|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 Ka
controlfield tag 001 002004903
007 cr mnu|||uuuuu
008 090527s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002691
Modeling and growth of the 3C-SiC heteroepitaxial system via chloride chemistry
h [electronic resource] /
by Meralys Reyes-Natal.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 120 pages.
Dissertation (Ph.D.)--University of South Florida, 2008.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Co-advisor: Stephen E. Saddow, Ph.D.
Co-advisor: John T. Wolan, Ph.D.
ABSTRACT: This dissertation study describes the development of novel heteroepitaxial growth of 3C-SiC layers by chemical vapor deposition (CVD). It was hypothesized that chloride addition to the "traditional" propane-silane-hydrogen precursors system will enhance the deposition growth rate and improve the material quality via reduced defect density. Thermodynamic equilibrium calculations were performed to obtain a criterion for which chloride specie to select for experimentation. This included the chlorocarbons, chlorosilanes, and hydrogen chloride (HCl) chloride containing groups. This study revealed no difference in the most dominant species present in the equilibrium composition mixture between the groups considered. Therefore, HCl was the chloride specie selected to test the hypothesis. CVD computerized fluid dynamic simulations were developed to predict the velocity, temperature and concentration profiles along the reactor.These simulations were performed using COMSOL Multiphysics and results are presented. The development of a high-temperature (1300 ¨C -1390¨C) 3C-SiC growth process resulted in deposition rates up to ~38 m/h. This is the highest value reported in the literature to date for 3C-SiC heteroepitaxy. XRD FWHM values obtained varied from 220 to 1160 arcsec depending of the process growth rate or film thickness. These values are superior or comparable to those reported in the literature. It was concluded from this study that at high deposition temperatures HCl addition to the precursor chemistry had the most significant impact on the epitaxial layer growth rate. Low-temperature (1000-1250¨C) 3C-SiC growth experiments evidenced that the highest deposition rate that could be attained was ~2.5 m/h.The best quality layer achieved in this study had a FWHM of 278 arcsec; which is comparable to values reported in the literature and to films grown at higher deposition temperatures in this study. It was concluded from this work that at lower deposition temperatures the HCl addition was more beneficial for the film quality by enhancing the surface. Surface roughness values for films grown with HCl additive were 10 times lower than for films grown without HCl. Characterization of the epitaxial layers was carried out via Nomarski optical microscopy, FTIR, SEM, AFM, XRD and XPS.
Computerized fluid dynamics
x Chemical Engineering
t USF Electronic Theses and Dissertations.
Modeling and Growth of the 3C-SiC Heteroepitaxial S ystem via Chloride Chemistry by Meralys Reyes-Natal A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Chemical Engineering College of Engineering University of South Florida Co-Major Professor: Stephen E. Saddow, Ph.D. Co-Major Professor: John T.Wolan, Ph.D. Norma A. Alcantar, Ph.D. Andrew M. Hoff, Ph.D. Ryan Toomey, Ph.D. Olle Kordina, Ph.D. Date of Approval: October 24, 2008 Keywords: homogeneous nucleation, hydrogen chloride computerized fluid dynamics, thermodynamic equilibrium, epitaxial layers Copyright 2008, Meralys Reyes-Natal
Dedication I would like to dedicate this dissertation to my f amily, especially to my parents for raising me with strong moral values and an enor mous will to become the best I can be. I would also like to dedicate this work to my h usband, for supporting me these past years and helping me achieve my most important grow th experiment, Sofia. Without all of you I wont be the person I am today. Thank you.
Acknowledgments There are many people I would like to acknowledge for their support and guidance throughout my dissertation work. First, I would like to express my gratitude to my co-major professors Dr. Stephen E. Saddow and Dr John T. Wolan. They have been my most loyal mentors and friends throughout my gra duate research studies. Thank you for making this experience possible for me. Next, I would like to thank Dr. Norma Alcantar, Dr. Ryan Toomey, Dr. Andrew M. Hoff, and Dr. Olle Kordina for accepting to be part of my committee. I am truly grateful for th eir interest in my research and their continued advice to ensure the success of my disser tation. Next, I wish to thank Suzie Harvey for assisting m e in my research and Dr. Y. Shishkin for his continued advice and support durin g process development. I extend my appreciation to past and current members of the SiC group; especially to Chris Frewin, I. Haselbarth, and Chris Locke for ensuring reactor op erations ran effortlessly. I would also like to thank Dr. C. Colletti and Alexandra Olivero s for film characterization support. I would also like to show appreciation to Dave Edward s, USF-COT at Largo, FL for providing SEM and AFM support. Additionally I would like to acknowledge the use of the services provided by Research Computing at the University of South Florida. This work was supported by grants from the Office of Naval Research under grant No. W911NF-05-2-0028 (C.E.C Wood) and the Army Rese arch Laboratory under grant No. DAAD19-R-0017 (B. Geil), which are gratefully a cknowledged.
i Table of Contents List of Tables ... .. iii List of Figures ... ..v Abstract ... ix Chapter 1: Introduction ........................... ................................................... ..................1 1.1 Overview....................................... ................................................... ............1 1.2 SiC crystallography............................ ................................................... .......2 1.3 SiC properties and applications................ ................................................... 4 1.4 3C-SiC hetero-epitaxy.......................... ................................................... ....7 1.4.1 Growth process............................... ................................................7 1.4.2 CVD growth precursors........................ ..........................................8 1.4.3 3C-SiC epitaxial film defects................ ........................................11 1.5 Summary........................................ ................................................... .........14 Chapter 2: Chemical Vapor Deposition............... ................................................... ..........15 2.1 Overview....................................... ................................................... ..........15 2.2 Chemical vapor deposition...................... ..................................................1 5 2.2.1 USF hot-wall CVD system...................... .....................................16 2.2.2 CVD thermodynamics........................... .......................................19 2.2.3 CVD kinetics................................. ................................................27 2.2.4 CVD transport and fluid dynamics............. ..................................30 2.2.5 Computerized Fluid Dynamic (CFD) simulations. .......................33 2.3 Summary........................................ ................................................... .........43 Chapter 3: High Temperature 3C-SiC Heteroepitaxial Growth.......................................44 3.1 Overview....................................... ................................................... ..........44 3.2 3C-SiC without HCl additive process development using Geometry I.....45 3.2.1 Carbonization stage.......................... .............................................45 3.2.2 Second thermal ramp.......................... ..........................................49 3.2.3 Growth stage................................. ................................................52 3.3 HCl additive process development using Geometry I...............................59 3.3.1 HCl addition to second thermal ramp.......... .................................59 3.3.2 HCl addition to the growth stage............. .....................................59 3.3.3 CVD reactor characterization................. ......................................61 3.4 HCl additive process optimization using Geometr y I................................65 3.4.1 Optimization of the carbonization step....... ..................................66
ii 3.4.2 Optimization of second thermal ramp.......... .................................68 3.4.3 Optimization of growth stage................. .......................................68 3.5 Process transfer to Geometry II using DOE...... ........................................70 3.5.1 DOE results.................................. .................................................71 3.6 Growth on 50 mm substrates using Geometry II... ....................................75 3.7 Summary........................................ ................................................... .........77 Chapter 4: Low Temperature 3C-SiC Heteroepitaxial G rowth........................................78 4.1 Overview....................................... ................................................... ..........78 4.2 Low-temperature 3C-SiC growth process developme nt............................79 4.2.1 Carbonization................................ ................................................79 4.2.2 Growth stage................................. ................................................79 4.2.3 Growth rate dependence on HCl mole fraction.. ..........................82 4.2.4 Growth rate as a function of temperature..... ................................84 4.2.5 Growth rate as a function of silane mole frac tion.........................85 4.3 Summary........................................ ................................................... .........87 Chapter 5: Summary and Future Work................. ................................................... .........88 5.1 Dissertation summary........................... ................................................... ..88 5.2 Future work and current work................... .................................................91 5.2.1 Species concentration profile simulation..... .................................91 5.2.2 Temperature profile simulation............... ......................................94 References .. 95 Appendices .. .. ... 102 Appendix A Reactions for the gas phase model ... 103 Appendix B Reactions for the surface reaction model .109 Appendix C Simulation procedure . ..113 Appendix D Statistical Design of Experiments (DOE) ... .117 About the Author ....End Pag e
iii List of Tables Table 1.1 Comparison of SiC and Si basic parameter s at 300 K.................................5 Table 2.1 Description of the reactors geometries c onsidered in this study................19 Table 2.2 Summary of chloride species considered i n the thermodynamic simulations........................................ ................................................... ......23 Table 2.3 Process parameters for 3C-SiC deposition process with and without HCl addition............................... ................................................... 23 Table 3.1 Summary of parameter ranges considered d uring the carbonization stage development.................................. ................................................... .47 Table 3.2 Summary of parameters ranges considered during the second thermal ramp development........................... .............................................50 Table 3.3 Summary of growth rate data for the calc ulation of baseline process repeatability using ~8.6 m/h process....... ...................................56 Table 3.4 Summary of process parameters during sec ond thermal optimization....................................... ................................................... .....67 Table 3.5 Summary of factors range considered to p erform 25-2 fractional factorial DOE...................................... ................................................... ....72 Table 3.6 Runs and experimental results for 25-2 DOE..............................................7 2 Table 3.7 Center point runs and experimental resul ts for the 25-2 DOE....................72 Table 3.8 DOE model results comparison with experi mental values for processes performed using Geometry I............... .......................................73 Table 3.9 Thickness measurements taken on 3 repres entative samples grown at a speed of 20, 30 and 38 m/h................... ............................................76
iv Table 3.10 XRD FWHM summary for films grown at 20, 30 and 38 m/h...............77 Table 4.1 Summary of factors range considered to p erform 26-2 fractional factorial DOE...................................... ................................................... ....80 Table 4.2 Experimental matrix and response values for 26-2 DOE............................80 Table 4.3 Center point runs and response values fo r 26-2 DOE.................................81
v List of Figures Figure 1.1 Illustration of a SiC tetrahedron that forms the basis for all SiC crystals........................................... ................................................... ...........3 Figure 1.2 Stacking sequence for the most common S iC polytypes.............................3 Figure 1.3 Schematic representation of a misfit di slocation......................................12 Figure 1.4 Schematic representation of the formati on of an edge dislocation............12 Figure 1.5 Illustration of stacking faults which a re defects caused by the misalignments of the crystal planes................ ...........................................13 Figure 1.6 Representation of a twin boundary defec t.................................................1 3 Figure 2.1 USF horizontal hot-wall CVD reactor.... ................................................... 17 Figure 2.2 Cross-section view sketch of the USF CV D reactor..................................18 Figure 2.3 Predicted product equilibrium mixture c omposition for (a) SiH4C3H8-H2 and (b). SiH4-C3H8-HCl-H2 precursor systems...........................25 Figure 2.4 Schematic of CVD steps................. ................................................... ........28 Figure 2.5 Boundary layer development near a flat surface........................................31 Figure 2.6 Representation of the rate limiting ste ps in a CVD reaction (a) surface reactionlimited and (b) mass transport limi ted..............................32 Figure 2.7 Reactor 2D modeling domain............. ................................................... ....33 Figure 2.8 Gas temperature profile across the CVD reactor configured for (a) Geometry I and (b) Geometry II..................... ...........................................37
vi Figure 2.9 Gas temperature variation along the CVD reactor for (a) Geometry I (b) Geometry II.................................. ................................................... ...38 Figure 2.10 Gas velocity profile across the CVD re actor configured for (a) Geometry I and (b) Geometry II..................... ...........................................40 Figure 2.11 Parabolic velocity fields for (a) Geom etry I and (b) Geometry II at susceptors inlet, center and outlet............... ..............................................41 Figure 2.12 Streamline plot of the velocity profil e illustrating fluid is flowing in parallel lines.................................. ................................................... ......42 Figure 3.1 Carbonization process schedule develope d for Geometry I......................47 Figure 3.2 XPS high resolution spectra of (a) C1S a nd (b) Si2p peaks for a representative carbonized layer.................... ..............................................48 Figure 3.3 Second thermal ramp process schedule de veloped for Geometry I, including the carbonization stage.................. .............................................50 Figure 3.4 Plan view SEM image of a representative layer after carbonization and second thermal ramp at best process conditions. ................................51 Figure 3.5 AFM micrographs of a representative lay er after the carbonization and second thermal ramp processes.................. .........................................52 Figure 3.6 XPS high resolution spectra of (a) C1S and (b) Si2p peaks of a representative layer after the carbonization and se cond thermal ramp processes..................................... ................................................... ...53 Figure 3.7 3C-SiC growth process schedule using C3H8-SiH4-H2 chemistry developed for Geometry I........................... ...............................................55 Figure 3.8 Plan-view SEM image shown at a magnific ation of 5.0k for a representative 3C-SiC layer grown with the no HCl p rocess at a rate of ~8.6 m/h grown.......................................... ..................................56 Figure 3.9 AFM micrograph (10 m x 10 m) taken in contact mode of a representative 3C-SiC layer grown at a rate of 8.6 m/h...........................57 Figure 3.10 XRD powder diffraction of a 3.3 m thick 3C-SiC epitaxial layer...........58
vii Figure 3.11 XRD rocking curve of a 3.3 m thick 3C-SiC epitaxial layer performed at the 3C-SiC (002) diffraction peak..... ...................................58 Figure 3.12 3C-SiC HCl additive growth process sch edule developed for Geometry I......................................... ................................................... .....60 Figure 3.13 3C-SiC film growth rate vs. SiH4 mole fraction..................................... ...62 Figure 3.14 Plan-view SEM image showing protrusion defects...................................62 Figure 3.15 3C-SiC film growth rate vs. process pr essure............................................6 3 Figure 3.16 3C-SiC film growth rate vs. growth tim e.................................................. .65 Figure 3.17 Optical microscope images at a magnifi cation of 20X of (a) 5.8 m, (b) 11.7 m and (c) 23.3 m 3C-SiC epitaxial la yers........................65 Figure 3.18 XRD rocking curve of a 3.3 m thick 3C-SiC epitaxial layer performed at the 3C-SiC (002) diffraction peak..... ...................................66 Figure 3.19 Surface AFM micrographs taken on tappi ng mode of the carbonized surface during the initial stages of gro wth..............................67 Figure 3.20 Optimized 3C-SiC HCl additive process developed for Geometry I.........69 Figure 3.21 Pressure dependence experiments perfor med on Geometry I () compared to the predicted values from the DOE model for Geometry II ().................................................. ......................................74 Figure 3.22 Growth rate dependence on SiH4 mole fraction comparison for the 20, 30 and 38 m/h grown using Geometry I () and Geometry II ().................................................. ................................................... .........75 Figure 3.23 Position of the 5 different measuremen t points considered for film properties evaluation.............................. ................................................... .76 Figure 4.1 3C-SiC growth process schedule for opti mum process predicted by ANOVA analysis..................................... ..................................................8 2 Figure 4.2 Growth rate dependence on HCl mole frac tion.........................................83
viii Figure 4.3 Plan view SEM images for representative films grown with HCl addition at mole fractions (a) 0, (b) 0.25 x 10-4 and (c) 0.75 x 10-4...........84 Figure 4.4 AFM micrographs taken in contact mode f or representative films grown with HCl addition at mole fractions of (a) 0, (b) 0.25 x 10-4 and (c) 0.75 x 10-4................................................... ...................................84 Figure 4.5 Growth rate dependence on temperature.. .................................................85 Figure 4.6 Growth rate dependence in SiH4 mole fraction..................................... ....86 Figure 4.7 XRD rocking curve of a 2 m thick epitaxial layer grown at a rate of 2.5 m/h................................................ .................................................87 Figure 5.1 Calculated species concentration as a f unction of time for hydrocarbon species based on a perfectly mixed reac tor kept at 1385C............................................. ................................................... ........92 Figure 5.2 Calculated species concentration as a f unction of time for silicon containing species based on a perfectly mixed react or kept at 1385C............................................. ................................................... ........92 Figure 5.3 Calculated species concentration as a f unction of time for chlorocarbon species based on a perfectly mixed rea ctor kept at 1385C............................................. ................................................... ........93 Figure 5.4 Calculated species concentration as a f unction of time for chlorosilane species based on a perfectly mixed rea ctor kept at 1385C............................................. ................................................... ........93
ix Modeling and Growth of the 3C-SiC Heteroepitaxial S ystem via Chloride Chemistry Meralys Reyes-Natal ABSTRACT This dissertation study describes the development of novel heteroepitaxial growth of 3C-SiC layers by chemical vapor deposition (CVD) It was hypothesized that chloride addition to the traditional propane-silane-hydrog en precursors system will enhance the deposition growth rate and improve the material qua lity via reduced defect density. Thermodynamic equilibrium calculations were perfor med to obtain a criterion for which chloride specie to select for experimentation This included the chlorocarbons, chlorosilanes, and hydrogen chloride (HCl) chloride containing groups. This study revealed no difference in the most dominant species present in the equilibrium composition mixture between the groups considered. Therefore, HCl was the chloride specie selected to test the hypothesis. CVD computerized fluid dynamic simulations were de veloped to predict the velocity, temperature and concentration profiles al ong the reactor. These simulations were performed using COMSOL Multiphysics and result s are presented. The development of a high-temperature (1300 C -13 90C) 3C-SiC growth process resulted in deposition rates up to ~38 m/h. This is the highest value reported in the literature to date for 3C-SiC heteroepitaxy. XR D FWHM values obtained varied from 220 to 1160 arcsec depending of the process growth rate or film thickness. These values are superior or comparable to those reported in the literature. It was concluded from this study that at high deposition temperatures HCl addi tion to the precursor chemistry had the most significant impact on the epitaxial layer growth rate. Low-temperature (1000-1250C) 3C-SiC growth experi ments evidenced that the highest deposition rate that could be attained was ~2.5 m/h. The best quality layer achieved in this study had a FWHM of 278 arcsec; wh ich is comparable to values
x reported in the literature and to films grown at hi gher deposition temperatures in this study. It was concluded from this work that at lowe r deposition temperatures the HCl addition was more beneficial for the film quality b y enhancing the surface. Surface roughness values for films grown with HCl additive were 10 times lower than for films grown without HCl. Characterization of the epitaxial layers was carri ed out via Nomarski optical microscopy, FTIR, SEM, AFM, XRD and XPS.
