 11 Figure 1.5 Schematic drawing of the quasi-hot-wall reactor. 12 Figure 1.6 Schematic drawing of a verti cal radiant reactor showing gas inlet from bottom. 13 Figure 2.1 Thermodynamic predictions of chemical specie (a) mole fraction and (b) mass fraction as a func tion of temperature for the H2C3H8-SiH4 system. 29 Figure 2.2 Schematic diagram of mechan istic steps which occur during the CVD process which are (1) Gas inlet, (2) dissociation of reactants, (3) diffusion of reactants to the surface, (4) adsorption of reactants to the surface, (5) heterogeneous surface reaction, (6) desorption of by-products, (7 ) diffusion of by-products back into the bulk gas. 32 Figure 2.3 Sketch of boundary layer s howing an initially uniform gas velocity. 34 Figure 2.4 Schematic diagram of the growth process for CVD where Cb is the reactant concentration in the bulk gas, Cs is the reactant concentration at the surface, Jbs is the flux from the bulk gas to the surface, Js is the flux at the surface and is the boundary layer. 38
v Figure 2.5 Sketch of the 75 mm horiz ontal hot-wall reactor including graphite foam insulation, susceptor, graphite adapter, RF coil, gas inlet liner, ss head plate and coldwater jacket. 41 Figure 2.6 Sketches of the 75 mm hot-w all reactors (a) susceptor, placed inside graphite half of the foam insulation and (b) susceptor top and bottom portions. 44 Figure 2.7 Sketch of the 200 mm horizont al hot-wall reactor including head plate, foam insulation, adapter, RF coil, inlet and foam liners, drum (back plate). 48 Figure 2.8 Cross-section schematic dr awing of 200 mm horizontal hot-wall CVD reactor hot-zone, showing ga ses entering on the left and exiting the right. 49 Figure 2.9 SiC polycrystalline plates with (a) one 12 mm x 12 mm recess, (b) thirteen 10 mm x 12 mm r ecesses, and (c) one 60 mm diameter recess. 50 Figure 2.10 Gas velocity distributi on for the 200 mm hot-wall reactor showing (a) the entire reactor cross-section and (b) an enlarged area of the hot-zone with the 50 mm wafer area and susceptor indicated. 53 Figure 2.11 Velocity profile of the ma ximum gas velocity along the 200 mm hot-wall susceptor. 54 Figure 3.1 4H-SiC epitaxial film gr own in the 75 mm LP horizontal hotwall reactor for 15 min. 59 Figure 3.2 Sketch of insulating foam cross-sections for the 200 mm reactor with (a) initial circular geometry and (b) modified design to enable RF coupling of foam and susceptor. 60 Figure 3.3 Cross-section sketch of the hot-wall CVD reactor hot-zone including graphite foam, flat graphite susceptor, SiC polycrystalline plate (SiC polyplate), RF coils, and SiC substrate. 61 Figure 3.4 Plan-view SEM micrographs of a 4H-SiC film grown in the 200 mm reactor at a magnificati on of (a) 2500x and (b) 10000x. 63 Figure 3.5 Plan-view SEM micrographs of a 4H-SiC epitaxial film grown for 15 minutes at a magnificat ion of (a) 2500x and (b) 30000x. 65
vi Figure 3.6 Cross-section sketch of the hot-wall CVD reactor hot-zone including graphite foam, angled graphite suscep tor, graphite adapters, SiC polycrystalline plate (SiC polyplate), RF coils, and SiC substrate. 66 Figure 3.7 Sketch of the boundary layer a bove the susceptor with a (a) flat bottom and (b) tilted bottom. 67 Figure 3.8 Simulated gas velocity di stribution for the 200 mm hot-wall reactor with angled susceptor ceiling showing (a) the entire reactor cross-section and (b) an enlarged view of the hot-zone with the 50 mm wafer area and susceptor indicated by vertical dashed lines. 68 Figure 3.9 Velocity profile of the ma ximum gas velocity along the 200 mm hot-wall susceptor with the angled ceiling. 69 Figure 3.10 SEM micrographs of the first 4H-SiC epitaxial film grown in the hot-wall CVD reactor with angled ceiling. 71 Figure 3.11 Plan-view SEM micrographs showing poor film morphology caused from excess precursor flow rates at a magnification of (a) 500x and (b) 10000x. 72 Figure 3.12 (a) Photograph of the gas in let port of 200 mm hot-wall reactor showing diffuser plate. 73 Figure 3.13 4H-SiC epitaxial growth pr ocess schedule developed for the 200 mm LP horizontal hot-wall reac tor with angled ceiling. 74 Figure 3.14 4H-SiC epitaxial layer grow n with the baseline process for 45 minutes in the 200 mm reactor. 76 Figure 3.15 X-ray diffraction rocking curv e of the (0004) di ffraction peak for a 4H-SiC epitaxial film grown fo r 45 min at a temperature of ~ 1520C and a process pressure of 150 Torr. 77 Figure 3.16 Near bandedge LTPL spectrum (taken at 2K) of a 23 m thick 4H-SiC film grown at 31 m/h. 78 Figure 3.17 4H-SiC epitaxial growth rate dependence on silane flow rate in the 200 mm LP hot-wall reacto r with angled ceiling. 79
vii Figure 3.18 Plan-view SEM micrograph of 4H-SiC grown for 15 minutes with a silane flow rate of 12 sc cm and a Si/C ratio of 1.0 at a magnification of (a) 500x and (b) 5000x. 80 Figure 3.19 Plan-view micrographs of a 4H -SiC epitaxial sample grown at a growth rate of 39 m/h using a flow rate of SiH4 at 39 sccm. 81 Figure 3.20 4H-SiC epitaxial growth rate vs. process pressure in the 200 mm horizontal hot-wall reactor with angled ceiling. 82 Figure 3.21 Growth rate of 4H-SiC as a function of growth time in the 200 mm LP hot-wall CVD reactor with angled ceiling. 84 Figure 3.22 Dependence of surface r oughness of 4H-SiC films on film thickness as measured by AFM. 85 Figure 3.23 AFM scan of a 65 m thick 4H-SiC film grow n for 120 minutes at a rate of 32 m/h. 86 Figure 3.24 X-ray diffraction rocking curv e of the (0004) peak for the 65 m thick 4H-SiC epitaxial film. 87 Figure 3.25 Net carrier concentration of 4H-SiC epitaxial films as a function of Si/C ratio as measured with the C-V technique. 88 Figure 3.26 Picture of the polycrystalline plate used to simulate growth on a 50 mm wafer. 90 Figure 3.27 X-ray diffraction rocking cu rves of 4H-SiC epitaxial films grown for 30 minutes placed on the polycrystalline plate in (a) sample placement A and (b) sample placement B from Figure 3.26. 92 Figure 4.1 Sketch of the first pr oposed HCl mechanism where (a) homogeneous nucleation occurs wi th low etch rate of the Si clusters on the substrate surface and (b) high etch rate with the addition of HCl into the gas mixture resulting in a smooth surface and no Si clusters. 97 Figure 4.2 Sketch of second proposed HCl mechanism to suppress homogeneous nucleation. 99
viii Figure 4.3 Thermodynamic simulations of the mole fraction of gaseous species for the chemistry of H2-C3H8-SiH4 in the 200 mm CVD reactor (a) without the addition of HCl gas and (b) with the addition of HCl. Simulations were performed using the NASA CEA code. 100 Figure 4.4 HCl purge panel near the HCl gas bottle which was used to replace HCl with Ar in the gas li nes to help prevent corrosion. 102 Figure 4.5 4H-SiC epitaxial film grow n using the baseline process (from Section 3.2) with the addition of 0.5 sccm of HCl in the LP horizontal hot-wall reactor. 104 Figure 4.6 Plan-view micrographs of 4H-S iC epitaxial growth run in the 75 mm horizontal hot-wall reactor with the addition of 1.0 sccm of HCl to the baseline process (see Section 3.2 for details) showing step-bunching at a magnification of (a) 2500x and (b) 10000x. 105 Figure 4.7 SEM micrographs of 4H-S iC epitaxial films grown at QSiH4 and QC3H8 (3% in H2)) of 160 sccm with the addition of HCl. 107 Figure 4.8 Plan-view SEM micrographs at a magnification of (a) 100x and (b) 2500x of an epitaxial film grown for 15 minutes in the LP hot-wall reactor. 108 Figure 4.9 Growth rate of 4H -SiC as a function of silane flow rate for three separate experimental sets with the addition of HCl. 109 Figure 4.10 HCl flow rate as a function of silane flow rate for the experiments displayed in Figure 4.9. 110 Figure 4.11 Growth rate of 4H-SiC as a function of HCl flow rate in the 75 mm hot-wall CVD reactor. 111 Figure 4.12 Growth rate of Si versus mole fraction of SiCl4 in H2. 112 Figure 4.13 XRD 2 / diffraction scan of the sample shown in the SEM micrograph of Figure 4.7(c). 113 Figure 4.14 AFM surface scans of 4H-SiC epitaxial films grown for 15 minutes (a) without the addition of HCl and (b) with the addition of 1.0 sccm of HCl, resulting in surface roughness of ~ 0.30 and ~ 0.25 nm RMS, respectively. 115
ix Figure 4.15 X-ray rocking curve of the (004) diffraction peak for 4H-SiC epitaxial films. 116 Figure 4.16 X-ray rocking curve of the (0004) diffraction peak of a 4H-SiC epitaxial film grown at a QSiH4 (3% in H2) of 140 sccm with the addition of QHCl at 2.25 sccm. 117 Figure 4.17 Near bandedge LTPL spectra of 4H-SiC epitaxial layers (collected at 2K). 118 Figure 4.18 2K LTPL spectrum at longer wavelengths, compared to Figure 4.13, showing (a) the presence of the L1 defect line in the sample grown without HCl and (b) no L1 defect line present for 4H-SiC epitaxial film grown with the a ddition of 0.5 sccm of HCl. 119 Figure 4.19 Near bandedge LTPL spectra ta ken at 2K of 4H-SiC epitaxial films grown using HCl as a growth additive. 120 Figure 4.20 Plan-view SEM micrographs of a 4H-SiC epitaxial film grown in the 200 mm hot-wall reactor with 17 sccm of HCl additive at a magnification of (a) 2500x and (b) 10000x. 122 Figure 4.21 Plan-view SEM microgra phs of 4H-SiC grown with a QSiH4 of 45 sccm. 123 Figure 4.22 200 mm reactor growth rate as a function of SiH4 flow rate. 124 Figure 4.23 Thickness results for a 4H-SiC epitaxial film grown with the 200 mm reactor baseline process with 30 sccm of SiH4 and 7.5 sccm of HCl. 125 Figure 4.24 Net carrier concentration of 4H-SiC films grown with the addition of HCl at elevated growth rates. 127 Figure 4.25 AFM surface scans of 4H-SiC epitaxial films grown in the 200 mm LP hot-wall CVD reactor. 128 Figure 4.26 X-ray diffraction rocking curv e of the (0004) di ffraction peak for the 4H-SiC film with the high est growth rate (49 m/h), indicated in Figure 4.22. 130 Figure 5.1 Plan-view SEM micrographs of 4H-SiC grown with (a) HCl additive and (b) CH3Cl additive to the st andard chemistry of H2C3H8-SiH4. 136
x Figure 5.2 Thermodynamic simulations of the standard CVD growth chemistry with the addition of (a) HCl additive and (b) CH3Cl additive. 138 Figure 5.3 Cross-section schematic drawing of the 200 mm reactor susceptor bottom with a portion of the (a) bottom removed and (b) top removed to produce a backend thickness of 12 mm, (c) bottom removed additionally from (a) to produce a backend thickness of 5 mm. 141 Figure 5.4 Temperature profile measured at the bottom of the susceptor in the 200 mm horizontal hot-wall reactor. 144 Figure 5.5 Computed 2D temperature profile in the 200 mm horizontal hotwall CVD reactor. 145 Figure 5.6 Mass fraction distributions of the dominant species a) CH4, (b) SiH4, (c) SiH2, and (d) Si at a temperat ure of 1565 C for the 200 mm horizontal hot-wall reactor geometry. 147 Figure 5.7 Calculated (a) growth rate and temperature profiles along the reactor hot-zone for a maximum reactor temperature of 1565 C and (b) Si/C ratio across the reactor. 150 Figure A.1 Screen image of the CEA program display. 161 Figure A.2 CEA display screen for the TP problem. 162 Figure A.3 Display window for the Reactant tab. 164 Figure A.4 Display window for the Onl y tab which allows a specified species list to be created. 165 Figure A.5 Display screen for the Omit tab in the CEA program. 166 Figure A.6 Display window for the Output tab. 168 Figure B.1 Sketch of electron dist ribution from an electron beam bombardment onto a sample surface. 177 Figure B.2 Plan-view SEM images of (a) a sample with step bunching and (b) a sample with smooth morphology. 178
xi Figure B.3 Cross-section SEM microgra phs of a 4H-SiC cleaved sample edge showing the epitaxial layer and substrate for (a) top, (b) middle, and (c) bottom positions of the sample. 179 Figure B.4 AFM beam sample detection system. 180 Figure B.5 AFM surface scans of 4H-SiC epitaxial films grown in the 200 mm LP hot-wall CVD reactor. 181 Figure B.6 Demonstration of Braggs la w, showing the angle of incidence, the atomic plane spacing, and two X-rays being diffracted. 183 Figure B.7 X-ray diffractometer schematic for rocking curve measurements. 185 Figure B.8 X-ray diffraction rocking cu rves for 4H-SiC films of (a) 23 m thickness and (b) ~ 3.4 m thickness. 186 Figure B.9 Data plots of I-V and C-V curves used to extract the doping concentration of an n-type 4H-SiC epitaxial layer. 188 Figure B.10 Diagram of the FTIR spectroscopy set-up. I ( ) is the interference signal, BS is the beam splitter, M1 is the fixed mirror, and M2 is the movable mirror. 189 Figure B.11 Schematic drawing of the ra diation reflections during epitaxial thickness measurements using FTIR. 190 Figure B.12 Flow diagram for epitaxial film thickness determination using FTIR spectroscopy. 192 Figure B.13 Fourier transform of the reflectance spectrum showing a thickness of ~ 8 m. 193 Figure B.14 Diagram of photoluminescen ce arrangement which includes a source laser, a mirror, a monochromator, a detector, and a sample placed inside a cryostat. 194 Figure B.15 Recombination processes th at may create radiative photons which can be detected for PL analysis. 195 Figure B.16 Near bandedge LTPL spectrum of a 23 m thick 4H-SiC film. 197 Figure D.1 HCl purge panel near the HCl gas bottle which was used to replace HCl with Ar in the gas li nes to help prevent corrosion. 204
xii Figure E.1 FTIR spectra of sample USF-05-060A (a) before sample was cleaved and (b) after sample was cleaved. 209 Figure E.2 (a) Cross-section SEM micrograph of USF-05-060A cleaved edge and (b) plan-view SEM mi crograph of 10 m calibration sample. 210 Figure E.3 Sample ID USF-05-347 (a) FT IR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 8.12 m. 212 Figure E.4 Sample ID USF-05-348 (a) FT IR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 8.97 m. 212 Figure E.5 Sample ID USF-05-350 (a) FT IR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 10.4 m 213 Figure E.6 Sample ID USF-05-360 (a) FT IR spectrum and (b) cross-section SEM micrograph of middle edge showing thickness of 12.3 m. 213
xiii HIGH GROWTH RATE SIC CV D VIA HOT-WALL EPITAXY Rachael L. Myers-Ward ABSTRACT This dissertation research focused on the growth of 4H-SiC ep itaxial layers in low-pressure horizontal hot-wall chemical va por deposition (CVD) reactors. The goal of the research was to develop a growth pro cess that maximized the growth rate and produced films of smooth morphology. The epit axial growth of SiC was carried out in two different reactor sizes, a 75 mm r eactor and a 200 mm reactor. The maximum repeatable growth rate achieved was 3032 m/h in the 200 mm reactor using the standard chemistry of hy drogen-propane-silane (H2-C3H8-SiH4) at growth temperatures 1600 C, which is the highest grow th rate reported to date in a horizontal hot-wall reactor at these temperatures. This growth rate was ac hieved with a silane flow rate of 30 sccm. The process development and characterization of 4H-SiC epitaxial films grown using the standard chemistry are presented. There are many ways to increase the growth rate, such as changing the pressure, increasing the reactant flow rates, or increa sing the temperature. The method of choice for this dissertation work was to first increase the reactant flow rates, i.e. silane flow rate, and then to alter the growth chemistry by usi ng a growth additive. When the silane flow is increased, while maintaining a specific growth temperature, supersaturation of silicon
xiv may occur. When this happens, particulat es may form and deposit onto the sample surface during growth which degrades the f ilm morphology of the epitaxial layers. In order to overcome this severe limitation in the growth of SiC, hydrogen chloride (HCl) was added to the standard chemistry of H2-C3H8-SiH4 during growth when the SiH4 flow was increased beyond 30 sccm. With the additi on of HCl, the Si supersaturation was suppressed and the growth rate was increased from ~32 m/h to ~ 49 m/h by increasing the silane precursor up to 45 sccm, while maintaining the Si/C ratio of the standard chemistry process. The addition of HCl to the standard chemistry for growth of SiC films was pioneering work that has since been duplicated by several research groups.
xv PREFACE There are many people whom I would like to acknowledge for their utmost assistance during my dissertation work. Without them, I would not have succeeded. I would like to first and foremost give thanks and gratitude to my a dvisor Dr. Stephen E. Saddow for his guidance and loyalty to my re search. He has been my mentor throughout my post graduate studies by being an inspir ation and has encouraged me throughout the years. Without his guidance and constant suppo rt, I would not have be en able to achieve all that I have. He has taught me a treme ndous amount and has given me much support. Dr. Saddow has provided me the opportunity to study abroad and attend many conferences throughout the country and in Europe. For all of these things, I am truly grateful. I would also like to thank Dr. A. M. Hoff, Dr. J. T. Wolan, and Dr. G. Centeno for being part of my committee. They have s upported me and given me guidance throughout my research. I wish to give gratitude to Dr O. Kordina, who has helped and advised me with my dissertation work. Much appreciation goes to Dr. Y. Shishkin, who has provided me with insight about my re search and spent a considerab le amount of time helping to refine my dissertation. Next I wish to tha nk Professor W. J. Choyke for providing LTPL analysis on many of the epitaxial film s grown during this dissertation work.
xvi Next, I wish to thank Dr. S. P. Rao, fo r being my best friend and colleague. She has always been there for me and has always provided help and encouragement. I would also like to thank T. Fawcett for his assist ance with simulations. I extend my appreciation to the SiC group, especially I. Haselbarth, C. Frewin, M. Reyes, B. Grayson, C. Coletti, J. Walker and S. Harvey, who have been of grea t assistance to me during this research. My appreciation also goes to R. Ev erly and R. Tufts of the NNRC. I would also like to thank Dr. U. Starke of the Max-Planck Institute, Stuttgart, Germany, for allowing me to visit and perform research in his laboratory. My gra titude goes to Dr. C. Wood of the Office of Naval Research (ONR) for his support of this research. Th is work has been supported under Dr. Wood by the Defense Univers ity Research Initiative on Nanotechnology (DURINT) program, Grant N00014-0110715. Finally, I would like to thank my parents, aunt, sister, brothers and most of all, my husband. They have been very patient w ith me and have given me support and encouragement throughout my education. Even though I have not been able to see them much over the past few months, they have understood and been there for me when I needed them. They are very loving and caring, and always have wanted the best for me. For this, I am grateful, more than words can express.
1 CHAPTER 1 INTRODUCTION 1.1 Silicon Carbide Overview Silicon carbide (SiC) is a semiconductor ma terial which is being developed for high-temperature, microwave and high-power switching device ap plications. Unlike silicon-based electron ics, which can only withstand temp eratures up to 250 C , SiCbased devices can endure elevated temperat ures up to 650 C due to its excellent chemical and physical stability . Some of the physical and electrical properties that make this semiconductor superior to Si and GaAs are a large energy bandgap, high critical electric field, high saturated electron drift veloc ity, and high thermal conductivity . From Table 1.1 it can be seen that the material properties (exc luding mobility) of SiC are outstanding compared to those of Si. High-temperature device applications are no t generally possible using Si due to its narrow bandgap of 1.12 eV. At elevated temper atures, the thermal ge neration of electronhole pairs exceeds the number of dopant-provi ded free carriers, causing the device to fail . The thermal generation of electron-hole pairs is much lower in SiC than in Si due to its wider energy bandgap (3.2 eV at room te mperature for the 4H-SiC polytype). The difference in the intrinsic carrier concentration, ni, between Si and SiC, is 19 orders of magnitude (i.e., 1019), which enables higher temperature applications using SiC. As seen
2 in Table 1.1, the bandgap values for SiC are at least twice that of Si with the resulting values of intrinsic carrier concentration be ing several orders of magnitude lower. The intrinsic carrier concentration is proportional to minus the exponent of the bandgap value. Since the bandgap of SiC is higher than that of Si, ni is much lower for SiC than Si. For high-voltage, high-power devices, the semiconductor material must be able to withstand high electric fields (> 0.3 MV/cm ). Since SiC has a wide bandgap, the impact ionization energy, the impact energy which creates additional electrons into the conduction band, is higher for this material than it is for Si . Therefore, the electric field may be higher for SiC before impact-i onization driven breakdown occurs, caused by avalanche multiplication of ionized carriers [5 ]. SiC devices can be made to have ten times higher breakdown voltage compared to Si with the same depletion width , or the depletion region of SiC can be much thinner than that of Si with the same breakdown voltage. Table 1.1 Properties of commonly used SiC polytypes compared with Si [1,3,6]. Property 4H-SiC 6H-SiC 3C-SiC Si Energy bandgap at 300K (eV) 3.20 3.00 2.29 1.12 Intrinsic Carrier Concentration at 300K (cm-3) 5x10-9 1.6x10-6 1.5x10-1 1x1010 Critical breakdown electric field (MV/cm) 2.2 2.5 2.12 0.25 Saturated electron drift velocity (x 107 cm/s) 2.0 2.0 2.5 1.0 Thermal conductivity (W/cm-K) 4.9 4.9 3.2 1.5 Electron mobility (cm2/V-s) 1000 600 800 1450 Hole mobility (cm2/V-s) 115 100 40 470
3 At low electric fields (E < 10 kV/cm), the drift velocity (the velocity a particle reaches due to an electric field) is proportiona l to the applied electric field. However, at fields greater than 10 kV/cm, the velocity incr eases sub-linearly with the field. At fields substantially higher than 10 kV /cm, the drift velocity satu rates and becomes independent of electric field . The saturated drift veloc ity in SiC is much higher than that in Si, making SiC a candidate material for high-fr equency devices. With similar breakdown voltages for Si and SiC devices, the depletion width can be thi nner in SiC, thus the device will be much faster using SiC because the series resistance is lower, thus a faster RC time constant can be realized . The dissipation of heat in a high-power devi ce is critical to the reliability of that device . When the material is heate d, the physical properties often change. For example, the carrier mobility decreases with increasing temperature  and the intrinsic carrier concentration increases, which may cause the device to fail. Since the thermal conductivity of SiC is greater th an three times that of Si, th e heat can flow more readily from hot spots in the SiC-based device to the package  compared to Si and other semiconductors. SiC possesses the ability to crystallize in to different structures which are called polytypes. There are over 200 different know n polytypes of SiC which have been observed to date . The differentiation of these polytypes is based on the stacking sequence of the tetrahedrally bonded Si-C bilayers in the vertical (c-a xis) direction. There are three general categories in which the polytypes fall under : cubic (C), hexagonal (H), and rhombohedral (R). There is only one purely cubic struct ure which is labeled 3C-SiC. The only pure hexagonal structure is 2H-S iC. The rest of the polytypes have a
4 combination of cubic and hexagonal bonding [6 ]. The most commonly studied polytypes of SiC are 4H-, 6H-, and 3C-SiC. The desi gnation of each polytype is by a number and a letter. The number represents the number of la yers required to complete the unit cell and the letter designates the type of the structure . Figure 1.1 shows the stacking sequence of the three most common polytypes. Figure 1.1 Stacking sequences of the thr ee most common SiC polytypes. The larger atoms are Si, while the smaller atoms are C. Symbols k and h denot e lattice sites with cubic and hexagonal symmetry, respectively . The A, B, and C labels in Figure 1.1 denote the position of the carbon atoms within the bilayers of the hexagonal structur e . The designation of k and h are the type of crystal symmetry at that point, which is either cubic (k) or hexagonal (h), respectively. As seen in Figure 1.1, 4H-SiC has a stacking sequence of ABCB, or 4 layers, therefore the designation is 4H. This structure has an equal number of cubic and hexagonal lattice sites. The 6H -SiC structure has 6 stacking layers before the sequence repeats ABCACB, and, finally, 3C-SiC is a continuation of the ABC stacking sequence which has purely cubic symmetry. Due to differe nces in stacking sequence, the electrical, mechanical and optical properties vary for each polytype of SiC, as shown in Table 1.1.
5 3C-SiC has a high electron mobility which makes it a promising material for high-frequency devices. However, there is still not a commercially available growth process that produces bulk substrates for th is polytype . The growth of 3C-SiC typically takes place heteroepitaxially on Si s ubstrates. The lattice and thermal expansion coefficient mismatch between these two mate rials are ~ 22% and 8%, respectively . This often creates defects at the SiC/Si in terface which propagate through the epitaxial layer , thus limiting the quality of the f ilms. Therefore, this polytype was not the material chosen for this dissertation resear ch, although a considerable amount of 3C-SiC growth has and is being conduc ted in the USF SiC group [10,11]. The hexagonal close-packed structure is a main reason for the high stability of the hexagonal SiC polytypes. The 4H -SiC polytype has the highe st stability due to the alternating cubic and hexagonal layers  6H-SiC has a low, anisotropic electron mobility, while 4H-SiC has a much higher elec tron mobility and is less anisotropic (i.e. less directionally dependent) . Thus 4H-S iC is, at present, the most commonly used polytype for electronic devices . With all these factors outlined, the research for this dissertation was focused on the grow th of 4H-SiC epitaxial layers. 1.2 Motivation for Dissertation Research Growing epitaxial films is one of the key technological steps in the fabrication of SiC device structures . In order to pr oduce the desired electrical specifications for device applications, such as doping, epitax ial layers are grown on bulk SiC substrates. These epitaxial layers are required to have high crystalline quality, low surface
6 roughness, low point and extended defect de nsity, while exhibiti ng a high degree of doping and thickness uniformity . A limita tion for the success of SiC as a device material is the high extended defect density wi thin the epitaxial films. Defects reduce the minority carrier lifetime, which is important for high voltage device performance. The crystal defects tend to decrease the carrier life time of the films, which in turn reduces the performance of the devices . A few SiC-based devices ha ve recently emerged as the leading candidates for power electronic applications . Some of these devices are Schottky diodes and metal semiconductor field effect transistors (M ESFETs). The MESFET device requires thin epitaxial layers with different doping concen trations, both nand p-type, with abrupt doping profiles . Other high power devices, PiN diodes (where a ptype region and an n-type region are separated by an intrinsic i region), require low-doped drift layers that are approximately 40-50 m thick (or even up to 100 m thick ) with a low n-type doping concentration of 1 x 1014 cm-3 . It is gene rally understood that epitaxial layers of SiC greater than 80 m thick are needed for electrical switching devices in order to outperform Si-based devices . In order to produce thick (> 40 m) uniform epitaxial films, high growth rates are needed for economical reasons. For example, if the growth rate is only 5 m/h and a film thickness of 100 m is desired, it will take a pproximately 20 hours to grow the film. This results in very high production costs, thereby increasing the co st of the actual device. It would obviously be more advantageous to grow a film for 2 hours at 50 m/h. The primary goal of growing at such high speed s then becomes maintaining a high quality epitaxial deposition in a shorter amount of time.
7 The main concern with high growth rates is being able to maintain the electrical and structural quality of the films. Many of the methods deve loped to date, in attempt to achieve this purpose, take place at elevated temperatures, which limit the life of the reactor parts and can create defects in the grown films. The work conducted during this research was focused on using growth temp eratures around 1600 C with the goal of growing thick, high-quali ty 4H-SiC films. 1.3 SiC CVD Epitaxy Background The most common technique used to gr ow epitaxial SiC is chemical vapor deposition (CVD) . The typi cal chemistry used to grow SiC CVD films is hydrogen, propane, and silane, or H2-C3H8-SiH4. In this mixture, hydrogen serves as the carrier gas and silane and propane are the precursors fo r growth. Growth of SiC to produce devicequality thick films has been the focus of many groups over the years. Nakazawa, et al  grew 4H-SiC in a hor izontal hot-wall CVD reactor. For reference, a schematic drawing of a horizontal hot-wal l reactor is shown in Figure 1.2. A detailed description of this type of reactor is pres ented in Chapter 2, Section 2.2.1. The growth rate Nakazawa  produced was 14 m/h, and f ilm surfaces grown using this rate displayed growth pits on the surface. The growth rate was then reduc ed to 6-7 m/h and the films were smooth without such defects. A 50 m thick film was grown with this process. An X-ray diffraction (XRD) rocking curve of the sample proved the film to be of very high quality with a full-width at half maximum (FWHM) of the (0004) diffracti on peak of 9 arcsec. Low temperature photoluminescence ( LTPL) was also performed and no L1 defect line at
8 2.902 eV or Ti peak (Co at 2.790 eV) were observed. The background doping level was 1-2 x 1013 cm-3. Clearly, this process produced high quality films, but at such a low growth rate as to render the process econom ically infeasible to produce thick films. Kimoto, et al  also reported growth in a horiz ontal hot-wall reactor with a growth rate of 5 m/h. A 50 m thick 4H-SiC epita xial film of smooth morphology was grown. The unintentionally doped layers had a low carrier concentration of 2 x 1013 cm-3. Again, at a rate of 5 m/h, the growth rate was too low to economically produce thick films. Zhang, et al  recently documented that growth of 4H-SiC epitaxial layers in a horizontal hot-wall reactor could be carried out with a growth rate of 20 m/h, however, repeatable results were only achievable at 15 m/h. A 50 m film was grown at 8.3 m/h, resulting in a smooth morphologica l film surface as determined by Nomarski microscopy. At thicknesses of 50 m and a bove, the films displayed step-bunching. A film was also grown at 15 m/h, to pr oduce a thickness of 120 m. The thickness uniformity was 5.2% with a p-type background doping of 7 x 1014 cm-3. Clearly, thick films are attainable with horizontal hot-wall reactors; however, the growth rates need to be improved along with the surface mo rphology of the thick grown films.
9 A vertical cold-wall reactor has been us ed to grow 4H-SiC films of 20 40 m thickness at a growth rate of 5-6 m/h by Rupp, et al . In this stai nless steel vertical reactor, the substrate was plac ed on a holder and the gases we re injected from the top and flow down and then laterally over the subs trate. The reactor was water cooled. The susceptor was heated using a radio frequency so urce. A schematic drawing of this type of reactor is shown in Figure 1.3. For the film s that were grown beyond a thickness of 30 m, any substrate surface damage became very pronounced on the surface of the epitaxial layers. To reduce this effect, the substrates were re-pol ished and dry-etched prior to growth. The resulting thick films did not have any severe surf ace defects. Again, the produced films were of reasonable surf ace morphology, but the growth rate was too low to achieve films of >40 m w ithin a reasonable time frame. Figure 1.2 Schematic drawing of a horizont al hot-wall CVD reactor growth zone showing gas inlet from the le ft and exit on the right. Dr awing includes RF coil, SiCcoated graphite susceptor, graphite felt insulation, and SiC substrate .
10 Figure 1.3 Schematic drawing of a vertical cold-wall CVD reactor : (1) double-walled chamber, (2) gas diffuser, (3) SiCwafer, (4) substrate holder, (5) RF coil, and (6) graphite susceptor. The gases enter from the top and exit out th e bottom sides . Another CVD technique used to grow SiC is the vertical hot-wall, or chimney, reactor [23-25]. Figure 1.4 shows a schematic drawing of such a reactor. It is similar to the cold-wall vertical reactor however, the samples are pla ced on the side walls of the growth zone and there is no c ooling water jacket. The gases flow vertically up the reactor parallel to the sample surface. Growth was car ried out in a vertical hot-wall reactor by Zhang, et al  to produce 30 120 m thick 4H-S iC epitaxial layers. The temperature at which these films were grown ranged between 1650 and 1850 C. Growth of a 74 m film took place at 1710 C at a rate of 25 m/h, resulting in a step-bunched surface. In order to reduce or eliminate the macros copic step-bunching, the temperature was increased to 1850 C for the growth conditions used. Epitaxial growth in a vertical hotwall reactor have also been reported by Ellison, et al . N-type 4H-SiC epitaxial
11 layers with thicknesses of over 100 m at a gr owth temperature of ~1750 C were grown. A thickness uniformity was 5% and a growth rate of 23 m/h was achieved. The lowest doping concentration for the growth process was 6 x 1013 cm-3. Masahara, et al  was able to accomplish 4H-SiC epitaxial growth at rates up to 70 m/h at a growth temperature of 1800 C in a vertical-type, quasi -hot-wall, reactor. A schematic drawing of this quasi-hot-wall r eactor is shown in Figure 1.5. The gas nozzle and susceptor may be independently heated by RF power. The gas nozzle heats the substrate surface by radiative heating and it also controls the temperature of the reactants. There are radiation reflectors in this reacto r instead of graphite foam insulation. The surface morphology and structural quality of f ilms grown at two different speeds (26 m/h and 70 m/h) were compared. The thic knesses of the films we re 26 m and 31 m, respectively. The surface morphology of the 31 m thick film was rougher than the 26 m thick. Even so, an X-ray rocking curve of the 31 m thick film had a FWHM value of Figure 1.4 Schematic drawing of a vertical hot-wall, or chimney, reactor showing gas inlet from the bottom and exhausts out of the top. Drawing includes quartz tube, RF coils, susceptor, insulation, and SiC substrate .