1 Chapter 1: Introduction 1.1 Overview Silicon carbide (SiC) is a group IV-IV compound se miconductor that is highly regarded as a suitable material for a myriad of hig h-voltage, high-frequency and hightemperature device applications under which convent ional semiconductors cannot adequately perform. To date, silicon is the semicon ductor material of preference for a majority of electronic devices. However, Si-based t echnology is limited in electronic device performance to temperatures below 250C and to temperatures below 600C in mechanical device performance.1,2 SiC is expected to overcome the limitations impose d by silicon-based (Si) technology, mainly due to its excellent physical, chemical, mechanical and electrical properties. Despite the p romising potential of SiC and the theoretical studies that suggest its advantages, it s technological widespread use has being hindered mainly by challenges associated with the m aterial fabrication. Typical technological barriers that must be overcome includ e: high growth temperatures, low growth rates, high defect density and a lack of hig h quality crystalline substrate material. This dissertation research explores the heteroepit axial growth of 3C-SiC layers by CVD using chloride addition to the SiH4-C3H8-H2 chemistry. The hypothesis being that chloride based chemistry will aid to increase the e pitaxial layers growth rate and material quality via reduced defects thus addressing two of the technical challenges mentioned above. Two deposition temperature range (1000-1250C) and (1300-1390C) were studied during this work. Typically, high deposition temper atures ( 1350C) are required to ensure high quality films and high deposition rates However, the implementation of low deposition temperatures would be beneficial for dev ice process fabrication. Lower process temperatures will eliminate or decrease pro blems due to interdependencies with other process steps during device fabrication proce sses. This will help to avoid problems
2 related to auto-doping, solid state diffusion and a lleviate stresses in the epitaxial film. In addition, lower deposition temperatures are attract ive for selective epitaxial growth (SEG) applications where lower deposition temperatu res are needed to avoid damage to the required silicon dioxide (SiO2) mask.3 In order to meet the intended goals of this work, the use of theoretical CVD calculations coupled with statistical design of exp eriments (DOE) techniques were implemented as major experimental strategies. The C VD theoretical calculations include: (1) a thermodynamic analysis of the product composi tion under equilibrium conditions to provide an insight in to the role of the chloride s pecie on deposition rate, (2) computerized fluid dynamic (CFD) calculations which provide information regarding the velocity, temperature and species concentration pro files along the CVD reactor. By using this approach, theoretical and empirical models wit h adjustable parameters were derived. Such simulations minimize the large number of expen sive and time-consuming growth experiments that are typically required for the opt imization of reaction chemistry. To provide the reader with a better understanding of the main material under study, the remainder of this chapter describes the basic properties of SiC and the potential applications for this semiconductor material. In ad dition, a brief survey of SiC epitaxial growth methods along with an introduction to common crystal defects will be presented. 1.2 SiC crystallography The basic building block of a SiC crystal consists of a stacking of tetrahedral units composed of four carbon atoms covalently bonded to a silicon atom positioned at the center as shown in Figure 1.1 (alternatively one ca n view this as four silicon atoms bonded to a single carbon atom but the result is eq uivalent). SiC crystals are then formed when multiple corners of this basic tetrahedron are joined forming crystal planes. However, disorder in the stacking periodicity of th e planes during crystal formation may occur resulting in defective material formed by num erous dissimilar crystal structures called polytypes. In the case of SiC about 170 poly types have been identified to date.4 Among all polytypes only three possible crystal lat tice structures are known to exist, namely cubic (C), hexagonal (H) and rhombohedral (R ).1,2
3 Figure 1.1 Illustration of a SiC tetrahedron that f orms the basis for all SiC crystals. Four carbon atoms are covalently bonded to a silicon ato m located at the center.5 (Note an equivalent situation involves a single C atom bonde d to four Si atoms). While the variety of SiC polytypes is extensive, o nly a few are commonly used for electronic applications: 3C-SiC, 4H-SiC and 6H-SiC The designation of each polytype follows a widely adopted nomenclature that identifi es both crystalline symmetry (letter) and stacking periodicity (number). Figure 1.2 shows the stacking sequence of these common SiC polytypes. A more comprehensive study of SiC crystallography can be found elsewhere.1,6,7 Figure 1.2 Stacking sequence for the most common Si C polytypes.8
4 This work focuses exclusively on the 3C-SiC polyty pe; commonly known as -SiC, the only purely cubic polytype known to exist. Deno ting each of the three bilayers within the SiC hexagonal frame with the arbitrary letters A, B and C, the stacking sequence for the 3C-SiC polytype is observed to be ABC-ABC as s hown in Figure 1.2. This specific stacking sequence results in a cubic zinc blende cr ystal structure. 3C-SiC potentially offers superior electrical prop erties compared to 4H-SiC and 6H-SiC; which include higher electron mobility and a higher electron saturation drift velocity.4 These properties are of great advantage for the de velopment of high-frequency and high-power switching devices.9-11 In addition, 3C-SiC is isotropic which has inheren t advantages in device operation compared with the hi ghly anisotropic hexagonal polytypes. However, no suitable homoepitaxial techn ique is commercially available for 3C-SiC, thus making device realization more difficu lt. This is mainly due to challenges encountered during the 3C-SiC growth process (refer to section 1.4 for a more detailed discussion on 3C-SiC growth and process challenges) Therefore, the growth of 3C-SiC on Si substrates is of great importance in order to obtain high-quality material that can potentially be used as a substrate in the developme nt of bulk 3C-SiC crystals as well as hetero-structure device fabrication. 1.3 SiC properties and applications Although all the SiC polytypes have the same atomi c composition, namely bilayers of Si and C, each have a characteristic set of electrical properties due to the differences in the stacking sequence of the crystal planes. A comparison of some basic properties of the most common SiC polytypes compare d with Si is presented in Table 1.1. As can be seen, many properties of SiC are superior to those of Si except for the mobility parameter. The properties mentioned above; among others, are justification to choose SiC as a semiconductor material, but the importance of eac h property will depend on the intended application i.e., high-temperature, high -power, or high-frequency. Many of these applications are possible for the most part b ecause SiC possesses a wide bandgap, a
5 property in semiconductors that dictates the energy needed to break covalent bonds in the material and thus generate electrons in the conduct ion band.12 Table 1.1 Comparison of SiC and Si basic parameters at 300 K.1,2,4,8 Property 3C-SiC 4H-SiC 6H-SiC Si Melting point (C) 2827* 2827* 2827* 1415 Physical stability Excellent Excellent Excellent G ood Thermal conductivity (W/cm-C) 3.6 3.7 4.9 1.5 Thermal expansion coefficient (10-6/C) 3.8 N/A 4.3 ^ c-axis 4.7 c-axis 1.0 Energy gap (eV) 2.4 3.2 3.0 1.1 Electron mobility (cm2/V-s) 800 900 400 1400 Hole mobility (cm2/V-s) 320 120 90 471 Critical breakdown electric field (MV/cm) 2.1 2.2 2.5 0.25 Saturated electron velocity (107 m/s) 2.5 2.0 2.0 1.0 *Sublimation temperature. The wide bandgap of SiC makes it possible for high -temperature device operation. High-temperature operation is mainly att ributed to the thermal ionization of electrons from the valence band to the conduction b and.13 At elevated temperatures the concentration of electron-hole pairs can be higher than the free carrier concentration from intentional impurity doping.7,14 When this occurs, the material becomes intrinsic r esulting in device failure because voltages can not be block ed due to the lack of a p-n junction.7 It is believed that replacing Si technology with SiC w ill help to increase device operating temperatures and thus decrease the size of power-el ectronics modules. A powerful argument implies that SiC technology allows for a 5 0% increase in power and a 90 % decrease in weight and volume in power modules made of SiC vs. Si.15 A simple example could be applied to hybrid electric vehicles (HEVs) where smaller, lighter and simpler electrical systems for HEVs would result in reduced vehicle weight and operational costs.
6 This may make HEVs more attractive and affordable s o that greener and more efficient energy utilization can be realized.16 The high breakdown voltage and high thermal conduc tivity are perhaps the most significant properties of SiC for high-power, highvoltage and high-frequency devices. The breakdown voltage determines the maximum field that can be applied before the material breaks down.13 Conversely the thermal conductivity is a measure o f the materials ability to conduct and dissipate heat, w hich is of great importance for device reliability.17 In SiC, the breakdown voltage is about an order of magnitude higher and the thermal conductivity is about 2-3 times higher than Si as shown in Table 1.1. The combination of such properties allow lower losses a nd higher power densities with a smaller on-resistance for high-power devices.7,18 For high-frequency devices the high electric field strength implies that devices can be made smaller and therefore faster but still be able to hold a large voltage thus achievin g high power output.7,18 High-frequency devices may include metal oxide semiconductor field effect transistors (MOSFETs) and bipolar transistors, among others. However, power M OSFETs exhibit an advantage over bipolar transistors due to their high switching spe eds, excellent safe operating area, and better output characteristics for device parallelin g.18 Although most of the above discussion focuses on t he electrical properties of SiC, this material is also attractive due to the combina tion of its electrical properties with its physical and mechanical properties. Microelectromec hanical systems (MEMS) are a principal focus area which is being developed to ta ke advantage of most of the properties of SiC. SiC is a very hard material with hardness v alues comparable to those of diamond and topaz.2 In addition, SiC presents a high level of chemical inertness. This will be beneficial, for instance, in NASA space probes and landers that must operate under extreme conditions of high temperatures and pressur es (~ 460C and 92 bar) and chemically harsh atmospheres such as that on Venus which is composed of highly concentrated sulfuric acid.19 However, this advantage is also a drawback for dev ice fabrication. Due to its chemical inertness, no effi cient wet etchant exists for SiC that could be viable for manufacturing purposes. Consequ ently, the research community has adopted fabrication techniques based on bulk Si sub strates for which fabrication
7 processes are fully developed.2 SiC-based MEMS fabrication and device testing have been demonstrated for a variety of sensors for the measurement of temperature, gases, pressure and other parameters. In addition other st ructures such as resonators and atomizers have been achieved.2,20 These structures could potentially be used in the fabrication of military and commercial gas turbines There is no doubt that SiC technology development could open the door to new systems that could impact a myriad of application s ectors such aerospace, military defense, automotive, nuclear power instrumentation, satellites, etc. Therefore, the continued study of SiC growth processes as well as device fabrication techniques and testing is crucial to achieve the much needed scien tific breakthroughs that will launch SiC as the preferred semiconductor material for har sh environment applications. 1.4 3C-SiC hetero-epitaxy 1.4.1 Growth process Chemical vapor deposition (CVD) is the primary dep osition technique for the growth of 3C-SiC epitaxial layers. A detailed theor etical background of CVD is discussed in Chapter 2. 3C-SiC has been hetero-epit axially grown on Si substrates for many years due to the initial lack of commercially available bulk substrates, as mentioned earlier. Unfortunately, the heteroepitaxial growth techniques still fail to yield sufficiently high quality single crystal material for electronic device applications. Two of the main reasons for this are the mismatch in lattice coeffi cient (~20%) and mismatch in the coefficient of thermal expansion (CTE, ~8%) existing between 3C-SiC and Si. The large mismatches are the main cause of highly defective 3 C-SiC/Si interfaces. This problem is typically reduced by introducing a carbonization pr ocess to the growth sequence as described below but, unfortunately, a highly defect ive interface remains.21 The 3C-SiC on Si hetero-epitaxial growth sequence is commonly performed in two stages: (i) carbonization of the silicon substr ate and (ii) growth of the SiC layer. Sometimes an initial etching of the substrate is ca rried out before conducting these two stages.21 This is done primarily to ensure removal of both t he Si native oxide and surface damage caused in polishing processes. Therefore, an etching process provides a cleaner,
8 smoother surface before the growth sequence starts. The substrate surface is commonly etched by conducting a H2 etching process. A more aggressive alternative util izes diluted HCl in a hydrogen carrier gas, while a milder alter native includes a mixture of propane and hydrogen. The latter is sometimes preferred bec ause the likelihood of forming silicon droplets is reduced.22 However in this work an etch step was not employed as the material quality achieved was outstanding as will be discuss ed in later chapters. The heteroepitaxial approach to grow 3C-SiC was no t possible until the carbonization process was introduced by Nishino et al.23 During the carbonization process a buffer layer formed on the Si surface by heating the substrate in the presence of a hydrocarbon diluted in the hydrogen carrier gas b efore growth.24 This stage was performed at temperatures ranging from 700 C to 13 00 C and yielded buffer layers several nanometers thick.25-27 Even though the carbonization process is not fully understood, there has been speculation that this bu ffer layer aids to reduce the effect of the large lattice mismatch at the SiC/Si interface.23,24 After the carbonization process is performed, the substrate surface is ready for the growth stage at which time a Si containing precurso r is added to begin the SiC deposition process. Typically 3C-SiC is grown at temperatures ranging from 1250 C to 1390 C (as a point of reference the Si substrate melting tempe rature is (1410 C). Studies performed on hot-wall CVD systems usually yield growth rates up to 13 m m/h.26,28-30 1.4.2 CVD growth precursors The typical chemistry used for the epitaxial growt h of 3C-SiC consists of hydrogen (H2), propane (C3H8) and silane (SiH4). In this gas mixture, hydrogen serves as a carrier gas. Other carrier gases used may include argon (Ar) and nitrogen (N2). However, H2 is typically preferred, mostly because of cost, hi gh thermal conductivity, low viscosity, and low density which aid to ensure the laminar flow conditions necessary during the growth process. Also, it is believed tha t H2 aids in the reaction process functioning as a light surface etchant during growt h thus allowing for smoother and cleaner surfaces.22
9 The SiH4 and C3H8 components of the gas chemistry provide the silico n and carbon growth precursors needed to form the desired film. Although these are the most popular compounds, alternative sources have been in vestigated by numerous research groups mainly due to the highly flammable and toxic nature of SiH4.31-35 To solve the above challenges numerous organic com pounds of silicon have been studied, which are often referred to as single prec ursors. These single precursors are of interest since they are safer to handle than SiH4. It has also been suggested that since the original molecule already contains Si-C bonds, the Si-C bond formation at the substrate surface is more efficient.31 Additionally, single precursors are known to offer better stoichiometric control of the gas mixture since the y contain both silicon and carbon atoms.32 Nakasawa et al. reported the formation of an interfacial buffer la yer for 3CSiC/Si(100) heteroepitaxy using monomethylsilane (M MS, or CH3SiH3) at temperatures as low as 450C-650C. It was also reported in this study that lower-temperature single crystal 3C-SiC deposition was possible at 900C yie lding films without the formation of voids at the interface.33 Ferro et al. reported 3C-SiC deposition rates of up to 7 m/h at process temperatures of 1350C by using a hexamethy lthysilane-propane (HMDSpropane, or the Si2(CH3)6-C3H8) precursor system. However, they could not achieve lower deposition temperatures when using HMDS due t o its stable nature.31 Other single precursors used in 3C-SiC growth studies include: t etramethylsilane (TMS, or Si(CH3)4)34, and methyltrichlorosilane (MTCS, or CH3SiCl3).35 However all of these suffer from one drawback the dual precursor syste m allows for accurate doping control via the gas inlet manipulation of the Si/C ratio us ing the site-competition effect pioneered by Larkin et al .36 Thus a single-precursor, while attractive especial ly for MEMS applications, is less attractive for electronic dev ices where precise doping control is critical. In addition to single source alternatives, the stu dy of chlorosilanes has also been well explored. Chlorosilanes, specifically dichloro silane (SiH2Cl2), has been employed in silicon homoepitaxy. From these studies it was obse rved that SiH2Cl2 provided a higher sticking coefficient on silicon surfaces, produced higher purity layers at lower deposition temperatures and provided a safer, less toxic envir onment than silane.37 Ban et al.
10 conducted a thermodynamic analysis for silicon depo sition using SiH2Cl2.38 This study revealed that the main reactions occurring with dic hlorosilane lead to the formation of other chlorosilanes and intermediate species such a s SixCly. These intermediate species aid to increase the silicon atomic content in the g as mixture, making it more available to react thus raising deposition rate values.38 The use of SiH2Cl2 has been applied to 3C-SiC heteroepitaxy yielding similar results.25,28 Wang et al. produced 3C-SiC films on Si(100) substrates by using the SiH2Cl2-C2H2-H2 precursor system at 750 C and 800 C, however the resulting films were amorphous and microcrystal line, respectively. However, Yagi et al. were able to obtain single crystalline layers usin g the same precursor chemistry at 1020 C by applying a layer-by-layer type of growth .25 A similar effect can be obtained when hydrogen chl oride (HCl) is added to the standard H2-C3H8-SiH4 precursor system.39 It has also been suggested that the addition of HCl allows for the enhancement of both deposition r ates and surface morphology. As in the case of SiH2Cl2, chloride ions (Cl-) preferentially attaches to the silicon species resulting in increased silane mole fractions in the gas mixture. This phenomenon is believed to reduce homogeneous nucleation of silico n in the gas phase which creates particulate precipitates. These particles limit the film growth rate by reducing the available silicon in the reaction chemistry.40,41 In addition, it has been suggested that the presence of Clin the gas chemistry improves the surface morpholo gy of the epitaxial layers by etching high energy surface atoms during deposition due to the formation of HCl.42 Gao et al. were able to achieve 3C-SiC growth on Si(100) surf aces using H2SiH4-C2H4-HCl precursor chemistry at temperatures as low as 1000 C.43 They reported that HCl improved the film structure and quality; f or instance, the dislocation density decreased from 1.1 x 1010 cm-2 to 4.27 x 109 cm-2 when the Cl/Si ratio was increased from 0 to 50, respectively. A study performed as pa rt of this research where HCl was added to the standard H2-SiH4-C3H8 gas chemistry allowed for a growth rate increase o f 3 times the highest 3C-SiC deposition rate value repo rted in literature for hot-wall epitaxy (~13 m/h).44 X-ray diffraction (XRD) measurements of the full w idth at half maximum (FWHM) of these films are within 220-360 arcseconds ; this is as good or better than values reported elsewhere (refer to Chapter 3).45
11 1.4.3 3C-SiC epitaxial film defects Defects are undesireable in semiconductor films be cause they disrupt the crystal lattice periodicity which alters the band structure and scatters electrical carriers, and provide paths for electrical leakage and impurity d iffusion.46 The large lattice coefficient and CTE mismatches between SiC and Si are the main cause of defect generation and propagation in 3C-SiC films. Typically defects of z ero dimensions (0D) through three dimensions (3D) are encountered in 3C-SiC epitaxial films. Some of these include interfacial voids, threading dislocations, stacking faults and precipitates. Since it is beyond the scope of this work to cover every defect observed, a brief discussion of some of the major defects is provided below. Dislocations are one-dimensional (1D) defects whic h represent linear imperfections in the atomic array.47 In general the introduction of dislocations into epitaxial layers may cause elastic distortions and band bending. In addition, dangling bonds are created along the core of the dislocation Different types of dislocations include: misfit, threading edge and screw dislocati ons. Misfit dislocations occur when a missing or dangling bond is present between the sub strate and the underlying layer as seen in Figure 1.3 .48 A consequence of this defect is the formation of t wo threading dislocations at the end of each misfit dislocation. In addition, the presence of misfits induces stress into the layer as it grows.49 The layer becomes unconstrained causing the insertion or removal of extra partial lattice plane s that terminate in the dislocation line or misfit dislocation.49,50 This is typically known as an edge dislocation as illustrated in Figure 1.4. External forces cause internal stress i n the crystal which results in the movement of a plane. After the dislocation disappea rs, the crystal is completely stress free and plastically deformed leaving behind an ele mentary step.50 The density of misfit dislocations is dictated by the materials under consideration, therefore it can not be altered by the growth condi tions. However the density of threading dislocations may be altered by using a buffer layer .49 In the case of 3C-SiC this is achieved through the before mentioned carbonization process.