12 13.5 arcsec indicating good crystal linity. As described in the lit erature, the vertical hotwall reactor is a technique that produces thick films much faster than the horizontal hotwall system. However, the films are grown at high process temperat ures which increases the wear and tear on the reactor. Figure 1.5 Schematic drawing of the quasi-hot -wall reactor. The reactor contains an outer water-cooled quartz tube as shown . Vertical radiant reactors have also been developed to increase the growth rate to permit the production of thick films . This type of reactor is very similar to the vertical, or chimney, reactor except that th e susceptor is V-shaped with the samples placed on the side of the V (see Figure 1.6). In this geometry, the gases flow upward over the substrate. Induction heating is applied dir ectly to the hot-wall while the susceptor and substrate are heated by radiation from the hot -wall. The temperature of the substrate is higher than the susceptor for th is type of reactor. Tsuchida, et al  was able to achieve growth rates up to 18 m/h in a vertical radiant reactor at growth temperatures between 1530 and 1560 C. Thick 4H-SiC epitaxial fi lms of 150 m were grown with smooth surfaces. The net carrier concentration of one of the 150 m thick films was 1 x 1014
13 cm-3. A film was also grown by Tsuchida, et al  with a growth rate of 15 m/h, resulting in a thickness of 250 m. The surf ace roughness increased slightly for films of higher thicknesses. An X-ray rocking curve measurement of a 110 m film was taken, and the FWHM was 6.0 arcsec . A 4H-SiC epitaxial layer was grown by Tsuchida at 16 m/h resulting in a thickness of 246 m. The Nomarski contrast of the film showed no features except for occasional growth pits. As pointed out, the vertical radiant growth technique shows promising resu lts for epitaxial growth, how ever, the growth rate is relatively low compared to the vertical reactor. Figure 1.6 Schematic drawing of a vertical radiant reactor showing gas inlet from bottom. Drawing includes RF coils, sus ceptor, hot-wall, and substrate . Overall, the processes described in this section resulted in epitaxial layers that were thick and of good quality. However, us ing the horizontal hot-w all reactor or the vertical cold-wall reactor to produce grow th rates around 5 15 m/h, the process time would be very long to achieve films up to and beyond 50 m in thickness, which are needed for power devices. The vertical radian t reactor is able to produce films within reasonable amounts of time, however, the load ing of the samples is more difficult than for a horizontal reactor. Finally, the vertical reactors are capable of producing such film
14 thicknesses, but at high process temperatures. This limits the lifetime of the reactor parts and increases the probability of particulate contamination in the films for long growth times. This dissertation focuses on the fast growth of 4H-SiC ep itaxial films in a horizontal hot-wall CVD reactor. This type of reactor allows for easy loading/unloading of samples and, more importantly, operates at a growth temperature lower than those possible in vertical reactors. 1.4 Organization of Thesis The growth of 4H-SiC at elevated growth rates using a horizontal hot-wall chemical vapor deposition (CVD) reactor will be demonstrated. There are four main chapters. Chapter 2 discusses the general theo ry of CVD with a look into the kinetics, fluid flow and thermodynamics of SiC C VD. Two CVD reactors were used in the dissertation research and they will be descri bed in this chapter. The development of a standard H2-C3H8-SiH4 chemistry 4H-SiC epitaxial proce ss to achieve a high growth rate of 32 m/h will then be discussed in Chapter 3. The addition of chlorine to the process via hydrogen chloride (HCl) will be describe d in Chapter 4. It will be seen that an increase of growth rate from 32 to 49 m/h was achieved using HCl as a growth additive. Chapter 5 will provide a summary of the dissert ation research as well as suggestions for future work.
15 1.4.1 Publications and Presentations The following are a list of publications a nd presentations resulting from research conducted at USF in the area of SiC CVD technology. The items related to the research presented in this dissertation are annotated with (PhD) afterwards: Publications (3 Journal, 6 conference papers) R. L. Myers, Y. Shishkin, O. Kordina, and S. E. Saddow, "High growth rates (>30m/h) of 4H-SiC epitaxial layers using a horizontal hot-wall CVD reactor", Journal of Crystal Growth 285 (2005) 483-486. (PhD) R. L. Myers, Y. Shishkin, O. Kordina, I. Haselbarth, and S. E. Saddow, "High Epitaxial Growth Rate of 4H-SiC using Horizontal Hot-Wall CVD", Materials Science Forum, in press (2005). (PhD) R. Myers, O. Kordina, Z. Shishkin, S. Rao, R. Everly, and S.E. Saddow, "Increased growth rate in a SiC CVD reactor using HCl as a growth additive", Materials Science Forum 483-485 (2005) 73-76. (PhD) S. E. Saddow, R. L. Myers and S. P. Rao, Use of a CVD Reactor for Advanced SiC Processing, Proc. of International Sympos ium on Advanced Materials and Processing ISAMAP2K4, Kharagpur, India (2004) 1002-1009. (PhD) R. L. Myers, K. D. Hobart, M. Twigg, S. Rao, M. Fatemi, F. J. Kub, and S. E. Saddow, Structural characterization of 3C-SiC films grown on Si la yers wafer bonded to polycrystalline SiC substrates Materials Research Society Symposium. Proceedings 815 (2004) 145-148.
16 R. L. Myers, S.E. Saddow, S. Rao, K.D. Hobart, M. Fatemi, F.J. Kub, Development of 3C-SiC SOI structures using Si on polycrystalline SiC wafer bonded substrates, Materials Science Forum Vols. 457-460 (2004) 1511-1514. T. Fawcett, J.T. Wolan, R.L. Myers, J. Walker, and S.E. Saddow, Hydrogen gas sensors using 3C-SiC/Si epitaxial layers, Materials Science Forum Vols. 457-460 (2004) 14991502. T. J. Fawcett, J. T. Wolan, R. L. Myers, J. Walker, and S. E. Saddow, Wide-range (0.33%100%) 3CSiC resistive hydrogen gas sensor development, Applied Physics Letters Vol. 85 (3) (2004) 416418. S. Dogan, A. Teke, D. Huang, H. Morkoc, C. B. Roberts, J. Parish, B. Ganguly, M. Smith, R. L. Myers and S. E. Saddow, 4HSiC photoconductive switching devices for use in high-power applications , Applied Physics Letters Vol. 82 (18) (2003) 3107-3109. Presentations R. L. Myers, Y. Shishkin, O. Kordina, I. Haselbarth, and S. E. Saddow, High Epitaxial Growth Rate of 4H-SiC using Horizontal Hot-Wall CVD, International Conference on Silicon Carbide and Related Materials, Pittsburgh, PA, Sept ember 2005. (PhD) R. Myers, O. Kordina, Z. Shishkin, S. Rao, R. Everly, and S.E. Saddow, "Increased growth rate in a SiC CVD reactor using HCl as a growth additive", European Conference on Silicon Carbide and Related Materials, Bologna, Italy, September 2004. (PhD) R. L. Myers, K. D. Hobart, M. Twigg, S. Rao, M. Fatemi, F. J. Kub, and S. E. Saddow, Structural characterization of 3C-SiC films grown on Si la yers wafer bonded to polycrystalline SiC substrates, Materials Research Society spring meeting San Francisco, CA April 2004.
17 R.L. Myers, S.E. Saddow, S. Rao, K.D. Hoba rt, M. Fatemi, and F.J. Kub, Development of 3C-SiC SOI structures using Si on polycrystalline SiC wafer bonded substrates, International Conference on Silicon Carbide and Related Materials Lyon, France, October 2003.
18 CHAPTER 2 HOT-WALL CHEMICAL VAPOR DEPOSITION In Chapter 1, SiC was introduced as a candidate semiconductor material for highfrequency, high-temperature, and high-power electronic device applications. One of the key factors in developing device s for such applications is th e growth of epitaxial layers on bulk SiC substrates. Epitaxy is the arrangeme nt of atoms on a crystalline substrate, which results in a deposited film with the same crystallographic orientation as that of the substrate [28,29]. There are two different t ypes of epitaxy, namely heteroand homoepitaxy. Heteroepitaxy is the term used for th e growth of an epitaxial layer on a crystal that is not of the same material (e.g. 3C-SiC on Si). Homoepitaxy is the term used for the growth of an epitaxial layer th at is of the same material as the substrate (e.g. 4H-SiC on 4H-SiC). The focus of this dissertation work is the homoepitaxial growth of 4H-SiC. The goal of epitaxial growth is to produce films of desired physical specifications, i.e. film thickness, electrical polarity and conduc tivity. The doping of the layers may be performed during the growth by the incorpor ation of dopant atoms provided by reactant gases, or after growth by ion implantation. For the later process, the crystal is bombarded with dopant ions, which damages the crysta l structure. Implant annealing is then performed to both activate th e implanted ions and repair the crystal damage . However, the highest quality doped films ar e always realized during film growth.
19 The focus of this dissertation research is the growth of device-quality homoepitaxial layers with a specific conductivity, normally low-doped n-type. 2.1 Chemical Vapor Deposition Chemical vapor deposition is the formation of a stable film on a crystal surface due to a chemical reaction from gaseous species . While this is a deposition technique, it is commonly referred to as a growth process. Th e basic process (i.e. procedure followed to achieve growth) of CVD begins by flowing precursors (or reactants), via a carrier gas, into a reaction chamber. Th e gases are heated so that chemical reactions can take place above and at the substrate surface, which results in the deposition of a film. The by-pr oducts of the reactions, along w ith any un-reacted species, are exhausted out of the chamber. The ga s phase reactions are activated by thermal energy supplied by either resist ive heating, a laser source, th e creation of a plasma, or radio frequency (RF) induction heating of the re actants . The latter RF heating, is the source that is most commonly used in Si C CVD and was used during this research. There are many variables that impact the deposition rate and film properties, such as the reactants selected, the temperature of th e reactor, the type of substrate, the ratio of reactants to each other, the gas flow rates, the process pressure, the geometry of the chamber, and, finally, the substrate surface preparation . With elevated process temperatures, the deposition rate (or growth rate) is typically increased. In addition, certain reactions can take place which might not occur at lower temperatures, and in-situ etching of the substrate simultaneously takes place . The CVD reactor may be
20 operated at atmospheric pressure (AP), or low pressure (LP) and, therefore, are commonly referred to as APCVD and LPCVD, respectively. SiC LPCVD is typically carried out in the pressure range of 10 150 Torr. Growth at low pressures results in improved atomic step coverage, with th e deposition rate gove rned by the surface reactions. As a consequence, the defect density is typically reduced . The reactor geometry is an important parameter in the growth of epitaxial films. There are many types of reactors, as described in Chapter 1, and the choice of reactor is governed by the requirements of the epitaxial films. The two main reactor geometries are vertical and horizontal, both of which can be further classified as hot-wall, cold-wall, or intermediate configurations. The cold-wall de sign has a reaction tube that is surrounded by a water-cooled jacket which sets the temper ature of the reaction tube to approximately 100 C. The hottest component of the cold-wall reactor is the suscepto r which is the part that the substrate is placed on and heated during growth. The remaining areas of the reactor are kept cold , hence the name cold-wall re actor. A large temperature gradient from the surface of the susceptor to the walls of the reactor is produced. In this configuration, there is less deposition on the wa lls of the reaction tube due to the lower temperature of these surfaces. In the hot-wall reactor, there is a much lower temperature gradient because the side-walls, bottom and top of the susceptor are all heated to approximately the same temperature. This design is basically an isothermal furnace. This dissertation research was performed using a horizontal, hot-wall, low-pressure CVD reactor since this combination of parameters is most likely to result in high-quality epitaxial layers grown at fast growth speeds.
21 During CVD, a carrier gas transports the reactant gases to the surface of the substrate. The carrier gas also aids the reactio ns to take place. Typical carrier gases are argon (Ar), helium (He), and hydrogen (H2); however, H2 is the most commonly used, especially for SiC CVD . There are a numbe r of precursors that have been used to grow SiC epitaxial films  Some of the silicon based precursors include silane (SiH4), disilane (Si2H6), and silicon te trachloride (SiCl4). Some of the carbon source precursors that have been reported are acetylene (C2H2), propane (C3H8), methyl chloride (CH3Cl), methane (CH4), and carbon tetrachloride (CCl4). Precursors containing both Si and C have also been investigated such as hexamethyldisilane (C6H18Si2 or HMDS) methyltrichlorosilane (CH3SiCl3 or MTS) and dimethyldichlorosilane ((CH3)2SiCl2). Propane and silane, C3H8 and SiH4, are the most commonly used precursors for SiC growth and are the precursors used for the growth of 4H-SiC epitaxial films in this dissertation research. As stated in the beginning of this chapte r, the epitaxial films are grown to have specific doping concentrations depending upon the device application. The desired doping density may be achieved during the grow th process by contro lling the amount of precursors along with the addition of dopant gases. The dopants for SiC are aluminum (Al) and boron (B) for p-type conductivity and nitrogen (N) and phosphorus (P) for ntype conductivity. Typical dopant pr ecursors are trimethylaluminum ((CH3)3Al or TMA), diborane (B2H6), nitrogen gas (N2), and phosphine gas (PH3) . Al is the most common p-type dopant because it has a low acceptor io nization energy, while N is the preferred ntype dopant over P because it ha s a higher probability to be in corporated into the lattice . These dopants are incorporated into th e Si or C lattice si tes as substitutional
22 impurities. Al typically competes with Si and N competes with C for lattice sites (positions) during growth . An inlet gas Si/C ratio is typically specified when growing SiC films. This ratio is the amount of moles of Si atoms to the amount of moles of C atoms in the gas mixture. For a low Si/C ratio, the incorporation of Al into the lattice is enhanced as its relative concentration in the gas mixture increases, resulting in a p-type layer. Likewise, for a high Si/C ratio, there is an ample supply of Si atoms. This enables N to occupy the C lattice sites which produces an n-type epitaxial layer. With the ability to control the doping of the ep itaxial films, CVD is an excellent technique to dope various layers for speci fic device applications. For control of the CVD process, knowle dge of the gas flow dynamics, equilibrium thermodynamics, mass transport, and chemical kinetics is needed. Thermodynamics may determine the feasibility of th e reaction to occur while kinetics determines the rate of the reaction . These will be discussed separately in the following sections. 2.2 CVD Thermodynamics Thermodynamic calculations are often used to provide insight into the CVD process [31-33]. The analysis can predict the feasibility of a CVD process chemistry and what gaseous species may be produced during the process. In order to obtain this information, data is needed such as the temper ature and pressure of the system, as well as the reactant flow rates (w hich will be denoted as Qi, where i denotes the ith specie). The thermodynamic equilibrium state of a CVD system can be calculated for the identification of the condensed phases, par tial pressures of all the gaseous species,
23 possible equilibrium deposition rates, and the theoretical reaction efficiency. It must be noted that the thermodynamic calculations are us ed as a guide for the CVD process as the calculations are performed under equilibrium conditions, while the actual CVD process is clearly not in equilibrium . There are two ways to calculate thermodynamic equilibrium in CVD: the nonlinear equations method using partial pressure s of the species (see section 2.2.1) and the Gibbs free energy minimization method (see se ction 2.2.2) . Both methods use the Gibbs free energy to evaluate the thermodyna mics of CVD. The Gibbs free energy, G is a thermodynamic state function which can be used to dete rmine whether a reaction is likely to occur. The change in Gibbs fr ee energy is the maximum amount of work obtainable for a reaction , given temperature and pressure remain constant. If the associated change in G G is a negative value, the react ion will occur spontaneously. If G is positive, then the reaction is not fa vorable (nonspontane ous). The system has reached equilibrium when G is zero. If several reactions occur simultaneously, the most prominent reaction would be the one resulti ng in the lowest Gibbs free energy because it would be the most stable reaction . Howe ver, this prediction is not always correct since the equilibrium thermodynamics does not consider the reaction kinetics, so care must be used when drawing conclusions from these calculations only . 2.2.1 Non-linear Equation Method The non-linear equation method uses independe nt equations to express the partial pressures of each gaseous specie in the system. These may be expressed as the
24 equilibrium constants of each reac tion in terms of partial pressure of each species, or as a function of the partial pressures. The total pr essure of the system is found by the addition of all partial pressures . This met hod uses the law of mass action to find the equilibrium constant of each reaction . As an example, consider the following reaction, D v C v B v A vd c b a (2.1) where i is the stoichiometric coefficient for ea ch specie. The law of mass action can be described in terms of partial pressures, pi, by : T v b v a v d v cK p p p pb a d c (2.2) where KT is an equilibrium constant of the reaction and is a f unction of temperature. The Gibbs free energy is related to the equilibrium constant by : TK RT G log (2.3) where R is the universal gas constant and T is the temperature. The standard state free energy of formation, G is the change of free energy in the system obtained at the reactants a nd products standard states (i.e. when T = 298 K and P = 101 kPa) . This may be used instead of G and is found by adding the G of each specie for each reaction . For example, the reaction in equa tion (2.1) would have a total G as follows: B A C DG G G G G (2.4) Note the G for elements in their reference state is conventionally set equal to zero , i.e. GH2=0. The change in Gibbs free energy of a reaction is given by: R R RS T H G (2.5)
25 where the subscript R stands for the reaction, HR (the change in enthalpy), is the change of energy in the form of heat, SR (the change in entropy), is disorder in the system and T is the temperature. Equation (2.5) may also be written in the standard states as : R R RS T H G (2.6) The enthalpy of the reaction may be calcul ated from the enthalpy of formations Hf , which is the heat absorbed or released in a reaction . Hf can be expressed by: reactants productsf f RH H H (2.7) The standard heat of formation (i.e. Hf at 25C) for elements is equal to zero by convention. The standard entropy, SR may be found using a similar equation as equation (2.7), however, one must replace H with S Using the non-linear method, the equilibri um constant is found using equation (2.3). The Gibbs free energy may be found using equations (2.4) (2.7 ), while the values of enthalpy and entropy are found from data ta bles . Then, the partial pressures of each specie can be calculated by solving the equilibrium constant equation for all reactions, equation (2.2). This would leave n unknowns, with (n-1) equations. To solve for all the partial pressures in the system, another equation would be needed to make n equations with n unknowns, which is: the tota l pressure is equal to all the partial pressures of the species. The equilibrium constant may be used to determine whether or not a reaction is feasible. The higher the equilibrium constant the more thermodynami cally feasible the reaction is . Using this method, one must predict all of the appropriate reactions and
26 which species are significant and their phase for the reaction . Clearly, it becomes quite tedious to calculate the equilibrium sp ecies concentrations when there are numerous reactions to evaluate; such is the case for SiC CVD. 2.2.2 Minimization of Gibbs Free Energy Another method has been developed to predict the species concentrations at thermodynamic equilibrium during CVD growth. Using the free energy minimization method, equilibrium is reached when the Gibbs free energy reaches a minimum. In order to calculate the equilibrium concentrati on, the following equation is used [31,33,39]: ii i i i i ia RT n n G ln i g i i iP P N n RT n ln s i is s i m i g g i g i ig g in P P N n RT n n1 1ln or s i is m i g g i g i g i igG N n RT n P P RT n G G1 1ln ln (2.8) where G is the Gibbs free energy of the system, Gig and Gis are the free energy of formation of gaseous and solid species, ni g and ni s are the moles of gaseous and solid species, ig and is are the chemical potentials of gaseous and solid species which is the contribution of that substance to th e total Gibbs energy of a mixture, Ng is the total moles in the gas phase, m is the number of gaseous species, s is the number of solid species, R is the universal gas constant, and P and T are the process pressure and temperature,
27 respectively and P is a standard pressure of 750 Torr. Finally ai is the activity , which is the effective mole fraction. Th is is the partial pressure of specie i to the standard pressure of 750 Torr . Activity, ai is therefore equal to 1 for pure solids . The values of ni must be found in order to minimize G which is subject to the mass balance constraint . Taking the mass balance relation where the number of atoms of a particular element are conserved: j s i s i s ij m i g i g ijb n a n a 1 1 (2.9) where the number of atoms of element j in a molecule of species i is aij. For example, for the molecule C3H8, let i=1 and j=1 represent carbon, C, and j=2 represent hydrogen, H. Therefore, a11 = 3 and a12=8. The total number of moles of the element j is bj, m is the number of gaseous species, and s is the number of solid phase species. The solution of the equilibrium state is found by minimizing the Gibbs free energy subject to the constraints of equatio n (2.9) . Using free energy minimization, there is no need to predict the number of r eactions which may take place in the reactor. This technique enables one to consider a ll possible species, without speculating which species are important and without determining the r eactions of the system a priori. 2.2.3 Thermodynamic Simulations Thermodynamic numerical calculations, ba sed on the theory of the minimization of the Gibbs free energy just presented, were performed to predict the equilibrium concentrations of the chemical species in the USF CVD reactor. The calculations were performed using the NASA-Lewis Chemical E quilibrium Computations and Applications
28 (CEA) computer code . This program us es the minimization of Gibbs free energy method to determine the composition of species in the system. An outline of how to use the program is given in Appendix A. Informa tion about the software can be found in the literature [41,42]. Two thermodynamic state functions are required to determine the equilibrium compositions; in this case, temperature and total pressure were supplied. Along with the thermodynamic functions, T and P, the chemical reactants and their flow rates were input into the program. The species flow rates were to be specified in moles; however, since the pressure and temperature were constant, the volumetric flows were input into the program instead of the mole values. This was valid because the CVD process is assumed to follow the ideal gas law: PV = nRT. With the same system temperature and pressure, the volumetric flow is equal to the molar flow. Given the reactants, temperature and pr essure of the system, the program calculated the possible species present for all phases (gas, liquid, and solid). Based on this information, one could predict the dominant chemical species of the process. The input values for the calculations were based on typical process parameters for 4H-SiC epitaxial growth in the low-pressure horizontal reactor, which will be described in Chapter 3. The input gases were H2-C3H8-SiH4 and the total pressure was 150 Torr. The temperature range was chosen between 1500 and 2000 K for the reason that the growth of 4H-SiC typically takes place w ithin a temperature range of 1800 1900 K. Figure 2.1 shows a simulation of the mole a nd mass fractions of th e dominant computed species as a function of temperature. The input H2 flow was 30 slm (standard liters per minute), C3H8 flow was 10 sccm (standard cubic centimeters per minute) and SiH4 flow was 30 sccm giving a Si/C precursor flow ratio of 1.0.
29 (a) (b) Figure 2.1 Thermodynamic predictions of chem ical specie (a) mo le fraction and (b) mass fraction as a function of temperature for the H2-C3H8-SiH4 system. Input parameters were P = 150 Torr, QH2 = 30 slm, QC3H8 = 10 sccm, and QSiH4 = 30 sccm. The Si/C ratio was 1.0. The NASA CEA code was used to perform these simulations . The mole and mass fraction of H2 is higher than the ot her species in Figure 2.1 since it is the dominant chemical specie. Howe ver, it was excluded from the figure so that
30 the range of mole and mass fractions could be seen clearly. Atomic hydrogen has a higher mole fraction for temperatures beyond 1850 K, most likely due to the dissociation of hydrogen-containing molecules. As seen in Fi gure 2.1, the most dominant specie is solid SiC. The mole fraction of SiH4 decreases with temperature, as expected, due to the dissociation of SiH4. The C3H8 precursor is not seen on the figure as it has very low mole fractions at these high of temperatures. The mass fractions were computed to compare the thermodynamic simulations with simulatio ns used to predict the mass fraction distribution of species within the 200 mm r eactor (which are presented Chapter 5). The mass fractions have similar profil es as that of the mole fract ions, as seen in Figure 2.1. The only difference is that there are order of magnitude differences between the mole and mass fractions because the molecular weight is taken into account for the mass fraction, while it is not for the mole fraction. For example the difference between the mass and mole fractions of C2H2 at each temperature is appr oximately 12.7. This difference, however, is not the molecular weight of C2H2. The reason the difference between mole and mass fractions is not the actual molecula r weight of the specie is that when converting between the two fractions, the molecu lar weight of all species considered in the simulation is taken into account along with the molecular weight of the specific specie. From the thermodynamic simulations, the dominant species in the system are predicted to be as follows: CH4, C2H2, SiH2, SiH, Si, Si2, Si2C, SiC and H.
31 2.3 Kinetics In order to interpret what is occurring in the CVD process, it is also necessary to study the chemically reacting gas flows . It is important to study the reactions occurring at the surface which results in film deposition. In the gas above the substrate, homogeneous, i.e. gas phase, reactions take place, while at the surface, there are heterogeneous surface reactions (reactions occurring at the gassolid interface) . To accurately describe the CVD process, chemical kinetics is required. The epitaxial growth of SiC depends on the surface mobility of atoms on the surface (adatoms) and the number of available growth sites . A series of steps must occur for the deposition of epitaxial laye rs by CVD to take place, which are shown schematically in Figure 2.2. In step (1), gase s enter the reaction tube in a bulk flow. In step (2), intermediate reactants are formed from dissociation of reactant gases. In step (3), the reactant gases diffuse through a gaseous boundary layer, of thickness to the substrate. The boundary layer is a stagnant layer above the surface of the substrate. Adsorption of the species onto th e surface occurs in step (4). In step (5), a heterogeneous surface reaction takes place, which results in the deposition of the film on the substrate. Finally, the by-products desorb from the subs trate surface and then diffuse through the boundary layer back into the ma in gas stream where they ex it the reactor, steps (6) and (7), respectively [31-33].
32 Figure 2.2 Schematic diagram of mechan istic steps which occur during the CVD process which are (1) Gas inlet, (2) dissociati on of reactants, (3) diffusion of reactants to the surface, (4) adsorption of reactants to the surface, (5) heterogeneous surface reaction, (6) desorption of by-products, (7 ) diffusion of by-products back into the bulk gas . 2.3.1 Fluid Dynamics In order to quantify the CVD growth pr ocess, the fluid dynamics of the system (which includes gases) must be investigate d. For a CVD process, the fluid flow must be of laminar fashion to avoid intermixing of gas concentrations  For example, when switching from a p-type dopant source to an n-type dopant source during growth to produce an abrupt pn-junction, one does not want th e two dopant source s to intermix. If they do, then there will not be an abrupt j unction . In laminar flow, the fluid moves along straight, parallel paths in layers . For turbulent flow, the particles move in all directions and it becomes difficult to follow the motion of individual particles. In order to provide an indication of the type of fluid flow, the Reynolds number, Re, is introduced, which is a dimensionless number and is shown in equation 2.11. For laminar flow, Re
33 must be below 2000; above this value the fl ow is turbulent  Typically in a CVD process the Re number ~ 200 , which falls well inside the laminar flow regime. When a fluid flows over a solid surface, th e velocity at the surface is assumed to be equal to zero. The velocity increases from zero at the surface, to a maximum velocity, which is the bulk gas velocity. The boundary la yer is defined as the distance from the surface of the substrate to the maximum veloc ity of the bulk gas. With high velocities and laminar flow, the boundary layer, can be thin and is propor tional to the inverse of the square root of the Reynolds number, Re : Rex (2.10) where ux Re (2.11) and is the mass density of the fluid, u is the flow veloci ty of the fluid, x is the distance along the susceptor in th e flow direction, and is the dynamic viscosity of the fluid. The kinematic viscosity, is related to the dynamic velocity by: (2.12) Note that the viscosity and density of the fluid changes with temperature. The distance along the susceptor, x, may be replaced by the hydraulic diameter, dh (which is 2 times the radius of the tube), for flow w ithin in a circular tube. However, the flow is assumed to be across a plate since the area of focus is on the susceptor where the sample is placed.
34 A sketch of the boundary layer is shown in Fi gure 2.3. The velocity is initially constant, however, when the fluid comes into contact with the surface, a velocity gradient is generated, where the thickness of th e gradient is the boundary layer, An estimation of the Reynolds number for the 200 mm reactor is approximately 300, which is well within the laminar flow regime. This was determined by using equation (2.11) and assuming hydrogen as the dominant gas since it has a much higher flow rate compared to the precursors. The ve locity of the gas was found from the flow rate of H2, 30 slm, divided by the area of the in let. The distance al ong the susceptor for the calculation was 100 mm. Figure 2.3 Sketch of boundary layer showing an initially uniform gas velocity. Once the fluid comes into contact with the surface, th ere is a velocity gradient with thickness, . There are two types of regimes that CVD processes are typically classified into. One is surface reaction controlled and the other diffusion controlled CVD process . The surface controlled regime, more commonly referred to as reaction rate limited, is based on surface kinetics while the diffusion controlled regime, more commonly called mass transport limited, is based on diffusion of the reacting species through the boundary layer. The determination between the two regimes depends on the slowest step for the deposition to occur, whic h is the overall rate controlling (or limiting)
35 step . In the surface controlled regime, r eactants easily diffuse to the surface, where an over abundant supply of r eactants are available for the re action. The rate of the surface reaction therefore depends on how fast the reaction can take pl ace and not the availability of reactants. This regime typically occurs when the process pressure is low, producing large diffusion coefficients, and the surface temperature is low . At higher pressures (small diffusion coe fficients) and high temperatures, the growth process is governed by th e diffusion regime. In this regime, if the reactants can diffuse to the surface, the surface reaction will occur quickly. The diffusion through the gas adjacent to the surface governs the rate of deposition. Diffusion is weakly affected by temperature compared to the surface reaction regime, theref ore, the surface temperature does not play as strong a role in this regime until the reac tion rate slows down and starts to limit the process . Typically it is pref erred to have the grow th process controlled by diffusion, or mass transport , since the growth rate is not highly sensitive to temperature changes. To fully understand the CVD process, the fluid mechanics, chemical kinetics and thermodynamics must a ll be examined, preferably together. Until recently, this has been a daunting task but commercial software packages are now available for this purpose [46,47]. These will be described later in this chapter. 2.3.2 Mass Transport Limited Versus Surface Reaction Limited CVD If the process is mass transport limited, it is typically the diffusion of reactants through the boundary layer that is limiting the rate of reaction . Assuming there is a mixture of ideal reacting ga ses (although in the actual system it is not ideal) and a
36 concentration boundary layer, the diffusion of the species from the bulk gas to the surface is determined by Ficks Law : dy d D Ji i i (2.12) such that J is the mass flux, i is the density of the species i, Di is the diffusion coefficient of the specie i in the mixture, i is the mass fraction of specie i and y is the direction perpendicular to the substrate surface. The ne gative sign shows that the initial substances are consumed. The mass fraction, i, is: i i (2.13) The flux may then be written as : dy d D Ji i (2.14) Since density is mass divided by volume an d assuming an ideal gas, because the temperature is high and the pressure is low, the mass flux can be written in terms of pressure: R T PM (2.15) where M is the molecular weight of the specie an d R is the universal gas constant. So the mass flux becomes dy dP RT DM J (2.16) Assuming pressure varies linearly across th e boundary layer because we are assuming steady state conditions, e quation (2.16) becomes:
37 s i b ip p RT DM y P RT DM J (2.17) where pib is the partial pressure of specie i in the bulk gas, pis is the partial pressure of specie i at the surface, and is the boundary layer thickness. As seen from equation (2.17), for diffu sion controlled deposition, the boundary layer thickness is inversely proportional to the mass flux, ther eby affecting the ability for reactants to reach the surface to undergo the surface reaction. Note that when there is a higher velocity, due to constrictions of fl ow or by lower pressures, the boundary layer thickness decreases as predicte d by equations (2.10) and ( 2.11). When the boundary layer decreases, the ability for the reactants to reach the surface should increase. When reacting species are diffusing thr ough the boundary layer, reactions may be taking place which are considered homogene ous gas phase reactions. Once the species reach the surface, they may adsorb to the surf ace, react with the su rface, and become a film. This type of reaction is considered a heterogeneous surface reaction which will have a characteristic r eaction rate . Figure 2.4 show s a schematic diagram of the concentration gradient of reactants going from the bulk gas, through the boundary layer and approaching the solid film.
38 Figure 2.4 Schematic diagram of the growth process for CVD where Cb is the reactant concentrati on in the bulk gas, Cs is the reactant concentration at the surface, Jbs is the flux from the bulk gas to the surface, Js is the flux at the surface and is the boundary layer. The solid line is the concentration gradient. In Figure 2.4, Cb is the reactant concentration in the gas bulk phase, Cs is the reactant concentra tion at the surface, is the boundary layer, Jbs is the flux of reactants from the bulk gas to the surface, and Js is the flux of reactants at the surface. The phenomena may be described in terms of a m ass transfer coefficient. The flux of the reactants from the bulk gas to the surface may given by : s b g bsC C h J (2.18) where hg is the mass transfer coefficient, whic h is used when the diffusion takes place across an interface and increases with increa sing pressure and decreasing temperature. Assuming first order kinetics, where th e flux depends on the surface concentration to the first power, the flux at the surface is determined by [32,39]: s s sC k J (2.19) where ks is the surface reaction rate constant for the slowest surface reaction. For steady state conditions, the flux from th e bulk gas to the substrate, Jbs, and the flux at the substrate, Js, are equal. Therefore:
39 bs sJ J (2.20) Combining equations (2.18) and (2.19) gives: s b g s sC C h C k (2.21) which can then be written in terms of concentration at the surface by: g s b sh k C C 1 (2.22) From equation (2.22), if ks >>hg (which means a low gas tr ansport rate through the boundary layer), then Cs goes to zero and the system is mass transport limited. However, if ks<
40 where A is the pre-exponential factor, R is the universal gas constant, T is the temperature and ks is the rate consta nt. By plotting log ks vs. 1/T, the activation energy and preexponential factor may be found as the slop e and intercept, respectively. When the growth rate increases with increasing te mperature (in the low temperature region) according to the Arrhenius equation, the proce ss is deemed to be surface reaction limited. However, when the growth rate varies litt le with increasing temperature (in the high temperature region) according to the Arrhen ius equation, the process is mass transport limited. 2.4 Reactor Hardware The research for this dissertation was performed using two different horizontal hot-wall reactors which were of two different sizes. The first reactor had an inside tube diameter of 75 mm and it will be referenced as the 75 mm reactor. This reactor was suitable for epitaxial growth, however, there was one drawback to the reactor design. It was only able to accommodate 20 mm diameter samples. In order to grow on larger substrates, the reaction tube was enlarged to 200 mm in diameter. This reactor will also be designated by the inside tube diameter size, i.e. the 200 mm reactor. With the 200 mm reactor, substrates up to 100 mm in diameter are capable of fitti ng into the reactor. The hardware and components of these re actors are discussed in the following subsections.
41 2.4.1 75 mm Hot-wall System Originally, the 75 mm reactor was a horiz ontal atmospheric-pressure cold-wall reactor. The quartz reaction tube was surrounde d by a cooling water jacket. Details of the cold-wall reactor are given in references [9,49]. With the cold-wall design, there was a large temperature gradient between the suscep tor and the water-cooled quartz. In order to achieve a more uniform temperature distribu tion in the direction perpendicular to the susceptor length, the 75 mm reactor was c onverted from a cold-wall to a hot-wall configuration by redesigning the hot-zone. Fi gure 2.5 shows a sketch of the USF 75 mm hot-wall SiC CVD reactor configuration. Figure 2.5 Sketch of the 75 mm horizontal hot-wall reactor includ ing graphite foam insulation, susceptor, graphite adapter, RF co il, gas inlet liner, ss head plate and coldwater jacket. (1) Denotes the quartz adapter, (2) denotes water-cooled quartz reaction tube, and (3) denotes the quartz end-cap. Sketch provided by I. Haselbarth, University of South Florida. The total length of the 75 mm hot-wall reactor was 900 mm. The reactor was separated into three parts, as indicated by nu mbers in Figure 2.5; (1) a quartz adapter, (2) a water-cooled quartz reaction tube, and (3 ) a quartz end-cap. Th e quartz adapter was attached to a stainless steel head plate, us ing an o-ring (viton) a nd a clamp. The stainless
42 steel head plate was fixed to the reactor housing and had a V CR to conflat inlet port from which the gases were injected into the reac tion tube. The quartz end-cap was connected to the exhaust by a 50 mm (outer diameter) ball and socket joint. The end-cap had a quartz pyrometer view port whic h was used to focus the pyro meter onto the susceptor to read the temperature. The view port had an extended shape to prev ent parasitic reaction by-product deposition, which woul d give inaccurate temperat ure readings. The reaction tube was inserted between the quartz adapter and end-cap, and sealed with viton o-rings using bake-o-lite clamps. The reaction tube and end-cap were supported by bake-o-lite stand-offs to reduce stress at the reaction t ube connections caused by the weight of the tube. With this set-up, the reaction tube c ould be easily removed for cleaning purposes. The samples were loaded by removing the clam ps and detaching the end-cap. As seen in Figure 2.5, the gases entered the quartz reaction tube from the left and exited to the right were they were exhausted to a scrubber. A specially designed quartz inlet liner directed the gases from the inlet port into the susceptor slot. A SiC-coated adapter was placed between the quartz liner and the susceptor to provide a temperat ure transition which reduced thermal wear of the qua rtz liner. Since th e hot-wall reactor wa s converted from a cold-wall reactor, the coolingwater jacket surrounding the qua rtz reaction tube was still present and used as an extra safety precaution. With the hot-wall design, the substrat e was surrounded by a graphite susceptor which consisted of a bottom part with integr al side-walls, and a t op, or ceiling, part. Figure 2.6 shows sketches of th e susceptor and foam insulation. The only openings to the susceptor were at the inlet side for gases to enter and at the outlet side for the gases to exit. The top part of the susceptor was SiC coated dense graphite while the bottom was
43 TaC coated dense graphite. The coatings were used to reduce chemical etching of carbon during growth, impurity out-diffusion and provide a longer life span of the susceptor. The bottom was TaC coated to reduce the nucleatio n of SiC on its surface. The top portion of the susceptor was not TaC coated, as particles would tend to drop from the ceiling onto the sample since the particles do not adhere well to TaC. With the SiC coating, the particles nucleate on the surface of the ceiling. The particles, however, do not tend to fall onto the substrate surface unless there is t oo much buildup of particles on the surface. Another reason the top part of the susceptor was SiC coated was that the emissivity of SiC does not vary over time. However, the emissivity of TaC changes because SiC deposits on the surface. Th e emissivity of SiC is ~ 0.8  while that of TaC is ~ 0.3 . The higher emissivity of SiC provides more accurate pyrometer readings because a higher amount of radiation is emitted from SiC compared to TaC. Also, the smooth surface of the TaC reduces the accuracy of the pyrometer readings. Therefore, the top portion of the susceptor was SiC coated. The susceptor was enclosed in porous graphite foam which acted as both a support as well as ther mal insulation. Ar flow was estab lished outside of the liner to provide additional thermal insulation for the susceptor. The suscepto r and foam insulation comprise the hot-zone of the reactor as this is the heated area of the system. The graphite susceptor was heated by radio freque ncy (RF) radiation by an RF coil that was wrapped around the reaction tube. This desi gn essentially provided an isothermal environment, however, there were still temper ature gradients along the flow direction of the susceptor. In the hot wall reactor, the amou nt of power needed to achieve a specific temperature is lower compared to the cold-w all reactor . In a cold-wall reactor,
44 radiation losses are higher than in a hot-w all reactor due to the susceptor not being insulated. Lower radiative losses in a hot-wal l reactor are achieved by using the porous graphite foam insulation. Not only is there more uniform heating within the hot-wall reactor, there is also a higher cracking effi ciency of reactants w ith this design [43,52]. (a) (b) Figure 2.6 Sketches of the 75 mm hot-wall reactors (a) susceptor, placed inside graphite half of the foam insulation and (b ) susceptor top and bottom portions. The width and length of susceptor were 50 and 80 mm respectively, and the thickness of the susceptor top and bottom were 15 mm. The py rometer position hole is where the infrared pyrometer was focused. Sketches provided by I. Haselbarth, University of South Florida. The temperature of the susceptor was esti mated by focusing an infrared pyrometer at a machined recess, denoted as pyromet er position hole in Figure 2.6(b), in the top part of the susceptor. The temperature of the substrate was estimated by performing Si melt tests. This was accomplished by placing a piece of Si on top of a SiC substrate, which was itself placed onto the graphite susceptor. The test consisted of heating the susceptor until the Si was observed with the naked eye to melt. This temperature was recorded and compared to th e actual melting temperature of Si (1410 C). The difference between the two temperatures was used to control the growth te mperature during all subsequent CVD experiments. This was done using the following equation:
45 T = Tpyrometer T (2.25) where T is the temperature desired, Tpyrometer is the pyrometer temperature reading, and T is the temperature difference between the melt temperature of Si (1410 C) and the temperature reading of the pyrometer at th e time of Si melt during the melt test. For example, if the Si melted at a pyrometer reading of 1310 C, and since the true melting temperature of Si is 1410 C, then the T, the temperature error, is -100 C. Therefore, when performing experiments with a desired growth temperature of, say, 1560 C, the set point temperature should be 1460 C due to the temperature difference between the pyrometer reading and the actual temperature. Note that the melt test method is used only as an approximation of the actual temperat ure. Since the melt te mperature varies for different pressures and H2 flow rates, the temperature error, T, is a function of the pressure and H2 flow rate. To improve the accuracy of the temperature set point, melt tests were conducted for spec ific process flows and pressures, for each hot-wall configuration used. In additi on, whenever the hot-zone wa s changed, the melt test was performed to optimize the temperature accuracy. In order to supply the process gases to the CVD system, a gas handling system was used. The gases (H2, C3H8, SiH4, N2, Ar, HCl and CH3Cl (methyl chloride)) were each fed separately from their respective gas bottles, through welded stainless-steel lines into a gas manifold, where the gases were distributed through individual gas mass flow controllers (MFCs). The gas flows were controlled with the MFCs and dual actuated pneumatic valves. The valves were controlled at the CVD reactor operator control panel. The reactor control system is detailed in M. Smiths Masters thesis . The gases could be directed into the reactor, or they could be directed to a vent. The vent enabled gases to
46 be purged without entering the reac tion tube if necessary. The SiH4, HCl, and methyl chloride (CH3Cl) lines each had the option to flow Ar gas through the lines and out to the vent. This option was implemented to purge th e lines of any moisture or air that may have seeped into the lines after reactor shut down. Therefore, the lines were purged at the end of each day and backfilled with Ar. The H2 was fed into a palladium diffusion cell purifier before entering the gas manifold to purify the gas beyond VLSI grade (< 1ppb of O2, H2O, CO, CO2, N2 and all hydrocarbons ). Multiple sensors were incorporated within the system for safety purposes. When a sensor was activated, the system would shut down by shutting down all process gases and the heating source and Ar would be purged through the system. The different sensors monitored the presence of H2 and HCl gas, cooling water fl ow, and the security of the chamber door lock. In addition, an emer gency override button was installed which allowed the operator to shut down the process at any desired moment The safety sensors were also incorporated into the 200 mm hot-wall system, which is detailed in the next section. 2.4.2 200 mm Hot-wall System Although modifications were made to th e hardware, the 200 mm horizontal hotwall CVD reactor was similar to the 75 mm r eactor in the basic design of the hot-zone. Figure 2.7 shows a sketch of the reaction tube and its components. Like the 75 mm reactor, a gas inlet liner was used to guide the process gases into the slot in this susceptor via a SiC coated graphite adapter which was placed between the liner and the susceptor.
47 The susceptor was again surrounded by porous gr aphite foam insulation and was made of two molded parts. The reacti on tube was a single wall tube, i.e. no cooling jacket, which was placed between two stainless-steel plates, one at the front (gas inlet) end and one at the back (gas exhaust) end. An exhaust port was bui lt into the back pl ate. In order to prevent the quartz tubing from cracking at the ends when in contact with the stainlesssteel plates, o-rings were pl aced in between the quartz and the plates. The reaction tube was sealed by the compression of viton o-rings by steel flanges, denot ed as (1) in Figure 2.7. The stainless-steel plates were water cool ed and cooling fans were directed on them to prevent excessive heating. The steel back plat e (denoted as (2) in Figure 2.7) also had a hinged door (denoted as (3) in Figure 2.7) to allow access into the reaction chamber for loading/unloading of samples and cleaning of the reaction tube. The door had an Arpurged stainless steel tube with a quartz view port (denoted as (4) in Figure 2.7) attached which served as the pyrometer window. The temperature of the susceptor was again estimated using the Si melt test procedur e outlined in Section 2.4.1. Unlike the 75 mm reactor, the 200 mm reactor did not have a water-cooling jacket bu ilt around the reaction tube. This was not needed as the reactor de sign was meant to be a hot-wall system only, which does not require additional reaction tube cooling.
48 The susceptor for the 200 mm reactor was similar to that of the susceptor for the 75 mm reactor shown in Figure 2.6, except th e dimensions were larger. The width and length of the susceptor were 95 and 150 mm, respectively, and the thickness of the susceptors top and bottom parts were 15 mm. The distance from the bottom of the susceptor to the ceiling was 10 mm. This susc eptor also consisted of a top and bottom part; the top being SiC coated dense graphite and the bot tom being TaC coated dense graphite. There was a machined hole in the top portion of the susceptor to provide a more isothermal environment for the pyrometer read ings. A cross-sectional sketch of the hotzone is shown in Figure 2.8. Figure 2.7 Sketch of the 200 mm horizontal hot-wall reactor incl uding head plate, foam insulation, adapter, RF coil, inlet and foam liners, drum (back plate). (1) Denotes steel flanges, (2) denotes back plate, (3) denotes hinged door, and (4) denotes view port. Sketch provided by I. Haselbarth, University of South Florida.
49 Note that a polycrystalline plate was loaded into the growth zone, as shown in Figure 2.8. A few different types of polycrystalline plates with machined recesses were used to accommodate samples of different sizes as shown in Figure 2.9. The dimensions of the polycrystalline plates and the recesses are detailed in the fi gure. The depth of each recess is 0.5 mm (typical 4H-SiC substrates have a thickness of 0.4 mm). The use of the polycrystalline plates allows easy loading/re moval of the sample and enables repeatable sample placement from growth run to growth run. Figure 2.8 Cross-section schematic draw ing of 200 mm horizontal hot-wall CVD reactor hot-zone, showing gases entering on the le ft and exiting the right. The top part of the susceptor has a machined hole for pyrometer readings. The inlet gases enter through a quartz tube which guides the gases to the susceptor.
50 (a) (b) (c) Figure 2.9 SiC polycrystalline plates with (a ) one 12 mm x 12 mm recess, (b) thirteen 10 mm x 12 mm recesses, and (c) one 60 mm diam eter recess. The total length and width of each polycrystalline plate is 145 mm and 73 mm, respectively, which roughly corresponds to the dimensions of th e bottom portion of the susceptor. The foam insulation was placed inside a liner tube, which acted as a sacrificial quartz liner. The liner prevented the reaction tube from getting worn down due to the reaction between the graphite foam and the qua rtz tube. When the foam was heated in the 75 mm system, the quartz tube started to vitrif y. With the sacrificial liner included into the 200 mm reactor design, the reaction tube remained intact without damage from the foam.
51 2.5 Computational Fluid Dynamic (CFD) Simulations of the 200 mm Reactor The gas flow velocity distribution for the 200 mm horizontal hot-wall CVD reactor has been simulated and is presented in this section. The simu lations are needed to validate any design changes to the hot zone so that uniform deposition can be realized across a 50 mm wafer. The program used to perform the simulations was Femlab , which is a computational fluid dynamics (C FD) simulation software package that has been used throughout this work . The velo city profile is based on the equation of motion (momentum balance) and the equation of continuity (mass balance), which are : I u u u PI u u 3 2* (2.26) and 0 u (2.27) where is the mass density, u is the gas velocity vector, P is the system pressure, I is the identity matrix (diagonal matrix with ones on the main diagonal and zeroes everywhere else), is the dynamic viscosity, and superscript identifies the transpose vector. In order to perform the simulations, the reactor geometry, carrier gas and boundary conditions of the system must be sp ecified. The reactor geometry used for the CFD simulations was shown in Figure 2.7. The boundary conditions specified for the simulations were zero velocity at all surfaces except at the inlet and exhaust outlet ports. The pressure boundary condition was set to the process pressure of 150 Torr at the exhaust, since the pressure in the system is set downstream of the process. The boundary
52 condition at the gas inlet port was set to the maximum veloc ity , which was 30 m/s, as determined by: avgu u* 2max (2.28) where umax is the maximum gas velocity and uavg is the average gas velocity. The average gas velocity is found by: 24 r Q A Q uavg (2.29) where Q is the flow rate of the gas, A is the area of the reactor where the gas enters, and r is the radius. The gas enters the inlet from a stainless steel VCR fitting; therefore, the area was calculated based on the ar ea of the tube. The gas designated for the simulations was H2, with a flow rate of 30 slm, since the ca rrier gas has a much higher concentration compared to the precursors. The temperat ure was set to 1565 C. The gas velocity simulations are shown in Figure 2.10. The polycrystalline plate was included in the reactor geometry. The plate is the entire length of the susceptor and is 1 mm thick. Simulations were also performed without the po lycrystalline plate, wh ich resulted in no substantial difference between the two velocity distributions.
53 (a) (b) Figure 2.10 Gas velocity di stribution for the 200 mm hot-w all reactor showing (a) the entire reactor cross-section and (b) an enla rged area of the hot-z one with the 50 mm wafer area and susceptor indicated. The conditions for the simulations were QH2 = 30 slm, P = 150 Torr, and T = 1565 C and zero velocity fo r all boundaries except inlet gas flow and outlet pressure. The simulations were performed using Femlab . As seen in Figure 2.10, the gas velocity along the gas flow direction remains constant over the 50 mm wafer length. This in dicates that the reacta nts should diffuse to
54 the surface at approximately the same rate since the velocity is uniform. Velocity profiles were also computed at specific distances al ong the susceptor and the maximum velocities were plotted versus distance. This plot is shown in Figure 2.11. Figure 2.11 Velocity profile of the maxi mum gas velocity along the 200 mm hot-wall susceptor. The gas was H2 with a flow of 30 slm, T = 1565 C, and P = 150 Torr. As seen from the velocity profile in Figure 2.11, the maximum velocity remains constant along the susceptor. With the high H2 gas flow of 30 slm and the narrow area for the gases to flow, the velocity is at a ma ximum throughout the entire susceptor region for this particular susceptor design. Theref ore, a linear boundary layer is predicted throughout this area. The results from the CFD simulations will be compared to the experimental results within th e 200 mm hot-wall reactor in th e following chapter. (In the next chapter, it will be shown that due to severe process issues, the susceptor was redesigned.
552.6 Summary Deposition of SiC by mean s of CVD has been discussed and the theoretical underpinnings of this method have been reviewed in this chapter. Many different aspects must be considered so that predictions can be made which is helpful during CVD process development. In order to determine which chem ical species may be present in the system, the minimization of Gibbs free energy method was selected as the individual reactions need not be assumed a priori. Simulations were performed using the NASA CEA algorithm to determine the most dominant species in the SiC CVD reaction system studied during this dissertation work and were predicted to be H, CH4, SiH, SiC(s) and Si. The design of the 200 mm hot-wall reacto r was similar to that of the 75 mm reactor hot-wall design which ha d been implemented before th is dissertation research was initiated. The quartz reaction tube for the 200 mm reactor, however, was designed without a cooling water jacket and changes to the glass-to-metal seals were made to improve system reliability. The gas veloc ity distribution with in the 200 mm hot-wall reactor was predicted using a 2D CFD simula tion tool called Femlab. The predictions indicated that the velocity remained constant along the susceptor. The growth of 4H-SiC epitaxial layers in both hot-wall reactors will be presented in the following chapter.
56CHAPTER 3 4H-SIC EPITAXIAL GROWTH EXPERIMENTS The reactors used during this dissertation research work were horizontal hot-wall CVD reactors capable of process temperat ures up to 1850C. The typical growth temperature was from 1560C to 1585C. This chapter describes the development of the 4H-SiC epitaxial process us ing two reactor configurati ons: a 75 mm tube and a 200 mm tube. The latter is cap able of growth on 100 mm substrates however at the time of this work, the reactor was configured for 50 mm substrates, as that was the most common wafer size commercially available. 3.1 Characterization Techniques Before going into the experimental devel opment of the 4H-SiC growth process, the characterization techniques used will be de scribed. To develop the growth process, characterization of the 4H-SiC films was needed to determ ine the morphology, thickness, doping, and structural quality of the mate rial. The surface morphology and thickness were analyzed using a Hitachi S-800 S canning Electron Micros cope (SEM). The thickness was found by cleaving the samples a nd looking at the cro ss-section of the epitaxial layer using the SEM. The surface r oughness of the films was measured using a Digital Instruments Nanoscope Dimension 300 Atomic Force Microscope (AFM). The
57 tool was used to determine whether the films had any surface defects such as stepbunching, which are large steps on the surface whic h occurs when atomic steps combine , or 3D features. The structural quality was analyzed using a Phillips Analytical Xpert Pro X-ray Diffracti on (XRD) System. The film qu ality was assessed based on FWHM values of X-ray rocking curves of the (0004) diffraction peak. The electrical characteristics of the epitaxial films were analyzed by performing current-voltage (I-V) measurements, using a HP 4145B Semiconduc tor Parameter Analyzer and capacitancevoltage (C-V) measurements taken using a HP 4280A C-V meter. The net carrier concentration of the films was found using th ese two techniques. Finally, select samples were sent to the University of Pittsbu rgh where they were analyzed using low temperature photoluminescence (LTPL) (Professo r W. J. Choykes research group). The characterization tools used during this research are detailed in Appendix B. 3.2 75 mm Hot-Wall Reactor Experiments The growth of 4H-SiC epitaxial films wa s initially developed in the 75 mm coldwall reactor capable of atmosphe ric pressure (AP) growth. Th e details can be found in M. Smiths Masters Thesis . Growth could be carried out on a full 50 mm wafer in the cold-wall reactor system. The sy stem was then modified to grow epitaxial films at low pressure (LP) . The reactor was then modifi ed with a hot-wall in sert to realize a hotwall design (refer to Figure 2.5 for reference) Due to the size constraints, the hot-wall reactor was capable of holding a maximum sample size of 2 cm in diameter. The 4H-SiC
58 process for the hot-wall reactor was established starting with the process schedule developed for the cold-wall design, which is given in Table 3.1. Table 3.1 4H-SiC baseline process in th e 75 mm LP horizontal cold-wall reactor. Pressure (Torr) Temperature (C) QH2 (slm) QSiH4 (sccm) QC3H8 ** (sccm) Si/C Growth rate (m/h) 150 1550-1580 10 120 80 0.33 2.0 2% in H2; ** 3% in H2 Secondary electron microscopy (SEM) micr ographs from a representative sample for the 4H-SiC growth process in the horiz ontal hot-wall 75 mm reactor are shown in Figure 3.1. The morphology of the film wa s observed to be smooth and specular. A typical growth rate for this process was approximately 16 m/h. By comparison, growth rates presented in the literature for horizontal hot-wall reactors operating at similar temperatures, around the time of development of the 16 m/h growth process in the 75 mm hot-wall CVD reactor (late 2003), were between 5 and 20 m/h [19,56]. The morphology of a film from refere nce  was very similar to the morphology of a typical sample shown in Figure 3.1.
59 (a) (b) Figure 3.1 4H-SiC epitaxial film grown in the 75 mm LP horizontal hot-wall reactor for 15 min. (a) Plan-view SEM micrograph show ing a smooth, specular film surface, and (b) cross-section SEM micrograph of cleaved edge showing film thickness of ~4.3 m, yielding a Vg ~ 17 m/h (t = 15 min). In order to grow films on 50 mm wa fers, and potentially up to 100 mm in diameter, the area of the reactor needed to be increased. The reactor was thus scaled-up to allow for the possibility to accommodate larger samples. The development of the growth process in the 200 mm reactor is detailed next. 3.3 200 mm Reactor Process Development The hot-wall hot-zone insert (graphite susceptor and insulating foam) were scaled-up in size from the 75 mm hot-wall r eactor to permit growth on 50 mm substrates (i.e., the 200 mm reactor). The foam was orig inally designed with a cylindrical geometry to fit the inside diameter of a sacrificial qua rtz liner (a quartz tube cut to match the hotzone length), as shown by the sketch in Figure 3.2(a). With the full circular cross-section foam insulation design, there were RF impe dance matching issues which prevented the RF generator from turning on (it was believe d that the coil inducta nce was too large and
60 beyond the tuning range of the Ameritherm L-80 RF Generator ). The foam geometry was re-designed by shaving the mass of the cy lindrical foam into the shape shown in Figure 3.2(b), which allowed the RF generator to be able to match the load impedance. The foam design in Figure 3.2(b) was used for the thermal insulation of the 200 mm reactor and all subsequent process development was conducted with this design. (a) (b) Figure 3.2 Sketch of insulating foam cro ss-sections for the 200 mm reactor with (a) initial circular geometry and (b) modified design to enable RF coupling of foam and susceptor. Two different susceptor designs were explored in the 200 mm hot-wall CVD reactor. The first version of the susceptor ha d the same design as the one used in the 75 mm hot-wall reactor. This susceptor had a flat ceiling and was essentially a scaled-up replica of the 75 mm reactor susceptor. It was quickly determined that in the new 200 mm tube design, the susceptor needed to be modified due to precurs or depletion at the inlet to the hot-zone. The modified susceptor had an angled ceiling intended to reduce the depletion at the inlet. The experimental gr owth process developed using the flat and angled ceiling designs is desc ribed in the next section.
613.3.1 Flat Ceiling Susceptor Design The initial design of the 200 mm hot-wall reactor was based on the 75 mm hot-wall reactor, which was described earlier in Chapte r 2, Section 2.4.2. Replicating the susceptor design from the smaller reactor, the susceptor originally had a flat ceiling as shown in Figure 3.3. Figure 3.3 Cross-section sketch of the hot-wall CVD reactor hot-zone including graphite foam, flat graphite su sceptor, SiC polycrystalline pl ate (SiC polyplate), RF coils, and SiC substrate. The flat ceiling design in the 75 mm react or produced a uniform deposition across the hot-zone. However, the flat ceili ng design in the 200 mm reactor produced nonuniform deposition within the larger hot-zone. The reactants were being depleted at the entrance (inlet) of the susceptor, as there were lots of deposits on th e polycrystalline plate at the front end. No growth occurred in the center or back end of the susceptor, as no deposits were seen in these areas. Looking at the gas velocity distributions in Chapter 2, Figure 2.10, the velocity is high at the entrance to the sus ceptor. Therefor e, the boundary layer would be thin at this point and the reac tants could easily diffuse to the surface at the
62 front of the susceptor. Also, the efficiency of SiH4 cracking was very high at the entrance due to the high temperature at the front of the susceptor, enabling the reactants to be easily consumed by the reaction. This in turn left little reactants for growth downstream of the hot-zone. The samples had to be placed at the front of the susceptor, as shown in Figure 3.3, in order to realize growth. With the flat ceiling design, growth on a 50 mm wafer could not be carried out since the de position only took place at the front of the susceptor, which defeated the purpose of scaling up the reactor for growth on 50 mm wafers. The modification, which was made to eliminate the excessive depletion of reactants at the growth zone entran ce, is discussed in Section 3.3.2. The development of the growth proce ss of 4H-SiC in the 200 mm horizontal hotwall reactor was initiated by using the proces s parameters of the 75 mm reactor as described in Section 3.2. Ho wever, the flow rates of H2, C3H8, and SiH4 were doubled compared to what was used for the 75 mm r eactor process. This was done because the area of the growth zone in the 200 mm react or was double that of 75 mm reactor, thus the flow rates were scaled accordingly so that a similar gas velocity would be achieved. The pressure and temperature were maintained at 150 Torr and ~1550C, respectively. With these parameters and the doubled flow rates, there was not any epitaxial film growth. This was an indication that only et ching was occurring due to excess H2 flow. It is known that H2 etches SiC  and the t echnique is often used to remove surface damage on a sample prior to deposition [20,27]. Therefor e, to eliminate etching effects, the H2 flow rate was decreased to 7slm and QSiH4 (2% in H2) and QC3H8 (3% in H2) were increased to 300 sccm and 220 sccm, respectively. Along with reducing the H2 flow to 7 slm, Ar was introduced (10 slm) to mainta in a high carrier gas flow. Ar gon is an inert gas and does
63 not etch SiC as much as hydrogen. With th ese process parameter changes, growth was achieved and the resulting films were highly mo saic, if not polycrystalline, in structure with growth rates of around 2.4 4.0 m/h. An example of the resulting morphology is shown in Figure 3.4. As seen in the plan-view micrographs, the film surface was very rough with large grain boundaries and many 3D f eatures (which are th e white features in the micrographs), indicating further de velopment of the process was needed. (a) (b) Figure 3.4 Plan-view SEM micrographs of a 4H-SiC film grown in the 200 mm reactor at a magnification of (a) 2500x and (b) 10000x. The process pressure and temperature were 150 Torr and ~1550C, respectively. The QH2, QAr, QSiH4 (2% in H2) and QC3H8 (3% in H2) were 7 slm, 10 slm, 300 sccm, and 220 sccm, respectively. Data taken by Dr. S. Rao, University of South Florida. The precursor flow ra tes were increased no t only to suppress H2 etching, but to supply an excess amount of reactants to the surface. This was done to achieve sufficient amounts of reactants at the inle t of the hot-zone (which were being depleted at the inlet with the lower flow rates) as well as have enough reactant left over to supply reactions further along the growth zone. Even with incr eased precursor flow ra tes, there was still no deposition observed anywhere in the hot-zone except at the front entrance. This was determined by inspecting the polycrystalline pl ate for the deposition pattern in the hot-
64 zone. In order to achieve uniform deposit ion through the hot-zone, the pressure was decreased from 150 Torr to 90 Torr to allow a higher gas velocity which would also increase the diffusion of the reactants to the substrate surface at th e back end area of the hot-zone. Decreasing the pressure did in fact bring the reactants farther downstream as deposition was observed near the center of th e polyplate. However, the film morphology was still polycrystalline in structure. The argon was then eliminated, leaving 7 slm of hydrogen as the carrier gas. This produced a film with less text ure on the surface than that shown in Figure 3.4. Clearly, radical changes were required to realize an acceptabl e 4H-SiC epitaxial growth process in the 200 mm reactor. Once the morphology was improved by eliminating Ar as the carrier gas, the next step was to increase the growth rate. The QC3H8 (3% in H2) and QSiH4 (2% in H2) were increased from 220 sccm and 300 sccm, to 300 sccm and 800 sccm, respectively. Since the pressure at 90 Torr di d not produce good film morphology, the pressure was set back to 150 Torr. The H2 flow was also increased from 7 to 24 slm. The increase in H2 was used to push the reacta nts further into the hot-zone. With the large precursor flow rates, H2 etching of the substrate instead of growth should not have been an issue. The flows describe d above produced an improved film with step bunching (shown in Figure 3.5) and a growth rate of 16-24 m/h. The thickness uniformity varying across the sample was 40 %. This value was calculated by taking the difference between the high and low thic kness values and dividing by the mean.
65 (a) (b) Figure 3.5 Plan-view SEM micrographs of a 4H-SiC epitaxial film grown for 15 minutes at a magnification of (a) 2500x a nd (b) 30000x. Micrograph of part (b) shows step bunching. Data taken by Dr. Y. Shis hkin, University of South Florida. With the severe depletion of reactants at the front of the susceptor, growth on a full 50 mm wafer could not be carried out to produce a uniform film thickness. Therefore, the susceptor was modified to push the react ants farther downstream in the hot-zone, enabling a more uniform deposition. The susceptor top was exchanged with a tilted ceiling design. The 4H-SiC epitaxial growth process development then continued as described next. 3.3.2 Angled Ceiling Susceptor Design For CVD systems, the bottom portion of the susceptor is usually angled so that precursor depletion downstream can be ove rcome . However, the bottom portion remained flat, while the top portion was redesign ed to have an angled ceiling. The angled ceiling design ( ~ 23 from the horizontal) of the 200 mm reactor susceptor is shown in
66 Figure 3.6. This design would allow for futu re wafer rotation which is commonly used in SiC CVD to ensure epitaxial film uniformity over a full wafer . Figure 3.6 Cross-section sketch of the hot-wall CVD reactor hot-zone including graphite foam, angled graphite susceptor, graphite adapters, SiC polycrystalline plate (SiC polyplate), RF coils, and SiC substrate. With the angled ceiling, the boundary layer above the susceptor surface becomes more uniform across the entire susceptor. This is accomplished by having a lower gas velocity at the entrance of the susceptor due to the larger vol ume, and a higher gas velocity at the outlet from the constriction due to the tilt . This ultimately produces a thinner boundary layer downstream of the su sceptor for reactants to diffuse through [5,33], thus increasing the deposition rate downstream. Figure 3.7 shows sketches of the boundary layer without a tilted su sceptor and with a tilted sus ceptor. Note that the bottom of the susceptor is angled, however, the same principle should apply with the angle being on the ceiling of the susceptor.
67 (a) (b) Figure 3.7 Sketch of the boundary layer abov e the susceptor with a (a) flat bottom and (b) tilted bottom. The gas velocity for the tilted bottom susceptor increases farther downstream along the flow direction producing a thinner boundary laye r at the back end of the susceptor . CFD simulations were performed to predic t the velocity distribution within the 200 mm hot-wall reactor with a tilted suscep tor ceiling. CFD simulations for the flat ceiling susceptor design were described in Ch apter 2, Section 2.5. The reactor geometry for the simulations was similar to the reactor sketch shown in Chapter 2, Figure 2.7; however, the susceptor geometry simulated wa s that of the sketch in Figure 3.6. The simulated gas velocity distribution for the 200 mm hot-wall CVD reactor with the tilted ceiling is shown in Figure 3.8. As seen in Figure 3.8(b), th e velocity increases along the length of the susceptor, as expected, due to na rrowing of the flow cross section. With the lower velocity at the inlet of the susceptor, the boundary la yer would be slightly thicker there compared to that the back end of th e susceptor, which has a higher gas velocity.
68 (a) (b) Figure 3.8 Simulated gas velocity distri bution for the 200 mm hot-wall reactor with angled susceptor ceiling showing (a) the entir e reactor cross-section and (b) an enlarged view of the hot-zone with the 50 mm wafer area and susceptor indicated by vertical dashed lines. The boundary conditions were u = 0 for all surfaces, except inlet flow of QH2 = 30 slm, P = 150 Torr, and T = 1565 C. The simulations were performed using Femlab .
69 Velocity profiles were also computed at specific distances along the susceptor and the maximum velocities were plotted versus distance. The plot is shown in Figure 3.9. The velocity increases from a velocity of ~ 14 m/s to a velocity of ~ 30 m/s along the susceptor. With this susceptor design, the bou ndary layer will be thicker at the front of the susceptor than at the back (exhaust) end. This should theoretically help to reduce the depletion of reactants at th e front of the susceptor, as was observed to be the case experimentally for the flat ceil ing design. With the higher gas velocity at the back end of the susceptor, the reac tants will be able to diffuse thr ough the stagnant layer faster, which will provide more reactants at the surface for the reaction to take place. Therefore, the titled susceptor ceiling design should improve the uniformity of the deposition and allow growth downstream of the sus ceptor. Indeed this was found to be the case as will be described shortly. Figure 3.9 Velocity profile of the maxi mum gas velocity along the 200 mm hot-wall susceptor with the angled ceiling. The gas velocity profile was carried out with a H2 flow of 30 slm, T = 1565 C, and P = 150 Torr. Trend line to aid eye only.
70 Another benefit to the angled susceptor design is that silane is less efficiently cracked at the inlet, thus allowing for uncracked reactants to be available farther downstream for deposition. This is likely due to the susceptor ceiling and bottom being farther apart, resulting in less efficient gas heating at the sus ceptor inlet compared to that of the flat ceiling design. The susceptor modifi cation was one of the ma in factors used to improve the growth rate and surface morpholog y of the 4H-SiC epita xial process in the new 200 mm reactor. Along with changing the susceptor geometry the precursor gas mixture was also changed from 2% and 3% in H2 to 100% silane and propane concentrations, respectively. This change was made to extend the life of the gas bottle by using a higher concentration gas source. Another reason is that the precurs or MFCs have a specific maximum value. When the precursor flows (at the low mixture percentages (2% and 3% in H2 for SiH4 and C3H8, respectively)) were increased, the majority of the increase was due to H2 gas. By having 100% concentration of pr ecursors, a larger range of the actual precursor gas flow was permitted. With the new susceptor ceiling and gas concentrations, the experiment that produced the best result before these changes took place (Vg = 16-24 m/h) was repeated. As a note, the sample placement, in both the flat ceiling and the first few experiments in the tilted ceiling design, for all experiments described was towards the entrance of the susceptor. The resulting film was smooth and specular. The growth rate was found to be 6.4 7.2 m/h, with a much higher degree of thickness uniformity (7.5%) across the sample surface. SEM mi crographs of the surface morphology and the thickness of the film are shown in Figure 3.10.
71 (a) (b) Figure 3.10 SEM micrographs of the first 4H-S iC epitaxial film grown in the hot-wall CVD reactor with angled ceiling. (a) Plan -view micrograph of smooth surface and (b) cross-section micrograph of film grown for 15 minutes resulting in a Vg of ~7 m/h. Note the growth parameters were not yet optimized Data collected by Dr. S. Rao, University of South Florida. The growth parameters clearl y needed to be optimized to account for the angled ceiling. The carrier gas (H2) was increased to 30 slm and the precursors (SiH4 and C3H8) were each increased to 10 sccm, producing a Si/C ratio of 0.33. The morphology of the films produced using this process was smooth a nd the growth rate was determined to be 9.2 m/h. In an effort to enhance the growth rate further, the silane flow rate was increased while maintaining a constant propane flow rate. A sequence of films was grown with increasing silane flow rate until th e surface morphology deteriorated. The maximum silane flow rate that resulted in a smooth f ilm surface was 30 sccm, giving a growth rate of 22 m/h. The Si/C ratio was 1.0 for this ex periment. For Si/C ratios higher than 1.0, the surface degraded. The surface morphology of a film with excess SiH4 flow is shown in Figure 3.11. The surface of the film was r ough and severely step-bunched. This is likely due to the silane flow rate being too high for the grow th temperature used. That is, the partial
72 pressure of the precursor may be too high to supply reactants fo r growth at this temperature . The growth temperatur e used for the experiment was 1535 C. However, later experiments were performed at higher growth temperatures of ~1560 C. The morphology may have been improved by in creasing the growth temperature beyond 1560 C, however, the higher temperature was no t pursued at the time of the experiment shown in Figure 3.11. (a) (b) Figure 3.11 Plan-view SEM micrographs showing poor film morphology caused from excess precursor flow rates at a magnification of (a) 500x and (b) 10000x. The silane flow rate was 48 sccm, with a Si/C ratio of 1.0. Process temperature and pressure were ~ 1535C and 150 Torr, respectively. Data taken by Dr. Y. Shishkin, University of South Florida. Once the process parameters were determ ined as described above, the samples were placed in the center of the polycrysta lline plate as deposition was observed to take place across the entire susceptor. At the cente r, a growth rate of 28 m/h was observed. A diffuser plate was then incorporated into the sy stem at the inlet port of the gas stream, as shown in Figure 3.12, to potentially eliminat e gas jetting effects. Jetting may occur when the gas flows out of the inlet without dispersion. The diffuser was meant to help disperse the gases throughout the inlet liner, shown in Figure 3.12(b). By implementing this
73 diffuser into the system, there was a slight growth rate increase from 28 m/h without the diffuser, to 30 32 m/h with the diffuser. The addition of the diffuser may have been the reason for the increase in growth rate. (a) (b) Figure 3.12 (a) Photograph of the gas in let port of 200 mm hot-w all reactor showing diffuser plate. (b) Cross-section sketch of diffuser plate demonstrating the dispersion of the inlet gas into the inlet liner. With all of the modifications to the 200 mm system described in Section 3.3.2, the growth rate increased from 4 m/h (with the sample placed at the front of the susceptor entrance) to 30 32 m/h (with the sample place in the center of the susceptor). The modification of the susceptor from a flat cei ling to an angled ceiling, the increase in precursor and carrier flows, and the additio n of a diffuser plate were crucial to the development of the 4H-SiC baseline process. This baseline process will be summarized next along with characterization performed on the films grown using this process.