12 Figure 1.3 Schematic representation of a misfit dis location. This defect is caused by the presence of a dangling bond between the substrate a nd the underlying layer due to a lattice coefficient mismatch.48 (a) (b) (c) Figure 1.4 Schematic representation of the formatio n of an edge dislocation. External forces cause internal stress in the crystal which results in the movement of a plane.50 (a) A biaxial force is present on the top of the crystal which ca uses a broken bond which (b) then continues to propagate through the crystal until ev entually (c) an edge is produced resulting in the lowest system energy. Stacking faults, twin boundaries, grain boundaries and anti-phase domain boundaries are typical examples of two-dimensional defects. High densities of stacking faults and twins have been typically observed in 3C -SiC layers regardless of the growth conditions used.51-53 Stacking faults and twins are known to form due to facetted growth and misfit stress-induced deformation.49 Stacking faults are misalignments of the crystal planes and they can be classified as intrinsic or e xtrinsic depending on how they are formed. An intrinsic stacking fault is produced by vacancy agglomeration (Figure 1.5
13 (a)), while the extrinsic stacking fault is formed by interstitial agglomeration (Figure 1.5 (b)). Figure 1.5 Illustration of stacking faults which ar e defects caused by the misalignments of the crystal planes: (a) intrinsic and (b) extrinsic stacking faults.50 Twin boundaries occur when two crystals of the sam e type inter-grow in such a way that a slight misorientation exists between the m. Both crystals are often the mirror image of each other and atoms are shared among them as observed in Figure 1.6.54 Twin formation is detrimental for growing layers since i t may lead to misoriented or polycrystalline phases.51 Yun et al were able to successfully suppress twin formation by performing a two-step epitaxial process consisting of a nucleation step followed by the growth of 3C-SiC. This nucleation stage appeared to be more efficient than the normal carbonization process.51 Figure 1.6 Representation of a twin boundary defect Two slightly misoriented crystals of the same type inter-grow and share atoms.50 Note that there are no dangling bonds associated with this form of defect but they still impact carrier transport in the films and are therefore detrimental to device performance.
14 1.5 Summary SiC is a robust material with many properties supe rior to Si. In order to develop SiC growth and fabrication techniques, it is crucia l to achieve the much needed scientific breakthroughs that will launch SiC as the next gene ration semiconductor system. However, despite the knowledge of the potential of SiC and the theoretical studies that suggest its numerous advantages, its technological widespread use has been hindered mainly by challenges associated with material fabri cation. Typical technological barriers that must be overcome include: high growth temperat ures, low growth rates, high defect density and resulting lack of high crystalline qual ity material. Therefore, this work was undertaken to investigate the growth of SiC, specifically 3C-SiC, with the aim to overcome some of these tech nological barriers. A chloride-based CVD precursor chemistry was applied in the heteroep itaxial growth of 3C-SiC on Si (001) surfaces via horizontal hot-wall CVD. To assi st and guide the reader, Chapter 2 discusses CVD theory and theoretical simulations pe rformed. In Chapter 3 the development of a high-temperature 3C-SiC process vi a HCl as a growth additive is presented. The same precursor chemistry was then us ed to demonstrate low temperature growth of 3C-SiC layers which is presented in Chapt er 4. Finally, Chapter 5 provides a summary of the research performed as part of this w ork followed by experimental trends and suggestions for future work.
15 Chapter 2: Chemical Vapor Deposition 2.1 Overview Among the main objective of this study was to dete rmine the effect of chloride addition to the SiH4-C3H8-H2 precursor chemistry system via thermodynamic equil ibrium calculations and to develop CVD process simulations in order to predict the velocity, temperature and species concentration profiles alon g the reactor; the ultimate goal being prediction of film deposition rates. The aim of thi s type of calculation is to obtain a deeper understanding of the SiC deposition process and to facilitate process optimization resulting in improved film material. In addition, t he simulation work will enable modeling of changes in the reactor hardware and eff ects of changing process parameters without direct experimentation. In order to obtain such models, in depth understan ding of process thermodynamics, chemical kinetics and transport phe nomena are needed. The following sections are intended to introduce the theoretical aspects of these disciplines applicable to CVD, developing the different models and to present the model results. 2.2 Chemical vapor deposition CVD is a deposition technique in which gases decom pose and chemically react near or on a surface with the aim to synthesize a s olid product.42 CVD can be performed in a closed or open system reactor. However, nowada ys most deposition processes are carried out in open reactors where the effluent spe cies are removed from the chamber after the reaction takes place. Various open system reactor designs have been used for the development of CVD process including horizontal vertical, semi-pancake, barrel, and multiple wafer.42,55,56 Among the most popular geometries are the vertical and horizontal tube reactors.42 Both of these geometries can be further grouped in to cold-wall and hotwall designs. In the cold-wall design the sample is kept at the required temperature while
16 all other surfaces bound to the reacting gas flow p athway are at a greatly reduced temperature. This is achieved by surrounding the re actor tube with a cooling jacket which, in theory, causes the reaction to occur only on the hot sample keeping all the remaining surfaces as free as possible of deposits due to a slower reaction rate. However, severe natural convection may occur due to the stee p temperature gradient around the substrate. Therefore, one drawback in the cold-wall reactor is the difficulty in maintaining a uniform temperature over the sample a rea. Such concerns can be eliminated or significantly reduced if the entire c hamber is heated at a uniform temperature. This type of heating is achieved if a hot-wall design is implemented as is the case for this work. Numerous parameters control and affect the CVD pro cess and, hence, the properties of the deposited film. These parameters can be classified into reactor design variables and operator variables. Reactor design va riables can include the susceptor tilt angle, gas inlet geometry, wafer/carrier configurat ion, wafer/reactor wall configuration, exhaust configuration, among others. However, opera tor variables are the primary control factors for any reactor geometry; these include gas flow rate, gas composition, temperature, reactant chemistry and temperature pro file. For instance, the temperature is crucial as it controls the thermodynamics and the k inetics of the process. Optimal temperature must be achieved and maintained in orde r for gas and surface reactions to overcome activation barriers. Any variance in tempe rature may lead to inferior material morphology or quality due to variation in the react ions or kinetics.56 2.2.1 USF hot-wall CVD system The USF CVD reactor was designed to be horizontal with hot-walls as illustrated in Figure 2.1. This reactor was custom built and modif ied by members of the USF SiC research group.57,58 The reactor chamber consists of a main quartz tube supported in place by two water cooled stainless steel end plates. The main growth components used during the deposition process are then loaded into this qu artz tube. The gas line supply connects
17 to the stainless steel plate located at the reactor inlet. The gases then exit the other side of the plate through a gas diffuser. The diffusers fu nction is to provide uniform flow in which laminar flow conditions along the gas path pr evail. The gas path is composed of an inner quartz liner that connects the diffuser with the reactor deposition area (i.e., hot zone) via graphite adaptors. The adaptors provide t he necessary connections between the quartz liner and the hot zone; in addition they hel p to avoid overheating of the quartz liner. The hot zone is composed of an angled ceilin g designed SiC coated graphite susceptor in which the sample is loaded using a pol ycrystalline plate. The hot zone is surrounded by a graphite foam insulating material t o help maintain a fairly uniform temperature during the deposition process by minimi zing heat losses to due radiation. Figure 2.2 illustrates a cross-section view of the USF CVD reactors inlet quartz liner and hot-zone areas for visualization purposes. Figure 2.1 USF horizontal hot-wall CVD reactor. The necessary heating to achieve the desired depos ition temperatures is provided by an RF generator which produces radial heating vi a RF induction copper coils wrapped around the hot zone. An infrared pyrometer, which i s connected to a computer interface, is used to monitor and control the temperature. The same computer interface is used to control mass flow controllers (MFCs) that regulate the gas flows introduced into the
18 reactor. Finally after deposition the effluent is t ransported out the chamber via negative pressure at the exhaust line via a vacuum pump whic h also is used to control the deposition pressure. Figure 2.2 Cross-section view sketch of the USF CVD reactor. Sketch provided by I. Hasselbarth, University of South Florida. At the time this work was conducted the reactor us ed in this research, named MF1, supported processes such as the epitaxial grow th of 3C-SiC and 4H-SiC, H2 etching, implant annealing and epi doping with nitr ogen (N2) gas. A second reactor named MF2 solely dedicated for 3C-SiC processing wa s also available. The standard dual precursor chemistry (C3H8-SiH4) with H2 as the primary carrier gas was available; in addition Ar was also accessible as a secondary carr ier gas option or cooling gas. The system also supported halocarbon chemistry includin g HCl and CH3Cl. The reactor is capable of process temperatures up to 1800C and pr essures from 75 Torr to 760 Torr. Two different reactor geometries were used to cond uct the deposition experiments in this work using reactor MF1; for simplicity they will be referred as Geometry I and Geometry II. Geometry I included the use of three a daptors, a 140 mm long graphite susceptor and a hexagonal shape insulating foam. A 3C-SiC growth process on 8 mm x 10 mm Si die samples was developed using Geometry I (refer to Chapter 3). However, initial experimentation indicated that large temper ature gradients were being formed at the susceptor causing film quality degradation and, on occasions, substrate melting when the process was transferred for growth on 50 mm sub strates.59 In order to solve this problem a study was performed by S. Harvey and Dr. Y. Shishkin of our group in which it was concluded that a reactor geometry change was necessary to minimize/solve the temperature gradient problem in the reactor.59 As a result Geometry II was applied.
19 Geometry II consisted of a round shape insulating f oam with an elongated graphite susceptor (210 mm). In addition, Geometry II used t wo adaptors instead of three. In both geometries an angled ceiling susceptor design was u sed. Table 2.1 summarizes the properties of the two different reactor geometries used during this work. Table 2.1 Description of the reactors geometries co nsidered in this study. Reactor Component Geometry I Geometry II Insulating foam shape hexagonal round Susceptor design angled ceiling angled ceiling Susceptor length 140 mm 210 mm Number of adaptors 3 2 2.2.2 CVD thermodynamics The main objective of this study was to investigat e the introduction of chloride species into the growth chemistry; the hypothesis b eing to increase the epitaxial layers deposition growth rate and to improve the resulting material quality via reduced defects. In order to determine the chloride specie to be sel ected for the experiments, a theoretical study was performed first to establish a criterion for gas source selection. As such, a thermodynamic equilibrium study of the CVD product mixture composition was performed to monitor the effect of the chloride spe cie on the major effluent species composition. 188.8.131.52 Thermodynamic equilibrium Thermodynamic equilibrium calculations applicable to CVD systems are based on the fact that the total Gibbs free energy (G) of a closed system should decrease during an irreversible process when the system is operating a t constant temperature (T) and pressure (P). Therefore, at equilibrium conditions, the change in Gibbs free energy of reaction (Grxn) attains a minimum or the differential change in G ibbs free energy approaches zero ((dG)T,P 0). As a reference, the Gibbs free energy is define d as the
20 thermodynamic potential which measures the "useful" or process-initiating work obtainable from an isothermal, isobaric thermodynam ic system.60 The CVD reactor can be described as an open system which typically involves rapid changes. But when long reaction times are con sidered the expressions developed for closed systems can also be applied to open syst ems given the assumption that once equilibrium is reached, no further changes occur an d the system still continues to operate at the same constant T and P. Two different approac hes are often used to perform the Gibbs free energy minimization analysis: (1) the no n linear method and (2) the more generalized method based on the fact that at equili brium the total Gibbs free energy has a minimum value. The non linear method utilizes equilibrium constan ts to obtain the partial pressure of the species; this method is useful for less comp lex systems where there are a small number of known significant species and when inform ation of the reaction pathways, as well as what phases are formed, is known. However, the generalized method is independent of the reaction pathways and is more ap plicable to computer routine solution techniques. Hence the latter method is generally mo re suitable especially for complex chemical systems like CVD. Therefore calculations i n this work will be performed by using the generalized method. 184.108.40.206 Gibbs free energy minimization The minimization routine assumes that the Gibbs fr ee energy for a single phase system is defined as in Equation 2.1: (G)T,P = g(n1,n2, ,nN) Equation 2.1 where T is temperature, P is pressure and n represe nts the number of moles. The solution approach involves finding the set ni at constant T and P such that (dG)T,P 0. The minimization procedure is subject to the conservati on of mass. That is, the number of atoms of a specific element must be conserved. For example, a gas mixture containing one mole of SiH4 and one mole of H2 will contain one mole of Si atoms and six moles of
21 H atoms. The total number of atomic masses of the kth element in the system can be defined as Ak. Then the number of atoms of the kth element present in each molecule of chemical specie i is aik. As a result, the material balance can be expresse d as: ) ,..., 2,1 ( 0 w k A a ni k ik i = = Equation 2.2 where w is the total number of elements comprising the system. Upon applying the method of Lagranges undetermined multipliers to th e materials balance constraint, a new function (F) is formed by adding to G the materials balance sum over the kth element. Then, + =ki k ik i kA a n G Fl Equation 2.3 This equation is identical to Equation 2.2 since t he summation term is equal to zero. However, the partial derivatives of G and F a re different due to the second term in the right hand side of Equation 2.3 The minimum of the functions G and F occurs when (F/ni)T,P,nj 0. Then ) ,..., 2,1 ( 0, , ,N i a n G n Fik k nj P T i nj P T i= + = l Equation 2.4 The first term in the right hand side of Equation 2.4 is known to be the chemical potential, which for gas phase reactions and standa rd states as pure ideal gases can be expressed as: + =o o iP f RT G i^lnm Equation 2.5
22 where R is the ideal gas constant and P is the pre ssure for the standard state. o iG can be set equal to zero for all elements in their standar d states and o f o iiG G D = for compounds. Finally, ^f is the fugacity, which when expressed in terms of the fugacity coefficient leads to P y fi i ^ ^f=, then Equation 2.5 becomes: ) ,..., 2,1 ( 0 ln^N i a n P P y RT Gk ik i o i i o fi= = + + Df Equation 2.6 Equation 2.6 represents N equilibrium equations an d Equation 2.2 represents w material balances for a total of N + w equations (unknowns are ni and k). The values of ^ ifcan be estimated depending if the phase can be cons idered ideal or real.60-62 220.127.116.11 Thermodynamic equilibrium simulations results In this study NASA Glenn's Chemical Equilibrium wi th Application (CEA) computer program was used to perform the minimizati on routine.61-63 The program allows the equilibrium composition calculation of c omplex mixtures through the minimization of Gibbs free energy by using the meth od described in section 18.104.22.168. The specific computer program used allows for the simul ation of 90 chemical species if SiH4C3H8-H2 gas system is used and 120 chemical species for th e SiH4-C3H8-H2-HCl gas system. It also allows for the addition of 5 conden sed species in both cases. In order to determine the effect of the addition o f chloride species into the SiH4C3H8-H2 equilibrium mixture composition; simulations were p erformed including the following groups: chlorocarbons, chlorosilanes and hydrogen chloride (HCl). Table 2.2 lists all the species considered within these group s. In the following discussion thermodynamic simulation results will be presented using the process parameters of two 3C-SiC deposition processes, namely with and withou t chloride additive, developed as part of this work.
23 Table 2.3 summarizes the process parameters of thes e two processes which were used as input parameters to obtain the thermodynamic equili brium results. The process parameters for the gaseous species represent the re spective gas mole fraction. Table 2.2 Summary of chloride species considered in the thermodynamic simulations. Species Chlorocarbons Chlorosilanes Other CCl4 SiCl4 HCl CHCl3 SiHCl CH2Cl2 SiH2Cl2 CH3Cl SiH3Cl C2HCl C2Cl2 C2Cl4 C2Cl6 Table 2.3 Process parameters for 3C-SiC deposition process with and without HCl addition. Temperature (C) Pressure (Torr) yH2 yC3H8 ySiH4 yHCl Si/C Si/Cl 1385 100 0.99 2.0 x 10-4 5.3 x 10-4 -0.9 -1385 100 0.99 2.0 x 10-4 5.3 x 10-4 0.97 x 10-4 0.9 6.5 Figure 2.3 (a) illustrates the predicted equilibri um composition as a function of temperature for the SiH4-C3H8-H2 precursor chemistry. The formation of solid beta-Si C over the entire temperature range considered for a total molar composition of one in the solid phase was predicted in the calculations. The gas phase composition in both cases evidenced that the most dominant species were H2 (not shown) and H The presence of H2 is expected since it is the process carrier gas wh ich is present in the inlet mixture in a much higher concentration than the precursors. The presence of atomic hydrogen may be
24 attributed to disassociation of H2 due to reaction with the precursor molecules. It c ould also be the product of the propane and silane decom position reactions. CH4, 3CH, C2H2, and C2H4 are predicted to be the major carbon species. CH4 being the most dominant at temperatures below 1400 C for which deposition process temperatures are typically carried out for 3C-SiC h eteroepitaxy. Since C3H8 is not present as a thermodynamically possible chemical specie in the equilibrium gas phase mixture; it can be suggested that most of the molecular C3H8 will mainly crack to 3CH, CH4, C2H2, and C2H4. However, the assumption that C3H8 will fully crack or that these are the only carbon species will be misleading since other carbo n containing species, as well as the presence of C3H8 itself, might be favorable at molar fractions below 10-10. The most dominant silicon species were -SiH, Si, and SiH2. Note that SiH4 cracked to some extent and then its concentration r emained fairly constant over the entire temperature range. This may indicate that the therm al decomposition of SiH4 may restrict the silicon content in the system thus limiting the growth rate. In addition, the presence of the Si3 and Si2 species can be noted at much lower mole fractions. The presence of these species may indicate the evolution of solid or liqu id clusters of homogeneous nucleation which are a known cause of problems during growth. However, the simulation did not predict that they condensed either as a solid or a liquid. Finally, Si2C and SiC2 are present at low temperatures and their concentration increas es with increasing temperature. These are recognized as a primary species for growth. The study of chloride addition to the SiH4-C3H8-H2 precursor chemistry started with HCl addition since it is the simplest specie, containing only hydrogen and chlorine. The equilibrium composition mixture for the SiH4-C3H8-HCl-H2 precursor chemistry (Figure 2.3(b)) is not much different than that wit hout HCl. The formation of solid silicon carbide as the only condensed specie over the tempe rature range studied was also predicted. In addition, the presence of H2 (not shown) and H as major species in the gas mixture was noted, but in this case HCl and -Cl were also present at high mole fractions. The simulation predicted that the presence of HCl in the gas mixture does not have a dramatic effect on the carbon containing spe cies. As can be seen in Figure 2.3(b), the same carbon containing species as in the proces s without HCl were observed and their
25 (a) (b) Figure 2.3 Predicted product equilibrium mixture co mposition for (a) SiH4-C3H8-H2 and (b). SiH4-C3H8-HCl-H2 precursor systems. Simulation performed using the NASA Glenns CEA code.61-63
26 molar composition did not vary dramatically. Howeve r, it can be seen that chlorine preferentially bonds to silicon. This can be observ ed by the formation of species containing silicon and chlorine (SiHCl, SiCl, and S iCl2) and suggests that HCl provides a different mechanistic reaction path that is going t o affect mainly the Si containing reactions. As a result, the HCl addition is hypothe sized to make Si species more available to react in the deposition mixture, resulting in a growth rate increase. This result is in agreement with other studies reported in literature .40,41 Adding HCl to the SiH4-C3H8-H2 precursor chemistry has been attributed to the suppress of homogeneous nucleation in SiC CVD epita xial growth.64 Therefore a reduction in the Si3 and Si2 species mole fraction is expected. The simulation results indicated that the Si3 and Si2 species are still present in the equilibrium mixtur e composition. However they only started to form at s lightly higher temperatures compared to the no HCl process. This suggests that the intro duction of chloride species suppressed homogeneous nucleation for a larger temperature ran ge compared to that of the SiH4C3H8-H2 precursor chemistry. Finally, no chlorocarbon spec ies were observed via the simulation. Because the minimization procedure is subject to t he constraints of the mass balance; simulations using any of the chlorocarbon, chlorosilane or HCl species at the same process conditions, i.e. same process Si/C and Si/Cl, will lead to the same equilibrium mixture composition. Therefore, the sim ulation results are not shown to avoid redundancy. In summary for the chlorocarbon system the followi ng trends were observed. When higher C/Si ratios were explored the chlorine mole fraction increased and the Si3 and Si2 species mole fractions decreased dramatically. It was noted that chloride is not attached to any of the carbon containing species bu t only bound to Si or hydrogen once the precursor has cracked. It is not until higher C /Si ratios are simulated, i.e. using CH2Cl2 or chloroform, that some amounts of CH3Cl can be seen. A final observation on the higher chlorinated chlorocarbons such as CHCl3 is that at higher C/Si ratios carbon clusters such as C3 begin to be seen in very low concentrations.