743.4 200 mm Reactor Process The work performed during this resear ch for the development of a 4H-SiC epitaxial growth process has led to two publi cations [60,61]. The growth rate of 32 m/h is the highest reported rate to date for a horizontal hot-wall CVD reactor using standard process chemistry of H2 SiH4 C3H8 at low (~ 1600C) growth temperatures. The process parameters used to achieve this high growth rate are as follows: P = 150 Torr, T ~ 1600C, QH2 = 30 slm, QSiH4 = 30 sccm (100%), QC3H8 = 10 sccm (100%), and Si/C = 1.0. Note the temperature of the process ge nerally shifts. Therefore, when growing the epitaxial layers using the same process flows and pressure described above, the temperature may be somewhere within the range shown above to produce smooth films. The process schedule developed is shown in Fi gure 3.13. The step-by-step experimental procedure to grow the epitaxial layers using this process is detailed in Appendix C. Figure 3.13 4H-SiC epitaxial growth pr ocess schedule developed for the 200 mm LP horizontal hot-wall reactor with angled ceiling. Note, the reactor was configured for growth on a 50 mm wafer.
75 Comparatively speaking, the amount of silane and propane needed for the baseline growth process in the 200 mm reactor is higher than that needed in the 75 mm reactor. The mole fractions of the precursors, found by taking the flow rate of the desired specie divided by the total flow rate (added fl ow rates of carrier gas and precursors), for the 75 mm reactor baseline process were 2.36 x 10-4 for both silane and propane. In the 200 mm reactor baseline process, the mole fractions are 9.99 x 10-4 and 3.33 x 10-4 for silane and propane, respectively. This may be due to the difference in reactors. Typically the growth process of one react or can not be used to produce the same results of surface morphology and growth rate for another reac tor. Therefore, the difference in mole fractions between the two reactors may be due to the reactors being different. The different reactors need different process pa rameters to achieve smooth, specular films. Planand cross-section view micrographs of a representative epitaxial film grown for 45 minutes using the developed baselin e process in the 200mm reactor are shown Figure 3.14. As seen in the figure, the mor phology of the film was smooth and specular, and the growth rate was approximately 31 m/h.
76 (a) (b) Figure 3.14 4H-SiC epitaxial layer grown w ith the baseline process for 45 minutes in the 200 mm reactor. (a) Plan-view SEM micr ograph showing a smoot h, specular 4H-SiC film and (b) cross-section SE M micrograph of the same sample showing a thickness of 23 m for a growth rate of 31 m/h. Data take n by Dr. Y. Shishkin, University of South Florida. Structural and optical analysis was perf ormed on the sample shown in Figure 3.14 using XRD and LTPL, respectively. Defect s influence the diffraction of X-rays, therefore, X-ray diffraction measurements may be used to evaluate the quality of the films. Using LTPL, donor and acceptor impurit ies may be investigated within the film. The X-ray rocking curve of the (0004) re flection peak for sample USF-05-529C, grown in the same run as the sample shown in Fi gure 3.14, is shown in Figure 3.15. The full width at half maximum (FWHM) of the ( 0004) diffraction peak was approximately 36 arcsec. This indicates the film is of reasona ble quality, however, hi gher quality films have been produced with FWHM values between 5.2 and 10 arcsec, grown in vertical and horizontal hot-wall reactors as well as vertic al radiant-heating reac tors [18,24,27]. While the quality of the film in Figure 3.14 is not as high as that repor ted for the films in [18,24,27], other films have since been produ ced in the 200 mm reactor with higher quality. A reason for the high FWHM value ( 36 arcsec) of the (0004) diffraction peak
77 may be that defects from the substrate propa gated into the epitaxi al layer, producing a lower quality film. The characterization results of these films will be shown in Section 3.5.3. Figure 3.15 X-ray diffraction rocking curve of the (0004) diffraction peak for a 4HSiC epitaxial film grown for 45 min at a temp erature of ~ 1520C a nd a process pressure of 150 Torr. Sample ID: USF-04-529C. The film quality was also analyzed us ing LTPL (collected at 2K), which was provided by Professor W. J. Choykes group at the University of Pittsburgh. The near bandedge spectrum of film US F-04-529B (refer to Figure 3.14) is shown in Figure 3.16. In the spectrum, it can be seen that the intrinsic peak, labeled I75, is strong which is an indication of a high quality film . The dopi ng of the epitaxial layer was estimated by taking the ratio of the Qo peak to the I75 peak . The net carrier concentration, Nd-Na, was found to be approximately 1.75 x 1015 cm3, which is comparable to the doping concentration found using C-V measurements conducted. A four-particle aluminum complex peak is seen between P0 and Q0 lines, indicating there is a source of Al contamination present in the CVD reactor. Th e source of the Al contamination may be
78 caused from a high concentration of Al in the graphite adapters. The graphite adapters used in this dissertation work had recently been replaced, and the films grown with the new adapters contained lower concentrations of Al. Figure 3.16 Near bandedge LTPL spectrum (tak en at 2K) of a 23 m thick 4H-SiC film grown at 31 m/h. The process temperat ure ~ 1520C, pressure = 150 Torr, and Si/C = 1.0. Data courtesy of Professor W. J. Choyke, University of Pittsburgh. With the baseline 4H-SiC epitaxial film process developed, the growth parameter space was explored to characterize the react or. The only parameter which was left unchanged was T. The growth rate was st udied as a functio n of silane flow, pressure, and growth time. The results of the reactor char acterization will be presented in the next section of this dissertation. 3.5 200 mm Reactor Growth Characterization The characterization of the 200 mm reactor is important as there are many parameters (i.e. pressure, silane flow rate, et c) that influence the growth of the epitaxial
79 layers. By investigating growth parameters such as the silane flow rate and the Si/C ratio, the ability to grow at differe nt speeds or vary the doping c oncentration (within a certain range) may be examined. This work has recently been published in the Journal of Crystal Growth . 3.5.1 Growth Rate as a Function of Silane Flow The ability to grow epitaxial films at diffe rent speeds in the hot-wall CVD reactor was investigated by varying the SiH4 flow rate while maintaining the same Si/C ratio. The baseline process described in Section 3.4 was the starting point for this experiment. The silane flow was varied from 12 sccm to 39 sccm. The pressure, temperature, and H2 carrier gas flow rate were maintained at 150 Torr, ~ 1600C, and 30 slm, respectively. Figure 3.17 shows the resu lt of the experiment. Figure 3.17 4H-SiC epitaxial growth rate dependence on silane flow rate in the 200 mm LP hot-wall reactor with angled ceiling. Carrier gas flow rate and the Si/C ratio maintained at 30 slm and 1.0, respectively. The solid trend line shown to aid eye only [60,61].
80 As seen in Figure 3.17, the gr owth rate is a linear function of silane flow rate, in which it increased from 16 to 39 m/h. The lowest flow rate for silane (12 sccm) produced a film with black features on the surface, shown in Figure 3.18, and a growth rate of 16 m/h. The features may be due to the substrate or from the low silane flow rate. If there were defects or there was surf ace damage on the substrate prior to growth, the epitaxial film could be affected by it. If the features were cau sed by the lower silane flow, there may not have been enough precu rsors to support growth of a smooth film. (a) (b) Figure 3.18 Plan-view SEM micrograph of 4H -SiC grown for 15 minutes with a silane flow rate of 12 sccm and a S i/C ratio of 1.0 at a magnifi cation of (a) 500x and (b) 5000x. When the silane flow rate was increase d to 15 sccm, the growth rate increased slightly to ~ 17 m/h, resulting in decent f ilm morphology without th e black features as those seen in Figure 3.18. With a silane flow increased from 15 sccm up to 30 sccm, the growth rate increased. At the same time the morphology of the films was found to be smooth, similar to one in Figure 3.14. The firs t sign of film deterioration began when the silane flow rate was increased from 30 to 36 sccm. Occasional triangular defects were observed on the surface by plan-view SEM ( not shown). When the flow was increased
81 further to 39 sccm, the surface morphology b ecame wavy, which may have indicated the growth rate was too fast for this temperat ure  (without the addition of any other precursors, which will be discussed in Chap ter 4). The morphology of the film grown at 39 sccm of SiH4 flow is shown in Figure 3.19. (a) (b) Figure 3.19 Plan-view micrographs of a 4H-S iC epitaxial sample grown at a growth rate of 39 m/h using a flow rate of SiH4 at 39 sccm. The magnifications are (a) 500x and (b) 10000x. When excess SiH4 is added to the process, homogeneous nucleation in the gas phase is likely to occur. Homogeneous gas phase nucleation can re sult in Si clusters which may descend onto the surfac e of the film. These clusters become detrimental to the film by incorporation of large 3D particles [17,21]. With the film deterioration at the higher silane flow rates, this may be an indi cation the supersaturation limit for this Si/C ratio and temperature is being approached. In Chapter 4, research using HCl as a growth additive to reduce the homoge neous gas phase nucleation will be discussed. Adding HCl to the precursor mixture allowed the possibility of increasing in the silane flow up to 45 sccm without significant detrimenta l effects on the film morphology.
823.5.2 Growth Rate as a Function of Process Pressure The pressure plays a crucial role in chem ical vapor deposition. If the pressure is too high, the growth rate can decrease due to a thicker boundary layer making it more difficult for the reactants to diffuse through  When the pressure is excessively low, a crowning effect can occur, in which the epit axial deposition takes place mostly at the edges of the sample. Therefore, the eval uation of the reactor based on the process pressure is important. Figure 3.20 shows the obtained growth rate dependence on process pressure, which was varied from 100 to 450 Torr. Figure 3.20 4H-SiC epitaxial growth ra te vs. process pressure in the 200 mm horizontal hot-wall reactor with angled ceiling. The H2 flow and the Si/C ratio maintained at 30 slm and 1.0, respectively. Solid trend li ne shown to aid eye only. Vertical dashed line indicates baseline pro cess pressure of 150 Torr. As demonstrated in the Figure 3.20, the growth rate was at a maximum (33 m/h) at P = 100 Torr. The process pressure could not be decreased much below 100 Torr as the lowest achievable pressure with the existing vacuum system was 90 Torr for the experimental conditions of interest. Th e growth rate at 150 Torr (32 m/h) was approximately the same as that at 100 Torr. The growth rate may be highest at these process pressures because the boundary layer is thinner, allowing the reactants to diffuse
83 quickly to the surface where the reaction may take place. The resulting morphologies at both pressures were very similar to the mo rphology of the sample shown in Figure 3.14. As the pressure was increased to 250 Torr, the growth rate re duced by more than 25% to approximately 22 m/h. With a furt her increase in pressure to 450 Torr, no growth was observed. An explanation could be that homogeneous nucleation occurs at the higher pressures thus reducing the amount of Si available to supply the reaction. At elevated pressures, homogeneous nucleation of the SiH4 precursor limits the deposition rate . It may also be that at the higher pressure, the gas flow velocity is decreased and the reactants may not be getting to the substrate fast enough for growth to take place. A final explanation for the decrea se in growth rate at the 450 Torr pressure may be that, the process parameters are not optimized for this pressure. The temperature may need to be adjusted to allow growth at this high of pressure. The temperature was not adjusted during these experiments. 3.5.3 Analysis of Thick Films When growing thick epitaxial films, it is im portant to verify that the growth rate is maintained for extended amounts of time. This is an essential piece of information in the characterization of the reactor as it provides insight into process stability and reliability. For example, if a film thickness of 75 m is desired and the growth rate is 15 m/h, then it is important to know that gr owing for 5 hours will truly result in this thickness. If the growth does not produce a film of preferred thickness and quality, then the growth process is not stable or reliable. Therefor e, experiments were conducted in which the
84 baseline process was used and films were grow n for various lengths of time, as shown in Figure 3.21. The Si/C ratio was, however, 1.2 in stead of 1.0 in order to produce an n-type epitaxial layer during these experiments (f or reasons explained in Section 3.5.4). Figure 3.21 Growth rate of 4H-SiC as a function of growth time in the 200 mm LP hot-wall CVD reactor with angled ceiling. The process pressure and the silane flow rate were 150 Torr and 30 sccm, respectively. Solid trend line shown to aid eye only. Note the growth rate is invariant with respect to growth time. The substrates were polished using chemical mechanical polishing (CMP) by NovaSiC prior to deposition . The growth of epitaxial films was conducted for a few different periods of time between 15 and 120 minutes. The epitaxial layer thicknesses were determined via cross-section SEM and th e average growth rates were calculated as described in Section 3.1. As seen in Figure 3.21, the growth rate remained constant (Vg ~ 32 m/h) for all growth runs. Thus, this da ta showed that longer growth runs should produce the desired film th ickness based on the nominal growth rate of 32 m/h. The morphology of the films was simila r to that shown in Figure 3.14(a). However, occasional triangular and carrot-like defects were observed in the films thicker than 30 m. When thick epitaxial layers ar e grown, the defects from the substrate may
85 propagate into the film and increase in size . Triangular defects are inclusions of 3CSiC in the 4H-SiC films which may be deve loped from imperfections in the substrate . Carrot-like defects are extended line defects which also propagate from the substrate interface . Defects within the epitaxial layers reduce the minority carrier lifetime, which affects the device performance. The quality of the film surface is typically evalua ted by a quantity known as surface roughness RMS which is obtained by performing AFM scans of the surface. The RMS roughness is the root mean square value of vertical deviati ons from the mean horizontal surface . The surface roughness of the films from the experiment in Figure 3.21 was investigated. The plot of surface r oughness as a function of film thickness is shown in Figure 3.22. Figure 3.22 Dependence of surface roughness of 4H-SiC films on film thickness as measured by AFM. Growth conditions were maintained at 30 slm of H2, 30 sccm of SiH4 and a Si/C ratio of 1.2. Data taken by Dr. Y. Shishkin, University of South Florida. The surface roughness for the 30 m film wa s 0.27 nm RMS, while that of the 65 m film was 0.32 nm RMS. This is only a slight increase in surface roughness when the film thickness was doubled, indicating that th e epitaxial growth in hot-wall CVD for
86 prolonged periods of time should not signif icantly affect the surface roughness of the grown films. An AFM scan of the 65 m thick film is shown in Figure 3.23. In the literature, films were grown at 11 15 m/h resulting in 60 m thick films which had a surface roughness between 0.2 and 0.7 nm RMS . In another work , approximately 100 m thick films were grow n with surface roughness values from 0.2 and 0.3 nm RMS. Therefore, the surface roughn ess of the thick films grown in the 200 mm reactor is of reasonable value. Figure 3.23 AFM scan of a 65 m thick 4H-S iC film grown for 120 minutes at a rate of 32 m/h. Surface roughness is 0.32 nm RMS. The scan area was 10 m x 10 m. Data taken by Dr. Y. Shishkin, Univ ersity of South Florida. The structural quality of the 65 m fi lm was analyzed using XRD. A rocking curve taken of the (0004) re flection peak is shown in Fi gure 3.24. The X-ray rocking curve is customarily used to measure the semiconductor crysta l perfection. In order for a film to be considered high quality, the diffraction peak must be narrow and result in a low FWHM value. If the crystal was perfect, di ffraction would only occur at one specific angle, the Bragg angle. However, for re al crystals, the peak broadens due to
87 imperfections in the lattice called defects. Therefore, the lower the FWHM value, the more perfect the crystal structure. The FWHM for the 65 m thick film is approximately 11 arcsec which is comparable to other hi gh quality films grow n using hot-wall CVD [18,24]. Figure 3.24 X-ray diffraction rocking curve of the (0004) peak for the 65 m thick 4HSiC epitaxial film. The FWHM is approximately 11 arcsec. 3.5.4 Analysis of Doping Concentration The results of the experiments aimed to determine the carrier concentration, |NdNa|, as a function of Si/C ratio in the gas phase are shown in Figure 3.25. The films grown during these experiments were not inte ntionally doped. However, Al and N atoms are known contaminants in the reactor, as dete rmined by LTPL. It is speculated that the contaminants may be out-diffusing from the graphite adapters during growth. The doping concentration of film s grown in the 200 mm reactor was explored by changing the Si/C ratio, which was performed by varying the C3H8 flow rate while maintaining a constant SiH4 flow rate. During the growth of SiC, the Al (p-type dopant) atoms compete with Si atoms and N (n-type dopant) atoms compete w ith C for vacant latt ice sites . To
88 produce an n-type doped epitaxial layer, the Si/C ratio is increased, providing a lower concentration of C atoms which presents a greater chance for N atoms to occupy the vacant lattice sites. On the other hand, to create a p-type epitaxial laye r, the Si/C ratio is decreased. This is accomplished by reducin g the amount of Si in the system, which enables the Al atoms to more easil y occupy the vacant lattice sites. Figure 3.25 Net carrier concen tration of 4H-SiC epitaxial films as a function of Si/C ratio as measured with the C-V technique. The 200 mm LP hot-wall reactor with angled ceiling was used. H2 and SiH4 flow rates maintained at 30 slm and 30 sccm, respectively. Doping concentrations of p-type layers indi cated by circles while doping concentrations of n-type layers indicated by triangles. Da ta provided by Dr. S. Rao and J. Walker, University of South Florida. The Si/C ratio was varied between 0.8 and 1.8 by fixing the amount of available silicon and varying the amount of carbon. As sh own in Figure 3.25, the Si/C ratios of 0.8 and 1.0 resulted in p-type epitaxial layers. When the Si/C ratio was increased from 1.0 to 1.2, the conductivity type conversion occurred. At and beyond a Si/C ratio of 1.2, the films were n-type with increasing carrier c oncentrations as the Si/C ratio increased. The lowest value of |Nd Na| obtained was ~ 2 x 1015 cm-3 at a Si/C ratio of 1.2. The p-type epitaxial layers were grown at the lower Si/C ratios, where more C atoms are available to
89 occupy the vacant lattice sites, resulting in a lower concentra tion of N atoms. With less ntype dopants, the material became p-type since there was the same amount of Al contaminants in the system as with a higher concentration of C atom s. As the Si/C ratio was increased, by decreasing the C concentratio n, the doping concentration of the p-type films decreased and type conversion occurred With the higher Si/C ratios, there were less C atoms to compete with the N atoms, resulting in films with higher n-type conductivity. The morphology of the films was also exam ined as a function of Si/C ratio. With a Si/C of 0.6, the resulting film surface was wavy which may have been caused by a carbon-rich process  or too low of a grow th temperature . With an increase in carbon, films may roughen at one specific gr owth temperature. However, smooth film surfaces may be produced when there is a higher amount of carbon present by increasing the growth temperature . Wh en the Si/C ratio was increase d to 0.8, triangular defects were observed on the surface. The triangular def ects, which are 3C inclusions in the film, are common when Si/C ratios are low . To reduce this problem, the Si/C ratio may be increased. The morphology was smooth and specu lar for the Si/C rati os between 1.0 and 2.0. At a Si/C ratio of 2.5, the film cont ained Si droplets on the surface caused by excess Si not consumed by the reaction. The growth rate was not aff ected by variation of the Si/C ratio be tween 0.8 and 1.4, resulting in growth rate s of approximately 30 m/h. With a Si/C ratio above 1.4, the growth rate started decreasing. The growth rate was approximately 26.5 m/h for the Si/C ratio of 1.6, and decrea sed down to 17 m/h for the S i/C ratio of 2.0. Since the
90 amount of Si was fixed, this may have b een caused by a carbon-limited process in which not enough carbon is being s upplied to the reaction. 3.5.5 Growth Variation in Hot-Zone In order to determine the growth rate, film quality, and doping variation along the hot-zone, an experiment was carried out to simulate growth on a 50 mm wafer. Five samples were loaded onto the polyplate as shown in Figure 3.26. The size of each sample was 8 mm x 11 mm. This was done to reduce th e substrate cost of growing on an actual whole wafer. The growth was carried out usi ng the baseline process wi th a Si/C ratio of 1.2 to produce n-type epitaxial film at 30 32 m/h. The films were grown for 30 minutes which resulted in ~ 17 m thick film s. The results of the experimental analysis are shown in Table 3.2. Figure 3.26 Picture of the polycrystalline plate used to simulate growth on a 50 mm wafer. Note the lettering is used to desi gnate placement of sample s. The circle shown indicates the simulate d 50 mm wafer position. The thickness of the films across the simulated wafer was observed to be fairly uniform in both the gas flow direction and perp endicular to it. It must be noted that the reactor does not currently have wafer rotati on. Without rotation, the thickness uniformity was found to be approximately 4%. Th is value was calculated by using:
91 % 100 %mean Uniformity (3.1) where N i ix x N1 2) ( 1 (3.2) such that x is the average of all thicknesses (mean) and xi is the value of the individual thickness measurement, and N is the number of measurements. Typical thickness uniformities (with wafer rotation) given in the literature are approximately 5% for 50 mm and 35 mm diamet er wafers [19,21]. With the addition of wafer rotation, the thickness uniformity of the process should improve. Table 3.2 Experimental results of growth over a 50 mm wafer area. T ~ 1600 C, P = 150 Torr, QSiH4 = 30 sccm, and Si/C ratio = 1.2. Sample ID USF-05-346 Thickness (m) Growth Rate (m/h) | Nd Na| (cm-3) n-type layers (0004) XRD FWHM (arcsec) A 17.2 34.5 4.55 x 1015 8.0 B 17.3 34.7 5.16 x 1015 16.0 C 18.0 36.0 7.09 x 1015 12.0 D 17.0 34.0 7.53 x 1015 9.7 E 16.0 32.0 4.30 x 1015 10.5 Typical reported doping uniformities across 35 mm diameter wafers [21,67] with wafer rotation are 10 12%. The doping concen tration of the films across the simulated 50 mm wafer slightly increased along the gas flow direction from ~ 5 x 1015 to ~ 7 x 1015 cm-3 as shown in Table 3.2. The doping unifo rmity was found to be approximately 24%
92 using equations (3.1) and (3.2); however, x and xi were the average value and the values of individual measurements of the net carrier concentration, respectively. From the LTPL data presented in Figure 3.16, Al contamina tion is present in the reaction system. Recent investigations show that the graphite adapters are the likel y source of the contamination. It is speculated that Al impurities might be c onsumed at the inlet of the reactor, reducing the amount of Al downstream. This would in turn reduce the acceptor concentration and produce a higher n-type doping concentration. The structural quality of all the films in the simulated 50 mm growth run is very good. The FWHM values of the rocking cu rve of the (0004) X -ray diffraction peak ranged between 8 and 16 arcsec. The films placed in the A and B positions on the polycrystalline plate resulted in the larg est difference in the structural quality. Representative rocking curves from the simu lated wafer experiment are shown in Figure 3.27. (a) (b) Figure 3.27 X-ray diffraction rocking curves of 4H-SiC epitaxial films grown for 30 minutes placed on the polycrystalline plate in (a) sample placement A and (b) sample placement B from Figure 3.26. Data provided by Dr. Y. Shishkin, University of South Florida.
93 In Figure 3.27(a), a rocking curve of the film placed in position A on the polyplate (Figure 3.26) is show n. The FWHM (~ 8 arcsec) of this film was the lowest value of the five samples. The FWHM fo r the sample in Figure 3.27(b) (sample placement B) was the highest value of 16 ar csec. The lower quality of the film placed in position B may have been caused by the s ubstrate. If there were initially a higher amount of defects on this substrate, which could be possible sinc e micropipes tend to cluster, they could have propa gated through the epitaxial la yer, causing a lower quality film compared to the other samples in th is experiment. While the FWHM value for sample placement B is twice th at of the film set in placem ent A, the peak is still fairly narrow indicating a good quality film. 3.6 Summary The development of a baseline 4H-SiC ep itaxial growth process in the USF 200 mm hot-wall CVD reactor has been completed. The obtained growth rate of 32 m/h is the highest reported for this type of reactor to date at temperatures around 1600 C . It was demonstrated that a tilted susceptor ceiling was necessary to enable a uniform thickness deposition (~4 %) across the hot-z one. Simulated velocity distributions predicted the boundary laye r to be higher at the entrance of the tilted susceptor design to that at the back end. With an increase in silane flow, the growth rate increased, as expected; however, the morphology de graded when excessive flows ( 36 sccm) were used. In addition to silane flow, the growth rate increased with a decrease in process pressure, ultimately producing the best surf ace morphology at a process pressure between
94 100 and 150 Torr. 150 Torr was c hosen as the baseline the pr ocess pressure since this would allow for higher process flows in the future (e.g. 100 mm wafer growth process development). The pump is only capable of ac hieving a certain process pressure with certain flow rates. A process pressure of 150 Torr would be achievable with the pump for higher flow rates, however, a pressure of 100 Torr may not be obtainable. A 65 m thick film was grown and the surface roughness was found to be 0.32 nm RMS, which is a comparable surface roughness value to films, of approximately the same thickness, in the literature. The structural quality of the 65 m thick film was evaluated by an X-ray diffraction rocking curve of the (0004) peak The FWHM value was 11 arcsec, indicating a high quality film. The uniformity in carrier concentration and thickness was 24% and 4%, respectively, for films of ~ 17 m thickne ss grown to simulate a 50 mm wafer. The thickness uniformity is hi gh given the reactor does not have wafer rotation.
95CHAPTER 4 4H-SIC EPITAXIAL GROWTH USIN G HCL AS A GROWTH ADDITIVE The main goal of this dissertation research has been to increase the growth rate of 4H-SiC in the horizontal hot-w all CVD reactor. It was detailed in Chapter 3 that the growth rate could be increased from 4 m/h to 32 m/h by optimizing the input parameters and tilting the susceptor ceiling. Th ere are other ways to increase the growth rate further while maintaining smooth mor phology, for example, by elevating the growth temperature, changing the process pressure, or changing the carrier gas flow rate of the system. The most effective method is to increase the SiH4 flow rate, which, in turn, provides more Si species at the substrate su rface to supply the r eaction which increases the growth rate. When excessive amounts of SiH4 are present, gas phase homogeneous nucleation commonly occurs as discussed in Section 3.5.1. To overcome gas phase homogeneous nucleation, HCl was added to the process chemistry. The details and results of the addition of this chlorinated precursor to the growth process will be presented in this chapter.
964.1 Proposed HCl Mechanism In order to increase the growth rate, th e supply of the silicon containing precursor, in this case SiH4, may be elevated to provide more Si vapor to feed the growth reaction. However, it was shown in Sec tion 3.5.1 that with higher SiH4 flow rates, the film morphology deteriorated. As the SiH4 flow rate is increased, gas phase homogeneous nucleation may occur producing a supersaturation of Si which may result in Si clusters or precipitates . Si clusters do not aid in the growth of SiC. Instead they reduce the amount of Si reactants available for 2D film growth, thus limiting the growth rate. The precipitates form and may deposit onto the surf ace of the film, crea ting 3D features. One way to prevent homogeneous nucleation from occurring is to increase the growth temperature. By doing this, the Si clusters may evaporate into Si vapor, thus providing more of the reactant for growth . Another way to reduce the gas phase hom ogeneous nucleation of Si is to add chlorine (Cl) into the H2-C3H8-SiH4 gas mixture . For this dissertation work, hydrogen chloride (HCl) was selected as the Cl precursor. Two mechanisms were proposed where HCl may be used, in addition to an increase of silane flow rate, to produce elevated epitaxial growth rates wh ile maintaining smooth films. The first proposed mechanism is that the HCl etches aw ay the 3D particles which may have fallen onto the surface due to homogeneous nucleatio n. This would result in a smooth film surface, albeit at a reduced growth rate. HCl diluted in H2 has been used to pre-treat the SiC surface prior to growth to suppress di slocation-related growth pits caused by selective H2 etching . Xie, et al.  performed a study on the difference between H2 etching and H2 etching with the addition of HCl of SiC, proving that HCl does in fact
97 etch SiC at elevated temper atures of ~ 1400 C. Excess si licon may create Si clusters, which in turn produce a deteriorated film surf ace. Knowing that HCl etches Si and SiC, it is possible that HCl vapor may be etching the particles that fall onto the film surface from the Si clusters, leaving a smooth surface behind. A sketch of this proposed mechanism is shown in Figure 4.1. (a) (b) Figure 4.1 Sketch of the first proposed HCl mechanism where (a) homogeneous nucleation occurs with low etch rate of th e Si clusters on the substrate surface and (b) high etch rate with the addition of HCl into the gas mixture resulting in a smooth surface and no Si clusters. Figure 4.1(a) is a sketch of what may occur during homogeneous nucleation. Without the addition of HCl, Si clusters form in the gas phase and deposit onto the growth surface and create 3D features With the standard chemistry of H2-C3H8-SiH4 at elevated SiH4 flow rates, there may be a low etch rate of the particles, thereby deteriorating the surface of the film. Howeve r, when HCl is added to the standard gas mixture (shown in Figure 4.1(b)), there may be a high etch rate of the precipitates on the surface, due to etching with HCl. As detailed in , H2 with the addition of HCl etches SiC faster than H2 alone. This may explain why the 3D particles are not etched away fast
98 enough when in the presence of the standard ch emistry, but are etched away when HCl is added to the mixture. Also, since there is a higher surface area with the 3D features, the HCl may be able to etch the features at a faster rate th an the flat surface of SiC. The second proposed mechanism is th at the HCl suppresse s the homogeneous nucleation from occurring in the first place. Fi gure 4.2(a) shows a sk etch of the reactions occurring during growth. The precursors diffu se through the boundary layer as described in Chapter 2, Section 2.2. Pyrolysis occu rs above and throughout the boundary layer, in which case the precursors decompose into the separate species providing the Si and C containing reactants. With excess Si precurs ors in the system, homogeneous nucleation is likely to occur. Homogeneous gas phase nucleation may be seen with the naked eye and is viewed as white smoke within the reactor when growth takes place at elevated SiH4 flow rates or at very low temperatures. When HCl is introduced into the gas mixture, or other chlorinated precursors are added, this phenomenon may be suppressed. When the HCl and SiH4 molecules decompose, it is proposed th at the chlorine attaches to the Si atoms to create SixCly compounds. This behavior is sket ched in Figure 4.2(b). Indeed simulations performed by M. Reyes showed th is to be the case thermodynamically . With the Cl combining with the Si, the Si cluster compounds, such as Si2, Si2C, etc, formation is reduced. This eliminates the gas phase homogeneous nucleation of Si at higher partial pressures of SiH4, thereby allowing growth to pr oceed at high growth rates and result in smooth surface morphology films.
99 (a) (b) Figure 4.2 Sketch of second proposed HCl mechanism to suppress homogeneous nucleation. (a) Homogeneous nucleation o ccurs, creating Si clusters and (b) HCl suppresses the gas phase hom ogeneous nucleation of Si. Thermodynamic simulations were used to predict the dominating chemical species within the CVD system when HCl wa s added to the standard chemistry of H2C3H8-SiH4. The simulations were performed using the NASA CEA code . The process parameters of the baseline pr ocess for the 75 mm hot -wall CVD reactor, described in Chapter 3, Table 3.1, were used as the input parameters for the simulations. The standard chemistry thermodynamic simulations without the addition of HCl are shown in Figure 4.3(a) and those with the addition of HCl are shown in Figure 4.3(b). The H2 flow was input as 10 slm and the precursors SiH4 (2% in H2) and C3H8 (3% in H2) were input as 120 and 80 sccm, respectively. The process pressure was 150 Torr and Si/C was 0.33 for both cases. The Si/Cl ratio for th e simulations including HCl was chosen to be 1.0. It was assumed that the same amount of chlorine moles were needed as that of silicon moles to suppress homogeneous gas phase nucleation.
100 (a) (b) Figure 4.3 Thermodynamic simulations of the mole fraction of gaseous species for the chemistry of H2-C3H8-SiH4 in the 200 mm CVD reactor (a) without the addition of HCl gas and (b) with the addition of HCl. Simulations were performed using the NASA CEA code . In Figure 4.3(a) and (b), it can be seen that the dominating species at the growth temperature for the baseline process, ranging between 1800 and 1860 K (1520 and 1580
101 C), are H, solid SiC, C2H2, and CH4. The Si specie has the highest mole fraction for the silicon source. Molecular hydrogen is again no t shown as it has a higher mole fraction than all of the species s hown, ~ 1. The specie Si2C is present in the standard chemistry simulation, however, with the ad dition of HCl, there is no Si2C specie. In the simulations with HCl additive, there are two species w ith both Si and Cl which are SiCl and SiCl2. In Si CVD, the main specie of interest when chlorinated precursors are added is SiCl2 .The Cl may be combining with the Si2C to create these species. Based on the results of the thermodynamic simulations, the CVD gr owth experiments were carried out at elevated precursor partial pressures with the addition of HCl. These experiments will be discussed in the following sections of this chapter. 4.2 HCl Manifold With the prediction that HC l would be effective in a llowing for high growth rates of SiC, an HCl gas source was incorporated into the CVD gas handling system. Careful planning was needed as HCl is a corrosive gas when it comes into contact with water vapor. When hydrogen chloride gas combines w ith moisture, hydrochloric acid (a liquid) is produced, which is also highl y toxic and corrosive. Therefore, the gas lines that would be in contact with the HCl gas were welded to prevent moisture from getting into the lines. Another preventative measure to reduce moisture in the lines was adding a moisture purifier. This was put in the system to le ssen any moisture that could have possibly seeped into the lines from the VCR connecti ons. To reduce the corros ion rate of the gas lines, electropolished stainless steel (SS) tubi ng was used. Electropolished stainless steel
102 is different from standard stainless steel in that it has a much smoother surface, thereby reducing the surface area that the HCl can attack In addition to welding the lines, a purge panel, shown in Figure 4.4, was implemented into the system near the HCl bottle. The panel was used to remove HCl gas from the ga s line and replace it with Ar when the lines were not utilized for extended periods of time. The next paragraph describes the operation of the HCl panel. Figure 4.4 HCl purge panel near the HCl gas bottle which was used to replace HCl with Ar in the gas lines to help prevent co rrosion. Numbers in the picture represent valve numbers. The HCl gas flows from the bottle (not shown; bottom of photograph) through the gas line (labeled HCl gas in Figure 4.4), through a dual-stage pres sure regulator into the moisture purifier and out to the proce ss line. The diaphragm valve was used as a safety valve to isolate the HCl gas panel from the mainstream gas lin e and is actuated by the user by closing a switch on th e CVD control panel. It is closed at all times when the system is not in operation. Detailed operati on of the HCl panel is described in Appendix D.
1034.3 HCl Growth Additive Experiments: 75 mm Reactor The 75 mm reactor, described in Chapter 2, Section 2.4.1, was used for the first experiments using HCl as a growth additive. With all the modifications to the system finished to allow HCl gas to be safely supplied to the reactor, the experimental evaluation of the HCl additive was initiated. The experime ntal study of increasing the growth rate by increasing the precursor flows a nd adding HCl is discussed next. 4.3.1 Experimental Results The goal of adding HCl to the standard chemistry of H2-C3H8-SiH4 was to increase the growth rate, by increasing the SiH4 flow rate, while maintaining a smooth surface morphology with a low amount of surface defects. The theory of adding HCl to the standard chemistry to reduce homogeneous nucleation and increa se the growth rate was investigated. A small amount of HCl gas (100%) was added to the baseline process described in Section 3.2. The flow rate of HCl was set to a very low, but controllable, value of 0.5 sccm with the HCl MFC. This fl ow rate produced a Si/C l ratio of 4.8, which is the ratio of the number of moles of Si and the number of moles of Cl in the inlet gas mixture. Planand cross-section view SEM mi crographs of the films grown are shown in Figure 4.5.