27 When the precursor system was studied using the ch lorosilane species as an additive, similar trends were obtained. This is due to the mass balance constraint or conservation of mass that is identical to the previ ous cases. In conclusion, the thermodynamic equilibrium simulations indicated no significant difference in the equilibrium composition mixture between adding HCl or using chlorocarbons or chlorosilane species as the growth precursors. Ther efore, this work will focus on HCl which is the simplest chloride studied. In addition the literature suggests that HCl promotes the enhancement of SiC growth in the 4H-Si C homoepitaxy so this work was focused on the 3C-SiC on Si heteroepitaxy system.41,42,64 2.2.3 CVD kinetics A typical CVD process follows the generalized react ion path involving the steps depicted in Figure 2.4. (1) The reactant gases are transported to the deposition chamber. (2) After the gas enters the chamber in bulk flow, thermal dissociation, or cracking, forms the intermediate reacting gaseous species. (2a) Pow ders can be formed as a result of homogeneous gas phase reactions at temperatures hig her than the decomposition temperature of the intermediate species. These powd ers may precipitate to the substrate surface and function as 3-D nucleation centers. As a result, defects are created in the film leading to lower quality material. This is typicall y the case when the precursor mole fractions reach the saturation point. (2b) Diffusio n/convection of the intermediate species then occurs across the boundary layer at temperatur es below the dissociation temperature of the intermediate species. The species eventually undergo steps 3-6. (3) Absorption of the reacting species occurs on the substrate surfac e so that surface reactions can take place. (4) The reacting species diffuse through the sample surface creating crystallization centers and eventually film growth. (5) The gaseous by-products desorb from the surface and travel across the boundary layer by diffusion/c onvection. (6) the unreacted species and gaseous by-products are transported from the d eposition chamber via bulk flow.42,55,56,65,66
28 Figure 2.4 Schematic of CVD steps. Adapted from K.L Choy.65 As explained above, the kinetics of a CVD process involves chemical reactions in the gas phase, on the substrate surface, chemisorpt ion and desorption. However, a definite mechanism for SiC growth has not yet been determined. Ideally, the chemical kinetics of a CVD process could be derived from the analysis of all possible reaction pathways. As a consequence, a multi-step chemical r eaction pathway is normally implemented when modeling deposition kinetics, as i s the case in this work. This multistep reaction pathway mainly includes the decomposi tion of the different precursors into numerous elementary reactions leading to a combined homogenous and heterogeneous reaction model. The reactions are described by thei r formula and rate equations. 22.214.171.124 Gas phase model In the present study, C3H8, SiH4 and HCl, with H2 as a carrier gas have been used as the growth chemistry for all experiments and sim ulations. Therefore the complete reaction model will consist only on the decompositi on of C3H8, SiH4, and HCl and the set of reactions between the products of all decomposit ion reactions. Since incorporating all possible species resulting from this growth chemist ry is outside the scope of this work; a selection of species was made where molecules with four carbon or silicon atoms were
29 excluded from the model. In addition, all reactions considered contribute to the model; therefore all considered species are either present or being produced during the reactions. Appendix A lists the complete gas phase reactions considered in this work. Decomposition of propane was taken from the work of Petrov et al. and Danielsson et al.67,68 The decomposition of silane and the formation of o rganosilicons was taken from the work of Danielsson which mainly combined the mo dels from Coltrin et. al and Ho as well as other relevant publications.67,69-71. Finally, the chlorinated species reactions were taken from the work of several authors.72-79 The reaction rates for the gas phase reactions are described by the law of mass action (Equation 2.7) which states that the rate of any given chemical reaction is proportional to the product of the reactant concent rations.80 ' -=prod i i r j react i i f j jij ijc k c k ru u Equation 2.7 Wheref jk and r jk denote the forward and reverse rate constants resp ectively. The concentration of species i is denoted ci and ij is the stoichiometric coefficients which are defined as negative for reactants and positive for products. The reaction rate constant is strongly dependent on temperature and will be model ed by means of the Arrhenius expression as given by Equation 2.8: = RT E AT kA nexp Equation 2.8 126.96.36.199 Surface reaction model Surface reactions are characterized by reaction me chanisms such as chemiadsorption, dissociation, diffusion and desorp tion. A definite description of the surface process is not available, however studies o f possible mechanisms have been reported.70,73,74,81,82 In this work, the surface reaction model was taken from Veneroni et al. which provided a well documented complete set of reactions including those with
30 -Cl containing species which are applicable to this wo rk.82 Appendix B lists the surface model considered. For the surface reactions, the rate expression wil l be described by the Langmuir rate law which states that chemisorption will occur when a gas phase molecule reacts with an empty active site at the surface. Consequen tly, the adsorption/desorption rate will be proportional to the number of empty active sites at the surface as given by Equation 2.9.80 s s sC k rq= Equation 2.9 where Cs is total number of active surface sites available, ks is the temperature dependent rate constant for the surface reaction and is the surface coverage defined as the number of adsorbed molecules on a surface divided by the n umber of molecules in a filled monolayer on that surface.80 2.2.4 CVD transport and fluid dynamics In addition to reaction mechanisms, a thorough unde rstanding of mass and heat transport is critical in the design and modeling of CVD processes. The transport of mass determines the species concentration at the substra te surface; the transport of heat determines both the gas and substrate temperatures. Both transport mechanisms are obviously critical for film deposition rate, compos ition and uniformity. During a CVD process it is desired to deliver the gas uniformly to the substrate in order to obtain uniform films and avoid intermixing of gas concentrations56 Therefore it is crucial to determine in what flow regime the sys tem is operating; laminar or turbulent. This type of flow is also known as streamline, whic h occurs when a fluid flows in parallel layers with no disruption between the layers. 83 The dimensionless Reynolds (Re) number is used to determine whether flow conditions lead to laminar or turbulent flow regime s. This dimensionless number is described as the ratio of inertial forces to viscou s forces.83 In the case of a circular pipe the Re number can be calculated by using Equation 2 .10.
31 m r uD = Re Equation 2.10 where n is the mass density, u the velocity, D the charact eristic diameter and the viscosity. For a flow to be considered laminar a Re number of less than 2100 must be obtained.83 In the case of laminar flow in circular pipe, in th eory the velocity of the gas changes from zero at the walls to that of the bulk gas. This region of velocity change is called the boundary layer. Figure 2.5 shows a ske tch of the boundary layer. As can be seen, the fluid approaches the substrate surface at a uniform velocity and once in contact with the substrate, a velocity gradient is formed. Figure 2.5 Boundary layer development near a flat s urface. A velocity gradient is formed once the fluid contacts the surface. Adapted from A Sherman.56 At high gas velocities the thickness of the bounda ry layer can be estimated using Equation 2.11, where D is the tube diameter for the case of a circular pipe. According to this equation the thickness of the boundary layer i ncreases with reduced gas velocities and increased distance from the tube inlet. Re D =d Equation 2.11 188.8.131.52 Rate limiting steps In the generalized reaction path illustrated in Fig ure 2.4 the steps can be classified into two categories, namely mass transport and surf ace reaction steps. The slowest of
32 these steps determines if the process is mass trans port or surface reaction limited as the surface reaction is considered the rate limiting st ep in the reaction mechanism.66 The surface reaction limited regime is dominated by the surface temperature rather than by what is occurring in the bulk gas. A t low pressures and low surface temperatures, a large flux of reactants to the surf ace exists. Because of the low pressure and the small thickness of the boundary layer, the diffusion coefficients are large (diffusivity inversely proportional to pressure). T herefore the reactants reach the surface rapidly. The reactants react slowly due to the lowe r temperature so that there is an oversupply of reactants waiting to be consumed as s een in Figure 2.6(a). In the mass transfer limited regime, the controllin g factors are the reactants diffusion rate through the boundary layer and the b y-product diffusion out through this layer. This situation typically occurs at high pres sures (smaller diffusion rate) and high temperatures. Figure 2.6(b) provides a sketch of th is situation. As a result of the high temperature, any molecule arriving at the surface w ill react quickly. In addition, the gas velocity is lowered and the boundary layer thickens making the diffusion of reactants more difficult. (a) (b) Figure 2.6 Representation of the rate limiting step s in a CVD reaction (a) surface reactionlimited and (b) mass transport limited.66
33 2.2.5 Computerized Fluid Dynamic (CFD) simulations COMSOL Multiphysics and COMSOL Reaction Engineerin g Lab computer software were used to develop the CFD simulations i n this work.84 COMSOL Multiphysics is designed to couple transport phenom ena, CFD or mass and energy transport to chemical reaction kinetics. While COMS OL Reaction Engineering Lab, solves the reaction kinetics material and energy ba lances. In addition, provides ready made expressions to calculate thermodynamic and tra nsport properties. 184.108.40.206 Modeling domain The simulation domain takes into account the gas p ath along the reactor. Since the depth of the flow domain is large, a 2D approximati on is valid. This approximation takes advantage of the systems symmetry and assumes that variation with temperature and flow along the depth of the domain are small or neg ligible. Figure 2.7 illustrates the simulation domain. Figure 2.7 Reactor 2D modeling domain.
34 220.127.116.11 Governing equations The Navier-Stokes equations for non-isothermal flow and the energy and mass balance equations describe the basis of the model. Equation 2.12 and Equation 2.13 represent the momentum balance and the continuity e quations, which provide information regarding flow velocity. Equation 2.14 is the energy balance which provides information of the gas temperature distribution acr oss the CVD reactor. Finally, Equation 2.15 represents the mass balance equation which wil l provide information of the species distribution along the reactor hot-zone. () ( ) ()F p u u u u t uT= + + + r h r Equation 2.12 ()0 = + u t r r Equation 2.13 ( ) Q Tu C kT t T Cp p= + + r r Equation 2.14 ()i i i i iR u c c D t c = + + Equation 2.15 In these equations; n denotes the gas density, u the gas velocity, p pre ssure, the viscosity, the thermal conductivity, T temperature, Cp is the heat capacity and F is a volume force field. 18.104.22.168 Boundary conditions The boundary conditions for the momentum balance i mpose a velocity at the reactor inlet. This velocity value refers to the ma ximum velocity which can be calculated from Equation 2.16 and Equation 2.17. uave is the average velocity, Q volumetric flow rate and Ac the cross section area. The volumetric flow rate w as set to be the same as the carrier gas flow rate and the cross sectional area was that corresponding to the inlet.
35 ( ) aveu u 2max= Equation 2.16 c aveA Q u = Equation 2.17 The flow in our CVD reactor is mostly dictated by the H2 carrier gas which will dominate the flow calculations performed. The input flow considered was 30,000 sccm which is the carrier gas flow used during the growt h step for the 3C-SiC deposition process with and without the HCl additive (refer to Chapter 3). The properties of H2 were obtained from the COMSOL Muliphysics software built in materials properties library.85 The no-slip condition was applied at the reactor w alls; this condition sets all components of the velocity vector to zero. The boun dary condition pressure with no viscous stress was imposed at the outlet. The press ure value for this calculation was 13.3 kPa (100 Torr) which is the pressure physically mea sured downstream in the reactor. Finally, the substrate will be treated as an interi or boundary. These conditions are summarized by Equation 2.18 to Equation 2.20. inlet u uo= Equation 2.18 outlet p po= Equation 2.19 () ()outlet n I u u uT n r= + 0 3 2h h Equation 2.20 The boundary conditions for the energy balance requ ired the inlet temperature, and the process temperature in the susceptor area as seen i n Equation 2.21 and Equation 2.22. Equation 2.23 describes the convective flux boundar y condition implemented at the reactors outlet. Finally, a temperature gradient w as considered at the substrate as summarized in Equation 2.24. inlet T To= Equation 2.21 walls T To= Equation 2.22
36 ( ) 0 = T k n Equation 2.23 substrate T Tsuf= Equation 2.24 The boundary conditions required for the mass bala nce are the species concentrations and the inlet, substrate and outlet of the system as summarized by Equation 2.25 to Equation 2.28. Convective flux was implemented at the outlet and insulation at the walls. For the adsorption and rea ction at the surface the flux discontinuity boundary condition was imposed. There fore; inlet c ci o i,= Equation 2.25 ( ) outlet c D ni i0= Equation 2.26 ( ) substrate N N N ni= -2 1 Equation 2.27 substrate u c c D Ni i i j+ = Equation 2.28 22.214.171.124 CFD simulations results In order to perform the CFD calculations, COMSOL M ultiphysics and the Reaction Engineering Lab module have to be used ite ractively.84 Appendix C explains the procedure followed to perform the CFD simulatio ns.84 The gas temperature distribution across the CVD re actor for both susceptor geometries is depicted in Figure 2.8. It can be see n that in both cases the higher temperature region is located at the susceptor area where the deposition process takes place. Figure 2.9 shows the gas temperature profile along the susceptor area and the temperature at the susceptor area remained constant A constant temperature is desired to avoid variations in the reaction kinetics that may lead to inferior material morphology or quality. Although large temperature gradients were observed experimentally via Si melt
37 (a) (b) Figure 2.8 Gas temperature profile across the CVD r eactor configured for (a) Geometry I and (b) Geometry II. H2 flow = 30,000 sccm, vave=0.0173m/s and P=13.3 kPa (100 Torr).
38 (a) (b) Figure 2.9 Gas temperature variation along the CVD reactor for (a) Geometry I (b) Geometry II. Note that the gas temperature remained constant at the 50 mm wafer area.
39 tests on the polyplate, it can be observed that for both geometries the simulation predicted that the gas surrounding the sample, if placed in t he 50 mm diameter wafer area, will be at the constant temperature of 1659K (1386C). This temperature is only one unit higher than the experimentally measured temperature of 165 8K (1385C) for the no HCl and HCl growth processes process (see Chapter 3). The simulation is not in accordance with the obtai ned experimental results that a temperature gradient was observed to be present alo ng the 50 mm wafer area. Therefore, a more detailed simulation in which radio frequency induction, material properties, and heat losses by radiation are taken in to considerat ion may offer a better prediction of the temperature profile in the system (see Chapter 5). Similarly, the gas velocity profile was computed fo r both reactor geometries as shown in Figure 2.10. The velocity increased at the adaptor and susceptor area due to the compression of the gas as a result of the narrowing susceptor since the susceptor ceiling is angled. Because not all species travel at the sa me velocity, the shape of the velocity profile at a given cross section depends on the typ e of flow under consideration. It can be seen that the velocity at the susceptor area remain ed constant indicating that the precursor species should diffuse to the substrate surface at similar rates. For Geometry I the maximum velocity along the susceptor was predicted to increase from 0.29 m/s to 0.39 m/s (Figure 2.10(a)). For Geometry II the maximum v elocity increased from approximately 0.14 m/s to 0.32 m/s (Figure 2.10 (b) ). Note that the velocities between Geometry I and Geometry II were significantly diffe rent at the inlet; this is as a result of the elongation of the susceptor which changes the a ngle of inclination at the top portion of the susceptor and thus the cross-sectional area throught its length. At these particular velocities the Re number is les s than 2100, therefore the flow can be considered to be laminar. As a consequence t he velocity profile along the reactors cross-section should be parabolic in shape as shown in Figure 2.11. Finally, since the flow at the susceptor was deter mined to be laminar it is assumed that the fluid will flow in parallel lines. This wa s verified by constructing streamline plots of the velocity profile as shown in Figure 2.12a-b. It could be observed that this was indeed the case along the entire flow path for both configurations. However minor
40 disruptions to the pattern are present for Geometry I possibly due to the higher velocity compared to Geometry II. These disruptions may caus e some back stream issues possibly leading to the introduction of particulates into th e epitaxial layer. (a) (b) Figure 2.10 Gas velocity profile across the CVD rea ctor configured for (a) Geometry I and (b) Geometry II. H2 flow of 30,000 sccm., vave=0.0173m/s and P=13.3 kPa (100 Torr).
41 (a) (b) Figure 2.11 Parabolic velocity fields for (a) Geome try I and (b) Geometry II at susceptors inlet, center and outlet.
42 (a) (b) Figure 2.12 Streamline plot of the velocity profile illustrating fluid is flowing in parallel lines, thus confirming the flow is of laminar natur e (a) Geometry I and (b) Geometry II. Unfortunately, a final numerical solution for the species concentration profile along the reactor has not been established in this work due to problems with the COMSOL platform (see Chapter 5).
43 2.3 Summary As a result of the work performed in this study th e following simulations are now possible: (1) a thermodynamic analysis of the produ ct composition under equilibrium conditions, and (2) computerized fluid dynamic (CFD ) calculations which provide information regarding the velocity and temperature profiles along the CVD reator. It was first determined in this work that thermodynamicall y there appears not to be any difference in the equilibrium composition mixture b etween adding HCl or using chlorocarbons or chlorosilane species as the growth precursors. Simulation results were presented for the SiH4-C3H8-HCl-H2 precursor system. The major species present in the equilibrium composition mixture were CH4, -3CH, C2H2, and C2H4 as carbon-containing species and SiH, Si, and SiH2 as silicon-containing species. It was noted that th e presence of HCl in the gas mixture does not significantly af fect the carbon containing species. However, it was observed that chloride species pref erentially bond to silicon suggesting that HCl exhibits a different reaction mechanism ma inly affecting the Si containing reactions. The predicted reactor temperature profile showed t hat the area where deposition occurs was within the experimentally measured tempe rature of 1385C for both reactor geometries used in this work. However, the simulati on failed to show temperature gradients at the susceptor area as observed experim entally. The velocity profile obtained for both reactor geometries showed that the reactor was operating under desired laminar flow. However, the streamline plot for Geometry I s howed flow disruptions at the susceptor outlet that may lead to species back stre am thus possibly affecting the deposited film properties and growth rate.
44 Chapter 3: High Temperature 3C-SiC Heteroepitaxial Growth 3.1 Overview The objective of this research was to increase 3CSiC deposition rates by adding a chloride additive to the C3H8-SiH4-H2 precursor chemistry. A thermodynamic equilibrium study performed on this system (section 126.96.36.199) revealed, for the first time, that there appears not to be any significant differ ence on the most dominant species present in the equilibrium composition mixture betw een adding HCl or using chlorocarbons or chlorosilane species as growth pre cursors. It was predicted that HCl promotes the enhancement of 3C-SiC growth by allowi ng higher Si mole fractions via homogeneous nucleation reduction and formation of a dditional Si-containing species. Therefore, the main focus of this work will be on t he addition of HCl to the CVD precursor system. Indeed, other approaches to incr ease deposition rates can be achieved by manipulating the CVD process parameters, i.e. te mperature, pressure, and input gas composition. This approach was also explored and re sults are presented. As such, the development of a 3C-SiC process via H Cl additive to the C3H8-SiH4H2 chemistry is described in this chapter. This depos ition process was developed in several stages. First, a repeatable 3C-SiC process without HCl additive (i.e. baseline process) was established. Second, once this process was found to be repeatable, it was then used as the starting point for HCl additive pr ocess development. During these experiments HCl was added to the standard chemistry at the growth stage. Third, the HCl additive process was optimized to yield the optimum deposition rate and film quality. Finally, the optimized 3C-SiC deposition process wa s applied on 50 mm diameter Si (001) substrates. Higher area substrates allowed fo r the assessment of epitaxial layer uniformity and film properties.
45 The deposition experiments were carried out using the USF hot-wall CVD reactor MF1 described in section 2.2.1. The process schedul es developed in this work consisted of a two stage carbonization and growth process as described in section 1.4. Two different reactor geometries were used to conduct t he deposition experiments in this work and for simplicity they will be referred as Geometr y I and Geometry II as explained in section 2.2.1. Therefore, distinction among them wi ll be made when describing the different processes developed. Ultra high purity hydrogen, purified via a palladi um diffusion cell, served as the carrier gas during the deposition process. The carb on and silicon precursors were provided by C3H8 (100%) and SiH4 (100%), respectively. Finally, 100 % HCl was used in the HCl additive experiments. Planar n-type Si ( 001) samples diced into 8 mm x 10 mm die were used in this study. The substrates were cleaned using an RCA cleaning procedure preceding deposition. A 30 second immersi on into a buffered hydrogen fluoride etch (HF, 50:1) was performed before loadi ng the sample to the reactor. The resulting epitaxial layers were characterized to monitor the deposition process and results will be presented. Nomarski optical mic roscopy and secondary electron microscopy (SEM) were used to qualitatively analyze the film surface morphology after growth. SEM and Fourier transform infrared (FTIR) R eflectance were used for film thickness determination. Atomic force microscopy (A FM) was used to qualitatively assess surface morphology. X-ray diffraction (XRD) provided information of the crystal quality and X-ray photoelectron spectroscopy (XPS) was used to evaluate the chemical composition of the surface and near surface regions 3.2 3C-SiC without HCl additive process development usi ng Geometry I 3.2.1 Carbonization stage The carbonization stage is crucial during the 3C-S iC hetero-epitaxy since it is believed that the formation of this initial buffer layer influences the crystallinity of the SiC crystal grown due to a reduced effect in the la rge lattice mismatch at the SiC/Si interface.23,24 During the carbonization process, carbon originati ng from the thermal
46 cracking of C3H8 reacts with the silicon surface; as a result SiC n ucleates on the substrate forming a film that is a few nanometers thick. The main problems to be avoided during the buffer layer deposition are the formation of vo ids, etch pits and hillocks. These defects are, for the most part, the result of insuf ficient carbon flux in the gas stream. At low C3H8 mole fractions there is not enough carbon to react with the silicon surface; as a result the vertical diffusion of silicon is favor ed as the temperature is increased creating voids at the buffer layer/substrate interface.86 In addition, low C3H8 mole fractions may also promote the formation of etch pits at the subs trate due to a preferential etching rate process rather than the desired deposition. Finally protrusions that are formed due to hillock defects or surface particles may also be pr esent in the carbonized layers. The presence of these defects in the carbonized layers has been attributed to the agglomeration of silicon at the surface due to vert ical diffusion. 86 A 3C-SiC growth process developed by Dr. R. L. Mye rs in a cold-wall reactor as part as her thesis work performed in our laboratory was used as the starting point for the process development in this study.57 Although this process provided a starting point, successive study of the process parameters was carr ied out to obtain the optimum conditions for the hot-wall system. The process tem perature, pressure, H2 and C3H8 flows and time were selected as the main controllable par ameters influencing the carbonization stage. C/H2 ratios from 1.5 x 10-4 to 48.0 x 10-4 were explored by varying the process parameters within the ranges listed in Table 3.1. E ach parameter was varied one at a time to study its effect on the overall growth process. The optimum carbonization process developed in thi s work is shown in Figure 3.1. A thermal ramp was performed to raise the samp le temperature from approximately 300C to 1170C in the presence of a gas stream composed of H2 and C3H8. The gas stream total flow was 10,006 sccm which had a C3H8 molar fraction of 6.0 x10-4. Finally the sample temperature was held at 1170C for two minutes. After this procedure was completed, the sample was cooled under Ar flow. Compositional surface characterization of a carbon ized layer was determined though peak area analysis using XPS. Prior to the m easurements the samples were submerged for 30 seconds in an HF solution (50:1) f or native oxide removal. XPS
47 Table 3.1 Summary of parameter ranges considered du ring the carbonization stage development. Process Parameter Range Units Molar Fraction Temperature 1050-1200 C --Pressure 150-760 Torr --H2 flow 5-20 sLm* 0.99 C3H8 flow 1-8 sccm** 0.49 x 10-4 16 x 10-4 time 2-5 minutes --* standard liters per minute ** standard centimeter cube per minute Figure 3.1 Carbonization process schedule developed for Geometry I. T=1170C, P=760 Torr and a total gas flow (H2 and C3H8) of 10,006 sccm. A C3H8 molar fraction of 6.0 x 10-4 yielded the optimum buffer layer morphology. measurements revealed that carbon, silicon and oxyg en (O), were the elements present at the buffer layer surface at atomic concentrations o f 42.1%, 35.0% and 22.8%, respectively. The high resolution XPS spectra of the C1s and Si2 p are shown in Figure 3.2(a) and Figure 3.2(b), respectively. Each high resoluti on spectra was deconvoluted by fitting
48 (a) (b) Figure 3.2 XPS high resolution spectra of (a) C1S a nd (b) Si2p peaks for a representative carbonized layer. Each high resolution spectra was deconvoluted by fitting Gaussian curves.
49 theoretical Gaussian curves. The C1s spectra displa yed carbidic(49.7%) and graphiticbound (50.3%) carbon at the surface and near surfac e regions. Graphitization of SiC surfaces is known to occur due to exposure with air or due to a preferential evaporation of silicon at elevated temperatures.87 The Si2p spectra suggested the presence of metalli c silicon (42.5%), carbidic-bound silicon (55.0%) and silicon bound to oxygen as silicon dioxide SiO2 (2.5%). The metallic silicon observed could be comi ng from the silicon substrate. Finally the O1s (not shown) revealed tha t the oxygen present was bound to silicon as SiO2. 3.2.2 Second thermal ramp Since 3C-SiC growth is typically performed at temp eratures higher than 1300C, a second thermal ramp was conducted to transition fro m the carbonization stage to the growth stage. The second thermal ramp is crucial si nce the resulting layer will provide the final surface template used during the growth s tep. Initial experimentation was performed by ramping to growth temperature (10C/mi n) using a gas mixture composed of H2 and C3H8 at C/H2 ratios from 6.0 x 10-4 to 10.0 x 10-4. However pits and rough surfaces where observed via AFM in the resulting la yers. The presence of these defects may be due to insufficient silicon to react with th e carbon precursor resulting in preferential etching process. Therefore SiH4 was introduced during this process. Other challenges presented were the formation of hillock defects and the sporadic observations of particles. In this case hillocks and/or particle s at the surface could be the result of carbon or silicon precipitates due to elevated C3H8 or SiH4 mole fractions. To evaluate this, a study of the process parameters was carried out to determine optimal conditions. The process parameters and ranges considered are li sted Table 3.2. This study was also performed by varying one factor at a time. Figure 3.3 displays the process schedule developed for the second thermal ramp including the carbonization step. SiH4 was introduced into the gas stream and the temperature was increased from 1170C to 1385C aft er performing the carbonization stage. The total gas flow (H2, C3H8 and SiH4) was 10,010 sccm. The C3H8 and SiH4 molar fractions were 6.0 x 10-4 and 4.0 x 10-4, respectively. Once the temperature reached
50 1385C the process was stopped, the sample was cool ed under Ar flow and the thin layers were analyzed. Table 3.2 Summary of parameters ranges considered d uring the second thermal ramp development. Process Parameter Range Units Molar Fraction Temperature 1170-1350 C --Pressure 150-760 Torr --H2 flow 10-20 sLm .99 C3H8 flow 1-8 sccm 0.5 x 10-4 -8 x 10-4 SiH4 flow 0-10 sccm 010 x 10-4 standard liters per minute ** standard centimeter cube per minute Figure 3.3 Second thermal ramp process schedule dev eloped for Geometry I, including the carbonization stage. Total gas flow (H2, C3H8 and SiH4) was 10,010 sccm. C3H8 and SiH4 molar fractions were 6.0 x 10-4 and 4.0 x10-4, respectively. A plan-view SEM image of a representative layer gr own following the process schedule is illustrated in Figure 3.4. The surface lacked hillock defects indicating the
51 process parameters were indeed approaching optimum ranges (Figure 3.4(a)). A crosssectional view of the same layer is shown in Figure 3.4(b). No voids were observed at the interface. The estimated layer thickness via SEM wa s approximately 147 nm. (a) (b) Figure 3.4 (a) Plan view SEM image of a representat ive layer after carbonization and second thermal ramp at best process conditions (no etch pits or hillock defects). (b) Cross-section view of the same layer indicated no v oids at the interface. The estimated layer thickness was 147 nm. Image courtesy of D. Ed wards, USF COT. The film topology was analyzed via AFM. Figure 3.5 (a-b) shows the growth progression of the already coalesced islands formed in the carbonization process due to the low growth rate deposition process performed in the second ramp. As can be seen, the layer is characterized by large features which were homogenously oriented (Figure 3.1). This is expected since the carbonized layer previou sly formed provided a better template for the second thermal ramp. The features were obse rved to be triangular in shape with a measured surface roughness of 3.9 nm RMS. Surface chemical composition analysis by XPS of a resulting layer after carbonization and the second thermal ramp was perfo rmed. The samples were also cleaned prior to the analysis with an HF solution a s explained earlier. As with the carbonized layer, C, Si and O were determined to be the species present at the surface and near surface regions. The atomic concentrations wer e measured to be 52.0%, 31.5% and 16.5%, respectively.
52 (a) (b) Figure 3.5 AFM micrographs of a representative laye r after the carbonization and second thermal ramp processes. (a) Note homogeneously orie nted features. (b) Closer inspection revealed features were triangular shaped. Measured surface roughness was 3.9 nm RMS. Sample ID USF-06-097. The high resolution spectra for the C1s peak shown in Figure 3.6(a) revealed that 55% of the carbon present at the surface was due to C bound as SiC and the remaining 45% was due to adventitious carbon. Similarly the h igh resolution Si2p peak showed that only silicon bound as SiC (96.9%) and SiO2 (3.1%) were present (Figure 3.6(b)). Each high resolution spectra was also deconvoluted by fi tting theoretical Gaussian curves. Note that contrary to the XPS analysis performed fo r the carbonized layer (section 3.2.1) metallic silicon was not found to be present at the surface. This was due to an increased layer thickness as a result of the low-rate deposit ion carried out during the second thermal ramp which yielded a layer approximately 147 nm thi ck. Therefore X-ray penetration to the substrate was not an issue. 3.2.3 Growth stage With the carbonization and second thermal ramp pro cess developed and optimized, the 3C-SiC growth step could proceed. In itial growth step conditions were also based on the initial 3C-SiC growth process dev eloped by Dr. R. Myers.57 However, polycrystalline films were obtained when similar pr ocess conditions were implemented.
53 (a) (b) Figure 3.6 XPS high resolution spectra of (a) C1S a nd (b) Si2p peaks of a representative layer after the carbonization and second thermal ra mp processes. Each high resolution spectra was deconvoluted by fitting Gaussian curves Therefore multiple experiments were conducted in wh ich the process parameters of temperature, pressure and precursor mole fractions were varied one at a time. As a result, a preliminary growth step process which yielded sin gle crystalline 3C-SiC films grown at a rate of 4 m/h was obtained. The preliminary growth step was c onducted at a process
54 temperature and pressure of 1375C and 200 Torr. Th e total input flow was set to 30,014 sccm which had C3H8 and SiH4 mole fractions of 2.0 x 10-4 and 2.6 x 10-4, respectively. Once this preliminary process was obtained, optimi zation of the process parameters was performed. First the process tempera ture was varied within the temperature range of 1330C 1395C. It was observ ed that at temperatures lower than and equal to 1350C, non-specular layers were obtai ned suggesting that the saturation point had been reached. The best surface morphology as assessed by AFM and SEM techniques was found to occur within the temperatur e window of 1375C-1385C; the temperature of 1385C was chosen to be kept constan t for the continued experiments. Pressure variation experiments over the range of 75 Torr 400 Torr revealed an improvement in surface morphology at decreased pres sures. Despite that better surfaces were obtained at decreased pressures, 100 Torr was chosen as the most favorable pressure since that was the lowest value that the CVD system could sustain efficiently. Having established that the most favorable surface morphology for our process had been obtained at 1385C and 100 Torr, the focus was shifted to determine the optimum process Si/C ratio at these conditions. Thi s was accomplished by varying the SiH4 mole fraction keeping the process temperature, pre ssure and H2 and C3H8 mole fractions constant during the epitaxial deposition process. From these experiments it was found that a SiH4 mole fraction of 4.3 x 10-4 provided the best surface morphology as assessed by SEM and AFM. Details of the final growth process schedule for t he 3C-SiC process without HCl additive are illustrated in Figure 3.7. The carboni zation and second ramp process described in Figure 3.3 were performed prior to con duct the growth stage. Once the growth temperature of 1385C was reached the proces s pressure was lowered to 100 Torr and the total gas flow (H2, C3H8 and SiH4) was set to 30,019 sccm. The C3H8 and SiH4 molar fractions at the growth stage were 2.0 x 10-4 and 4.3 x 10-4, respectively. At these process conditions the process Si/C ratio was 0.72 and the process growth rate increased to 8.6 m/h. Once the optimum process conditions were determine d experiments were conducted in order to obtain the maximum process gr owth rate. This was accomplished
55 Figure 3.7 3C-SiC growth process schedule using C3H8-SiH4-H2 chemistry developed for Geometry I. Mole fractions presented for 8.6 m/h process. At the growth stage the total input gas flow (H2, C3H8 and SiH4) was set to 30,019 sccm with C3H8 and SiH4 molar fractions of 2.0 x 10-4 and 4.3 x 10-4, respectively. Process Si/C=0.72 by setting the process temperature, pressure and Si /C constant while increasing the precursor mole fraction by approximately 20% after each run using the parameters shown in the process schedule illustrated in Figure 3.7. From this study it was determined that the maximum process growth rate that could be obtai ned leading to specular surfaces was approximately 12 m/h; this was achieved at a total input flow of 30,022 sccm. The propane and silane mole fractions were 2.0 x 10-4 and 5.3 x 10-4, respectively. At higher precursor mole fractions the deposited layers obtai ned were polycrystalline. In conclusion, the optimization of the 3C-SiC proc ess without HCl additive not only allowed for good film morphology but the growt h rate was increased by 3 times that of the preliminary growth process. In addition, sin ce this process was desired to be used as the starting point for the HCl additive experime nts it was imperative to ensure that the process growth rate was repeatable. Therefore, the growth rate variation was studied using the ~8.6 m/h process described above. Table 3.3 lists growth rate measurements taken for 6 different samples grown using the proce ss schedule as explained in Figure 3.7.
56 Table 3.3 Summary of growth rate data for the calcu lation of baseline process repeatability using ~8.6 m/h process. Run Growth Rate (m/h) Average (m/h) Standard Deviation (m/h) 1 9.0 2 8.9 3 8.6 4 8.6 5 8.4 6 9.0 8.8 0.2 It was determined that the average process growth rate was 8.8 m/h. For this particular data set the standard deviation was dete rmined to be 0.2 m/h. The average (-x) and standard deviation (r) values were calculated using their respective def initions from statistics. Once the process growth rate was found to be repea table the epitaxial layers grown using the optimized process were characterize d further. Figure 3.8 shows a planview SEM image of a representative 3C-SiC film afte r performing the growth schedule depicted in Figure 3.7. The surface was specular, a nd clean of protrusion type defects. Figure 3.8 Plan-view SEM image shown at a magnifica tion of 5.0k for a representative 3C-SiC layer grown with the no HCl process at a rat e of ~8.6 m/h grown. Image courtesy of D. Edwards, USF COT.
57 Figure 3.9 shows a 10 m x 10 m AFM micrograph taken in contact mode for the same epitaxial layer. The surface topography wa s observed to have a more crystallike surface due to the coalescence of the differen t nucleation islands. The typical antiphase domain boundaries characteristic of 3C-SiC fi lms are also observed. As a reference, the measured surface roughness was 1.6 n m RMS. Figure 3.9 AFM micrograph (10 m x 10 m) taken in contact mode of a representative 3C-SiC layer grown at a rate of 8.6 m/h. The surface roughness was measured to be 1.6nm RMS. The quality of the epitaxial layer was also assess ed via XRD. The powder diffraction technique was applied first in order to observe which reflections were detected in the crystal under study (Figure 3.10). The data revealed that the most intense peak was from the 3C-SiC (002). The presence of the Si (002) and 3C-SiC (004) diffraction peaks were also noted, however at much lower intensities. A XRD rocking curve was generated for the 3C-SiC ( 002) diffraction peak which revealed a full width at half maximum (FWHM) of 500 arcsec as seen in Figure 3.11. This value was found to be better and/or comparable to values reported elsewhere and is indicative of a single crystal quality material.45 It should be noted that the (004) plane reflection indicates epitaxial film growth since th is is part of the (002) family of crystal planes.