104 (a) (b) Figure 4.5 4H-SiC epitaxial film grown usi ng the baseline process (from Section 3.2) with the addition of 0.5 sccm of HCl in the LP horizontal hot-wall reactor. (a) Plan-view micrograph showing the resulting smooth film surface and (b) cross-section micrograph of cleaved edge showing a film thickness of 6.5 m. (t = 15 min). As shown in the micrograph in Figure 4.5(a), the film surface was smooth and specular. The growth rate usi ng this process was 26 m/h, wh ich is considerably higher than the growth rate of the baseline process of 16 m/h for the 75 mm reactor. With the addition of HCl, there may be an increase in Si vapor to supply the reaction, resulting in a higher growth rate. This would support the th eory that the Si cl usters are not being formed when HCl is added to the growth process, which would provide more Si for deposition. Further experiments were conducted on the baseline process with the addition of HCl to determine the effects on the surface morphology due to increasing the HCl flow rate beyond 0.5 sccm. The highest flow rate of HCl added to the baseline process was 5 sccm resulting in a Si/Cl ratio of ~ 0.5. Pl an-view micrographs of a film grown using the baseline process and the addition of 1.0 sccm of HCl to produce a Si /Cl ratio of 2.4, are shown in Figure 4.6.
105 (a) (b) Figure 4.6 Plan-view micrographs of 4H-S iC epitaxial growth run in the 75 mm horizontal hot-wall reactor with the addition of 1.0 sccm of HCl to the baseline process (see Section 3.2 for details) showing stepbunching at a magnification of (a) 2500x and (b) 10000x. Vg = 14.4 m/h. Data taken by Dr. S Rao, University of South Florida. From the micrograph in Figure 4.6(a), ther e is slight step-bun ching on the surface of the film which may indicate there is t oo much HCl in the gas mixture causing some etching effects. Nakamura, et al.  found that etching of 6H-SiC surfaces with 3 sccm of HCl in 1 slm of H2 at 1300 C for 10 minutes resulted in step bunching. Therefore, at elevated HCl flows during growth of 4H-SiC step-bunching may occur and could be the reason for the step-bunched surface. At a magnification of 10000x (Figure 4.6(b)), there are white dots on the surface which are of higher contrast than the flat surface. Unfortunately, the cause of these white spot s is unknown. The growth rate of the film (14.4 m/h) was slightly lower than the ba seline process growth rate of 16 m/h. The surface of a film grown with 5 sccm of HCl added to th e baseline process was similar to that shown in the micrographs in Figure 4. 6, however, there was a higher amount of stepbunching and no white spots were seen at the hi gher magnification. Al so, the growth rate was reduced even further to ~ 11 m/h. From these experiments, it was determined that
106 the amount of HCl needed to produce a smoot h film using the baseline process was 0.5 sccm. With the baseline process evaluated, the precursor flow rates were doubled from 80 sccm of SiH4 and C3H8 (both 3% in H2) to 160 sccm. This was done for two reasons. The first was to determine how high the growth rate could be increased. With increasing the silane flow rate and keepi ng a constant Si/C ratio, there might be a limit to how high the precursors can be increased while still in creasing the growth rate and maintaining a smooth surface morphology. The second was to determine whether HCl could improve the morphology at higher precursor flow rates. At a certain flow rate of precursors, the HCl might not suppress the gas phase homog eneous nucleation which would result in deteriorated film surfaces. The growth parameters for the increas ed precursor flow experiments were maintained at T = 1535 C, P = 150 Torr, and Si/C = 0.33. Th e initial growth was carried out using the same amount of HCl flow as the one found in the baseline process (0.5 sccm). The resulting film of the experiment was polycrystalline in structure and a planview image of the film surface is shown in Fi gure 4.7(a). The growth rate of the film was 33 m/h, which is double the baseline process growth rate. Although this is a high growth rate, the film is polycrystalline in nature.
107 (a) (b) (c) Figure 4.7 SEM micrographs of 4H-SiC epitaxial films grown at QSiH4 and QC3H8 (3% in H2)) of 160 sccm with the a ddition of HCl . (a) 33 m/h growth rate with 0.5 sccm of HCl, (b) 31 m/h growth rate with 2.0 sccm of HCl, (c) 28 m/h growth rate with 2.5 sccm of HCl flow. Data of USF-04-082 taken by Dr. S. Rao, University of South Florida. Since the epitaxial layer in Figure 4.7(a) was polycrystalline, the HCl flow rate was increased further to 2.0 sccm. The growth rate of the film decreased slightly from 33 to 31 m/h. With the increased HCl flow rate the surface of the epitaxial layer was much smoother (see Figure 4.7(b)). There was still a wavy texture to the su rface, but the film was not polycrystalline in stru cture. The amount of HCl was th en increased further to 2.5 sccm. The growth rate for this experiment was 28 m/h Although the growth rate slightly decreased from 33 to 28 m/h with the increase of HCl flow from 0.5 to 2.0 sccm, the surface morphology of the film grown was very smooth (see Figure 4.7(c)). Therefore, the HCl mechanism of reduci ng homogeneous gas phase nucleation at elevated SiH4 flow rates to produce smooth mor phology films was shown to be the reason behind the increased growth rate with HCl. The precursor flow rates could be increased to 160 sccm of SiH4 and C3H8 (both 3% in H2), resulting in an overall increase of growth rate from the base line process (16 m/h) in the 75 mm hot-wall CVD reactor to 28 m/h.
108 The SiH4 precursor was again increased by an additional 25% to a flow of 200 sccm (3% in H2), while maintaining the Si/C ratio of 0.33, in order to determine whether the growth rate could be improved further. The H2 carrier gas flow a nd process pressure were maintained at 10 slm and 150 Torr, respectively. The HCl flow rate was set to 2.5 sccm, however, the film resulted in large mosaic structures over the entire surface. As seen in Figure 4.8, the morphology was much di fferent than that of the film grown at a SiH4 flow rate of 160 sccm. The surface was hi ghly mosaic, if not polycrystalline, with large grain boundaries, as evident from th e 2500x magnification image in Figure 4.8(b). (a) (b) Figure 4.8 Plan-view SEM micrographs at a magnification of (a) 100x and (b) 2500x of an epitaxial film grown for 15 mi nutes in the LP hot-wall reactor. QSiH4 and QC3H8 were both 200 sccm (3% in H2) and QHCl = 2.5 sccm. Vg = 46 m/h. Data taken by Dr. S. Rao, University of South Florida. In an effort to improve the morphology of the film grown at a SiH4 flow rate of 200 sccm (3% in H2), the amount of HCl ga s was increased. Polycr ystalline films were produced for each HCl flow rate attempted, be tween 2.5 and 6.0 sccm. Therefore, it was determined that at high precursor flow rate s, the morphology could not be improved with the addition of HCl. This may indicate that the SiH4 flow rate limit at this temperature has been reached. The growth ra te was ~ 56 m/h with the SiH4 precursor flow rate at
109 200 sccm (3% in H2) and the HCl flow rate at 4.0 sccm. The experimental results of the HCl additive experiments are presented in a graph shown in Figure 4.9. Figure 4.9 Growth rate of 4H-SiC as a functi on of silane flow rate for three separate experimental sets with the addition of HCl . T = 1530 C and P = 150 Torr. Curve is used to aid the eye only. As shown in Figure 4.9, the silane flow rate was increased from 80 sccm to 200 sccm (3% in H2). The H2 carrier gas, process temperature, and Si/C ratio for the experiments were maintained at 10 slm, ~ 1530 C, and 0.33, respectively. As seen in Figure 4.9, with an increase in silane flow rate, the growth rate of the 4H-SiC also increased. All films had smooth, specular surface morphology except for the film grown with a SiH4 flow rate of 200 sccm (3% in H2). There are three experimental sets shown in Figure 4.9. The first experimental set was conducted to determine whether the HCl additive process could increase the growth ra te. The second set was to provide results for the SiH4 flow rates between 80 sccm and 140 sccm th at were not evaluate d in the first set.
110 The third experimental set was performed to investigate the reliability of the growth rates with higher precursor flow rates. The amount of HCl needed to prod uce smooth epitaxial layers at QSiH4 between 80 and 160 sccm (3% in H2) are shown in a graph in Fi gure 4.10. The amount of HCl needed to maintain smooth film surfaces increased when the SiH4 flow increased, as expected. The HCl flow rate for the 80 sccm flow rate of SiH4 (3% in H2) experiments remained relatively constant. At QSiH4 of 100 and 120 sccm (3% in H2), the HCl flow rate varied considerably between the three experiments for each SiH4 flow. The reason for the high error in HCl flow may have been that the HCl MFC was starting to corrode and the actual flow rate of HCl was di fferent than what was believed to be flowing. Experiments for the higher SiH4 flow rates of 140 and 160 sccm were not repeated as there were MFC problems due to corrosion and the HCl system was taken down for repair. Figure 4.10 HCl flow rate as a function of silane flow rate for the experiments displayed in Figure 4.9. Process temperatur e and pressure were ~ 1530 C and 150 Torr, respectively. Curve is used to aid the eye only.
111 The growth rate as a function of HCl flow rate was also investigated in the 75 mm hot-wall reactor and the results are shown in Figure 4.11. The process conditions were maintained for all experiments at QSiH4 = 80 sccm (3% in H2), T = 1535 C, P = 150 Torr, and Si/C = 0.33. The HCl flow rate was varied between 1 and 5 sccm. As seen in Figure 4.11, the growth rate increased with an initial increase of HC l flow from 1.0 to 2.5 sccm. The maximum growth rate of ~ 23 m/h wa s produced with an HCl flow rate of 2.5 sccm. As the HCl flow was increased above 2. 5sccm, the growth rate began to decrease, indicating the HCl was beginning to etch the film surface. The morphology of the films grown with an HCl flow rate between 1.5 s ccm and 2.5 sccm was smooth and specular. However, at an HCl flow rate of 3.0 sccm, gr owth pits were present on the film surface. Finally, at an HCl flow rate of 5.0 sccm, th e resulting film surface displayed severe stepbunching. Figure 4.11 Growth rate of 4H-SiC as a functio n of HCl flow rate in the 75 mm hotwall CVD reactor. QSiH4 = 80 sccm (3% in H2), T = 1535 C, P = 150 Torr, and Si/C = 0.33. Curve is to aid the eye only.
112 This data is in general agreement with wh at is normally encount ered in the growth of Si films using chlorinated chemistry. Figur e 4.12 shows an example of Si growth using silicon tetrachloride (SiCl4), where a clear boundary betw een growth and etching has been shown for reference. While a similar pl ot can be generated from this dissertation work with additional experiments, resources did not permit this to be achieved and it is comforting to note from Figure 4.11 that the same qualitative behavior in SiC epitaxy with Cl has been observed. Figure 4.12 Growth rate of Si versus mole fraction of SiCl4 in H2 . Note that the growth rate increases rapidly for mole fractions below 0.1, then increases steadily for higher mole fractions. Etching occurs for mole fractions above ~ 0.275. This data is in qualitative agreement with Figure 4.11 for 4H-SiC growth usi ng HCl as a growth additive. 4.3.2 Characterization (AFM, XRD, and LTPL) The structure of the sample shown in Fi gure 4.7(c) was analyzed using XRD to confirm that the film was single crystal. The samples, whose micrographs are shown in
113 Figure 4.7(a) and (b), had polycrystalline and wavy surface structures, respectively. If the epitaxial film shown in Figure 4.7(c) was actually polycrystalline and the HCl was only smoothing the surface, then the XRD scan would show a diffraction pattern for a polycrystalline film. The 2 / scan, which is an X-ray diffr action scan used to determine the phase of the crystalline f ilm, is shown in Figure 4.13. Figure 4.13 XRD 2 / diffraction scan of the sample shown in the SEM micrograph of Figure 4.7(c). The reference pattern for a polycrystalline 4H-SiC film is shown by black vertical lines and the labels indicate th e possible reflecting planes. This experiment indicates the film to be single crystal. The diffraction peak shown in Figure 4.13 is indicative of a single crystal film. The vertical straight lines in Figure 4.13 are the reference peaks for a polycrystalline 4HSiC film. As seen in the 2 / scan, there is only one peak which is seen at the (004) reflecting plane. The peak is broad because this was a low resolution scan to test the crystallinity of the film. If there were more diffraction peaks the sample would be polycrystalline in nature. It can be stated from these results that the HCl is not etching polycrystalline films into sm ooth, flat surfaces. The films are actually single crystal.
114 Comparisons between films grown with and without the addition of HCl as a growth additive were made using AFM anal ysis of the grown films. The process parameters for both samples were from the ba seline process described in Section 3.2, which were T = 1530 C, P = 150 Torr, and QSiH4 (3% in H2) = 80 sccm. The AFM surface scans of the films are shown in Figure 4.14. The surface roughness of the sample grown (4.4 m thick) without HCl was 0.30 nm RMS, while that of the sample grown (3.9 m thick) with 1.0 sccm of HCl a dditive was 0.25 nm RMS. These films were chosen to be analyzed as they were representative samples for the baseline process and the baseline process with HCl additive. From the surface roughness values, it may be concluded that the addition of HCl has little to no effect on the films surface roughness. In fact, it may be slightly improving the film su rface. It is important to note that this was a comparison of only one experiment from each process (with and without HCl additive). To make a definitive conclusion, more statis tical data would be needed. However, the resources were not available for additional experiments. The film thickness was thinner for the fi lm grown with 1.0 sccm of HCl compared to that grown without the addition of HCl to th e baseline process. This is contradictory to the results of the film presented in Figure 4.5, in which the growth rate increased from 16 m/h for the baseline process to 26 m/h when HCl was added.
115 (a) (b) Figure 4.14 AFM surface scans of 4H-SiC ep itaxial films grown for 15 minutes  (a) without the addition of HCl and (b) with the addition of 1.0 sccm of HCl, resulting in surface roughness of ~ 0.30 and ~ 0.25 nm RM S, respectively. The AFM scan area was 10 m x 10 m. AFM scan of (a) was taken by Dr S. Rao, University of South Florida. The structural quality of films grown w ith and without the addition of HCl were investigated using XRD. X-ray rocking cu rves of the (0004) di ffraction peak were analyzed to determine whether the addition of HCl to the standard chemistry degraded the film quality. A representative 9 m thick film grown with the baseline process in the 75 mm reactor, described in Ch apter 3, Section 3.2, was analyzed. A sample (3.4 m thick) which was grown with the baseline pr ocess and the addition of 1.0 sccm of HCl was also investigated. The X-ray rocking curv es for both films are shown in Figure 4.15.
116 (a) (b) Figure 4.15 X-ray rocking curve of the ( 004) diffraction peak fo r 4H-SiC epitaxial films. (a) Sample grown (USF-04-088) with the baseline process described in Section 3.2. (b) Sample grown with baseline proce ss and the addition of 1.0 sccm of HCl. The FWHM of the (0004) diffraction p eak was 9 arcsec for the film grown without HCl, indicating a high quality film The film grown with HCl had a FWHM value of the (0004) X-ray diffraction peak of 22 arcsec. The thickness of the film grown with HCl is much thinner than that of the film grown wit hout HCl. This may have been the reason for the higher FWHM value of the diffraction peak. As seen in Figure 4.15(b), there is a double diffraction peak, which may be due to diffraction from both the thin film and the substrate. The structural analysis of a film grown at a higher silane flow rate of 140 sccm (3% in H2), which resulted in a growth rate of ~ 21 m/h, was also analyzed using XRD. The amount of HCl needed to produce a smooth film morphology was 2.25 sccm. An Xray rocking curve of the film is shown in Figure 4.16. The FWHM value of the (0004) diffraction peak was approximately 19 arcsec. This is comparable to the FWHM value (22 arcsec) from the diffraction peak of the sample grown with the baseline process and the addition of 1 sccm of HC l presented above, which had a growth rate of ~ 13 m/h.
117 The increase in SiH4 flow (which in turn increased th e growth rate) and HCl flow rates appear to have little effect on the structural quality of the 4H-SiC ep itaxial films from the two experiments conducted. The growth rate increased from 13 m/h to 21 m/h by increasing the silane flow rate and increasing the HCl flow ra te to produce smooth films. With this increase, the quality remained relatively the same. Figure 4.16 X-ray rocking curv e of the (0004) diffraction pe ak of a 4H-SiC epitaxial film grown at a QSiH4 (3% in H2) of 140 sccm with the addition of QHCl at 2.25 sccm. The film thickness was 4.7m. Low temperature PL, collected at a temp erature of 2K at the University of Pittsburgh (Professor W. J. Choykes group), wa s used to characterize the films as well. The 9 m thick film grown using the baseline process of the 75 mm reactor without the addition of HCl, whose XRD rocking cu rve was shown in Figure 4.15(a), was characterized using LTPL. A 6.4 m thick film grown using the baseline process with the addition of 0.5 sccm of HCl was also invest igated. The near bandedge spectra of the two films were overlayed on one another and s hown in Figure 4.17. The two near bandedge
118 spectra are similar within this wavelength ra nge, indicating that th e addition of HCl may not change the quality of the epitaxial film. Figure 4.17 Near bandedge LTPL spectra of 4H-S iC epitaxial layers (collected at 2K). Sample USF-04-088 was grown using the base line process described in Section 3.2, producing a thickness of a 9 m. The spectru m was overlaid on the spectrum of sample USF-04-089 (6.4m thick) grown with 0.5 sccm of HCl. Data courtesy of Professor W. J. Choyke, University of Pittsburgh. The LTPL spectrum at longer wavelengths, between 3980 and 4300 , for samples USF-04-088 and USF-04-089 ar e shown in Figure 4.18. The L1 defect line is present in the spectrum of the 4H-SiC epitaxi al film grown with the baseline process from Section 3.2. However, when 0.5 sccm of HCl gas was added to the same process, the L1 defect line was not observed in the LTPL spectrum. The L1 defect line is a zero phonon line that appears in the spectra at 2.901 eV in 4H-SiC. This defect complex may be formed from antisites, interstitials, or Si vacancies . Without the incorporation of HCl into the gas mixture, there may be more Si vacancies. This may be due to the high
119 concentration of Si2C present, shown in the thermodynamic simulations in Figure 4.3(b). However, when HCl was added to the gaseous mixture, the Si2C species were not present in the simulations, as shown in Figure 4.3(a). This may indicate that the Si vacancies are decreasing within the epitaxial films, thus resulting in a reduced/eliminated the L1 defect line in the LTPL spectrum. (a) (b) Figure 4.18 2K LTPL spectra at longer wavelengths, compared to Figure 4.17, showing (a) the presence of the L1 defect line in the sample grown without HCl and (b) no L1 defect line present for 4H-SiC epitaxial f ilm grown with the addition of 0.5 sccm of HCl. Data provided by Professor W. J. Choyke, University of Pittsburgh. Characterization was also performed on three samples to make a definitive analysis whether the growth rate and HCl additive deteriorate the quality of the films. The corresponding near bandedge spectra are shown in Figure 4.19. The 4H-SiC samples in Figure 4.19(a), (b) and (c) were grown with; a QSiH4 of 80 sccm with the addition of 1 sccm of QHCl, a QSiH4 of 100 sccm with the addition of 1.25 sccm of QHCl and a QSiH4 of 140 sccm with the addition of 2.25 sccm of QHCl, respectively. Comparing Figure 4.19(a), (b) and (c), the near bandedge emission does not change with an increase of growth rate.
120 Also noted was that the L1 defect line (spectra not shown) was reduced or eliminated with the addition of HCl for all samples. (a) (b) (c) Figure 4.19 Near bandedge LTPL spectra taken at 2K of 4H-SiC epitaxial films grown using HCl as a growth additive. (a) Sample USF-04-170 with a Vg of 13 m/hr, (b) sample USF-04-167 with a Vg of 18 m/hr, and (c) sample USF-04-163 with a Vg of 21 m/hr. Data courtesy of Professor W. J. Choyke, University of Pittsburgh. Although the results from the characteri zation of the 4H-SiC epitaxial films grown with the addition of HCl were very promising, modifications to the CVD reactor had to be made. The amount of HCl needed to produce smooth film surfaces varied from one experimental set to a nother when the same SiH4 flow rates were used. Therefore, a new HCl manifold had to be incorporated into the system to provide more reliable HCl flows. Along with adding the HCl gas mani fold, the reaction tube was enlarged to accommodate 50 mm wafers. 4.4 HCl Growth Additive Experiments: 200mm Reactor The 75 mm diameter reaction tube was enla rged to a 200 mm diameter. Details of the reactor modification were presented in Chap ter 2, Section 2.4. This enabled the ability to grown on wafers up to 100 mm in diameter, although the 200 mm reactor was
121 configured to accommodate 50 mm during th is research. While upgrading the reactor hardware, the HCl gas manifold was replaced an d a methyl chloride gas line was added to the same manifold (which was kept separate from the main manifold to reduce corrosion of the other gas lines). With the new HCl gas manifold and larger reaction tube, HCl experiments were continued and the results are detailed next. 4.4.1 Growth Rate as a Function of Silane Flow During these experiments, the SiH4 and C3H8 precursors were 100% concentrations instead of the 2% and 3% mixes in H2, respectively. Th e development of 4H-SiC epitaxial growth with the addition of HCl in the 200 mm reactor began by adding HCl to a process that was similar to the base line process detailed in Chapter 3, Section 3.4. The only difference was that the SiH4 precursor was increased from 30 sccm to 36 sccm, while all other process parameters were held constant. Initially, the amount of HCl added to the process was 15 sccm to ach ieve a Si/Cl ratio of 2.4, which was approximately what was used in the 75 mm react or process. The flow of HCl was varied from 5 sccm to 17 sccm, all resulting in de graded surfaces of the grown films compared to the films grown using the baseline proc ess without HCl gas. When an excessive amount of HCl was added to the process with 36 sccm of SiH4 flow, there was etching of the film surface. An example of a sample with excessive etch ing is shown in Figure 4.20. The morphology of the film is nearly the same as that shown in Figure 3.11, which was caused from excess precursor flows. Compar ing the two images, it seems as though the
122 surface morphology for excess etching and excess precursor flow rates result in the same morphology. (a) (b) Figure 4.20 Plan-view SEM micrographs of a 4H-SiC epitaxial film grown in the 200 mm hot-wall reactor with 17 sccm of HCl add itive at a magnification of (a) 2500x and (b) 10000x. The Si/C and Si/Cl ratios were 1.0 and 2.0 respectively. T ~ 1550 C. Data taken by Dr. Y. Shishkin, University of South Florida. It was then determined that it would be best to start out with the baseline process and then slowly add HCl to it, while maintaining a smooth surface morphology. After finding the 200 mm reactor process with HCl a dditive, then the precursor flows could be increased. This was how the experimental flow occurred for the 75 mm reactor, outlined in Section 4.3.1, so it was logical to follow the same approach for the new reactor. The amount of HCl added to the base line process (30 sccm of SiH4) that produced a smooth film morphology was 7.5 sccm, with a Si/Cl of 4.0. The SiH4 precursor flow rate was then incr eased to 35 sccm. The amount of HCl needed to produce a smooth morphology film was 8.75 sccm. The flow rates were incrementally increased by 5 sccm until a SiH4 flow rate of 45 sccm was achieved. When the precursor flows were increased beyond th is point, the film mo rphology could not be
123 improved with the addition of HCl. The resultant morphology of films grown beyond a flow rate of 45 sccm was similar to that in Figure 4.20. Plan-view images of samples grown with a SiH4 flow of 45 sccm are shown in Figure 4.21. (a) (b) Figure 4.21 Plan-view SEM microgr aphs of 4H-SiC grown with a QSiH4 of 45 sccm. The amount of HCl added was (a) 9.75 s ccm resulting in a rough surface with a Vg of 44 m/h and (b) 10.25 sccm resulting in smooth surface with a Vg of 49 m/h. Data taken by Dr. Y. Shishkin, University of South Florida. The micrograph in Figure 4.21(a) displays a film surface that has large grooves and is similar to that in Figure 3.11 i ndicating that not enough HCl was added. The surface of the film shown in Figure 4.21(a) ma y be caused from Si clusters in the gas phase which deteriorated the film morphology. When a slight ly higher flow of HCl (from 9.75 to 10.25 sccm) was added, the film morphology became smooth, as shown by a plan-view SEM micrograph in Figure 4.21(b). The growth rate also increased from 44 m/h to 49 m/h with the elevated HCl flow. This may be indicative that the suppression of Si clusters had been achieved.
124 The results of the growth rate as a function of SiH4 flow are displayed in the graph in Figure 4.22. All film morphologies were similar to the morphology shown in the plan-view micrograph in Figure 4.21(b). Figure 4.22 200 mm reactor growth rate as a function of SiH4 flow rate. The HCl flow rates needed to produce films of smooth morphology are displa yed on the right vertical axis. Solid trend line to aid eye only. As shown in Figure 4.22, the growth ra te increased with an increase in SiH4 flow rate. The amount of HCl flow needed to produce smooth morphology films also increased with an increase in silane flow rate, as expected. With an in crease of silane flow by 50%, the growth rate increased by ~50% and the HCl needed for the process increased by ~40%. The highest growth rate achieved in the 200 mm reactor with smooth morphology was 49 m/h. Unfortunately, the HCl MFC fa iled again, so the experiments were not pursued further. The growth rate might ha ve been increased beyond this value had the process been stable. With an increase in SiH4 flow rate, an increase in HCl flow was needed as well. The ratio of Si/Cl was not constant for the increase in SiH4 flow rate. The Si/Cl ratio was maintained at 4.0 for the experi ments performed with a silane flow rate of
125 30 and 35 sccm. However, the ratio increased to 4.1 for a SiH4 flow of 40 sccm and increased additionally to 4.4 for a SiH4 flow of 45 sccm. This implies that Si/Cl ratio needed to suppress the gas phase homoge neous nucleation varies for higher SiH4 flow rates. The thickness of each sample presented in the graph in Figure 4.22 was measured using FTIR to confirm the thickness of th e HCl samples grown in the 200 mm reactor. The method used to determine the correct film thickness is deta iled in Appendix E. A comparison of an FTIR spectra compared to a cross-section SEM micrograph, used to determine the thickness of a fi lm is shown in Figure 4.23. (a) (b) Figure 4.23 Thickness results for a 4H-SiC epitaxial film grown with the 200 mm reactor baseline process with 30 sccm of SiH4 and 7.5 sccm of HCl. (a) Cross-section SEM micrograph of bottom edge showing thickn ess of 8.12 m and (b ) Fourier transform of the reflectance spectrum showing thickness of ~ 8 m. The thickness of a 4H-SiC film, using the 200 mm baseline process with a SiH4 flow of 30 sccm and an HCl flow of 7.5 sccm, determined by cross-section SEM micrograph and FTIR are in very close agr eement. The percent e rror between the two measurements was approximately 3%. The same comparison was made for the other
126 three sample thicknesses presented in Figur e 4.22. The percent error of thickness comparing the cross section SEM micrograph and the FTIR spectrum techniques are ~ 8% for the 35 sccm of SiH4 sample, 2.5% for the 40 sccm of SiH4 sample, and ~ 1.5% for the 45 sccm of SiH4 sample. Therefore, from the comparison of these measurements, it can be concluded that the th icknesses of the HCl samples presented in Figure 4.22 are correct. Electrical, structural and surface character ization of the films presented in Figure 4.22 was performed. The results are presented in the next section. 4.4.2 Characterization To determine whether the increased growth rate and the addition of HCl had an affect on the doping of the samples, the net ca rrier concentration of the samples from the experiments described in S ection 4.4.1 was evaluated us ing the C-V technique. The results are shown in Figure 4.24 and summari zed in Table 4.2. From the graph, the doping of the films remained relatively consta nt with an increase in growth rate. The doping concentration was found to be in the low 1016 cm-3 range. Comparing the doping of a baseline process sample to a sample with the addition of HCl, the carrier concentrations are similar. This indicates th at the addition of HCl used to increase the growth rate has little effect on the net carrier concentration of the films.
127 Figure 4.24 Net carrier concentration of 4H -SiC films grown with the addition of HCl at elevated growth rates. Th e closed-circle data points ar e the experiments from Figure 4.22. The triangle data point is the doping dens ity of a baseline process sample. The trend line is to aide the eye only. The surface roughness of a film grown with the baseline process ( Vg = 32 m/h) was compared to that of a film grow n with HCl additive at a higher SiH4 flow rate of 45 sccm. The corresponding AFM scans are shown in Figure 4.25. The film grown with HCl ( Vg = 49 m/h) was produced from an HCl fl ow of 10.75 sccm. The surface is very smooth for the baseline process sample. The sample grown with added HCl does have a groove on the surface which could have been a defect extended from the substrate. The surface roughness values for the films grown w ithout and with HCl additive were 0.6 nm RMS and 0.9 nm RMS, respectively. Compar ing the two films, the surface roughness did not increase by much for the higher growth rate film.
128 (a) (b) Figure 4.25 AFM surface scans of 4H-SiC epitaxial films grown in the 200 mm LP hot-wall CVD reactor. Film grown (a) using baseline process (USF-05-099) and (b) with SiH4 flow of 45 sccm and HCl flow of 10. 25 sccm (USF-05-360). The resulting surface roughness values were 0.6 nm RMS and 0.9 nm RMS, respectively. Process T ~ 1600 C. The structural quality of the films was analyzed using XRD. The results of this analysis are displayed in Table 4.2. The FWHM values of the (0004) diffraction peak ranged between 8.3 and 13.6 arcsec for SiH4 flow between 30 and 45 sccm. The increased growth rate does not seem to be a large factor in the quality of the films. The (0004) diffraction peaks do broaden slightly with an increase in growth rate, however, the FWHM values prove the 4H-SiC epit axial films to be of good quality.
129 Table 4.1 200 mm CVD reactor film properties versus growth rate. Sample ID QSiH4 (sccm) QHCl (sccm) Vg 1.) (m/h) Net Carrier 2.) Concentration (cm-3) XRD3.) FWHM (arcsec) USF-05-347 30 7.5 31.2 1.89 x 1016 8.30 USF-05-348 35 8.75 36.0 2.24 x 1016 10.5 USF-05-350 40 9.75 40.0 2.76 x 1016 12.6 USF-05-360 45 10.25 49.0 2.08 x 1016 13.6 T ~ 1600 C, P = 150 Torr, and Si/C =1.0. 1.) As determined by cross-section SEM analysis 2.) As determined by the C-V technique 3.) As determined by XRD rocking curve analysis of the (0004) plane. The rocking curve of the film with the highest growth rate of 49 m/h is shown in Figure 4.26. The FWHM of the (0004) diffr action peak is 13.6 arcsec. This is a reasonable value compared to other films grown in hot-wall CVD reactors from the literature [18,24]. The FWHM va lues of the (0004) diffraction peak for other films grown with high growth rates during this work are slightly lower than 13.6 arcsec.
130 Figure 4.26 X-ray diffraction rocking curve of the (0004) diffraction peak for the 4HSiC film with the highest growth rate (49 m/h), indicated in Figure 4.22. The FWHM value for the 12 m thick film is 13.6 arcsec Data taken by Dr. Y. Shishkin, University of South Florida. 4.5 Summary The growth of 4H-SiC epitaxial film s with the addition of HCl has been demonstrated in a hot-wall CVD reactor. The growth rate has been increased from the baseline process of 30-32 m/h to 49 m/h in the 200 mm reactor. This has been accomplished by increasing the partial pressure of the precursors and adding HCl to the process. When the SiH4 flow rate was increased, the growth rate increased when the correct amount of HCl was added to the pr ocess (shown in Figure 4.21). The mechanism for this effect was proposed, in which the HCl suppresses gas phase nucleation of Si. Comparing the baseline process without and with the addition of HCl, the surface roughness only slightly increased. The film quality did not degrade with the introduction of HCl as shown by LTPL analysis (for th e 75 mm reactor process) and XRD (for both the 75 mm and 200 mm reactor processes). Over all, the growth rate increased by approximately 1.5 times the baseline process for both reactors (the 75 mm and the 200
131 mm). With the malfunctioning of the HCl MFC, the growth rate investigation could not be pursued further. Future work regarding the higher growth rates will be discussed in Chapter 5.