58 Figure 3.10 XRD powder diffraction of a 3.3 m thick 3C-SiC epitaxial layer. The diffraction peaks for Si <002>, 3C-SiC <004> and 3C -SiC <002> were observed. XRD performed by Dr. Y. Shishkin, USF SiC group. Figure 3.11 XRD rocking curve of a 3.3 m thick 3C-SiC epitaxial layer performed at the 3C-SiC (002) diffraction peak. The measured FWHM wa s 500 arcsec which is indicative of a high quality single crystal layer.45 XRD performed by Dr. Y. Shishkin, USF SiC group.
59 3.3 HCl additive process development using Geometry I The 3C-SiC without HCl process development was cru cial for the HCl additive experiments as it provided an optimized, repeatable and well documented baseline process. Since good quality single crystal 3C-SiC l ayers were obtained by performing the no HCl additive process, the addition of HCl was ap plied next to this process to determine if the process growth rate could be incre ased further and/or the epitaxial layer quality could be improved. In this section the resu lts of the HCl additive study are presented. The HCl addition was applied first to th e second thermal ramp and finally to the growth stage. 3.3.1 HCl addition to second thermal ramp The HCl additive experiments first focused on the second thermal ramp process illustrated in Figure 3.3 with the objective being to improve the surface morphology of the layer grown, thus providing a better surface te mplate. In this work HCl was added at mole fractions ranging from 0.5 x 10-4 to 2.0 x 10-4 after the carbonization process was completed. The epitaxial layers were analyzed via o ptical microscopy, SEM and AFM where the formation of etch pits and rough surfaces were noted. The presence of these defects was mainly attributed to HCl etching at the surface. Therefore the experimental results supported that the addition of HCl to the s econd thermal ramp was not beneficial for the process. As a result no changes were made t o the second thermal ramp process developed during the 3C-SiC process without HCl. 3.3.2 HCl addition to the growth stage HCl additive experiments performed next focused on the growth stage. During this series of experiments, HCl at mole fractions r anging from 0.2 x 10-4 to 2.7 x 10-4 were added to the 3C-SiC without HCl process result ing in deposition rates of ~8.6 m/h (Figure 3.3). It was observed that the epitaxial gr owth rate and surface morphology remained unchanged for HCl mole fractions up to 1.0 x 10-4 after which the surface morphology became rough and non-specular due to exc essive HCl in the gas stream
60 causing surface etching. It was concluded from thes e experiments that the maximum HCl mole fraction that the process could withstand with out compromising growth rate or surface morphology was 0.7 x 10-4. At these conditions the process Si/Cl ratio was determined to be 6.5. Similarly, experiments were c onducted by varying the SiH4 mole fraction while keeping the other process parameters constant to verify that the Si/C ratio used was the most beneficial for the HCl process. F inally, AFM measurements supported that a Si/C ratio of 0.9 provided a slight improvem ent in the surface morphology compared to that of the no HCl additive process. The final process schedule developed for the HCl ad ditive experiments is illustrated in Figure 3.12. When comparing the no H Cl additive process schedule to the HCl additive process schedule it can be noted that the only differences among them are the addition of HCl (Si/Cl=6.5) into the gas chemis try at the growth stage as well as an increased Si/C ratio from 0.7 to 0.9 for the HCl additive process. Figure 3.12 3C-SiC HCl additive growth process sche dule developed for Geometry I. Mole fractions presented for ~12 m/h process. At the growth stage the total input ga s flow (H2, C3H8 and SiH4) was set to 30,025 sccm. Process Si/C=0.9 and Si/C l=6.5. flow (H2, C3H8 and SiH4) was set to 30,025 sccm. Process Si/C=0.9 and Si/C l=6.5.
61 3.3.3 CVD reactor characterization From the work previously discussed it was discover ed that just adding HCl to the process did not result in a growth rate increase or a dramatical improvement of the epitaxial layer surface morphology or quality. Ther efore, continued study was conducted in order to determine which advantages to the proce ss the HCl addition will provide. This was accomplished by performing reactor characteriza tion via process parameters dependences as discussed next. 188.8.131.52 Growth rate as a function of silane mole fraction In section 3.2.3 it was discussed that increased p recursor mole fractions were not possible, thus the growth rate for the no HCl addit ive process was limited to 12 mm/h. Therefore a similar experimental series was impleme nted for the HCl additive process in order to determine if HCl would allow for increased precursor mole fractions resulting in increased growth rate while maintaining crystal mor phology/quality. As a reference, the process pressure and temperature used in this study were 1385C and 100 Torr, respectively. In addition, the process Si/C and Si/ Cl ratios were kept constant at 0.9 and 6.5, respectively; which were identified to be the optimum in section 3.3.2. Figure 3.13 illustrates the experimental growth ra te dependence on SiH4 mole fraction. The total input flow varied from 30,025 s ccm to 30,069 sccm. The SiH4 mole fractions ranged from 0.53 x10-3 to 1.49 x10-3 whereas C3H8 mole fractions varied from 0.19 x 10-3 to 0.55 x 10-4. Finally the HCl mole fraction on the gas stream f luctuated within the 0.09 x 10-3 to 0.29 x 10-3 range. As can be observed, the growth rate followe d a linear relationship increasing from ~8.6 m/h to 38 m/h. These values are the highest growth rates reported in the literature for the 3CSiC heteroepitaxy via horizontal, hotwall CVD to date. Specular surface morphology was obtained for every film grown in this set of experiments. Inspection of the 3C-SiC epitaxial lay ers suggested no void formation at the layer/substrate interface. Further inspection via S EM showed the sporadic growth of protrusions defect clusters as shown in Figure 3.14 It was observed that the size of these defects was more prominent at higher silane mole fr actions and as the epitaxial layer
62 thickness increased. They may occur as a result of homogeneous nucleation in the gas phase, which is known to occur at high Si mole frac tions when the precursor mixture is approaching the saturation limit at the given proce ss conditions. 0.5 1 1.5 2 x 10 -3 0 10 20 30 40 50 60 Silane mole fractionGrowth Rate (mm/h) 3C-SiC on Si(001)T=1385oC P=100 TorrSi/C=0.9Si/Cl=6.5 Figure 3.13 3C-SiC film growth rate vs. SiH4 mole fraction. Growth rates from 8.6 m/h to 38 m/h were obtained. Experiments conducted at precurs or Si/C=0.9 and Si/Cl=6.5. C/H2 varied from 6.0 x 10-4 to 16.7 x 10-4. Trend line to aid the eye only. Figure 3.14 Plan-view SEM image showing protrusion defects which were observed to increase in density and size as the SiH4 mole fraction and the epitaxial layer thickness increased. The epitaxial layer thickness is ~ 23 m grown at a speed of 20 m/h. Image courtesy of D. Edwards, USF COT.
63 184.108.40.206 Growth rate as a function of process pressure The process pressure effect on the deposition rate and film surface morphology was also studied. Experiments were carried out usin g a 12 mm/h growth process. The process Si/C, Si/Cl and C/H2 ratios were 0.9, 9.6 and 6.0 x 10-4, respectively. The process pressure was varied from 75 Torr to 760 Torr, where as the process temperature was fixed at 1385C. The experimental results are summarized in Figure 3.15. The deposition rate remained constant for process pressures from 75 Tor r to 250 Torr, after which the rate started to decrease. Although all the films present ed specular surface morphology, AFM measurements revealed the film surface roughness in creased from 3.7 nm RMS to 12.0 nm RMS over the pressure range studied (75 Torr to 450 Torr) which indicated film degradation with process pressure. This behavior is in accordance to theory which establishes that as the process pressure is decreas ed the deposition growth rate decreases due to lower absolute precursor concentrations and due to increased gas velocity. Similarly improved deposition uniformity is expecte d as the pressure is decreased due to an increased value in the diffusion length of the a dsorbed species on the growth surface. Based on the findings discussed above the optimum p rocess pressure for the deposition experiments was selected to be 100 Torr. Lower pres sures were not selected due to the inability of the vacuum system to maintain them. 0 200 400 600 800 0 5 10 15 20 25 30 Pressure (Torr)Growth Rate (mm/h) 3C-SiC on Si(001)T=1385oC Figure 3.15 3C-SiC film growth rate vs. process pre ssure. Films grown at a speed of 12 m/h with HCl additive at Si/C=0.9 and Si/Cl=6.5. Op timum film quality achieved at 100 Torr. Trend to aid eye only.
64 220.127.116.11 Process stability and reliability An important aspect during CVD reactor characteriz ation is to study if epitaxial layers grown at different deposition times lead to the same growth rate. If this could be attained then it can be ensured that the process is stable and reliable. In this study, deposition experiments were conducted using a 20 m/h deposition process. The process pressure and temperature were set to 100 Torr and 1 385C, respectively, and a total input gas flow of 30,038 sccm was maintained during the g rowth step. The C3H8, SiH4 and HCl mole fractions were 8.32 x 10-4, 3.10 x 10-4 and 1.33 x 10-4, respectively. Finally growth times from 5 minutes to 60 minutes were examined. F igure 3.16 illustrates the experimental dependence of growth rate with deposit ion time. As can be seen, the growth rate remained fairly constant suggesting that the p rocess will be able to sustain this growth rate at extended deposition times. Although the growth rate could be maintained and s pecular films were obtained in this study, the experimental data indicated degrada tion of the epitaxial layer due to an increased size of protrusion type cluster defects a t the surface as can be seen in the optical microscope images illustrated in Figure 3.17. Due t o this behavior, additional experimentation was performed at deposition times o f 75 minutes and 90 minutes to determine to what extent specular films could be ob tained. It was found that the maximum growth time that could be employed in this study was 60 minutes for films grown with the HCl additive process at a rate of 20 m/h. The deposited layers grown for 60 minutes (~ 20 m thick) were specular. At growth times greater tha n 60 minutes, film defect propagation was prominent and rough (non-spe cular) surfaces were often obtained. The film quality was also assessed via XRD. Figure 3.18 shows the XRD rocking curve collected on a 12 mm thick 3C-SiC film grown on the control Si (100) s urface at a rate of 20 mm/h under conditions described above. The full widt h at half maximum (FWHM) of the (200) diffraction peak was ~360 arcse conds. This value is comparable to the that reported in literature confirming the stru ctural integrity of the 3C-SiC films grown with this process.45
65 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Time (minutes)Growth Rate (mm/h) 3C-SiC on Si(001)T=1385oC P=100 Torr Figure 3.16 3C-SiC film growth rate vs. growth time The films were grown at a rate of 20 mm/h. The growth times were 5, 7, 15, 30 and 45 min. Trend line drawn to aid the eye only. Figure 3.17 Optical microscope images at a magnific ation of 20X of (a) 5.8 m, (b) 11.7 m and (c) 23.3 m 3C-SiC epitaxial layers sho wing increased size of protrusion defects at the surface as the epitaxial thickness i ncreases. 3.4 HCl additive process optimization using Geometry I Since growth at extended times ( > 60 min. @ 20m/h) was not possible due to surface morphology degradation, studies were conduc ted to investigate the growth of thicker films. The initial focus was directed towar d the carbonization and second thermal ramp process. It was hypothesized that providing a higher quality buffer layer before growth would decrease the propagation of defects fr om the buffer layer to the epitaxial
66 layer thus decreasing film degradation with growth time. To this end, optimization was carried out in the actual growth step where the fil m degradation with time was first noticed. Figure 3.18 XRD rocking curve of a 3.3 m thick 3C-SiC epitaxial layer performed at the 3C-SiC (002) diffraction peak. The measured FWHM wa s 500 arcsec which is indicative of a high quality single crystal layer.45 Before conducting the optimization procedure, samp les in which the process was stopped right after the second thermal ramp were re -examined. From this analysis it was confirmed that no voids were present at the 3C-SiC/ Si interface. However, more characterization of the films surfaces by AFM revea led the sporadic formation of protrusion defects several nanometers in size after the second thermal ramp was performed. 3.4.1 Optimization of the carbonization step This optimization focused on obtaining the optimum C3H8 mole fraction for the carbonization process and also explored the additio n of an H2 etch step after the 2 minute carbonization stage in order to solve the protrusio n defect problem. Table 3.4 summarizes
67 three of the numerous experiments conducted for thi s process optimization. Figure 3.19 (a) shows an AFM micrograph of the carbonized surfa ce when the amount of C3H8 in the gas stream during the second thermal ramp was reduc ed from that shown in Figure 3.12. As can be seen, the surface contains numerous grain s which are round in shape. Figure 3.19(b) illustrates the AFM film surface micrograph of a sample in which the same process described for Figure 3.19(a) was done after an H2 etch step of 1 minute at 1170C. It can be observed that by adding the etch step, the anti-phase domains at the surface are still rounded in shape but somewhat lar ger than those observed without the H2 etch step. Finally, Figure 3.19(c) shows the carbon ized surface using a process including the H2 etch step but with the same C3H8 mole fraction shown in Figure 3.12. Enlarged, more oriented and homogeneous cubic grains were obs erved. Table 3.4 Summary of process parameters during seco nd thermal optimization Run # C3H8 mole fraction H2 etch Surface Roughness (nm RMS) Notes 1 4.5 x 10-4 No 12.1 Figure 3.19(a) 2 4.5 x 10-4 Yes 5.5 Figure 3.19(b) 3 6.0 x 10-4 Yes 6.2 Figure 3.19(c) Figure 3.19 Surface AFM micrographs taken on tappin g mode of the carbonized surface during the initial stages of growth for (a) decreas ed C3H8 compared to process in Figure 3.12, (b) decreased C3H8 and 1 min H2 etch step compared to process in Fig.1, and (c) process in Fig.1 followed by 1 min H2 etch step. AFM courtesy of Dr. C. Colletti.
68 Comparing the initial surfaces after the carbonizat ion process with the surfaces in Figure 3.19 it was determined that the process of F igure 3.19(c) resulted in an improved surface morphology. The latter process provided a s urface with cubic-shaped mesas properties which are more in accordance with the cu bic properties of 3C-SiC. 3.4.2 Optimization of second thermal ramp Since the addition of HCl at the second thermal ra mp had been demonstrated to lead to surface morphology degradation, optimizatio n was performed by varying the Si/C ratio on the gas stream as well as adding or elimin ating steps to the process schedule. From this work it was found that the protrusion def ect formation (assessed via AFM) was reduced when the SiH4 mole fraction was ramped from 110-4 to 410-4 during the second thermal ramp. The optimum carbonization process described in sec tion 3.4.1 followed by the SiH4 ramp at the second thermal ramp were therefore app lied for subsequent deposition experiments which were carried out for extended gro wth times. A reduction in hillock defects was observed on thinner films with specular morphology. However, surface degradation resulted in non-specular films grown fo r times in excess of 60-70 minutes. These results suggest that the main problem with de fect generation may emanate from the growth step. 3.4.3 Optimization of growth stage The surface morphology optimization at the growth step was followed by performing Si/C and Si/Cl ratio studies. However th e protrusion defect problem persisted. Therefore changes were implemented to the growth st ep in order to improve the film morphology. It was discovered that when repeated 30 second H2 etch steps were added during growth after every 10 minutes during growth, the density of the protrusion defects was decreased. Figure 3.20 shows the growth process schedule incl uding the optimized carbonization followed by the grow/etch step for th e 3C-SiC HCl additive process. This
69 grow/etch step combination was repeated over the de sired growth time. Epitaxial films grown using the optimized growth process displayed specular surface morphology with comparable quality to films grown with the process described in Figure 3.12. Figure 3.20 Optimized 3C-SiC HCl additive process d eveloped for Geometry I. Mole fraction shown for the 20 m/h process. A total inp ut flow of 30,038 sccm was used. Plan view SEM images showed a reduced density of h illock defects that were very similar to films grown with the non-optimized HCl additive processes. Cross-section SEM showed no visible interface problems that may h ave emanated due to the H2 etch steps. Although this approach seemed to allow for t he growth of thicker films, 30 m compared to 20 m, clearly it did not eliminate the protrusion defe ct problem. Since it was determined that the growth/etch appro ach did not provide a solution for the protrusion defect problem, subsequent exper imentation was performed using the optimized carbonization and second thermal ramp wit h the original growth step as shown in Figure 3.12. This approach also yielded epitaxia l layers of similar properties to films reported in previous sections and at the same time provided a process with fewer disruptions of the precursor ratios during the grow th process. The latter are known to promote the formation of interfaces which could be detrimental for future device applications.