132CHAPTER 5 SUMMARY AND FUTURE WORK 5.1 Summary The development and characterization of a high-growth rate 4H-SiC epitaxial growth process in a 200 mm horizontal hot-w all CVD reactor was completed during this dissertation work. The process development fo cused on the growth rate and the surface morphology of the grown films. Characterization of the f ilms was carried out using optical (LTPL), electrical (CV-IV) and struct ural (X-ray, AFM and SEM) analysis. Using standard SiC CVD chemistry to develop a ba seline CVD process resulted in the highest growth rate (32 m/h) to date using a horizontal hot-w all reactor at te mperatures around 1600 C. Epitaxial layers were also grown by introducing hydrogen chloride into the gaseous mixture which permitted an increase in the silane flow rate and, thus, a maximum growth rate of 49 m/h. Thermodynamic simulations for the base line process were carried out to determine the dominant chemical species in the reacting system. The simulations were also performed on the growth process whic h included the addition of HCl gas. The simulations were used to predict if homogene ous gas phase nucleation of Si was expected in the system for the growth parameters studi ed. The analysis was able to predict where
133 the highest concentrations of the dominant chemical species (SiH2, Si, and CH4) were developing within the reactor. This type of information could be helpful if growth was not occurring within the hot-zone. By predicti ng which species are generated within the hot-zone, one can decide whether the temperatur e or pressure, or any other parameter is needed to be changed to improve the growth. A growth rate of 16 m/h was previously obtained using the 75 mm hot-wall reactor. The downside to this reactor size wa s that the maximum substrate size was only 2 cm in diameter. During this research the re actor diameter was scaled up to accommodate substrates up to 100 mm in diameter, howev er, the reactor was configured for 50 mm diameter substrate sizes for the experime nts performed. A growth process for 100 mm substrates will be developed in the future The development of the 4H-SiC baseline process for the 200 mm reactor was initiated by using the 75 mm hot-wall reactor process parameters developed previously. The proc ess development of the 200 mm reactor entailed changing various process parameters su ch as pressure, silane flow rate, and Si/C ratio. The characterization of the grown film s, by SEM analysis, was used to adjust the process parameters from run to run, to fi nally achieve films with smooth morphology at the highest growth rates possible. LTPL and XRD were later used to assess the optical and structural quality of the films, respect ively. The surface morphology of the films was smooth and specular, with little to no surface defects observed for films of thickness less than 30 m. AFM analysis was performed on thick films (30 and 65 m) and it was found that the surface roughness was approximately 0.3 nm RMS for both films which is quite respectable. When thicker films were produced (> 30 m), occasional defects were seen on the surface such as carrot-like and tria ngular defects which are known to develop
134 with thicker films. The LTPL res ponse was measured at 2K on a 23 m thick film. Near band-edge spectrum showed a strong intrinsi c peak, which is an indication of good crystalline quality. The four-p article Al complex peak was also seen via LTPL indicating Al contamination of the films was occurring during growth. The contamination source is still unknown and should be investigated in the future. The X-ray rocking curves of the (0004) diffraction peak were sharp with the lo west FWHM value being 9 arcsec, which is comparable to the best values reported in the literature (~ 5 arcsec ). After the development of the baseline pr ocess in the 200 mm hot-wall reactor, the process was altered to achieve even highe r growth rates. This was accomplished by increasing the silane flow rate, which in turn increases the growth rate. There are some issues with increasing the silane flow alone. When excessive amounts of SiH4 is injected into the growth zone, a supers aturation of gas phase reactions typically occur creating Si clusters which are detrimental to the epitaxial surface morphology. To reduce or eliminate this problem, HCl was introduced into the standard H2-C3H8-SiH4 chemistry. By increasing the silane flow rate and in troducing HCl gas, the growth rate was incrementally increased from 32 to 49 m/h. The morphology and structural quality of the films were found to be similar to that of the baseline process. Unfortunately, there were some limitations to this process. Th e HCl mass flow cont roller (MFC) and other plumbing parts were corroding due to the high ly corrosive HCl gas. Corrosion affected the repeatability of the growth process and the possibility to increase the growth rate further. During the HCl additive experiments, the silane flow rate was increased in increments, and, at each flow, a specific HCl flow rate was needed to produce films of smooth morphology. However, when the proces s was repeated at a later time using the
135 SiH4 and HCl flows found initially, the flow s did not produce the same quality of epitaxial layers. This indicated there wa s some problem with the HCl gas delivery system, mainly the HCl MFC. The 4H-SiC ep itaxial process was not evaluated beyond a flow rate of 50 sccm for SiH4 due to the HCl MFC failure. The maximum growth rate achieved during this research was therefore 49 m/h, for a silane flow of 50 sccm, which may not yet be the upper limit for the process. Due to the corrosion of the system, the research using HCl additive to the standard chemistry process for 4H-SiC epitaxial growth was terminated. 5.2 Recent and Future Work 5.2.1 Chemistry The growth of 4H-SiC epitaxial layers using HCl as a gas additive proved to be a promising technique to incr ease the growth rate. Howeve r, as explained above, the corrosion of the reactor system is a significan t problem not only to the reactor parts, but also to establishing a repeatable growth pro cess using this approach. An alternative to adding HCl gas, while still achieving the basi c principle of suppressing homogeneous gas phase nucleation via the presence of chlorine in the gas mixture, is to use methyl chloride (CH3Cl) instead of HCl. CH3Cl would not necessarily replace C3H8 as the carbon source. The reason is that in order to maintain both constant Si/C and Si/C l ratios, propane must also be added to the process along with the CH3Cl gas. Using CH3Cl instead of HCl may help to reduce the corro sion of the system (i.e. gas line s, etc.) since only cracked CH3Cl would lead to corrosion and might prove to be a precursor which could produce more
136 stable and repeatable results. Corrosion will s till occur, but at a much slower rate than when using HCl gas. Preliminary experiments had been perf ormed to determine the difference in surface morphology and growth rate for growth processes which included HCl and that which included CH3Cl. There was no substantial di fference in the surface morphology using the process parameters which produced the same Si/C a nd Si/Cl ratios. SEM micrographs of the surface morphology for the film grown with HCl and the film grown with CH3Cl are shown in Figure 5.1. The horiz ontal lines on the mi crograph are noise from the SEM tool and are not actual surface structures. The growth rate for the HCl additive sample was 33 m/hr, while that of the CH3Cl additive sample was 37 m/hr. There was a slight increase in growth rate ; however, only one experiment was performed due to resource limitations. Further analysis is needed to determine the differences between the two techniques. (a) (b) Figure 5.1 Plan-view SEM micrographs of 4H-SiC grown with (a) HCl additive and (b) CH3Cl additive to the standard chemistry of H2-C3H8-SiH4. P = 150 Torr, T = 1550 C, Si/C = 1.0 and Si/Cl = 4.0. The growth rates were 33 and 37 um/h, respectively.
137 Thermodynamic simulations were performed to predict differences in the mole fractions of the dominant species in the system when growing with HCl and CH3Cl as a gas additive. The results of the simulations are shown in Figure 5.2. The Si/C and Si/Cl ratios for each additive were the same at 1.0 and 4.0, respectively. In order to obtain the same values, C3H8 was added along with CH3Cl. The most dominant specie in the system is, as expected, hydrogen. However, H2 is not shown in the figur e as it has a much higher mole fraction, approximately 1, compared to the other species. As seen in Figure 5.2, the mole fractions for all species are approximately the same when using HCl as an additive versus CH3Cl. This should be the result because the same number of moles of carbon, silicon, and chlorine are being added into the system. The highest mole fraction below 1800 C is solid SiC. The following highest mo le fractions are atomic H and HCl for both cases. The dominant carbon source is CH4, while the dominant Si source species are Si and SiH. The simulations are an indication th at the growth process should be similar for either additive, as long as the S i/C and Si/Cl ratio s are maintained. Comparing the results of the simulations presented in Figure 5.2 to that presented in Chapter 2, Figure 2.1, the process with the added chlorinated gas has many more species than that of the standard 4H-SiC base line process. The reason for this is that either HCl or CH3Cl was added into the gas mixture, cr eating species that have Cl. This is what accounts for the added species in the list presented in Figure 5.2.
138 (a) (b) Figure 5.2 Thermodynamic simulations of the standard CVD growth chemistry with the addition of (a) HCl additive and (b) CH3Cl additive. The Si/C ratio and Si/Cl ratio were 1.0 and 4.0, respectively. H2 is excluded from graphs. Simulations performed using NASA CEA code . Koshka, et al  reported on the growth of 4H-SiC epitaxial layers using CH3Cl to replace C3H8 as the carbon source,. A comparison between the two carbon precursors was made at a temperature of 1600 C. The surface morphology for the film grown with
139 CH3Cl was smooth resulting in a growth rate of 7 m/h. However, growth conducted with C3H8 while using the same silane flow rate as with the CH3Cl experiment resulted in film degradation and a lower growth rate of 2.6 m/h. Koshka, et al  also reported on a 4H-SiC epitaxial gr owth process using CH3Cl, to again replace C3H8, at reduced growth temperatures of ~ 1300 C. The film su rface morphology was mirror-like and a growth rate of 2 m/h was achieved. LTPL was performed on the sample and the spectrum indicated a good film quality wh ich was as good as that of the films grown at 1700 C. From [73,74], it can be seen that CH3Cl may, in fact, be a pr ecursor worth investigating for future work on 4H-SiC halocarbon epitaxial growth. The epitaxial growth of 4H-SiC using CH3Cl as a growth additive may be carried out in the same fashion as the HCl additive experiments. A suggestion would be to start with the baseline process (standard chemistry H2-C3H8-SiH4). Then introduce CH3Cl into the baseline process at a lo w flow rate to a point wher e the surface morphology remains smooth and specular. Then increase the SiH4 flow rate to increase the growth rate while maintaining the CH3Cl flow rate from the first experiment. If needed, increase the CH3Cl flow rate until a smooth film surface is produ ced. Then proceed to a higher silane flow rate and perform the experiments by the method of increasing the SiH4 flow rate and then introducing more CH3Cl until a smooth morphology is achieved. The films should be characterized using FTIR for thickness measurements, AFM for surface roughness, XRD for structural quality and LTPL for optical qua lity. An interesting investigation would be to determine whether the samples using the same Si/C and Si/Cl ratios, but using the two different additives, would result in the same structural and optical quality.
1405.2.2 Temperature Uniformity Another issue to look into is the temp erature distribution ac ross the 50 mm wafer area within the hot-zone. Currently, th ere is approximately a 20 -25 C T across this area with a P = 150 Torr and QH2 = 30 slm, with a 120 C T across the entire susceptor and a 20 C T from the middle to the back end of susceptor. Experiments have been performed by M. Reyes and S. Harvey grow ing 3C-SiC on a 50 mm Si wafer at ~ 1380 C in the center of the growth zone. The b ackend of the wafer melted while the front end of the wafer resulted in polycrystalline film The only single-crystal material was in the center of the wafer. This is an indication that there is a temperature uniformity issue, the temperature being too low in the front and too high in the back of the growth zone. Based on these experiments, growth on a 50 mm wafe r could not be successfully carried out with this temperature gradient. Interestingly, characterization of the reactor during this research using 4H-SiC did not indicate such a high degree of growth non-uniformity but that is a different process and at a higher temperature. The temperature uniformity issue has b een studied recently due to the problems encountered as described above. The RF coils were adjusted to determine whether any significant improvement to the temperature pr ofile could be produced. The coils were put closer together at the front of the hot-zone a nd spread further apart in the back. This was done to increase the temperature at the front end of the susceptor while decreasing the temperature at the backend (this had been succ essfully used in the past to level out the temperature uniformity in the cold-wall reactor). Unfortunately this resulted in a higher T, increasing to 150 C across the entire su sceptor and 25 C from the middle to the backend of susceptor. Another experiment performed to improve the temperature
141 uniformity was to raise the polycrystalline pl ate off the susceptors backend which would reduce the temperature of the substrate since the plate would not be in intimate thermal contact with the susceptor. This improved the temperature profile by 5 C from the middle to the backend of susceptor. Since th is experiment improved the profile slightly, the bottom susceptor had some of the mass removed which was done to reduce the thermal mass in that area. This was accomplished by removing a portion of the bottom of the bottom susceptor, starting at the center and creating an angle upward toward the back end. The same was done on the top of the bottom susceptor. The backend was changed from a thickness of 15 mm to 12 mm, initially. A schematic drawing of this is shown in Figure 5.3. (a) (b) (c) Figure 5.3 Cross-section schematic drawi ng of the 200 mm reacto r susceptor bottom with a portion of the (a) bottom removed a nd (b) top removed to produce a backend thickness of 12 mm, (c) bottom removed add itionally from (a) to produce a backend thickness of 5 mm. The modifications shown in Fig. 5.3(a & b) resulted in a profile that was slightly improved with a measured T of 10 C from the middle to backend of the susceptor. Since this design improved the temperature di stribution, the susceptor had even more mass removed from the bottom of the bottom susceptor in both designs by extending the
142 shaved area to the front end of the suscep tor and making the thickne ss at the backend to be 5 mm (shown in Figure 5.3(c)). The temp erature distribution wi th this design had approximately a 15 C T from middle to back of the susceptor. Since the additional shaving of the susceptor from the bottom di d not improve the temperature distribution enough, then the mass of the top susceptor may need to be reduced. This trial and error process needs to be continued until there is at most a 5 C temper ature difference from front to back across the 50 mm wafer area. Another issue may be that the susceptor top and bottom portions are so close together at the backend of the susceptor that the area close to the backend will always be hotter w ith this design. A suggestion would be to increase the distance between them, i.e. d ecrease the angle of the top portion of the susceptor to increase the distance between t op and bottom portions at the backend of the susceptor. Once the temperature uniformity is improved, growth on a 50 mm wafer will be possible. The epitaxial films may then be characterized across the entire wafer to determine doping and thickness uniformity of the layer. 5.2.3 CVD Simulations One of the most important tasks that has yet to be fully addressed is the complete modeling of the hot-wall CVD process. At the 2005 ONR Electroni c Materials Program Review, Dr. Y. Makarov of Sem itech  presented simulatio ns of a CVD reactor, which were performed using Semitech software. At the same review, Dr. Saddow presented results of 4H-SiC epitaxial la yers grown in the 200 mm react or during this dissertation work (see Chapter 3). Dr. Y. Makarov was very interested in the epitaxial growth results
143 and offered to model the 200 mm reactor to pr edict the distribution of species and growth rate within the reactor. As stated in Chapter 2, a system of e quations must be developed that encompass all phenomena involved in order to model th is or any CVD process. The flow of the reactants, reactor temperature distribution, a nd reactant species distribution are required to simulate the process. CVD reactor mode ls involve non-linear partial differential equations that represent c onservation of momentum and energy, and track individual chemical species [46,47,76]. Along with the transport equations, the homogeneous reactions and surface reaction equations must be solved . Many computer programs have been created to solve such equations and to mode l CVD reactors. The final 200 mm horizontal hot-wall reactor design was mode led by Dr. Y. Makarov and Dr. A. Galyukov of Semitech  using a software program called CVD-Module . The geometry used for the simulations is that shown in the reac tor cross-section sketch, Figure 3.7. Since any hot graphite part can affect the reaction, it is critical to in clude the graphite adapters as well (e.g. silane may crack and SiC may fo rm on the adapters, thus robbing reactants from the growth zone farther downstream). The information needed to perform the computational analysis was the reactor geometry (Figure 3.6), the gas flow rates, the process temperature and pressure, and the hot-zone temperature profile. The temperatur e profile was determined by performing multiple Si melt tests along the susceptor in the 200 mm reactor. The Si melt test samples were placed 15 mm, 39 mm, 61 mm, 73 mm, 107 mm and 131 mm along the flow direction starting at the front (gas inlet side) of the susc eptor. Figure 5.4 shows the measured temperature profile as determined by the Si melt tests. The temperatures along
144 the vertical-axis correlate to the relative temp erature with respect to the center of the hotzone. As seen in Figure 5.4 there is a temp erature difference of about 20 25 C in the center of the hot-zone for an area corre sponding to a 50 mm wafer. The lateral temperature gradient (not s hown) was measured to be ~ 1-5 C, which is essentially negligible. Figure 5.4 Temperature profile measured at the bottom of the susceptor in the 200 mm horizontal hot-wall reactor. Vertical do tted lines indicate the position of a 50mm wafer and horizontal dotted lin es represent the correspondi ng temperature difference. A 20 25 C temperature gradient was measured with the hottest position being towards the hot-zone outlet. Data is normalized with re spect to the hot-zone center as indicated by the dashed lines. Along with the measured temperature profile the process parameters used for the simulations were those similar to the baseli ne process in the 200 mm reactor, Section 3.4. The parameters were: P = 150 Torr, T = 1565 C, QH2 = 30 slm, QC3H8 = 8 sccm, QSiH4 = 30 sccm (Si/C = 1.2). With the process para meters, reactor geometry, and temperature profile, the following simulations, shown in Fi gure 5.5 and Figure 5.6, were generated. The temperature profile of the reactor was us ed to input information about the actual
145 process into the simulations. With the measured data, the simulated temperature profile should be more realistic. Note that the te mperature of the ceiling of the susceptor was assumed to have the same temperature distri bution as the bottom since the Si melt test could only be performed on the bottom of th e susceptor. The computed temperature profile of the 200 mm horizontal hot-wa ll reactor is shown in Figure 5.5. Figure 5.5 Computed 2D temperature profile in the 200 mm horizontal hot-wall CVD reactor. Temperature profile sc ale is in Kelvin. Data provi ded by Dr. Y. Makarov and Dr. A. Galyukov . The 2D simulation represents the cros s section of the reactor hot-zone which includes the adapters and the susceptor. Th e geometry shown in the 2D temperature profile in Figure 5.5 is the shape of th e 200 mm reactor area where the gases flow through the graphite parts. The gases enter the left hand side from the gas inlet port (inlet) and exit the right side (outlet). The suscepto r position is also shown. The simulations predict that the temperature increases starti ng at the inlet of the hot-zone and reaches a maximum at the outlet. When looking at a spec ific distance along the reactor, there is a temperature gradient going from the susceptor bottom to the susceptor ceiling, in which the center is the coolest section. This is due to the ceiling and bottom portions being hotter than the gases between them. However, the ceiling and bottom are very close to
146 each other near the outlet (10 mm spacing), in which the simulations predict there is a small temperature gradient in that area. The mass fractions of the dominant species (CH4, SiH2, and Si) from the thermodynamic simulations in Figure 2.1, as well as SiH4, were also modeled for the reactor. The distributions are show n in Figure 5.6. The methane, CH4, mass fraction distribution shows that methane is not pres ent at the entrance of the reactor, but is generated approximately half way down th e hot-zone. This is likely caused by C3H8 cracking and dissociating into CH4 as the propane gas is heated along the direction of gas flow. From the temperature di stribution in Figure 5.5, the temperature when methane generation becomes significant is approximately 1250 K (~ 970 C). The highest mass fraction for CH4 is above the 50 mm wafer area, indi cated as wafer in Figure 5.6(a). Silane has a totally different profile as seen in Figure 5.6(b). The amount of SiH4 at the entrance is initially large, and as the ga s is heated up along the hot-zone, the SiH4 dissociates, resulting in a lower mass frac tion. Approximately 1/3 the distance along the wafer zone, there is a very small amount of SiH4 remaining, indicating that most of it has been dissociated, most likely into SiH2 and Si, or consumed in the surface reaction.
147 (a) (b) (c) (d) Figure 5.6 Mass fraction di stributions of the do minant species (a) CH4, (b) SiH4, (c) SiH2, and (d) Si at a temperature of 1565 C for the 200 mm horizontal hot-wall reactor geometry. Data provided by Dr. Y. Makarov and Dr. A. Galyukov . Silylene, SiH2, has a similar profile as CH4, which can be seen in Figure 5.6(c). There is little SiH2 gas at the entrance of the hot-zone because initially the silicon source
148 enters as SiH4. As silane is heated, it dissociates into SiH2. This occurs approximately 1/3 of the distance into the susceptor along the ga s flow direction. The highest concentration of SiH2 is around the area of the wafer. Finally, the Si specie is predicted to behave in a similar fashion to SiH2, as can be seen from Figure 5.6(d) The Si specie is not present at the beginning of the hot-zone. However, Si is generated further along the hot-zone. The difference between the SiH2 and the Si distributi on is that the Si sp ecie develops farther downstream in the hot-zone. This is most likely due to SiH2 dissociation. Comparing SiH2 and Si mass fraction distributions, the Si concentration increases close to the area where the SiH2 starts to decrease. Comparing the mass fraction results in Figure 5.6 to those from the thermodynamic simulations shown in Figure 2.1(b), there are some discrepancies between the two simulations. The mass fractions in Figure 5.6 are much higher than those from the thermodynamic simulations in Figure 2.1(b). Also, the CH4 mass fraction is higher than Si in the thermodynamic simula tions, however, the mass fraction is for CH4 than Si for the simulations in Figure 5.6. Th e temperature of course is different between the two simulations, however, the mass fractions at the lower temperature scale in Figure 2.1(b) are decreasing. Therefore, the mass fractions would be even lower, making the difference between the two simulations great er. An explanation could be that the simulations are performed using two differe nt software techniques. The thermodynamic simulations (Figure 2.1) use the minimization of Gibbs free energy to calculate the mass fractions, while the CVD-Module simulations take into account the mass transport, homogeneous and heterogeneous chemistry, and gas phase nucleation. Therefore, the
149 simulations in Figure 5.6 interpret the CVD pr ocess more realistically compared to the thermodynamic simulations in Figure 2.1(b). The growth rate of SiC acros s the hot-zone was predicte d and is shown in Figure 5.7. The growth rate of SiC within th e 50 mm wafer area is the most important information for growth purposes because this is where the samples were placed during growth experiments. The growth ra te was found using the equation : sr sN r r ir N i i gR M V1 1 (5.1) where Vg is the growth rate, Mi is the molecular weight of the ith specie, ir is the stoichiometric coefficient of the ith specie in the rth reaction, Rr is the rate of rth surface reaction, is the surface material density, Ns is the number of species, Nsr is the number of species involved in the reactions. The simulations show that the growth rate will be at a maximum near the center of the 50 mm area and then will dramatically decrease for the maximum reactor temperature of 1565 C. In order for this to happen looking at equation (5.1), the reaction rates must be increasing in that area. The gr owth rate may decrease after the center of the wafer because the reactants are being consumed in the reaction. The Si/C ratio for all cases shown in Figure 5.7(b) decreases across th e wafer zone. This is likely due to the Si being consumed. However, when looking at the mass fraction distribution of CH4 in Figure 5.6(a), the highest concen tration is in the wafer zone. Therefore, it may be that there is more C from the C3H8 cracking in this area causing the Si/C ratio to decrease.
150 (a) (b) Figure 5.7 Calculated (a) grow th rate and temperature pr ofiles along the reactor hotzone for a maximum reactor temperature of 1565 C and (b) Si/C ratio across the reactor. Case 1, 2 and 3 are for maximum reactor temperatures of 1410, 1565 and 1710 C, respectively. Data courtesy of Dr. Y. Makarov and Dr. A. Galyukov . Comparing the experimental thickness results across the simulated 50 mm wafer zone shown in Table 3.2 and the computa tional analysis of the growth rate (at T = 1565 C) (shown in Figure 5.7( a)), the results tend to correlate. However, the simulations were performed assuming the temperature profile on th e bottom of the susceptor is the same as
151 that at the ceiling of the susceptor. Recent experiments indicate that the temperature at the susceptor ceiling is approximately 70 C hotter than the bottom portion of the susceptor. Therefore, new simulations shoul d be carried out with the correct ceiling temperature to determine if the simulations a nd experimental results are still in agreement with one another. Comparing the doping concentration experime ntal results from Table 3.2 with the simulations presented Figure 5.7(b), the results are opposite to each other. The reason for this may be that the calculated simulations do not account for the impurities within the reactor. It is known, from th e LTPL data presented in Chapter 3, Section 3.4, that there are Al contaminants in the reaction system likely caused by the graphite parts. The Al impurities might be consumed at the inlet of the reactor, reducing the amount of Al present downstream. This would in turn re duce the acceptor concentration and produce a higher n-type doping concentration. Another expl anation for the decrease in growth rate may be that the Si is being consumed at the inlet of the susceptor. This would deplete the available Si, resulting in a lower de position farther along the susceptor. Work must be continued to find a revise d correlation between the simulations of the 200 mm hot-wall CVD reactor performed by Dr. Y. Makarov and Dr. A. Galyukov (Figure 5.7) and the experi mental results presented in Chapter 3. The corrected temperature profile for the bottom and ceiling po rtions of the susceptor must be included in the simulations. The CVD simulations may be used to predict the outcome of SiC growth if the 200 mm reactor is modified further, or a new reactor comes online.
152 REFERENCES  P. Masri, "Silicon carbide and silicon carbide-based structures The physics of epitaxy", Surface Science Reports 48 (2002) 1-51.  http://www.ecn.purdue.edu/WBG/Introduction/Index.html "Purdue Wide Band Gap Research", 2004.  A. R. Powell and L. B. Rowland, "SiC Materials Progress, Status and Potential Roadblocks", Proceedings of the IEEE 90 (2002) 942-955.  C. Zetterling, Process Technology for Silicon Carbide Devices INSPEC, London, 2002.  S. E. Saddow and A. Agarwal, Advances in Silicon Carbide Processing and Applications Artech House, Inc, Norwood, 2004.  J. B. Casady and R. W. Johnson, "Statu s of Silicon Carbide (SiC) as a widebandgap semiconductor for high-temper ature applications: A review", Solid-State Electronics 39 (1996) 1409-1422.  I. Khan and Jr. J. A.Cooper, ( http://www.ecn.purdue.edu/WBG/Basic_ Measurements/Transport/Index.html 2000).  K. D. Hobart, F. J. Kub, M. Fatemi, C. Taylor, E. Eshun, and M. G. Spencer, "Transfer of Ultrathin Silicon Layers to Polycrystalline SiC Substrates for the Growth of 3C-SiC Epitaxial Films", Journal of Electrochemical Society 146 (1999) 3833-3836.  R. L. Myers, CVD Growth of SiC on Novel Si Substrates MS Thesis, University of South Florida, Tampa, 2003.  R. L. Myers, S. E. Saddow, S. Rao, K. D. Hobart, M. Fatemi, and F. J. Kub, "Development of 3C-SOI st ructures using Si on polyc rystalline SiC wafer bonded substrates", Materials Science Forum 457-460 (2003) 1511-1514.  M. Reyes, M. Waits, S. Harvey, Y. Shishkin, B. Geil, J. T. Wolan, and S. E. Saddow, "Growth of 3C-SiC on Si molds for MEMS applications", Materials Science Forum, in press (2005).
153  C. H. Park, B. Cheong, K. Lee, and K. J. Chang, "Structu ral and electronic properties of cubic, 2H, 4H, and 6H SiC", The American Physical Society 49 (1994) 4485-4493.  T. Kimoto, S. Nakazawa, K. Fujihira, T. Hirao, S. Nakamura, Y. Chen, K. Hashimoto, and H. Matsunami, "Recent Achievements and Future Challenges in SiC Homoepitaxial Growth", Materials Science Forum 389-393 (2002) 165-170.  T. Tawara, H. Tsuchida, S. Izumi, I. Kamata, and K. Izumi, "Evaluation of Free Carrier Lifetime and Deep Levels of the Thick 4H-SiC Epilayers", Materials Science Forum 457-460 (2004) 565-568.  U. Forsberg, . Danielsson, A. Henry, M. K. Linnarsson, and E. Janzn, "Nitrogen doping of epitax ial silicon carbide", Journal of Crystal Growth 236 (2002) 101-112.  H. Tsuchida, I. Kamata, T. Jikimoto, and K. Izumi, "Growth of thick 4H-SiC epilayers in a vertical radiant-heating reactor", Materials Research Society Symposium 640 (2001) H2.12.1-H2.12.6.  A. Ellison, J. Zhang, A. Henry, and E. Janzn, "Epitaxial growth of SiC in a chimney CVD reactor", Journal of Crystal Growth 236 (2002) 225-238.  S. Nakazawa, T. Kimoto, K. Hashimoto, and H. Matsunami, "High-purity 4H-SiC epitaxial growth by hot-wall chemical vapor deposition", Journal of Crystal Growth 237-239 (2002) 1213-1218.  J. Zhang, J. Mazzola, C. Hoff, Y. Koshka and J. Cassady, "High Growth Rate (up to 20 m/h) SiC Epitaxy in a Horizontal Hot-wall Reactor", Materials Science Forum 483-485 (2005) 77-80.  A. O. Konstantinov, C. Hallin, O. Kord ina, and E. Janzn, "Effect of vapor composition on polytype homogeneity of epitaxial silicon carbide", Journal of Applied Physics 80 (1996) 5704-5712.  R. Rupp, A. Wiedenhofer, P. Friedrichs, D. Peters, R. Schrner, and D. Stephani, "Growth of SiC Epitaxial Layers in a Ve rtical Cold Wall Reactor Suited for High Voltage Applications", Materials Science Forum 264-268 (1998) 89-96.  R. Rupp, P. Lanig, J. Vlkl, and D. St ephani, "First results on silicon carbide vapour phase epitaxy growth in a new type of vertical low pressure chemical vapour deposition reactor", Journal of Crystal Growth 146 (1995) 37-41.  K. Masahara, T. Takahashi, M. Kushibe, T. Ohno, J. Nishio, K. Kojima, Y. Ishida, T. Suzuki, T. Tanaka, S. Yoshid a, and K. Arai, "High-Rate Epitaxial Growth of 4H-SiC Using a Vertical -Type, Quasi-Hot-Wall CVD Reactor", Materials Science Forum 389-393 (2002) 179-182.
154  K. Fujihira, T. Kimoto, and H. Matsunami, "Fast Epitaxial Growth of HighQuality 4H-SiC by Vertical Hot-Wall CVD", Materials Science Forum 433-436 (2003) 161-164.  J. Zhang, A. Ellison, and E. Janzn, "Morphology Control for Growth of Thick Epitaxial 4H SiC Layers", Materials Science Forum 338-342 (2000) 137-140.  A. Ellison, J. Zhang, W. Magnusson, A. Henry, Q. Wahab, J. P. Bergman, C. Hemmingsson, N. T. Son, and E. Jan zn, "Fast SiC Epitaxial Growth in a Chimney CVD Reactor and HTCVD Crystal Growth Developments", Materials Science Forum 338-342 (2000) 131-136.  H. Tsuchida, I. Kamata, T. Jikimoto, and K. Izumi, "Epitaxial growth of thick 4HSiC layers in a vertical radiant-heating reactor", Journal of Crystal Growth 237239 (2002) 1206-1212.  A. Elshabini-Riad and F. D. Barlow III, Thin Film Technology Handbook McGraw Hill, New York, 1987.  K. Seshan, Handbook on Thin-Film Deposition Processes and Techniques Second ed., Noyes Publications, Norwich, 2002.  S. Rao, Implant Annealing of Al Dopants in Silicon Carbide using Silane Overpressure Dissertation Thesis, University of South Florida, Tampa, FL, 2005.  C. E. Morosani, Thin Films by Chemical Vapour Deposition Elsevier, Amsterdam, 1990.  J. Park, Chemical Vapor Deposition Vol. 2 ASM International, Materials Park, 2001.  A. Sherman, Chemical Vapor Deposition for Microelectronics: Principles, Technology, and Applications Noyes Publications, Park Ridge, 1987.  Y. S. Park, SiC Materials and Devices Vol. 52 Academic Press, Chestnut Hill, 1998.  D. J. Larkin, "SiC Dopant Incorporat ion Control Using Site-Competition CVD", Physica Status Solidi(b) 202 (1997) 305-320.  http://hyperphysics.phy-astr.gsu.edu/hbase/thermo/helmholtz.html "HyperPhysics", 2005.  S. Sandler, Chemical and Engineering Thermodynamics Third ed., John Wiley & Sons, Inc., New York, 1999.
155  http://chemed.chem.purdue.edu/genchem/topicreview/bp/ch21/gibbs.html "Gibbs Free Energy".  P. Atkins, Physical Chemistry W. H.Freeman and Company, New York, 1998.  B. McBride and S. Gordon, NASA-Glenn Chemical Equilibrium Program CEA2, 1994.  S. Gordon and B. McBride, Computer Program for Calculation of Complex Chemical Equilibrium Compositions and Applications II, Report No. NASA Reference Publication 1311 (1994).  B. McBride and S. Gordon, Computer Program for Calculation of Complex Chemical Equilibrium Compositions and Applications I, Report No. NASA Reference Publication 1311 (1996).  O. Kordina, Growth and Characterisation of Silicon Carbide Power Device Material Dissertation Thesis, Linkping Un iversity, Linkping, Sweden, 1994.  R.V. Giles, J. B. Evett, and C. Liu, Fluid Mechanics and Hydraulics Third ed., McGraw-Hill, New York, 1995.  R. V. Giles, J. B. Evett, and C. Liu, Fluid Mechanics and Hydraulics Third ed., McGraw-Hill, New York, 1995.  W. Ji, MP. M. Lofgren, C. Hallin, C.-Y. Gu, and G. Zhou, "Computational modeling of SiC epitaxial grow th in a hot wall reactor", Journal of Crystal Growth 220 (2000) 560-571.  A. Veneroni, F. Omarini, D. Moscatelli, M. Masi, S. Leone, M. Mauceri, G. Pistone, and G. Abbondanza, "Modeling of epitaxial silicon carbide deposition", Journal of Crystal Growth 275 (2005) e25-e300.  A. L. Hines and R. N. Maddox, Mass Transfer Prentice Hall PTR, Englewood Cliffs, 1985.  Matt Smith, Design and Development of a Silicon Carbide Chemical Vapor Deposition Reactor Thesis, University of South Florida, Tampa, FL, Dec 2003.  S. Plum, P. Graf, M. Beckers, J. Boersm a, E. van den Dungen, M.Frijns, and F. Verbakel, Advanced heating tech niques for glass melting, (2002).  "Experimental and Theoretical studies of nitrogen containing HF-CVD gas-phase environment", http://www.chm.bris.ac.uk/pt/diamond /jamessthesis/chapter4.htm#_edn7
156  O. Kordina, C. Hallin, A. Henry, J. P. Bergman, I. Ivanov, A. Ellison, N. T. Son, and E. Janzn, "Growth of SiC by "Hot-Wall" CVD and HTCVD", Physica Status Solidi(b) 202 (1997) 321-334.  Johnson Matthey, "Gas Pu rification Technology", http://www.hydrogentechnology.com/htm l/pd_membrane_purification.html  Femlab, Comsol Multiphysics Version 3.2, 1997-2005.  R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena Second ed., John Wiley & Sons, Inc., New York, 2002.  E. Janzn, J. P. Bergman, . Danielsson, U. Forsberg, C. Hallin, J. ul Hassan, A. Henry, I. G. Ivanov, A. Kakanakova-Georg ieva, P. Persson, and Q. ul Wahab, "SiC and III-Nitride Growth in a Hot-wall CVD Reactor", Materials Science Forum 483-485 (2005) 61-66.  Ameritherm Inc., http://www.ameritherm.com/  Z. Y. Xie, Ch. H. Wei, L. Y. Li, Q. M. Yu, and J. H. Edgar, "Gaseous etching of 6H-SiC at relatively low temperatures", Journal of Crystal Growth 217 (2000) 115-124.  H. O. Pierson, Handbook of Chemical Vapor Deposition: Principles,Technology, and Applications Second ed., Noyes Publication, Norwich, 1999.  R. L. Myers, Y. Shishkin, O. Kordina, I. Haselbarth, and S. E. Saddow, "High Epitaxial Growth Rate of 4H-SiC using Horizontal Hot-Wall CVD", Materials Science Forum, in press (2005).  R. L. Myers, Y. Shishkin, O. Kordina, and S. E. Saddow, "High growth rates (>30m/h) of 4H-SiC epitaxial layers using a horizontal hot-wall CVD reactor", Journal of Crystal Growth 285 (2005) 483-486.  E. Janzn, A. Henry, J. P. Bergman, A. Ellison, and B. Magnusson, "Material characterization need for SiC-based devices", Materials Science in Semiconductor Processing 4 (2001) 181-186.  I. G. Ivanov, C. Hallin, A. Henry, O. Kordina, and E. Janzn, "Nitrogen doping concentration as determined by phot oluminescence in 4Hand 6H-SiC", Journal of Applied Physics 80 (1996) 3504-3508.  NovaSiC, "NovaSiC Silicon Carbide and Related Materials", http://www.novasic.com/ 2003.  http://www.schmitt-ind.com/supporttechnical-papers-scatter.shtml
157  H. Tsuchida, I. Kamata, T. Jikimoto, and K. Izumi, "Growth and Electrical Characterization of Lightly-Dope d Thick 4H-SiC Epilayers", Materials Science Forum 389-393 (2002) 171-174.  J. Zhang, U. Forsberg, M. Isacson, A. Ellison, A. Henry, O. Kordina, and E. Janzn, "Growth Characteristics of SiC in a Hot-Wall CVD Reactor with Rotation", Materials Science Forum 389-393 (2002) 191-194.  R. Rupp, Y. N. Makarov, H. Behner, and A. Wiedenhofer, "Silicon Carbide Epitaxy in a Vertical CVD Reactor: Experi mental Results and Numerical Process Simulation", Physica Status Solidi(b) 202 (1997) 281-304.  M. Reyes and S. E. Saddow, "Phase I fina l report to Caracal, Inc., ONR STTR Phase I", December, 2004.  R. C. Jaeger, Introduction to Microel ectronic Fabrication Vol. 5, Second ed., Prentice Hall, Upper Saddle River, 2002.  S. Nakamura, T. Kimoto, H. Matsunami, S. Tanaka, N. Teraguchi, and A. Suzuki, "Formation of periodic steps with a un it-cell height on 6H-SiC (0001) surface by HCl etching", Applied Physics Letters 76 (2000) 3412-3414.  A. Fissel, W. Richter, J. Furthmller, a nd F. Bechstedt, "On the nature of the D1defect center in SiC: A photolumines cence study of layers grown by solid-source molecular-beam epitaxy", Applied Physics Letters 78 (2001) 2512-2514.  Y. Koshka, H. Lin, G. Melnychuk, M. S. Mazzola, and J. L. Wyatt, "Homoepitaxial growth of 4H-SiC using CH3Cl carbon precursor", Materials Science Forum 483-485 (2005) 81-84.  Y. Koshka, H. Lin, G. Melnychuk, and C. Wood, "Lower-temperature epitaxial growth of 4H-SiC using CH3Cl carbon gas precursor", ICSCRM 2005 (2005 (in press)).  www.semitech.us "Semiconductor Technology Research, Inc." 2005.  S. Carr and M. Masi, "Kinetic Approach to Materials Synthesis by Gas-Phase Deposition", Progress in Crystal Growth and Characterization (1998) 1-46.  J. Meziere, M. Ucar, E. Blanquet, M. Pons, P. Ferret, and L. Di Cioccio, "Modeling and simulation of SiC CVD in the horizontal hot-wall reactor concept", Journal of Crystal Growth 267 (2004) 436-451.  A. N. Vorob'ev, S. Yu. Karpov, M.V. B ogdanov, A. E. Komissarov, O. V. Bord, A. I. Zhmakin, and Yu. N. Makarov, "N umerical study of SiC CVD in a vertical cold-wall reactor", Computational Materials Science 24 (2002) 520-534.
158  D. Schroder, Semiconductor Material and Device Characterization Second ed., John Wiley & Sons, Inc., New York, 1998.  J. H. Wittke, "Effects of Electron Bombardment", http://jan.ucc.nau.edu/~wittke/Microprobe/Interact-Effects.html 2003.  S. L. Elliot, R. F. Broom, and C. J. Humphreys, "Enhanced SEM Doping Contrast on an H-passivated Silicon Surface", http://wwwhrem.msm.cam.ac.uk/research/Dopant/Dopant.html  Building and Fire Research Laboratory, "Atomic Force Microscopy (AFM)", http://www.bfrl.nist.gov/nanoscience/BFRL_AFM.htm  NanoScience Instruments, "Atomic Force Microscopy", http://www.nanoscience.com/education/AFM.html 2005.  "X-ray Diffraction and Braggs Law", http://www.cartage.org.lb/en/themes/Scie nces/Physics/SolidStatePhysics/Atomic Bonding/XRAY/XrayDiffrac tion/XrayDiffraction.htm  X. Huang, M. Dudley, and R. S. Okojie, "Characterization of SiC epilayers using high-resolution X-ray diffraction a nd synchrotron topography imaging", Materials Research Society Symposium Proceedings 815 (2004) J7.3.1-J7.3.6.  P. F. Fewster, X-ray Scattering from Semiconductors Imperial College Press, River Edge, 2000.  Phillips Analytical, X'Pert Pro User's Guide, (2002).  P. A. Heiney, "Instrumentation for XR D: Diffractometers and Detectors", http://www.physics.upenn.edu/~heiney/talk s/hires/instrument.html#SECTION000 33000000000000000 1996.  S. Perkowitz, Optical Characterization of Semi conductors: Infrared, Raman, and Photoluminescence Spectroscopy Academic Press, London, 1993.  F. Walder and R. Boyle, FT-IR Meas urement of Epitaxial Film Thickness Applications, Report No. A pplication Note: 0015 (2004).  R. P. Devaty and W. J. Choyke, "Opti cal Characterization of Silicon Carbide Polytypes", Physica Status Solidi(a) 162 (1997) 5-38.  M. Ahoujja, H. C. Crocket, M. B. Scott, Y. K. Yeo, and R. L. Hengehold, "Photoluminescence Characterization of Defects Introduced in 4H-SiC During High Energy Proton Irradiation a nd Their Annealing Behavior", Materials Research Society Symposium Proceedings 815 (2004) J5.21-J5.26.
160Appendix A: NASA-Lewi s Chemical Equilibrium Computations and Applications (CEA) The following instructions were used to determine the equilibrium composition of chemical species within the CVD reactor using the NASA-Lewis Chemical Equilibrium and Applications (CEA) program. The pr ogram was written by B. McBride and S. Gordon in ANSI standard Fortran . As stat ed in Chapter 2, Section 2.2.3, the program calculates chemical compositions by minimizing the Gibbs free energy of the system. In order to run the program, two sets of inputs must be provided. The first set of inputs is the thermodynamic and thermal transp ort property data (i.e. enth alpy, entropy, etc.). This data is provided in a library within the program which contains the properties for the chemical species. However, if the specie specif ied is not available, one must the program. The second input is the thermodynamic st ate functions, which are temperature and pressure for this case, the reactants used, and the reactants flow rates. The input files that are generated from the input data must be saved as .inp files. After executing the program, an output file is generated as a .out file and a plot file is generated as a .plot file. The output file will provide the informa tion from the input file the results of the computations and the species which were in cluded in the calculations. The plot file generates the data into columns which may be easily plotted within a graphical software program. The CEA executable file name is CEAexecute-win.bat, which is used to open the program. When the program is first opened, the screen shown in Figure A.1 is first displayed, with the Problem tab alr eady highlighted. The thermodynamic state
161 Appendix A: (Continued) functions which will be provided needs to be ch osen next. In this cas e, that would be the Assigned Temperature and Pressure tp problem. Double click on this problem. Figure A.1 Screen image of the CEA program display. The problem is specified by chosing which thermodynamic state functions will be used as the input values. After selecting the problem (thermodynamic state functions), a screen will appear which directs the user to input the temperat ure and pressure, which is shown in Figure A.2. There are drop down fields which one ca n specify the units of the thermodynamic state functions of the input values. The choices for the temperature unit are Kelvin, Rankine, Celsius, and Fahrenhe it, while those of the pressure are bar, atm, psia, and mmHg. The units specified for this exampl e are Kelvin and mmHg for temperature and pressure, respectively. The input values for the temperature and pressure are then
162 Appendix A: (Continued) specified under each column. The total number of input values allowed are 16. For this example, which is the 4H-SiC epitaxial gr owth baseline proces s in the 200 mm reactor, the temperature ranged between 1500 and 2000 K in increments of 50K. The process pressure was specified to be 150 mmHg which corresponds to the proc ess pressure of 150 Torr. Figure A.2 CEA display screen for the TP problem. Under the Temperature heading, the desired temperature range is input. Under the Pressure heading, the desired pressure is specified. Once the input values have been specified the Save button is pressed to save the input data. The Reactant tab is then chosen, which wi ll bring up the screen shown
163 Appendix A: (Continued) in Figure A.3. In the Ident column, there are three choices to identify the reactant which are name, fuel, and oxide. Choose th e name identity. In the Name column, enter the appropriate reactants which are fl own during growth in the CVD system. For this example, H2, C3H8, and SiH4 from the standard process. In the Amount field, input the flow rates used for the growth process. The Amount may be specified in moles or in relative weight. Since the temperature and pressure are the same, the flow rates are equal to the number of moles for each species. In this case, the H2 was set to 30,000 moles (representing 30,000 sccm), the C3H8 was set to 10 moles (representing 10 sccm), and the SiH4 was set to 30 moles (representing 30 sccm). The temperature and energy units may also be specified in the Reactant field. Kelvin and kJ/mol were chosen for this problem. The Save button is th en pressed after specifying all fields.
164 Appendix A: (Continued) Figure A.3 Display window for the Reactant tab. This is where the reactant information is input into the CEA program. The Only tab may be chosen if one want s to consider specified species only for the problem. The display window for this tab is shown in Figure A.4. All possible gaseous and condensed species possible from th e input reactants are displayed on the left top and bottom columns in Figure A.4. For this example problem, no species were specified for the Only tab. If one does want to enter a specie, it is done by selecting the specie from the left hand columns and clicki ng the Add button. If a mistake was made, a specie may be removed from the Selecte d List by selecting the specie and clicking Remove. The file is then saved by pressing Save.
165 Appendix A: (Continued) Figure A.4 Display window for the Only tab which allows a specified species list to be created. This will allow one to choose which species only are to be included in the calculation. Another option is the Omit tab, which is shown in Figure A.5. If one wishes to omit certain species from the simulation, this ta b is the place to choos e the specie that is to be omitted. The tab specified as Insert is used to insert certain condensed species to be tried in the initial iteration. There ma y be some cases where no condensed species were entered, and a convergence problem occurred. This may indicate that a condensed species must be entered into the Insert ta b. For the baseline pr ocess example problem,
166 Appendix A: (Continued) no species have been selected for either the Only, Omit, or Insert tab. When this occurs, the program will consider all possible species from the reactant species input list, which is on the left hand column in th e display screen shown in Figure A.5. Figure A.5 Display screen for the Omit tab in the CEA program. This allows the user to omit certain species that one specifies. The last set of information that needs to be input into the CEA program is under the Output tab. The display screen for this is shown in Figure A.6. The species composition unit may be chosen as mole fractions or mass fractions. For this example,
167 Appendix A: (Continued) the unit was chosen to be mole fraction. On e may also specify the trace species value, which is set for 10-10 for this problem. By putting in the trace species value, species below this value will not be written to the plot file. The chemical species and thermodynamic properties which are to be presente d in the plot file mu st be specified in the box above the Reset button and the in the Sel ected Plot List column, respectively. For this case, the temperature and pressure were selected to be displayed in the plot file, along with the relevant chemical species. The plot file is a file which is generated from the calculations which only disp lays what has been chosen in the Selected Plot List and the species list which is th e box above the Reset button.
168 Appendix A: (Continued) Figure A.6 Display window for the Out put tab. The thermodynamic properties which are to be put into the output file are specified in this tab, which for this case is temperature and pressure. Once all of the information has been put in to the program, the da ta is saved as a .inp file. To execute the program, the user goes to the Activity tab at the top of the screen and selects Execute CEA. To view th e plot file or the output file, select View PLOT File or View OUTPUT File from th e Activity tab. The plot file may be
169 Appendix A: (Continued) imported into another program to plot the resu lts. An example of th e data generated by the CEA program is shown next. Output file: ******************************************************************************* NASA-GLENN CHEMICAL EQUILIBRIUM PROGRAM CEA2, MAY 21, 2004 BY BONNIE MCBRIDE AND SANFORD GORDON REFS: NASA RP-1311, PART I, 1994 AND NASA RP-1311, PART II, 1996 ******************************************************************************* problem tp t,k=1500,1550,1600,1650,1700,1750,1800,1850,1900,1950,2000, p,mmhg=150 react name=H2 moles=30000 name=C3H8 moles=10 name=SiH4 moles=30 output trace=1E-10 plot p t CH2 CH3 CH4 C2H2,acetylene C2 H4 H H2 Si SiC2 SiH SiH2 SiH3 SiH4 Si2 Si2C Si3 SiC(b) end OPTIONS: TP=T HP=F SP=F TV=F UV=F SV=F DETN=F SHOCK=F REFL=F INCD=F RKT=F FROZ=F EQL=F IONS=F SIUNIT=T DEBUGF=F SHKDBG=F DETDBG=F TRNSPT=F T,K = 1500.0000 1550.0000 1600.0000 1650.0000 1700.0000 1750.0000 1800.0000 T,K = 1850.0000 1900.0000 1950.0000 2000.0000 TRACE= 1.00E-10 S/R= 0.000000E+00 H/R= 0.000000E+00 U/R= 0.000000E+00 P,BAR = 0.199984 REACTANT MOLES (ENERGY/R),K TEMP,K DENSITY EXPLODED FORMULA N: H2 ********** 0.000000E+00 0.00 0.0000 H 2.00000 N: C3H8 10.000000 0.000000E+00 0.00 0.0000 C 3.00000 H 8.00000 N: SiH4 30.000000 0.000000E+00 0.00 0.0000 SI 1.00000 H 4.00000 SPECIES BEING CONSIDERED IN THIS SYSTEM (CONDENSED PHASE MAY HAVE NAME LISTED SEVERAL TIMES) LAST thermo.inp UPDATE: 9/09/04
170 Appendix A: (Continued) g 7/97 *C tpis79 *CH g 4/02 CH2 g 4/02 CH3 g 8/99 CH4 tpis91 *C2 g 6/01 C2H g 1/91 C2H2,acetylene g 5/01 C2H2,vinylidene g 7/01 C2H3,vinyl g 1/00 C2H4 g 7/00 C2H5 g 7/00 C2H6 tpis79 *C3 n 4/98 C3H3,1-propynl n 4/98 C3H3,2-propynl g 2/00 C3H4,allene g 1/00 C3H4,propyne g 5/90 C3H4,cyclog 3/01 C3 H5,allyl g 2/ 00 C3H6,propylene g 1/00 C3H6,cyclog 7/01 C3H7,n-propyl g 9/85 C3H7,i-propyl g 2/00 C3H8 g tpis *C 4 g 7/01 C4H2,butadiyne g 8/00 C4H4,1,3-cyclon10/92 C4H6,butadiene n10/93 C4H6,1butyne n10/93 C4H6,2butyne g 8/00 C4H6 ,cyclon 4/88 C4H8,1-butene n 4/88 C4H8,cis2-buten n 4/88 C4H8,tr2-butene n 4/88 C4H8,isobutene g 8/00 C4H8,cyclon10/84 C4 H9,n-butyl n10/84 C4H9,i-butyl g 1/93 C4H9,s-butyl g 1/93 C4H9,t-butyl g12/00 C4H10,n-butane g 8/00 C4H10,isobutane g 8/00 *C5 g 5/90 C5H6,1,3cyclog 1/93 C5H8,cyclon 4/87 C5H10,1-pentene g 2/01 C5H10,cyclon10/84 C5H11,pentyl g 1/93 C5H11,t-pentyl n10/85 C5H12,n-pentane n10/85 C5H12,i-pentane n10/85 CH3C(CH3)2 CH3 g 2/93 C6H2 g11/00 C6H5,phenyl g 8/00 C6H6 g 1/93 C6H10,cyclon 4/87 C6H12,1-hexene g 6/90 C6H12,cyclon10/83 C6H13,n-hexyl g 6/01 C6H14,n-hexane g 7/01 C7H7,benzyl g 1/93 C7H8 n 4/87 C7H14,1-heptene n10/83 C7H15,n-heptyl n10/85 C7H16,n-heptane n10/85 C7H16,2-methylh n 4/89 C8H8,styrene n10/86 C8H10,ethylbenz n 4/87 C8H16,1-octene n10/83 C8H17,n-octyl n 4/85 C8H18,n-octane n 4/85 C8H18,isooctane n10/83 C9H19,n-nonyl g 3/01 C10H8,naphthale n10/83 C10H21,n-decyl g 8/00 C12H9,o-bipheny g 8/00 C12H10,biphenyl g 6/97 *H tpis78 *H2 g 8/97 *Si tpis91 SiC tpis91 SiC2 tpis91 SiH g 3/01 SiH2 g 3/99 SiH3 tpis91 SiH4 tpis91 Si2 tpis91 Si2C g 7/95 Si3 n 4/83 C(gr) n 4/83 C(gr) n 4/83 C(gr) tpis91 Si(cr) tpis91 Si(cr) tpis91 Si(L) tpis91 SiC(b) tpis91 SiC(b) tpis91 SiC(L) O/F = 0.000000 EFFECTIVE FUEL EFFECTIVE OXIDANT MIXTURE ENTHALPY h(2)/R h(1)/R h0/R (KG-MOL)(K)/KG 0.00000000E+00 0.00000000E+00 0.00000000E+00 KG-FORM.WT./KG bi(2) bi(1) b0i *H 0.97283694E+00 0.00000000E+00 0.97283694E+00 *C 0.48480246E-03 0.00000000E+00 0.48480246E-03 *Si 0.48480246E-03 0.00000000E+00 0.48480246E-03 POINT ITN T H C SI 1 19 1500.000 -10.106 -1.967 0.113 ADD SiC(b) 1 5 1500.000 -10.105 -8.102 -3.553 ADD Si(cr) 1 2 1500.000 -10.105 -7.069 -4.586
171 Appendix A: (Continued) 1 2 1500.000 -10.105 -7.069 -4.586 2 3 1550.000 -10.153 -6.941 -4.668 3 4 1600.000 -10.200 -6.823 -4.748 4 3 1650.000 -10.246 -6.715 -4.826 REMOVE Si(cr) 4 2 1650.000 -10.246 -6.613 -4.928 5 4 1700.000 -10.291 -6.119 -5.400 6 4 1750.000 -10.335 -5.652 -5.852 7 4 1800.000 -10.379 -5.233 -6.261 8 4 1850.000 -10.421 -4.885 -6.605 THERMODYNAMIC EQUILIBRIUM PROPERTIES AT ASSIGNED TEMPERATURE AND PRESSURE CASE = REACTANT MOLES ENERGY TEMP KJ/KG-MOL K NAME H2 ************ 0.000 0.000 NAME C3H8 10.0000000 0.000 0.000 NAME SiH4 30.0000000 0.000 0.000 O/F= 0.00000 %FUEL= 0.000000 R,EQ.RATIO= 0.000000 PHI,EQ.RATIO= 0.000000 THERMODYNAMIC PROPERTIES P, BAR 0.19998 0.19998 0.19998 0.19998 0.19998 0.19998 0.19998 0.19998 T, K 1500.00 1550.00 1600.00 1650.00 1700.00 1750.00 1800.00 1850.00 RHO, KG/CU M 3.2965-3 3.1901-3 3.0903-3 2.9965-3 2.9082-3 2.8249-3 2.7460-3 2.6712-3 H, KJ/KG 17645.8 18438.8 19239.4 20048.6 20867.7 21699.0 22545.5 23411.2 U, KJ/KG 11579.2 12169.9 12768.1 13374.8 13991.2 14619.6 15262.8 15924.6 G, KJ/KG -122674.2-127364.6-132080.7-136821.9-141587.8-146378.0-151192.1-156030.0 S, KJ/(KG)(K) 93.5466 94.0667 94.5750 95.0730 95.5621 96.0440 96.5209 96.9952 M, (1/n) 2.056 2.056 2.056 2.056 2.056 2.055 2.055 2.055 MW, MOL WT 2.054 2.054 2.054 2.054 2.053 2.053 2.053 2.053 (dLV/dLP)t -1.00001 -1.00002 -1.00003 -1.00005 -1.00008 -1.00013 -1.00020 -1.00031 (dLV/dLT)p 1.0004 1.0006 1.0011 1.0017 1.0026 1.0040 1.0061 1.0090 Cp, KJ/(KG)(K) 15.7913 15.9327 16.0933 16.2760 16.4965 16.7667 17.1055 17.5382 GAMMAs 1.3446 1.3407 1.3366 1.3321 1.3270 1.3212 1.3144 1.3065 SON VEL,M/SEC 2856.1 2899.1 2941.0 2981.6 3020.8 3058.3 3094.0 3127.5 MOLE FRACTIONS CH2 1.216-13 3.961-13 1.198-12 3.750-12 1.479-11 5.414-11 1.807-10 5.389-10 CH3 3.1889-9 5.4734-9 9.0795-9 1.6171-8 3.7532-8 8.3289-8 1.7332-7 3.3068-7 CH4 3.9492-6 3.7820-6 3.6334-6 3.8772-6 5.5610-6 7.8455-6 1.0652-5 1.3580-5 C2H2,acetylene 5.229-10 1.360 -9 3.329 -9 9. 470 -9 4.606 -8 2.061 -7 8.156 -7 2.728 -6 C2H2,vinylidene 1.299-15 5.398-15 2.051-14 8.808-14 6.310-13 4.065-12 2.269-11 1.050-10 C2H4 3.011-11 4.826-11 7.511-11 1.396-10 4.548-10 1.396 -9 3.867 -9 9.237 -9 *H 3.9368-5 7.0440-5 1.2160-4 2.0319-4 3.2956-4 5.2017-4 8.0073-4 1.2046-3 *H2 9.9896-1 9.9893-1 9.9888-1 9.9880-1 9.9867-1 9.9848-1 9.9819-1 9.9778-1 *Si 5.2591-8 1.6599-7 4.8717-7 1.2085-6 2.1056-6 3.5388-6 5.8893-6 9.9851-6
172 Appendix A: (Continued) SiC2 8.518-15 6.194-14 3.972-13 2.516-12 1.926-11 1.315-10 7.899-10 4.106 -9 SiH 1.4405-7 3.6484-7 8.7101-7 1.7795-6 2.5826-6 3.6528-6 5.1648-6 7.5042-6 SiH2 1.0237-7 2.0225-7 3.8279-7 6.2919-7 7.4464-7 8.6947-7 1.0263-6 1.2579-6 SiH3 9.2489-9 1.5276-8 2.4447-8 3.4323-8 3.5020-8 3.5551-8 3.6766-8 3.9755-8 SiH4 1.8879-7 2.0073-7 2.1267-7 2.0277-7 1.4380-7 1.0364-7 7.7590-8 6.1827-8 Si2 3.331-10 1.431 -9 5.599 -9 1.641 -8 2.479 -8 3.623 -8 5.386 -8 8.590 -8 Si2C 1.603-10 7.704-10 3.347 -9 1.198 -8 2.943 -8 6.825 -8 1.537 -7 3.466 -7 Si3 7.502-11 3.598-10 1.560 -9 4.543 -9 5.052 -9 5.510 -9 6.331 -9 8.289 -9 Si(cr) 3.4556-6 2.8329-6 1.6553-6 0.0000 0 0.0000 0 0.0000 0 0.0000 0 0.0000 0 SiC(b) 9.9172-4 9.9187-4 9.9200-4 9.9167-4 9.8981-4 9.8703-4 9.8269-4 9.7539-4 THERMODYNAMIC PROPER TIES FITTED TO 20000.K PRODUCTS WHICH WERE CONSIDERED BUT WHOSE MOLE FRACTIONS WERE LESS THAN 1.000000E10 FOR ALL ASSIGNED CONDITIONS *C *CH *C2 C2H C2H3,vinyl C2H5 C2H6 *C3 C3H3,1-propynl C3H3,2-propynl C3H4,allene C3H4,propyne C3H4,cycloC3H5,allyl C3H6,propylene C3H6,cycloC3H7,n-propyl C3H7,i-propyl C3H8 *C4 C4H2,butadiyne C4H4,1,3-cycloC4H6,butadiene C4H6,1butyne C4H6,2butyne C4H6,cycloC4H8,1-butene C4H8,cis2-buten C4H8,tr2-butene C4H8,isobutene C4H8,cycloC4H9,n-butyl C4H9,ibutyl C4H9,s-butyl C4H9,t-butyl C4H10,n-butane C4H10,isobutane *C5 C5H6,1,3cycloC5H8,cycloC5H10,1-pentene C5H10,cycloC5H11,pentyl C5H11,t-pentyl C5H12,n-pentane C5H12,i-pentane CH3C(CH3)2CH3 C6H2 C6H5,phenyl C6H6 C6H10,cycloC6H12,1-hexene C6H12,cycloC6H13,nhexyl C6H14,n-hexane C7H7,benzyl C7H8 C7H14,1-heptene C7H15,n-heptyl C7H16,n-heptane C7H16,2methylh C8H8,styrene C8H10,ethylbenz C8H16,1-octene C8H17,n-octyl C8H18,n-octane C8H18,isooctane C9H19,n-nonyl C10H8,naphthale C10H21,n-decyl C12H9,o-bipheny C12H10,biphenyl SiC C(gr) Si(L) SiC(L) POINT ITN T H C SI 1 4 1900.000 -10.463 -4.621 -6.869 2 4 1950.000 -10.505 -4.428 -7.066 3 4 2000.000 -10.545 -4.284 -7.219 THERMODYNAMIC EQUILIBRIUM PROPERTIES AT ASSIGNED TEMPERATURE AND PRESSURE CASE = REACTANT MOLES ENERGY TEMP KJ/KG-MOL K NAME H2 ************ 0.000 0.000 NAME C3H8 10.0000000 0.000 0.000 NAME SiH4 30.0000000 0.000 0.000 O/F= 0.00000 %FUEL= 0.000000 R,EQ.RATIO= 0.000000 PHI,EQ.RATIO= 0.000000 THERMODYNAMIC PROPERTIES P, BAR 0.19998 0.19998 0.19998 T, K 1900.00 1950.00 2000.00 RHO, KG/CU M 2.6002-3 2.5325-3 2.4678-3 H, KJ/KG 24301.4 25223.5 26187.0 U, KJ/KG 16610.3 17326.8 18083.1 G, KJ/KG -160891.7-165777.1-170686.7 S, KJ/(KG)(K) 97.4700 97.9490 98.4368
173 Appendix A: (Continued) M, (1/n) 2.054 2.053 2.052 MW, MOL WT 2.052 2.051 2.050 (dLV/dLP)t -1.00046 -1.00067 -1.00096 (dLV/dLT)p 1.0132 1.0188 1.0266 Cp, KJ/(KG)(K) 18.0963 18.8194 19.7601 GAMMAs 1.2973 1.2865 1.2741 SON VEL,M/SEC 3158.7 3187.3 3213.3 MOLE FRACTIONS CH2 1.4236-9 3.3818-9 7.4073-9 CH3 5.7223-7 9.0987-7 1.3610-6 CH4 1.6051-5 1.7788-5 1.8895-5 C2H 1.761-10 7.185-10 2.535 -9 C2H2,acetylene 7.5 457-6 1.7705-5 3.6933-5 C2H2,vinylidene 3.948-10 1.239 -9 3.407 -9 C2H3,vinyl 2.035-10 5.281-10 1.212 -9 C2H4 1.8567-8 3.2185-8 5.0352-8 C3H3,2-propynl 5.440-11 1.921-10 5.690-10 C3H4,propyne 1.661-11 4.587-11 1.076-10 C4H2,butadiyne 2.261-11 1.251-10 5.471-10 *H 1.7740-3 2.5616-3 3.6322-3 *H2 9.9721-1 9.9641-1 9.9532-1 *Si 1.7529-5 3.1602-5 5.7389SiC 1.386-11 5.018-11 1.702-10 SiC2 1.8345-8 7.1694-8 2.5188-7 SiH 1.1380-5 1.7852-5 2.8400-5 SiH2 1.6244-6 2.1889-6 3.0153-6 SiH3 4.5585-8 5.4863-8 6.7863-8 SiH4 5.3105-8 4.8604-8 4.6359-8 Si2 1.5149-7 2.8984-7 5.7767-7 Si2C 7.9586-7 1.8485-6 4.2651-6 Si3 1.2906-8 2.3190-8 4.5172-8 SiC(b) 9.6224-4 9.3828-4 8.9493-4 THERMODYNAMIC PROPER TIES FITTED TO 20000.K PRODUCTS WHICH WERE CONSIDERED BUT WHOSE MOLE FRACTIONS WERE LESS THAN 1.000000E10 FOR ALL ASSIGNED CONDITIONS *C *CH *C2 C2H5 C2H6 *C3 C3H3,1-propynl C3H4,allene C3H4,cycloC3H5,allyl C3H6,propylene C3H6,cycloC3H7,n-propyl C3H7,i-propyl C3H8 *C4 C4H4,1,3-cycloC4H6,butadiene C4H6,1butyne C4H6,2butyne C4H6,cycloC4H8,1-butene C4H8,cis2-buten C4H8,tr2-butene C4H8,isobutene C4H8,cycloC4H9,n-butyl C4H9,i-butyl C4H9,s-butyl C4H9,t-butyl C4H10,n-butane C4H10,isobutane *C5 C5H6,1,3cycloC5H8,cycloC5H10,1-pentene C5H10,cycloC5H11,pentyl C5H11,t-pentyl C5H12,n-pentane C5H12,i-pentane CH3C(CH3)2CH3 C6H2 C6H5,phenyl C6H6 C6H10,cycloC6H12,1hexene C6H12,cycloC6H13,n-hexyl C6H14,n-hexane C7H7,benzyl C7H8 C7H14,1-heptene C7H15,nheptyl C7H16,n-heptane C7H16,2-methylh C8H8,styrene C8H10,ethylbenz C8H16,1-octene C8H17,noctyl C8H18,n-octane C8H18,isooctane C9H19,n-nonyl C10H8,naphthale C10H21,n-decyl C12H9,obipheny C12H10,biphenyl C(gr) Si(cr) Si(L) SiC(L)
174 Appendix A: (Continued) Plot File: P T CH2 CH3 CH4 C2H2,acetyle C2H4 H 2.00E-01 1.50E+03 1.22E-13 3.19E-09 3.95E-06 5.23E-10 3.01E-11 3.94E-05 2.00E-01 1.55E+03 3.96E-13 5.47E-09 3.78E-06 1.36E-09 4.83E-11 7.04E-05 2.00E-01 1.60E+03 1.20E-12 9.08E-09 3.63E-06 3.33E-09 7.51E-11 1.22E-04 2.00E-01 1.65E+03 3.75E-12 1.62E-08 3.88E-06 9.47E-09 1.40E-10 2.03E-04 2.00E-01 1.70E+03 1.48E-11 3.75E-08 5.56E-06 4.61E-08 4.55E-10 3.30E-04 2.00E-01 1.75E+03 5.41E-11 8.33E-08 7.85E-06 2.06E-07 1.40E-09 5.20E-04 2.00E-01 1.80E+03 1.81E-10 1.73E-07 1.07E-05 8.16E-07 3.87E-09 8.01E-04 2.00E-01 1.85E+03 5.39E-10 3.31E-07 1.36E-05 2.73E-06 9.24E-09 1.20E-03 2.00E-01 1.90E+03 1.42E-09 5.72E-07 1.61E-05 7.55E-06 1.86E-08 1.77E-03 2.00E-01 1.95E+03 3.38E-09 9.10E-07 1.78E-05 1.77E-05 3.22E-08 2.56E-03 2.00E-01 2.00E+03 7.41E-09 1.36E-06 1.89E-05 3.69E-05 5.04E-08 3.63E-03 P T CH2 CH3 CH4 C2H2,acetyle C2H4 H
175Appendix B: Characterization Techniques The films grown during this research ha ve been characterized by many different techniques. The film surfaces were analy zed using both scanning electron microscopy (SEM) and atomic force microscopy (AFM). Th e thickness of the films was determined by cross-section SEM imaging, while a select few were determined by Fourier transform infrared (FTIR) analysis. The electrical char acteristics of the films were investigated using the capacitance-voltage (C-V) techniqu e. The structural a nd optical quality was determined by X-ray diffraction (XRD) and low temperature photoluminescence (LTPL), respectively. The characterization technique s performed on the 4H-SiC grown epitaxial layers are discussed below. (1) Secondary Electron Microscopy (SEM): The surface morphology and thickness of 4H-SiC epitaxial films grown during this research were analyzed using a Hitach i S-800 field emission SEM. Magnifications as high as 300,000X are achievable with this Hi tachi SEM, however, for the use of the samples grown during this research, a magnification range of 100 30,000X was typically used. The image resolution for th e Hitachi is ~ 20 . The following is a description of how the SEM is used to analyze the surface of the films. The Hitachi SEM employed in this work uses a field emission electron source to produce an electron beam which is focused ont o the sample surface. Secondary electrons are then collected and a magnified image of the secondary electr on distribution is
176 Appendix B: (Continued) displayed on a cathode ray tube (CRT) . The electron beam is produced by applying a voltage (extraction voltage across a cathode and anode) in which el ectrons are emitted from the cathode. The typical extraction voltage used to analyze the samples was 2 kV. The electrons were then accelerated by an ap plied voltage called the acceleration voltage, which was typically set between 15 20 kV to analyze the 4H-SiC samples in this research. The electron beam is focused by a condensing lens, which reduces the diameter of the electron beam. A scan coil is used to electromagnetically scan the electron beam across the sample surface. An obj ective lens is also used to focus the electron beam. The SEM is operated under vacuum to prevent at mospheric interactions with the emitted secondary electrons. The incident electron beam on the sample surface stimulates the emission of secondary electrons from the sample surface. There are three types of electrons emitted: secondary electrons, auger electrons, and b ackscattered electrons. Figure B.1 shows a sketch of the emitted electrons from the sample created by the incident electron beam. The resolution of images usi ng the SEM is very high beca use the secondary electrons emerge from the first 50 500 of the sample surface. The secondary electrons emitted from the sample are then colle cted by a detector. The main co mponent of the detector is a scintillation material, which emits light wh en struck by electrons. The light is then amplified and displayed on a CRT. With a CCD camera, the images may also be captured electronically and viewed on a computer monitor.
177 Appendix B: (Continued) Figure B.1 Sketch of electron distribution from an electron beam bombardment onto a sample surface . An example of the surface morphology evaluated using the Hitachi SEM as a characterization tool is shown in Figure B.2. The film in Figure B.2(a) displays stepbunching, which is created when atomic steps co mbine. The result is a film with lines on the surface as seen in the micrograph. Th e surface of a good quality film should be smooth. Therefore, the growth process that resulted in the film in Figure B.2(a) was changed to produce a higher quality surface. Figu re B.2(b) displays a film with no surface structure, indicating a smooth morphology. The SEM therefore is a useful tool to help characterize the surface of the grown films and to improve th e 4H-SiC epitaxial films.
178 Appendix B: (Continued) (a) (b) Figure B.2 Plan-view SEM images of (a) a sample with step bunching and (b) a sample with smooth morphology. The thickness of the epitaxial layers was estimated by cleaving the samples and looking at the cross-section of the film using the SEM. Contrast between the epitaxial layer and the substrate can be seen with th e SEM, provided there is a difference in the doping densities of the two regi ons (which is typically the case). An n-type material requires a higher energy for secondary electr on emission compared to a p-type material . Therefore, the SEM image for p-type mate rial will be brighter compared to that of n-type material for the same voltage applied. The thickness of each epitaxial film was typically measured in three different positions along the sample edge and the valu es averaged. Based on the estimated film thickness and known growth time, the growth rate (defined as the average thickness divided by growth time and denoted by Vg in m/h) was estimated. Figure B.3 shows an example of three cross-section SEM microgr aphs taken on one sample at different positions. The sample was mounted vertically so that the sample edge could be studied. The measurements were taken on the top, middle, and bottom of the cleaved edge.
179 Appendix B: (Continued) Occasionally, difficulties were experienced in viewing the epitaxial film due to the lack of contrast in the images caused by cleavage lines on the crystal edge. Therefore, the thickness was based on only two positions inst ead of the typical three. For the sample shown in Figure B.3, the aver age thickness was approximat ely 4.2 m and the growth time was 15 minutes. The calculated average growth rate was therefore 17 m/h. (a) (b) (c) Figure B.3 Cross-section SEM micrographs of a 4H-SiC cleaved sample edge showing the epitaxial layer and substrate fo r (a) top, (b) middle, and (c) bottom positions of the sample. The measured thicknesses were 4.1 m, 4.3 m and 4.3 m, respectively, yielding an average Vg of 17 m/h (t = 15 minutes). (2) Atomic Force Microscopy (AFM) A Digital Instruments Nanoscope Dime nsion 300 AFM was used to analyze the surface morphology of the 4H-SiC epitaxial films. The AF M provides information of surface topography with angstrom scale re solution. The AFM produces images in 3dimensions by continuous XY raster scans acro ss a selected area of the sample. Surface features such as step bunc hing may be investigated with this tool. The RMS surface roughness may be obtained from AFM scans, wh ich is an indication of the vertical
180 Appendix B: (Continued) deviation of the AFM tip from a horizonta l reference surface. This information is beneficial in characterizing the morphology of the epitaxial layers. For a smooth, specular film morphology, the RMS surface roughness w ill be lower than that for a surface that contains surface defects or a rough morphology. A probe, which has an atomically sharp tip at the end of a cantilever (usually Si), is scanned over the surface of a sample using a piezoelectric scanner . When the tip is within an interatomic distance to the sample, van der Waals forces either repel or attract the tip to the sample surface. A laser is focu sed on the backside of the cantilever tip and is reflected into a position sensitive detector, which is shown in Figure B.4. When the cantilever changes position due to the attractiv e and repulsive forces between the tip and the sample, the laser position changes and is detected by the detector. This change in height determines the topography of the sa mple. The AFM images are produced through a feedback loop between the detect ion system and the scanner . Figure B.4 AFM beam sample detection syst em . The laser is directed on the backend of the cantilever and is reflected to a position sensitive detector. This is how the height information is detected from the sample surface.