70 3.5 Process transfer to Geometry II using DOE Once the 3C-SiC HCl additive process had been deve loped, characterized and optimized the next step was to apply it to larger a rea substrates, specifically to 50 mm wafers which is the largest area substrate for whic h the CVD reactor is currently configured (4 growth possible with a hot-zone chan ge). However, as mentioned before, initial application of the growth process schedule revealed that specular morphology was not achieved over the entire substrate area and on occasion, melting of the silicon substrate occurred. As stated, S. Harvey and Y. Shi shkin of our group suggested that the reason for this behavior was the formation of a lar ge temperature gradient of approximately 35C along the 50 mm substrate area.59 As a result of this study it was suggested that a change in the reactor geometry was needed in order to solve the temperature gradient problem. Therefore, Geometry I I was implemented (Refer to section 2.2.1 for a more detailed description of the reacto r geometries used in this work). These changes produced a more uniform substrate temperatu re across the 50 mm substrate area. As a result, the experimentally measured temperatur e variation along the susceptor area was decreased to 5 C. In addition to the reactor geometry change, the co ncentration of the silane precursor and the HCl additive gas were changed fro m 100 % each to more diluted mixtures of 10% SiH4 in H2 and 10% HCl in Ar. These changes were performed ma inly to gain more control of the mass flow controllers w hen lower flows were needed. In addition, a lower concentration of HCl was desired for safety considerations due to problems with corrosion of the gas lines. Because a CVD process is very sensitive to system changes it was imperative to prove that the already developed and optimized proc ess using Geometry I could be applied to Geometry II. Since a change in geometry may cause changes in the system fluid dynamics thus affecting the process growth ra te and surface morphology, CFD simulations were initially performed and these CFD simulation results were reported in section 18.104.22.168. After it was theoretically determined that process transfer from Geometry I to Geometry II may be viable due to the similar fluid dynamics, the focus was shifted to
71 designing a more efficient experimental design appr oach to corroborate the process growth rate and process morphology for Geometry II with a minimum number of runs. Thus far, the one point at a time experimental a pproach had been the main experimental approach followed. This method result s in numerous experimental runs and the data obtained is difficult to analyze systemati cally. Therefore the method of statistical design of experiments (DOE) was implemented. DOE is an experimental strategy for setting up a set of experiments in which all variab les are varied in a systematic manner in order to determine correlation between variables an d to predict results.88 Specifically, the fractional factorial experimental design method was employed in this work. The fractional factorial DOE is a variation of the fact orial design in which factors are varied together instead of one at a time but only a subset of the experimental matrix is performed.88 This experimental design enables the experimenter to investigate the individual effects of each parameter and determine whether the parameters interact by performing a limited number of experimental runs. A brief explanation of the DOE construction is provided in Appendix D. Further DOE construction and data analysis details can be found elsewhere.88 3.5.1 DOE results Five factors were taken into consideration for the DOE experiments. These were: temperature, pressure, C3H8, SiH4, and HCl mole fraction. If a full factorial (25) design of experiment is considered, then the experiment will require 64 runs if the data is replicated once. Since CVD is a very expensive and time consum ing process then a method that decreases the number of runs is very desirable. A o ne quarter fraction 25-2 of the 25 design was chosen since it decreased the number of runs to 8. In addition, a center point design was implemented in order to have an improved estima te of experimental error. The proposed parameter ranges considered are listed in Table 3.5; these were chosen based on the knowledge acquired during the process developme nt using Geometry I. In order to prevent or minimize the effects of nuisance variabl es from contaminating the results, a completely randomized experimental design was used. The random numbers were generated and sorted using the computer software MA TLAB.89 Finally, the experimental
72 growth rate was chosen as the response variable to be studied. A summary of the experiments performed with its respective measured response values is presented in Table 3.6 and Table 3.7, respectively. Table 3.5 Summary of factors range considered to pe rform 25-2 fractional factorial DOE. Factor Range Units Mole fraction Temperature (A) 1300-1380 C --Pressure (B) 100-400 Torr --C3H8 flow (C) 6-10 sccm 2.0 x 10-4 3.3 x 10-4 SiH4 flow (D) 11-18 sccm 3.7 x 10-4 6.0 x 10-4 HCl flow (E) 1-4 sccm 0.33 x 10-4 1.3 x 10-4 Table 3.6 Runs and experimental results for 25-2 DOE Table 3.7 Center point runs and experimental result s for the 25-2 DOE Run T P yC3H8/104 ySiH4/104 yHCl/104 Growth Rate (m/h) 1 7.2 2 7.2 3 1340 250 2.6 4.8 0.83 7.4 Run T P yC3H8/104 ySiH4/104 yHCl/104 Growth Rate (m/h) 1 1300 100 2 6 1.3 8.4 2 1380 100 2 3.7 0.3 6.0 3 1300 400 2 3.7 1.3 5.1 4 1300 100 3.3 6 0.3 8.7 5 1380 400 2 6 0.3 6.9 6 1380 100 3.3 3.7 1.3 6.3 7 1300 400 3.3 3.7 0.3 5.4 8 1380 400 3.3 6 1.3 4.8
73 After the screening procedure was performed and th e response variable was measured, the data was then analyzed by performing an analysis of variance (ANOVA).88. The ANOVA results indicated that the main factors affecting the CVD process growth rate were the process temperature, p ressure, the SiH4 and HCl mole fractions as well as the interaction between the pr ocess pressure with the C3H8 fraction and the interaction between the SiH4 mole fraction and that of HCl. It is interesting t o note that despite that the C3H8 fraction by itself does not have a significant eff ect on growth rate; its interaction with the process press ure provides an effect on growth rate. Based on the ANOVA analysis, an empirical model th at describes the process growth rate within the parameter ranges considered was determined as illustrated in Equation 3.1. In this equation T is temperature, P is pressure, and C3H8, SiH4 and HCl stand for the flow of the respective specie. The mo del was determined in terms of flows instead of molar fractions in order to facilitate i ts use during experimentation since these are the typical parameters used during experimentat ion. ()()()()4 8 3 8 3 4 8 304.0 001 .0 2.0 557 .0 80.0 002 .0 01.0 1. 14SiH H C H C SiH H CF F F P HCl F F P T GR + + + = Equation 3.1 Now that the DOE model was developed, focus was sh ifted on model validation. The experimental growth rate values for 3 represent ative runs performed in Geometry II compared to the model predicted growth rates are li sted in Table 3.8. As can be seen, the model underestimated the growth rate by approximate ly 25%. Table 3.8 DOE model results comparison with experim ental values for processes performed using Geometry I. Run yC3H8/104 ySiH4/104 yHCl/104 Experimental growth rate Geometry I (m/h) DOE predicted growth rate (m/h) 1 2.0 5.3 .57 10.5 8.0 2 2.0 5.3 0.8 12.0 9.0 3 2.6 5.3 1.0 9.5 7.4
74 Despite that the model underestimated the process growth rate, it was desired to corroborate if the deposition rate will follow the same experimental trends when compared to the dependences obtained for Geometry I as shown in section 3.3.3. This dependence was selected since most of the experimen ts had process parameters that fall within the model ranges. In order to accomplish thi s, the pressure dependence experiments performed on Geometry I were compared t o the predicted values from the DOE model for Geometry II. It was not expected that the epitaxial growth rates predicted by the DOE model were accurate since the model was developed in Geometry II which exhibits a different hydrodynamic flow than that of Geometry I. However, a similar pressure effect on deposition growth rate was desir ed in order to verify that the 3C-SiC process developed for Geometry II will provide simi lar dependences. As can be seen in Figure 3.21, this indeed was the case. It can be cl early seen that process pressure is inversely proportional to the film growth rate. Thi s result not only agrees with the results obtained for Geometry I but also agrees with CVD th eory.42 Figure 3.21 Pressure dependence experiments perform ed on Geometry I () compared to the predicted values from the DOE model for Geometr y II (). In order to verify if comparatively similar deposi tion rates could be attained for both reactor geometries; experiments were carried o ut for the 20, 30 and 38 m/h
75 processes using Geometry II and were compared to th ose obtained in Geometry I. As can be seen in Figure 3.22 Geometry II yielded growth r ates values approximately 15% lower than those obtained in Geometry I. Despite the redu ction in the deposition rate the data followed the same behavior; the growth rate increas ed as the silane mole fraction increased. Figure 3.22 Growth rate dependence on SiH4 mole fraction comparison for the 20, 30 and 38 m/h grown using Geometry I () and Geometry II (). 3.6 Growth on 50 mm substrates using Geometry II After transferring the process to Geometry II, gro wth was performed on 50 mm silicon substrates to verify if films could be depo sited uniformly on the larger substrate. Film thickness measurements were taken via FTIR at 5 different points on the 50 mm substrate as shown in Figure 3.23. Results are summ arized in Table 3.9. The epitaxial layers were deposited at the growth speeds of 20, 3 0 and 38 m/h. As a reference this process uses Si/C and Si/Cl ratios of 0.9 and 6.5, respectively. Variations in thickness across all measured points were found to be less th an 15 %. A slight decrease in growth rate was observed to occur along the gas path. This could be attributed to reactant depletion as species flow across the susceptor. It was also noted that thickness uniformity improved as the growth speed increased.
76 Figure 3.23 Position of the 5 different measurement points considered for film properties evaluation. Table 3.9 Thickness measurements taken on 3 represe ntative samples grown at a speed of 20, 30 and 38 m/h. Measurements taken via FTIR. Thickness (m) Growth Rate (m/h) A B C D E 20 10.6 10.0 10.3 10.0 9.9 30 10.8 10.4 10.5 10.0 10.3 38 10.8 10.5 10.7 10.5 10.5 Deposition experiments were carried out to evaluat e the quality of the epitaxial layers as a function of film thickness and as a fun ction of growth rate. XRD was used to asses the film quality. Powder diffraction was perf ormed first to determine the orientation of the deposited layers. This measurement revealed that the most dominant peaks were located at = 41 which corresponds to the 3C-SiC (200) plane confirming that the films were 3C-SiC (100).Other peaks were observed at the 69 and 90 positions at much lower intensities. This peaks correspond to (400) 3 C-SiC and (200) Si planes, respectively. Table 3.10 provides a summary of the X-ray rocking curve data collected for films grown at 20, 30 and 38 m/h. The measurem ents indicated that the film quality degraded as the growth rate and epitaxial thickness increased.
77 Table 3.10 XRD FWHM summary for films grown at 20, 30 and 38 m/h. Growth Rate (m/h) Thickness (m) FWHM (arcsec) 20 10.2 360 20 20.5 1157 30 10.4 733 38 10.4 854 3.7 Summary The development of a novel 3C-SiC HCl additive pro cess has been completed on a hot-wall CVD reactor. The growth rate was shown t o increase from 12 m/h for the C3H8-SiH4-H2 precursor chemistry to 38 m/h for the HCl additive experiments. The later is the highest reported value in the literature to date. HCl proved to be highly beneficial to the process growth rate. However, the quality of the epitaxial layers did not significant ly improve via the HCl process. Film degradation was observed to occur at increased film thickness and at increased deposition rates for films of the same thickness.
78 Chapter 4: Low Temperature 3C-SiC Heteroepitaxial Growth 4.1 Overview In Chapter 3, a 3C-SiC heteroepitaxial growth proc ess via the C3H8-SiH4-HCl-H2 precursor chemistry was reported. During this work epitaxial growth rates were increased up to 38 m/h leading to the highest reported value in litera ture to date for a hot-wall CVD system.44 Since the later chloride based chemistry demonstra ted to be highly beneficial to the process growth rate it was desire d to apply this chemistry for the low temperature regime of 1000C 1250C (1273K 1523 K). The hypothesis being that the implementation of this precursor chemistry will res ult in useful epitaxial growth rates and good film quality at the lower deposition temperatu res. Typically, high deposition temperatures ( 1350C) are required to ensure high quality films and high deposition rates. However, t he implementation of low deposition temperatures would be beneficial for device process fabrication. Lower process temperatures will eliminate or decrease problems du e to interdependencies with other process steps during device fabrication processes. This will help to avoid problems related to auto-doping, solid state diffusion and a lleviate stresses in the epitaxial layers.3 In addition, lower deposition temperatures are attr active for selective epitaxial growth (SEG) applications where lower deposition temperatu res are needed to avoid damage to the required silicon dioxide (SiO2) mask.3 The experiments described in this chapter were per formed using the USF CVD reactor configured with Geometry II as described in section 2.2.1. A two-step carbonization and growth process was also applied. Ultra high purity hydrogen, purified via a palladium cell, served as the carrier gas dur ing the deposition process. The carbon and silicon precursors were provided by C3H8 (100%) and a mixture of 10% SiH4 in H2, respectively. Finally, the mixture of 10% HCl in Ar was used as the chloride growth
79 additive. Planar n-type Si (001) samples diced into 8 mm x 10 mm die were used in this study. The substrates were cleaned using an RCA cle aning procedure preceding deposition. A 30 second immersion into a hydrogen f luoride (HF, 50:1) was performed before loading the sample to the reactor. Unfortuna tely, at the moment this study was being performed several malfunctions on the HCl man ifold occurred until it was completely disabled. Therefore, this study is somew hat limited because all experimentation including HCl additive had to be di scontinued for safety considerations. Nomarski optical microscopy and secondary electron microscopy (SEM) were used to qualitatively analyze film surface morpholo gy after growth. SEM and Fourier transform infrared (FTIR) reflectance were used for film thickness determination. Atomic force microscopy (AFM) qualitatively assessed the s urface morphology. X-ray diffraction (XRD) provided information on the crystal quality. 4.2 Low-temperature 3C-SiC growth process development 4.2.1 Carbonization Before conducting the growth experiments, the sili con substrates were carbonized using the already developed process described in se ction 3.2.1. As a reminder, during this process a thermal ramp was conducted to raise the s ample temperature from approximately 300C to 1170C in the presence of a gas stream composed of H2 and C3H8. The gas stream total flow was 10,006 sccm which h ad a C3H8 molar fraction of 6.0 x10-4. Finally the sample temperature was held at 1170C for two minutes. 4.2.2 Growth stage In order to determine the most favorable condition s for the growth stage, a resolution IV screening design of experiments (DOE) was implemented. For this, a twolevel one quarter (26-2) fractional factorial design which included 16 run s and four center points was carried out to investigate the process p arameter space. The center point design was applied in order to have an improved estimate o f experimental error since the data was not replicated. No main effects are aliased wit h any other main effect or with any other two-factor interaction. The variables include the process temperature, pressure and
80 the hydrogen, propane, silane and hydrogen chloride mole fractions. A constant growth time of 40 minutes was targeted for all runs. A com pletely randomized design was implemented to avoid or minimize the effects of nui sance variables. The random numbers were generated by using the computer software MATLA B.89 The response variable considered was the process deposition rate. A summa ry of the factor parameters and the experimental matrix with its respective measured re sponse are provided in Table 4.1 and Table 4.2, respectively. The center point data is p rovided on Table 4.3 Table 4.1 Summary of factors range considered to pe rform 26-2 fractional factorial DOE Factors Range Units Mole fraction Temperature 1150-1250 C --Pressure 100-400 Torr --H2 10-20 sLm 0.99 C3H8 3-5 sccm 1.5 x 10-4 5.0 x 10-4 10% SiH4 in H2 40-90 sccm 2.0 x 10-4 9.0 x 10-4 10% HCl in Ar 0-20 sccm 0 2.0 x 10-4 Table 4.2 Experimental matrix and response values f or 26-2 DOE Run Temperature Pressure H2 C3H8 SiH4 HCl Growth Rate (m/h) 1 1250 400 20 3 40 0 0.2 2 1150 400 20 5 90 0 1.5 3 1150 100 20 3 90 20 0.8 4 1250 100 10 5 90 20 0.4 5 1150 400 10 5 40 20 0.1 6 1250 400 10 3 40 20 0.1 7 1250 100 20 5 40 20 0.8 8 1150 100 10 3 90 20 0.6 9 1150 400 10 3 40 0 0.1 10 1250 400 10 5 90 0 0.1 11 1250 100 20 3 90 0 2.2 12 1150 100 10 5 90 0 0.6 13 1250 400 20 5 40 0 0.8 14 1150 100 20 5 40 0 0.6 15 1150 100 20 3 40 20 0.7 16 1250 400 10 3 90 20 0.1
81 Table 4.3 Center point runs and response values for 26-2 DOE Run Temperature Pressure H2 C3H8 SiH4 HCl Sample ID 17 0.9 18 0.7 19 0.8 20 1200 250 15 4 65 10 0.8 An analysis of variance (ANOVA) was performed afte r the experimental matrix was completed and the response variable was measure d. The ANOVA results indicated the process pressure and the molar fractions of hyd rogen and propane were among the most significant factors affecting the epitaxial gr owth rate. In addition, the interaction between the process pressure and the propane mole f raction as well as the process pressure interaction with the hydrogen and propane molar fractions played major roles. The ANOVA analysis elucidated the empirical model t hat describes the process growth rate as shown in Equation 4.1. In this equation P s tands for pressure, and FC3H8, FH2 stands for the flow of the respective specie. The m odel was determined in terms of flows instead of molar fractions in order to facilitate i ts use during experimentation since this are the typical parameters used. 2 8 3 8 3 234.0 25.0 31.0 23.0 61.0H H C H C HF PF F PF P GR+ + + = Equation 4.1 After the empirical model from the ANOVA analysis was obtained, it was then used to manipulate the process parameters with the aim to obtain the combination of variables that resulted in the highest deposition r ate and a specular surface morphology. It was determined that the maximum predicted growth ra te that could be attained for the process parameter space considered was 2 m/h. Figure 4.1 describes the growth process conducted to achieve that deposition rate target va lue. Once the carbonization process was performed the sample temperature was raised fro m 1170 C to the growth temperature of 1250C at a rate of 15 C per minute After the growth temperature was attained the process pressure was lowered to 100 To rr and the total gas flow (H2, C3H8, SiH4 and Ar) was set to 20,093 sccm. The C3H8 and SiH4 molar fractions at the growth stage were 1.5 x 10-4 and 4.5 x 10-4, respectively. At these process conditions the pro cess Si/C ratio was 1.0. It should be noted that the mod el predicted the highest deposition rate
82 occurred when HCl was not present in the precursor chemistry, confirming that HCl would not have a significant impact on the depositi on rate. Figure 4.1 3C-SiC growth process schedule for optim um process predicted by ANOVA analysis. Mole fractions presented for 2 m/h process. At the growth stage the total input gas flow (H2, C3H8, SiH4 and Ar) was set to 20,093 sccm. Process Si/C=1.0. 4.2.3 Growth rate dependence on HCl mole fraction The ANOVA findings suggested that no significant i ncrease in deposition rates will be obtained at the low temperature regime due to HCl not having a significant effect on this response variable. However, it was decided to perform a study on the growth rate dependence on HCl mole fractions. This was in order to investigate the effect of HCl on the film quality and surface morphology which in th is study was hypothesized to be enhanced by the presence of HCl. It was determined to utilize the growth process de scribed in section 4.2.2 as the starting point for the HCl additive experiments as this process was well documented. Figure 4.2 illustrates the growth rate dependence o n HCl mole fraction. For these experiments the HCl mole fraction was increased fro m 0 to 1 x 10-4, while the process temperature, pressure and Si/C ratio were kept cons tant at 1250C, 100 Torr and 1.0, respectively. As can be observed, the deposition ra te remained fairly constant up to mole
83 fractions of approximately 0.5 x 10-4 after which the rate began to decrease. The study was concluded at the HCl mole fraction of 1 x 10-4 where the epitaxial surface morphology became rough and non specular. This was attributed to preferential surface etching as a consequence of the high HCl mole fract ions. Figure 4.2 Growth rate dependence on HCl mole fract ion. The process temperature, pressure and Si/C ratio were kept constant at 1250 C, 100 Torr and 1.0; respectively. The epitaxial layers were characterized via SEM. P lan view images of representative films grown using 0, 0.25 x 10-4 and 0.75 x 10-4 HCl mole fractions to the process described in section 4.2.2 are shown in Fig ure 4.3. As can be observed, the process with no HCl additive yielded a more mosaic like structure typically observed for 3C-SiC layers due to the presence of antiphase doma in boundaries. However, as the HCl mole fraction was increased, the surface morphology became smoother. This result was confirmed when the same samples we re inspected via AFM as shown in Figure 4.4. The surface roughness of the f ilms decreased from 23.9 to 2.3 nm RMS with increased HCl mole fraction. It could also be observed that the HCl additive lead to larger and better oriented antiphase domain boundaries. In addition, the epitaxial layer thickness decreased with increased HCl mole f raction, further confirming that surface etching was possibly the predominant mechan ism that leads to improved surface
84 morphology. Finally, the experimental data suggeste d that adding HCl to the growth process shown in Figure 4.1 at a mole fraction of 0 .75 x 10-4 will result in the optimum surface morphology. At these conditions the process Si/Cl ratio was determined to be 6.0 and the epitaxial growth rate was calculated to be ~ 1.7 m/h. (a) (b) (c) Figure 4.3 Plan view SEM images for representative films grown with HCl addition at mole fractions (a) 0, (b) 0.25 x 10-4 and (c) 0.75 x 10-4. Note the surface morphology improvement as the HCl mole fractions are increased (a) (b) (c) Figure 4.4 AFM micrographs taken in contact mode fo r representative films grown with HCl addition at mole fractions of (a) 0, (b) 0.25 x 10-4 and (c) 0.75 x 10-4. Note the surface morphology improvement as the HCl mole frac tions are increased 4.2.4 Growth rate as a function of temperature After determining the optimum HCl additive process a study of growth rate dependence on temperature was conducted since lower temperatures were desired as explained earlier. The growth parameters for the te mperature study were maintained at
85 P=100 Torr, Si/C=1.0 and Si/Cl=6.0 while the proces s temperature was varied from 1000C to 1250C as illustrated in Figure 4.5. In t his temperature range process growth rates from 0.5 to 1.7 m/h respectively were obtained. At temperatures low er than 1150C polycrystalline films were obtained. Specula r surface morphology was obtained for all films grown at temperatures of 1150C and a bove. Figure 4.5 Growth rate dependence on temperature. S ingle crystalline films obtained from temperatures of 1150C. 4.2.5 Growth rate as a function of silane mole fraction In an effort to increase the deposition growth rat e over the lower temperature regime, experiments were conducted where the silane mole fraction was increased while keeping a constant temperature and pressure of 1250 C. This temperature was chosen to perform this study since at lower temperatures poly crystalline films were obtained at increased mole fractions. The process Si/C and Si/C l ratios were also kept constant at 1.0 and 6.0, respectively. The silane mole fraction was varied from 4.3 x 10-4 to 7.8 x 10-4 as shown in Figure 4.6. Polycrystalline films were obt ained for silane mole fractions from 6.7 x 10-4 and above indicating that the silane saturation po int had been reached at this process temperature. Films with specular surface mo rphology were obtained at silane mole fractions lower than 6.7 x 10-4. The highest growth rate achieved in this study wa s
86 2.5 m/h. This was achieved for a SiH4 mole fraction of 5.3 x 10-4. This study is in agreement with the ANOVA results which indicated th at the deposition rate would not be significantly improved by the presence of HCl. Figure 4.6 Growth rate dependence in SiH4 mole fraction. Pollycristalline films obtained at mole fractions from 6.7 x 10-4. In order to evaluate the quality of the films grow n at the highest deposition rate achieved for the low temperature growth, a 2 m thick film grown at a rate of 2.5 m/h was deposited and analyzed via XRD. Unfortunately, thicker films could not be grown due to malfunctions of the HCl manifold mentioned e arlier. Figure 4.7 shows the XRD rocking curve of the (200) 3C-SiC diffraction peak. The FWHM was determined to be ~ 278 nm arcsec. This value compares to values report ed in literature and it is also comparable with films grown in this study at elevat ed temperatures which indicates good quality film.