181 Appendix B: (Continued) The mode used for the AFM analysis was non-contact tapping mode. In this mode, the cantilever is positioned over the sample and taps the surface for a small portion of the oscillation time, instead of being in continuous contact with the sample. This mode helps to eliminate damaging forces that ma y occur in contact-mode. The amplitude is maintained constant at the operating frequency . The typical scan area for the 4H-SiC samples was 10 m x 10 m. The scan rate depends on the scan size, in that a smaller scan rate is needed for larger scan areas to reduce damage to the surface. The typical scan rate used, for scan areas of 10 m x 10 m, was 0.5 Hz. An AFM scan of a sample grown during this research is shown in Figure B.5. Figure B.5 AFM surface scans of 4H-SiC epitaxial films grown in the 200 mm LP hot-wall CVD reactor. The resulting surf ace roughness values was 0.9 nm RMS. The AFM scan shown in Figure B.5 was used to determine the surface morphology of a 4H-SiC film grown during this research. The scan displays a defect on the surface of the film in th e top right corner which may have been extended from the substrate. The surface roughness was determin ed to be 0.9 nm RMS, which is a representative of a relatively smooth film. As seen from the image in Figure B.5, the surface morphology of films may be determined by using the AFM tool.
182 Appendix B: (Continued) (3) X-ray Diffraction (XRD) X-ray diffraction was used to characteri ze the structural qual ity of the 4H-SiC epitaxial films grown. This is a non-destruct ive technique in which the epitaxial film is subjected to an X-ray beam. The X-rays ar e generated within an X-ray tube where a current is passed through a filament and el ectrons are thermally produced. The electrons are accelerated from a cathode towards an anode due to an applied potential, in which the electrons within the anode are bombarded out of their valence shell from the collision, creating vacancies. When electrons from a higher shell fill the vacancies of the lower shell, energy is released and is given off as an X-ray. These released X-rays are guided through a window within the X-ray tube and cr eate an X-ray beam. The X-ray beam is then guided toward the sample where diffrac tion may take place. The X-rays that are diffracted from the film surface are detect ed by a detector. The amount of X-rays collected are converted into in tensity and plotted ag ainst the angl e position of the detector (or sample stage, depending upon the type of scan performed). Diffraction is only possible when the wave s of the X-rays are in phase with each other. When the waves are out of phase (or destructive interference ), they cancel each other out, in which diffraction cannot occu r. However, when the waves are inphase(constructive interference), the waves reinforce each other, creating diffracted Xrays. For diffraction to occur, the Braggs law must be satisfied (shown in Figure B.6), which is: sin 2 d n (B.1)
183 Appendix B: (Continued) where n = order of diffraction, = wavelength of X-rays, d = distance between atomic, and = angle of incidence. Figure B.6 Demonstration of Braggs law, s howing the angle of in cidence, the atomic plane spacing, and two X-ra ys being diffracted . High-resolution XRD (HRXRD) may be used to investigate epitaxial layers lattice constants, dopant concentration, latt ice mismatch, epitaxial thickness, and crystal perfection . One concentration for char acterization of the epitaxial films grown during this research was the determination of crystalline quality. Within the epitaxial film, defects such as dislocations and point defects may be present, which may significantly influence the diffraction of X -rays . The defect s may contribute to complicated X-ray diffraction patterns. The typical scan performed for the 4H-SiC epitaxial st ructural characterization was an X-ray rocking curve ( -scan). To obtain a rocking curve, the sample is slowly rotated or rocked about an axis relative the sample normal . The diffraction peak is then collected as intensity versus the angle of diffraction. This type of scan was used to determine the structural quali ty of the film by measuring the FWHM value of the (0004)
184 Appendix B: (Continued) diffraction peak. The (0004) reflection plane wa s measured because the epitaxial layers were 4H-SiC. The higher quality (less crys tal defects) films produce a lower FWHM value. Peak broadening may be contributed to curvature of the crystal or imperfect epilayers created by defects [ 86]. If double diffraction peaks are observed, it may indicate that the epitaxial film is tilted against the substrate . The high-resolution X-ray diffraction m easurements were performed using a Phillips XPert Pro XRD system. Figure B.7 shows a schematic drawing of the diffractometer arrangement. The Cu K 1 source (at 45kV and 40 mA) was selected using a 4-bounce Ge (440) high resolution monochr omator, which was placed on the incident beam side between the X-ray tube and the sample. The Ge (440) monochromator, used for high quality semiconductor material, is capable of resolving closely spaced peaks . The beam size displaced on the sample was approximately 5mm x 5mm, which was controlled by adjustment knobs on the monochromator. The diffracted beam optics utilized during the XRD measurements was a tr iple axis attachment, which may be used for rocking curve measurements. This attachment was placed between the sample and the detector. The triple axis optics uses a Ge crystal analyzer which allows an angle of acceptance for the incoming beam of less than 12 arcsec . The combined incident and diffracted beam optics were employed to perform the rocking curve measurements.
185 Appendix B: (Continued) Figure B.7 X-ray diffractometer schematic for rocking curve measurements. The schematic contains the X-ray source, monochromator, sample, analyzer crystal, and detector . Figure B.8 shows XRD rocking curves of the (0004) peak of 4H-SiC samples. The diffraction peak shown in Figure B.8(a) is a single peak with a FWHM of 36 arcsec, indicating a decent quality film The (0004) diffraction peak of the 4H-SiC sample shown in Figure B.8(b) has two peaks. This may be caused from the epitaxial layer being so thin. When the film is thin, diffraction from the substrate may also be displayed. The FWHM of the peak is approximately 22 arcs ec, also indicating decent quality film. As shown from Figure B.8, the film quality of the epitaxial layers may be found by using XRD as a characterization technique.
186 Appendix B: (Continued) Figure B.8 X-ray diffraction rocking cu rves for 4H-SiC films of (a) 23 m thickness and (b) ~ 3.4 m thickness. (4) Current-Voltage/ Capacitance-Voltage (CV/IV) The net carrier concentration of the 4H-SiC samples was determined by performing capacitance-voltage measurements using a HP 4280A C-V meter. The grown samples were deposited with a backside contact of Ni-Ti-Au and a front side contact of Ni-Au. Before capacitance-voltage measurem ents were performed, current-voltage measurements were taken, using a HP 4145B Semiconductor Parameter Analyzer, to verify the device was a diode. Once determ ining the device was a schottky diode (this was the type of diode used for the CV-IV measurements during this research), C-V measurements were then be performed. The C-V measurements were based on the depletion approximation. The width of the space charge region, a region where there are no mobile carriers, depends on the applied voltage. The relationship between the space charge region and the doping density is found in the following equations. C A K Ws 0 (B.2)
187 Appendix B: (Continued) where W is the space charge region, Ks is the semiconductor dielectric constant 0 is the permittivity, A is the area of the diode, and C is the capacitance. and dV C d A qK W Ns A/ 1 2 ) (2 2 0 (B.3) where NA is the p-type doping density, q is the electron charge, and V is the voltage. Note that equation (B.3) has assumed that NA >> ND. A reverse-bias is applied to the meta l/semiconductor juncti on (schottky diode) and the capacitance is measured as the width of the space charge region is increased. The typical voltage range that was applied to an n-type epitaxial layer was -5V to 0V with a step size of 0.05V. The doping density can be found either by the slope (d C /d V ) of the plot, C versus V or by taking the slope (d(1/ C2)/d V ) of the plot 1/ C2 versus V . An example of the I-V and C-V pl ots used to obtain the doping concentration of an n-type, 4H-SiC epitaxial layer are s hown in Figure B.9. Note the C-V method measures the net carrier concentration. For an n-type material Nd-Na is found, while for a p-type film, the carrier concentration is determined as Na-Nd.
188 Appendix B: (Continued) A characteristic I-V curve for an n-type 4H-SiC epitaxial la yer (not grown during this dissertation research), is shown in Figure B.9(a). At ap proximately 0.5 V, the diode turns on. The capacitance measured as a func tion of voltage is shown in FigureB.9(b). The data is entered into a plot ting program which then plots (1/C2) vs. V, shown in Figure B.9(c), using the equation (B.3). Finally, th e doping concentration is extracted (Nd-Na ~1.5 x 1015 cm-3) and an example is shown in Figure B.9(d). (a) (b) (c) (d) Figure B.9 Data plots of I-V and C-V curves used to extract the doping concentration of an n-type 4H-SiC epitaxial layer. (a) Current-voltage curve, (b) capacitance-voltage curve, (c) inverse of capacitance squared-volta ge curve, and (d) net carrier concentration vs. depth curve. Data taken by J. Wa lker, University of South Florida.
189 Appendix B: (Continued) (5) Fourier Transform Infr ared Spectroscopy (FTIR) Selected epitaxial films thicknesses were determined using FTIR analysis, as discussed in Chapter 4, Section 4.3. 2. An Accent QS 1200 by Accent Optical Technologies, Inc FTIR spectrometer was used for the measurements. This thickness measurement method is non-destructive and ha s the added benefit of quick processing results, while that of SEM analysis is de structive due to samp le cleavage. The SEM analysis method is quite time consuming. The basic principle of FTIR spectroscopy is that an optical intensity is determined versus wavelength . With this technique, th ere is an interference of light waves which is measured. A schematic of an FTIR spectrometer is shown in Figure B.10. An infrared beam is directed to a beam splitter, where part of the beam passes th rough the splitter and is reflected from a fixed mirror (M1). The othe r portion is reflected off the beam splitter, and then is directed to a movable mirror (M 2), in which the beam is reflected back. Figure B.10 Diagram of the FTIR spectroscopy set-up. I ( ) is the interference signal, BS is the beam splitter, M1 is the fixed mirror, and M2 is the movable mirror .
190 Appendix B: (Continued) As M2 moves, the interference signal varies due to the recombining of the waves from M1 and M2. When M2 changes positions, the length of travel ( ) changes for the wave from M2, thus changing intensity of th e combined wave. The detector measures the interference signal I ( ) and an interferogram is develope d, which is the intensity versus wavelength. A sample is placed before the de tector in order to measure the reflectance from the epitaxial sample. Software is th en used to convert the wavelength to the epitaxial thickness. The data from the FTIR spectrometer is called an interferogram, which is a summation of cosine waves . The configur ation of the FTIR to perform epitaxial thickness measurements is reflectance. In this mode, a light is observed by the detector after being displaced by the sample surface. Figure B.11 schematically demonstrates the reflectance of incident radiat ion upon an epitaxial layer. Figure B.11 Schematic drawing of the radiat ion reflections during epitaxial thickness measurements using FTIR. The radiation exit s the interferometer and reflects from the surface of the epitaxial layer (R1) and the remaining radi ation penetrates through the epilayer and reflects from the epi/substrate interface (R2) .
191 Appendix B: (Continued) A portion of the infrared radiation from th e interferometer (which consists of the stationary and movable mirrors, M1 and M2 respectively (shown in Figure B.10)) is reflected from the top surface of the epitaxia l layer. This is the primary reflection (or primary interferogram), indicated as R1 in Figure B.11. The remaining radiation penetrates through the layer to the epi/substrat e interface and is reflected. This is the secondary reflection (or secondary interferogra m) and is indicated as R2 in Figure B.11. The difference between the primary and sec ondary reflections produc es an interference pattern. The interference pattern depends on the doping concentratio n of the epitaxial film, the cosine of the angle shown in Figur e B.11, and the thicknes s of the layer. The doping concentration difference between the epilayer and substrate determines the strength of the thickness measurement signal. In order to obtain the data in a desired format, a Fourier transf orm is applied.  The process steps used to determine the film thickness using FTIR spectroscopy is shown in Figure B.12. Prior to evaluating the sample, a reference is assessed to eliminate any signal that may not come from the di fference between R1 and R2 . Once the interferogram of the reference has been obtained, an interferogram of the epitaxial sample is taken. An example of the interferogram is shown in Figure B.12(a). The spectra are then transformed into single-beam spectra by Fourier transforms (Figure B.12(b)). The sample and reference signals are then subt racted from each other (Figure B.12(c)). A second Fourier transform is applied to the signal, at which time a phase correction is taken of the spectrum. The results are then reported as thickness versus intensity
192 Appendix B: (Continued) (Figure B.12(d)). A software package, FilmZ by Accent Optical Technologies, automatically performs these steps for each film thickness measurement. The epitaxial film thickness is displayed as the peak with the strongest signal in Figure B.12(d). (a) (b) (c) (d) Figure B.12 Flow diagram for epitaxial f ilm thickness determination using FTIR spectroscopy. (a) Reflectance interferograms from the epitaxial layer and epi/substrate interface, (b) single-beam spectra created by Fourier transform of the reflectance interferograms, (c) subtraction of the si ngle-beam spectra, and (d) second Fourier transform of the spectra (top spectra) followe d by a report of thickness value determined by phase correction .
193 Appendix B: (Continued) Figure B.13 shows a representative of a film characterized by FTIR. The film thickness was estimated to be ~ 8 m as determined from the Fourier transform of the reflectance spectrum. The vertical line in the figure indicates the thickness of the film. Figure B. 13 Fourier transform of the reflec tance spectrum showing a thickness of ~ 8 m. (6) Low-temperature Photoluminescence (LTPL) LTPL was performed on the grown 4H-SiC films to determine their optical quality. The analysis was performed at the University of Pittsburgh, by Professor W. J. Choykes group. PL is a non-de structive method which can de termine donor and acceptor impurities within the grown film. The impurities which can be detected are those that produce radiative recombin ation processes . In LTPL experiments, the sample is t ypically placed in a cryostat because low temperatures are needed to reduce thermal broadening of excited carriers and reduce nonradiative recombination proces ses [79,89]. For the LTPL analysis performed by Professor Choykes group, the measurements were ta ken at 2K. A laser (with a wavelength ~
194 Appendix B: (Continued) 4000 for the measurements performed during this dissertation) is used to excite electrons to create electron-hole pairs, which then r ecombine and give off photons. For LTPL, the photons should mostly be radiative. The laser must have a higher energy than the band-gap of the material to create the electron-hole pairs. A diagram of the PL arrangement is shown in Figure B.14. The emitted photoluminescence from the sample passes through a monochromator, which provid es one wavelength to the detector. The data is collected as photoluminescence inte nsity versus photon en ergy or wavelength. Figure B.14 Diagram of photoluminescence arrangement which includes a source laser, a mirror, a monochromator, a detector, and a sample placed inside a cryostat . The energy of the photons depends on the t ype of recombination process. There are five main radiative recombination pro cesses which are shown in Figure B.15. Bandto-band recombination is show n in Figure B.15(a), however, it is not typically seen in LTPL. Figure B.15(b) shows a free-exciton reco mbination,. A free-exciton is an electronhole pair loosely bound to each other. The free-exciton is free to move throughout the crystal.
195 Appendix B: (Continued) Figure B.15 Recombination processes that may create radiative photons which can be detected for PL analysis. (a) Band-to-band recombination, (b) free-exciton recombination, (c) bound-exciton recombin ation by a neutral acceptor, (d) bound-exciton recombination by a neutral donor, and (e) donoracceptor (D-A) recombination. EV is the valence band energy, EC is the conduction ba nd energy, EA is the acceptor energy and ED is the donor energy.  The bound-exciton recombination is show n in Figure B.15(c) and (d). A boundexciton is where a free exciton (or electron hole pair) is captured by a shallow donor. There are two types of bound excitons; the neutral bound exciton a nd the ionized bound exciton. The most common is the neutral boun d exciton where the free exciton forms a four particle complex together with the donor s electron (or acceptors hole) and with the defect core. The ionized bound exciton is a free exciton bound to an ionized donor or acceptor. In Figure B.15(c), the donor is trap ped in the shallow site, while in Figure B.15(d), the acceptor is trapped in the shallow site. Note that for less pure films, the bound-exciton recombination dominates that of the free-exciton recombination. Finally, Figure B.15(e) demonstrates an acceptor-dono r recombination (also referred to as a donor-acceptor pair recombina tion) . For heavily doped samples which contain both acceptors and donors, the exciton luminescence may not be strong enough for detection . For this type of materi al, the most dominant lumine scence is that from the donor
196 Appendix B: (Continued) acceptor pair (DAP) recombination. The inte nsity of the DAP recombination typically appears in the no-phonon distant pairs peak along with the phonon replicas . The distance between the donor and acceptor determines the DAP spectrum. In low temperature photoluminescence, the recombination of neutral boundexcitons (neutral four-particle donor (nitr ogen) or acceptor (aluminum) complexes) is measured to determine the photon energies [ 91]. In a donor complex, the four particles consist of the donor ion, two electrons, and a hole for donors. The presence of nitrogen within the epitaxial film can be determined using photolum inescence, in which the 4N (four-particle nitrogen donor complex) spectr um is associated with the radiative recombination of the neutral nitrogen donor four-particle bound-exciton complex . A neutral four-particle aluminum acceptor bound complex may be observed in 4Hand 6HSiC polytypes using LTPL as well. The four particles consist of the negatively charged acceptor ion, its hole, and the electron-hole pair which is attached to the exciton. The number of no-phonon lines present within the LTPL spectra is corresponds to the inequivalent sites on the C sub-lattice [ 91]. There are approx imately 15 no-phonon lines present in the 4Al LTPL spectrum for the 6H -SiC polytype. However, for the 4H-SiC, there are only a few no-phonon lines in the 4Al LTPL spectrum. High-quality material results in the recombination of free excitons during photoluminescence measurements . An ex ample of a near bandedge LTPL spectrum of a 4H-SiC epitaxial film is shown in Figure B.16. The free exciton peak is labeled I75 on the spectrum. For higher quality films, th e free exciton related line becomes stronger in the spectrum . The 4Al0 line indicates the no-phonon line of the neutral four-
197 Appendix B: (Continued) particle aluminum acceptor co mplex. In the 4H-SiC polytype, the P and Q lines are representative of the neutra l four-particle nitrogen donor co mplexes at different lattice sites. While that in the 6H-S iC polytype are indicated at P, R and S. The subscript for each line comprises of the phonon replica en ergy in meV . The strongest phonon replicas are those with the weakest no-phonon line . Th e no-phonon lines are indicated with the subscript 0. The P0 peak is characteristic of the neutral nitrogen boundexciton at the hexagonal site, while the Q0 peak is that from the cubic site . The nophonon lines are due to the loss of constant translation related to the binding of the exciton . There is not a no-phonon line fo r the free-exciton because it is not bound. Figure B.16 Near bandedge LTPL spectrum of a 23 m thick 4H-SiC film. Strong free exciton line is labeled I75. P0 and Q0 lines are the neutral f our-particle nitrogen donor complexes, while the 4Al0 peak indicates the no-phonon line of the neutral four-particle aluminum acceptor complex. LTPL may also be used to determine the presence of nitrogen, as well as evaluating the epitaxial film quality. The con centration of the dopant may be assessed by
198 Appendix B: (Continued) comparing the intrinsic component, i.e. free exciton line, to that of the extrinsic component, i.e. the bound-excit on line . The nitrogen c oncentration within 6H-SiC may be found by the ratio of the intensity of th e no-phonon line, P0, to the free exciton replica, I77. The I77 peak is the strongest free exciton line. The concentration of nitrogen within a 4H-SiC epitaxial layer may be found by the ratio of Q0 to I75. For films with low nitrogen doping concentrations, the free exciton line will become stronger .
199Appendix C: Operating Procedure for 4H -SIC Epitaxial Growth in the USF 200 mm Reactor The development of the 4H-SiC baseli ne growth process was presented in Chapter 3, Section 3.4. A process has been developed where high, repeatable growth rates are achieved. The process procedure whic h will be detailed next may be used to grow 4H-SiC in the 200 mm USF hot-wall C VD reactor. Parameters may need to be changed slightly from time to time, such as the temperature due to slight drifts in the apparatus. A sketch of the process schedule was shown in Figure 3.14. The step-by-step procedure for the epitaxial growth is detailed next. Sample Loading: 1. Verify that the reactor hot-zone is free of debris by vacuuming out the hot-zone. 2. Load SiC substrate onto the polyplate a nd load the plate into the hot-zone. 3. Close reactor door with uniform pressure (inspect o-ring seals prior to closing).
200 Appendix C: (Continued) Reactor Seal Verification: Purge Ar for ~ 2 min by pressing the PURGE button and the Ar button on the reactor control panel. This will purge moisture, air, and contaminants from the reactor. 1. Evacuate the reaction tube via manual pumping: press H2 LEAK and press MANUAL PUMPING. (The H2 LEAK flows hydrogen through the reaction tube. By pressing the H2 LEAK, the H2 will not flow, which will allow pumping to occur.) 2. Verify PURGE button is pressed and switch from Ar purge to H2 purge. 3. Once reactor pressure is below 500 mTorr, de-press H2 LEAK to allow H2 to start to pressurize the reaction tube. 4. After approximately 1 min (or when the pressure gauge reads atmospheric pressure), de-press MANUAL PUMPING. This will begin H2 purge. 5. Allow H2 to purge for ~ 2 minutes. 6. Open labview program from reactor computer. 7. Set a flow of 10 slm of H2 in the labview program. 8. Press PROCESS MONITOR and PROCESS H2 to enable a H2 flow for ~ 2 minutes. This will allow the H2 in the gas line to be purged from the manifold box through the reaction tube.
201 Appendix C: (Continued) Growth Process: 1. Make sure all process gases are ready fo r growth (by having each line purged with the process gases), cooling water is on, and RF generator is on. 2. Set the desired flows for H2, Ar, C3H8 and SiH4 within the labview program. 3. Set the desired growth temperature, whic h is the set point temperature based on the Si melt test, in the labview program. 4. Start foam and view port argon purge by pressing TOP Ar. 5. Set the desired process pres sure in the pressure c ontroller and press POPPET VALVE. Make sure process pressure is achieved. 6. Check and clear interlocks by pressing I RRESET. If no sensors go off, then proceed to next step. If a sensor is activated, check and fix problem before continuing. 7. Turn off interlock override by pressing SHUT. This will enable all sensors so that if they are activated. In the event of a sensor trip, the system will shut down and go into an Ar purge. 8. Turn on RF power by enabling the inverter on the RF generator. This is done by flipping the switch up on the pa nel of the RF generator. 9. Set the board voltage to 5V in the Labview program. 10. Once the pyrometer is reading temperature from the susceptor, increase the board voltage as needed to continue a steady in crease in temperature. Do not set board voltage above 7V.
202 Appendix C: (Continued) 11. Flow Ar, SiH4 and C3H8 within the vent line to establish a steady flow for each gas approximately 10 minutes before grow th by pressing VENT Ar first. Wait ~ 1 minute and then press VENT SiH4 and VENT C3H8. This allows the gases to go directly to vent. 12. Once temperature is ~ 20 C before set point temperature, go into automatic temperature control by pressing G o AUTO in labview program. 13. 15 C before set point temperature, increase H2 flow from 10 to 30 slm. 14. Once at growth temperature, introdu ce propane into process by de-pressing VENT C3H8 and press PROCESS C3H8. 15. After 30 seconds, introduce SiH4 into the process by pressing PROCESS SiH4 and de-pressing VENT SiH4. The growth time starts from the time SiH4 is added into the process. 16. Shut off PURGE Ar 30 seconds after VENT SiH4 has been de-pressed. 17. Maintain these flows and temper ature throughout the growth run. 18. After the growth run is over, turn off RF power by setting the set point temperature and board voltage to zero. Shut SiH4 and C3H8 off by de-pressing PROCESS SiH4 and then PROCESS C3H8. 19. Disable the inverter on the RF genera tor by flipping the switch down and reset alarm on the RF generator. 20. Switch the H2 flow from 30 slm to 10 slm. 21. When temperature drops below 1000 C under these conditions, press and hold PURGE until the light on the button flashes.
203 Appendix C: (Continued) Note: This step is an auto purge which initially flows H2 for 30 seconds, and then is followed by an Ar purge for 30 minutes. 22. De-press POPPET VALVE. 23. After ~ 30 seconds, de-press TOP Ar. 24. Press SHUT to turn off interlocks. Unload sample: 1. After auto purge is complete and the susceptor is no longer glowing, press SAFE button to de-activate the safety locks. 2. Open reactor door, pull out the polyplat e and remove sample from polyplate.
204Appendix D: HCl Purge Panel Operating Procedure As described in Chapter 4, it is imperati ve to purge all HCl pr ocess lines with Ar at the end of use to prevent corrosion within the gas lines. If water vapor enters the HCl gas line, hydrochloric acid forms which is not only corrosive but al so highly toxic. An HCl purge panel was implemented into the gas delivery system at the beginning of the HCl gas line to reduce corrosion. The stepby-step procedure for evacuating the HCl gas line is described next. A photograph of the HC l purge panel was shown in Figure 4.4 and is duplicated here for easy reference. Figure D.1 HCl purge panel near the HCl gas bottle which was used to replace HCl with Ar in the gas lines to help prevent co rrosion. Numbers in the picture represent valve numbers.
205 Appendix D: (Continued) To purge the HCl gas out of the line, the HC l gas tank is closed and the following steps followed: (1) Close valves 1-4. (2) Open valve 5 (safety valve). (3) Open valve 6. This allows N2 to flow through the venturi (s hown in Figure 4.4(b)). The N2 gas goes through the ventur i at a high flow rate a nd to the exhaust. The N2 creates a vacuum, where the gases from the gas line are vacuumed out of the gas lines through valv es 3 and 4 from Figure 4.4(a). (4) Open valve 4 until low pr essure (~ 20 mmHg) is read on the low pressure side of the regulator. By opening this valve first, the pressure will not exceed that which is set on the low pressure side of the regulator. (5) Open valve 3. This allows the HCl gas to be purged from the gas lines. (6) Close valves 3 and 4. (7) Open valve 1 to back-fill Ar into the gas lines. Note the Ar tank should always be open to prevent HCl from backfilling into the Ar line. (8) Once the maximum pressure has been r eached (~ 10 15 psi), which is set by the regulator, close valve 1. (9) Repeat steps 4 8 approximately 10 15. (10) At the last backfill of Ar, l eave valve 1 open and close valve 6.
206 Appendix D: (Continued) (11) Purge Ar through the gas line for 10 minutes by setting a maximum flow (20 sccm) in the labview program  and press VENT Ar on the control panel. (12) Turn off gas flow by closing the HCl MFC and close valves 1-6. It should be noted that Valv e 2 is a secondary exhaust valve and is opened only when necessary. This may occur when the venturi has been damaged and the line needs to be purged with argon to the exhaust. To introduce HCl back into the gas line s, the following steps are followed: (1) Follow steps 2 11 above, which entail ed purging of the gas lines with Ar. This is performed to ensure that no air is in the gas li ne after not being operational for a long period of time. (2) Close valve1. (3) Open the HCl tank slowly. (4) Set a flow of 20 sccm in the labv iew program. Flow HCl gas to vent by pressing HCl VENT for 15 minutes. Simultaneously flow Ar through the vent to dilute the exhaust line.
207Appendix E: Thickness Determination via Cross-Section SEM and FTIR Analysis The thickness of the films grown in the 200 mm hot-wall reactor with the addition of HCl was presented in Chapter 4. The film thickness was initially determined by crosssection SEM analysis of the cleaved edge of the samples, which was the method employed for all the epi film thickness meas urements reported in this dissertation. Results from LTPL analysis of the film s displayed in Figure 4.22, performed by Professor Choykes group at the University of Pittsburgh, were not in agreement with the HCl additive 4H-SiC epitaxial thickness values that were measured using this method at USF. That is to say the Pittsburgh group noted substrate luminescence from epi samples which should not have been present given th at the epi thickness was greater than the optical absorption depth. Therefore, a study wa s performed to verify the thickness of the films grown (Figure 4.22) with the addition of HCl, as determined by cross-section SEM analysis. Fourier Transform Infrared Spectrosco py (FTIR) analysis was used to verify the thickness of the films as a new FTIR syst em (Accent Model QS 1200) was recently acquired by the SiC research group. Before addressing the specific goal of valid ating the growth rate of 4H-SiC using HCl as a growth additive (growth rates disp layed in Figure 4.22), a comparison of both the SEM and FTIR methods was made. A co mparison of the estimated film thickness determined by both FTIR and cross-secti on SEM was performed on a sample grown using the standard growth process (no HCl additive, see Chapter 3) in the 200 mm hotwall reactor. The epilayer was subjected fi rst to thickness measurement by FTIR. The measurement was performed three times for accuracy purposes. The sample was then
208 Appendix E: (Continued) cleaved and the cross-sectio n studied under the SEM where the thickness was taken at three positions along the film edge. The aver age value of the three measurements was then computed and tabulated. The sample was then again analyzed by FTIR to compare both measurements (note that only a single point on the sample can be measured with the FTIR due to the small sample size of ~5 mm x 8 mm and concerns that the optical beam impinges on the sample edge which would reduce the accuracy of the measurement). Table E.1 gives the average thicknesses from these three experiments. Table E.1 Epi layer thickness as determined by FTIR and cross-section SEM analysis. Sample ID Thickness (m), 1st FTIR measurement Thickness (m) by cross-section SEM* Thickness (m), 2nd FTIR measurement USF-04-060A 8.2 7.2 8.2 *value shown is average value of 3 meas urements made along the film edge. The FTIR spectra (reflectance spectra of the epi and the re ference substrate ratioed and shown in m units) from measurements of sample USF-05-060A prior to and after cleavage are shown in Figure E.1. As seen in the figure, there is a dominant peak which indicates the thickness of the epitaxial layer. The peak centered at 0 m in this figure (as well as in all other figures in this Appendi x) represents low frequency noise and should be disregarded.
209 Appendix E: (Continued) (a) (b) Figure E.1 FTIR spectra of sample USF-05060A (a) before sample was cleaved and (b) after sample was cleaved. In both cases the FTIR estimated thickness is 8.2 m. A cross-section SEM micrograph for the same sample (USF-05-060A) is shown in Figure E.2(a). Although the thickness seen in Figure E.2(a) is 7.39 m, the average thickness from three positions along the sample edge was calculated to be 7.2 m. Along with the epi thickness measurement, a 10 m calibration sample was used to determine how accurately the SEM was calibrated. The pl an-view image of the calibration sample is shown in Figure E.2(b). The distance from one etched line to anothe r is known to be 10 m, however, as seen in Figure E.2(b), it is only 9.33 m. Therefore, the thickness was underestimated using the Hit achi S-800 SEM with an error of ~ 7.5 %. The thickness measurements were adjusted to give the co rrect values of the epitaxial thickness via cross-section SEM. The corrected average va lue is 7.7 m for sample USF-06-060A in Figure E.2(a).
210 Appendix E: (Continued) (a) (b) Figure E.2 (a) Cross-section SEM microgra ph of USF-05-060A cleaved edge and (b) plan-view SEM micrograph of 10 m calibration sample. An error in the measurement of ~7.5% was observed. Note the error resulted in an underestimate of the film thickness and, therefore, growth rate. The thickness values from the FTIR analysis and the cross-section SEM analysis are similar with approximately a 6.5% di fference between the two methods for this example. With the results of the comparis on between the two thickness measurements in reasonable agreement, the thickness of th e 4H-SiC epitaxial fi lms grown with HCl additive were determined using FTIR analys is. This was performed to compare the FTIR results with those of cross-secti on SEM analysis applied earlier. The thickness of the grown films using HCl as a growth additive were first determined by cross-section SEM analysis (results presented in the graph of Figure 4.22). The samples were cleaved and thickness measurements taken at different positions. The average thickness was then calculated from these three values. The sample thicknesses were then evaluated using the FTIR. The th ickness of each sample was measured three
211 Appendix E: (Continued) times for accuracy purposes. The results of the average thickness using cross-section SEM and FTIR analysis are presented in the Table E.2 below. Table E.2 HCl growth additive epi laye r thickness summary from Chapter 4. Sample ID Thickness (m) via cross-section SEM Thickness (m), via FTIR % Error USF-05-347 7.67 7.9 3.0 USF-05-348 9.06 9.8 8.2 USF-05-350 10.15 10.4 2.5 USF-05-360 12.4 12.2 1.6 Representative spectra from the FTIR measurements and cross-section SEM micrographs are shown below. Figure E.3 di splays the spectrum for sample USF-05-347 and an SEM micrograph in which the film has an average thickness of 7.9 m. The average thickness of sample USF-05-348 is 9. 8 m and a representative FTIR spectrum along with a cross-section SEM image is sh own in Figure E.4. The representative spectrum and SEM micrograph for sample USF-05-350 is shown in Figure E.5. The mean thickness measured is 10.4 m. Finally, sample USF-05-360 has an average thickness of 12.2 m. A representative FTIR spectrum and SEM image is shown in Figure E.6.
212 Appendix E: (Continued) (a) (b) Figure E.3 Sample ID USF-05-347 (a) FTIR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 8.12 m. (a) (b) Figure E.4 Sample ID USF-05-348 (a) FTIR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 8.97 m.
213 Appendix E: (Continued) (a) (b) Figure E.5 Sample ID USF-05-350 (a) FTIR spectrum and (b) cross-section SEM micrograph of bottom edge showing thickness of 10.4 m. (a) (b) Figure E.6 Sample ID USF-05-360 (a) FTIR spectrum and (b) cross-section SEM micrograph of middle edge showing thickness of 12.3 m. Cross-section SEM analysis was initially used to determine the thickness of the samples. However, LTPL measurements perfor med at the University of Pittsburgh cast doubt on these measurements. The samples thicknesses were therefore measured
214 Appendix E: (Continued) independently using an FTIR to corroborate the thickness measurements that had been made previously using cross-section SEM analys is. For each sample, th ere is less than a 5 % difference between the two measurements except for sample USF-05-350, which has an error of ~8%. Therefore, when compar ing the FTIR thickness measurements to the cross-section SEM measurements, there is ev idence that the thickness of the samples grown using the HCl additive were indeed what was presented in Chapter 4, Figure 4.22.
ABOUT THE AUTHOR Rachael L. Myers-Ward obtained a Master of Science degree in Chemical Engineering from the University of South Florida in 2003. The focus of the research was 3C-SiC growth on novel Si substrates. She received he r Bachelor of Science degree in Chemical Engineering from the same university. Her Doctor of Philosophy degree will be in Electrical Engineering. The research work performed during both her Masters and Ph.D. has been supported by federal as well as indus trial sponsors. She had studied abroad for six weeks at the Max-Planck Institute in St uttgart, Germany under Dr. U. Starke during her dissertation work and has presented her re search at the 2003 International Conference on Silicon Carbide and Related Materi als (ICSCRM) (Lyon, France), 2004 Spring Materials Research Society (MRS) meeti ng (San Francisco, CA), 2004 European Conference on Silicon Carbide and Related Materials (ECSCRM) (Bologna, Italy), and lastly at the 2005 ICSCRM (Pittsburgh, PA).