87 Figure 4.7 XRD rocking curve of a 2 m thick epitaxial layer grown at a rate of 2.5 m/h The FWHM was ~278 arcsec and compares with values r eported elsewhere.44 4.3 Summary The development of a 3C-SiC growth process at low deposition temperatures has been completed. The highest growth rate achieved wa s 2.5 m/h which was obtained at a temperature of 1250C. HCl proved to be highly beneficial to the process surface morphology. However, the epitaxial layer growth rate did not significant ly improve via the HCl process suggesting that etching of the film surface dominat es in the lower temperature regime.
88 Chapter 5: Summary and Future Work 5.1 Dissertation summary This dissertation research focused on the heteroep itaxial growth of 3C-SiC layers by CVD at two deposition temperature regimes using a chloride additive to the SiH4C3H8-H2 precursor chemistry system. The hypothesis under i nvestigation was to determine if chloride based chemistry will aid to i ncrease the epitaxial layers growth rate and material quality via reduced defects. Character ization of the films was carried out via Nomarski interference optical microscopy, FTIR, SEM AFM, XRD and XPS. Thermodynamic equilibrium calculations were perfor med to obtain a criterion as to which chloride specie should be used during grow th experimentation. During this study the following chloride containing groups of s pecies were considered: chlorocarbons, chlorosilanes, and hydrogen chloride It was concluded from this work that no differences in most dominant species presen t on the equilibrium composition mixture were observed between adding HCl or using c hlorocarbons or chlorosilane as the chloride additive source. Therefore, HCl was chosen since it has been suggested that the addition of HCl allows for the enhancement of both the growth rate and surface morphology. This is believed to be accomplished by the creation of intermediate species such as SixCly, which increases the silicon atomic content in the gas mixture. Work performed, particularly on the homo-epitaxial growt h of 4H-SiC, revealed that HCl addition to the standard chemistry increased the fi lm growth rate by allowing higher silane process mole fractions otherwise not possibl e with the standard chemistry.66 CVD CFD simulations were also performed to determi ne the gas velocity and temperature profiles along the reactor. In addition they aid in the study of the changes in flow and temperature profiles due to reactor geomet ry changes. The simulation predicted that gas surrounding the susceptor area was at a co nstant temperature. However, the
89 temperature profile did not agree with the experime ntally measure temperature profile along the susceptor which showed a 35C difference along the flow direction for Geometry I and 5C for Geometry II. This is mainly attributed to the fact that the simulation does not account for heat losses by radi ation. The velocity profile revealed that the reactor is operating at the desired lamina r flow conditions. Laminar flow is desired in order to deliver the gas uniformly to th e substrate in order to obtain uniform films and avoid intermixing of gas concentrations56 Once it was decided which chloride specie was to b e used to test the hypothesis, then experiments were performed. A 3C-SiC growth pr ocess was developed on Si (001) substrates at high temperatures (1385C). The proce ss development was performed in two stages. First, a baseline process without HCl add itive was developed. Second, the baseline process was used as the starting process t o develop the HCl additive experiments. The growth process was achieved by per forming a two-step growth consisting of carbonization of the silicon substrat e and growth of the epitaxial layer. The baseline process was developed at a temperatur e of 1385C and a pressure of 100 Torr. The process Si/C was set to be 0.72. At t his process conditions a maximum growth rate of 12 m/h was achieved. Surface roughness values of 1.6 n m RMS were obtained in films the grown. The resulting layers p resented specular surface morphology and SEM analysis revealed that no voids were presen t at the 3C-SiC/Si interface. The XRD rocking curve of the 3C-SiC (002) peak revealed a FWHM of 500 arcsec which compares to values reported elsewhere.45 After corroborating that the baseline process was repeatable, the process was then used as the starting point for the HCl additive exp eriments. The HCl addition to the baseline process allowed growth rates up to ~38 m/h to be achieved. This is the highest reported value reported in the literature to date f or 3C-SiC heteroepitaxy. During the development of this process film degradation was no ted at increased film thickness and at increased growth speeds for samples of the same thi ckness. The film degradation was attributed to the formation of protrusion or hilloc ks at the film surface. Despite much effort to optimize the growth process the surface d egradation could not be solved. The higher film thickness obtained in this work was 30 m. XDR characterization performed
90 showed FWHM values from 220 to 1160 arcsec dependin g on the process growth speed or film thickness. These values are better or compa rable to those reported in literature and to those obtained for the baseline process.45 Finally, it was concluded from this study that at high deposition temperatures the addition of HCl to the precursor chemistry impacted more significantly the epitaxial layers growth rat e. After finishing the high temperature experiments f ocus was shifted to test the hypothesis for the low-temperature (1000-1250C) gr owth of 3C-SiC. In addition a low temperature process could potentially be beneficial for device process applications. The growth process was also performed using a two step growth: carbonization of the silicon substrate followed by the deposition of the epitaxi al layer. To develop this process a 26-2 fractional factorial DOE was carried out. The facto rs considered were the process temperature, pressure and the propane, silane, hydr ogen and hydrogen chloride mole fractions. The response under consideration was the process growth rate. The experimental results were analyzed by performing an d ANOVA analysis. The ANOVA analysis suggested that the main factors controllin g the 3C-SiC growth at the lower temperature regime were the process pressure and th e molar fractions of hydrogen and propane. In addition the interaction between the pr ocess pressure and the propane mole fraction as well as the process pressure interactio n with the hydrogen and propane molar fractions. An empirical model was developed from th e ANOVA analysis to predict the epitaxial growth rates. This empirical model predic ted that the highest growth rate that could be obtained within the parameter space under consideration was 2 m/h; this was confirmed experimentally. The process parameters to achieve this growth rate where T= 1250C, P=100 Torr and a Si/C=1.0. Although the ana lysis suggested that HCl would not impact significantly the process growth rate; HCl e xperiments were carried out using the process described as the starting point. With the H Cl additive the epitaxial growth rate could only be increased to ~2.5 m/h. The effect of HCl addition to the surface morphology was studied. It was observed that smooth er and flatter surfaces were obtained at increased HCl mole fractions. AFM measu rement revealed that the surface roughness was 10 times smaller for the optimum HCl additive process compared to the process without HCl additive predicted by the ANOVA model. An XRD FWHM of 278
91 arcsec was measured on a 2 m thick representative layer. This value is compara ble to the best values reported in the literature as well as to films grown at higher deposition temperatures in this study. It was concluded from t his work that at lower deposition temperatures the HCl addition was more beneficial f or the film quality by enhancing surface morphology via surface etching. 5.2 Future work and current work 5.2.1 Species concentration profile simulation Unfortunately, a final numerical solution for the species concentration profile along the reactor has not been established in this work due to problems with the COMSOL platform. The work performed to date allows for the transient solution of the gas phase reaction kinetics considered in this work The species concentrations as a function of time are presented in the next set of f igures. The calculations are based on a perfectly mixed batch reactor kept at 1385C which is the process temperature of the 3CSiC HCl additive growth process at high temperature s develop in Chapter 3. For visualization purposes the species have been graphe d into four different sets: hydrocarbon species in Figure 5.1, silicon containi ng species in Figure 5.2, chlorocarbon species in Figure 5.3 and chlorosilane species in F igure 5.4. After the COMSOL platform issues are resolved and a numerical solution for the species concentration profile and the process growt h rate for CVD reactor could be obtained, the 3C-SiC growth processes developed in this work will be simulated. In addition, the 3C-SiC heteroepitaxial processes on S i (111) surfaces currently being developed by C. Locke and C. Frewin from our group will be simulated. Finally, additional experimentation has to be performed if n eeded to validate the model for the process growth rate prediction.
92 Figure 5.1 Calculated species concentration as a fu nction of time for hydrocarbon species based on a perfectly mixed reactor kept at 1385C. Figure 5.2 Calculated species concentration as a fu nction of time for silicon containing species based on a perfectly mixed reactor kept at 1385C
93 Figure 5.3 Calculated species concentration as a fu nction of time for chlorocarbon species based on a perfectly mixed reactor kept at 1385C. (b) Figure 5.4 Calculated species concentration as a fu nction of time for chlorosilane species based on a perfectly mixed reactor kept at 1385C.
94 5.2.2 Temperature profile simulation The design of the reactor chamber is crucial for m aintaining a uniform sample temperature profile and efficient heating. It is es pecially important that the quartz tube surrounding the susceptor remains at moderate tempe ratures. In this work a temperature profile of the gas system along the susceptor area failed to predict the temperature gradients observed experimentally. Therefore, a mor e in depth simulation that examines the RF inducted heating of the graphite susceptor a nd takes into account heat losses by radiation could provide a better estimate of the te mperature distribution over the sample area.
95 References  J. B. Casady and R. W. Johnson, Solid-State Electro nics 39, 1409-142 (1996).  M. Mehregany, C. A. Zorman, N. Rajan, and C. H. Wu, Proceedings of the IEEE 86, 1594-1610 (1998).  A. Gupta and C. Jacob, Progress in Crystal Growth a nd Characterization of Materials 51, 43-69 (2005).  Y. Goldberg, M. E. Levinshtein, and S. L. Rumbayant sev, in Properties of Advance Semiconductor Materials, edited by M. E. Levinshtein, S. L. Rumyantsev, and M. S. Shur (John Wiley & Sons, New York, NY, 2001), p. 93.  ACD/ChemSketch; 10.0 ed. (1997-2006).  S. E. Saddow and A. Agarwal, Advances in Silicon carbide: Processing and Applications (Artech House INC., Norwood, MA 2004).  C.-M. Zetterling and M. Ostling, in Process Technology for Silicon Carbide, 2 ed., edited by C.-M. Zetterling (INSPEC, London U.K., 2002).  A. R. Powell and L. B. Rowland, Proceedings of the IEEE 90, 942-955 (2002).  M. Abe, H. Nagasawa, P. Ericsson, H. Strmberg, M. Bakowski, and A. Schner, Microelectronic Engineering 83, 24-26 (2006).  H. Nagasawa, K. Yagi, T. Kawahara, N. Hatta, and M. Abe, Microelectronic Engineering 83, 185-188 (2006).  J. Wan, M. A. Capano, M. R. Melloch, and J. A. Coop er, IEEE Electron Device Letters 23, 482-484 (2002).  http://www.mdatechnology.net/techsearch.asp?article id=174#sec6
96  O. Kordina and S. E. Saddow, in Advances in Silicon Carbide Processing and Applications edited by S. E. Saddow and A. Agarwal (Artech House INC, Norwood, MA, 2004), p. 1-27.  P. G. Neudeck, R. S. Okojie, and L.-Y. Chen, Procee dings of the IEEE 90, 1065-1076 (2002).  SiC technology will meet the military's future need (2007).  Silicon carbide set to reduce the size of hybrid el ectricengines (2005).  http://www.semiconductorglossary.com  C. I. Harris, S. Savage, A. Konstantinov, M. Bakows ki, and P. Ericson, Applied Surface Science 184, 393-398 (2001).  J. Hornberger, A. B. Lostetter, K. J. Olejniczak, T McNutt, S. M. Lal, and A. Mantooth, IEEE Aerospace Conference Proceedings 4, 2534-2555 (2004).  Y. T. Yang, K. L. Ekinci, X. M. H. Huang, L. M. Sch iavone, M. L. Roukes, C. A. Zorman, and M. Mehregany, Applied Phy sics Letters 78, 162-164 (2001).  S. E. Saddow, M. E. Okhuysen, M. S. Mazzola, M. Dud ley, X. R. Huang, W. Huang, and M. Shamsuzzoha, in Characterization of Single-Crystal 3C-SiC on Si Epitaxial Layers, Boston, MA, 1999, p. 107-112.  N. Nordell, in Process Technology for Silicon Carbide, 2 ed., edited by C.M. Zetterling (INSPEC, London, U.K., 2002), p. 13-5 0.  S. Nishino, Y. Hazuki, H. Matsunami, and T. Tanaka, Journal of the Electrochemical Society 127, 2674-2680 (1980).  J. A. Powell, L. G. Matus, and M. A. Kuczmarski, Jo urnal of the Electrochemical Society 134, 1558-1565 (1987).  K. Yagi and H. Nagasawa, Journal of Crystal Growth 174, 653-657 (1997).  N. Nagasawa and Y.Yagi, Phys. Stat. Sol. B 202, 335-358 (1997).
97  S. I. Molina, F. M. Morales, and D. Arajo, Materia ls Science and Engineering B80, 342-344 (2001).  C.-F. Wang and D.-S. Tsai, Materials Chemistry and Physics 63, 196-2001 (2000).  O. Kordina, L.-O. Bjrketun, A. Henry, C. Hallin, R C. Glass, L. Hultman, J.-E. Sundgreen, and E. Janzn, J. Cryst. Growth 154, 303-314 (1995).  M. Reyes, M. Waits, S. Harvey, Y. Shishkin, B. Geil J. T. Wolan, and S. E. Saddow, Materials Science Forum 527-529, 307-310 (2006).  G. Ferro, J. Camasel, S. Juillaguet, C. Ballaud, E. K. Polychroniadis, Y. Stoemenos, J. Dazord, H. Peyre, Y. Monteil, S. A. R ushworth, and L. M. Smith, Semiconductor Science Technology 18, 1015-1023 (2003).  J.-H. Boo, K.-S. Yu, M. Lee, and Y. Kim, Applied Ph ysics Letters 66, 3486-3488 (1995).  H. Nakasawa and M. Suemitsu, Applied Physics Letter s 79, 755-757 (2001).  A. Figueras, S. Garelik, R. Rodrguez-Clemente, B. Armas, C. Combescure, and C. Dupuy, Journal of Crystal Growth 110, 528-542 (1991).  J. Hofmann and S. Veprek, Thin Solid Films 318, 18-21 (1998).  D. J. Larkin, P. G. Neudeck, J. A. Powell, and L. G Matus, Applied Physics Letters 65 (1994).  P. A. Coon, P. Gupta, M. L. Wise, and S. M. George, Journal of Vacuum Science Technology A 10, 324-333 (1992).  V. S. Ban and S. L. Gilbert, Journal of the Electro chemical Society 122, 1382-1388 (1975).  Private communication.  R. L. Myers, O. Kordina, Z. Shishkin, S. Rao, R. Ev erly, and S. E. Saddow, Material Science Forum 483-485 73-76 (2005)
98  F. L. Via, G. Galvagno, A. Firrincieli, F. Roccafor te, S. D. Franco, A. Ruggiero, M. Barbera, R. Reitano, P. Musumeci, L. C alcagno, G. Foti, M. Mauceri, S. Leone, G. Pistone, F. Portuese, G. Abbo ndanza, G. Abagnale, G. L. Valente, and D. Crippa, Material Science Foru m 527-529 163-166 (2006).  K. K. Schuegraf, Hanbook of Thin Film Deposition Process and Techniques: Principles, Methods, Equipment and Appl ications (Noyes Publications, Park Ride New Jersey, 1988).  Y. Gao, J. H. Edgar, J. Chaudhuri, S. N. Cheema, M. V. Sidorov, and D. N. Braski, J. Cryst. Growth 191, 439-445 (1998).  M. Reyes, Y. Shishkin, S. Harvey, and S. E. Saddow, Materials Research Society Spring Meeting Proceedings 911, 79 (2006).  T. Chassagne, A. Leycuras, C. Balloud, P. Arcade, H Peyre, and S. Juillaguet, Mat. Sci. Forum 457-460, 273 (2004).  D. L. Smith, Thin-film deposition: principles and practice (McGraw Hill, Inc, 1995).  http://composite.about.com/library/glossary/d/bldef -d1722.htm  http://www.sp.phy.cam.ac.uk/~SiGe/Misfit.html  S. Mahajan and K. S. S. Harsha, Principles of growth and processing of semiconductors (McGraw Hill, 1999).  http://www.tf.uni-kiel.de/matwis/amat/def_en/index. html  J. Yun, T. Takahaski, T. Mitani, Y. Ishida, and H. Okumura, Journal of Crystal Growth 291, 148-153 (2006).  J. Yun, T. Takahashi, Y. Ishida, and H. Okumura, Jo urnal of Crystal Growth 291, 140-147 (2006).  X. H. Zeng, B. Qu, Y. T. Wang, Z. Z. Dai, J. Y. Han H. Yang, and J. W. Liang, Journal of Crystal Growth 233, 40-44 (2001).  http://en.wikipedia.org/
99  K. Seshan, Handbook of thin film deposition processes and tech niques: Principles methods and Applications (Noyes Publications, Norwich, New York, 2002).  A. Sherman, Chemical Vapor Deposition for Microelectronics: Pri nciples Technology and Applications (Noyes Publications, Westwood, N. J. USA, 1987).  R. L. Myers, Thesis, University of South Florida, 2 003.  M. T. Smith, Thesis, University of South Florida, 2 003.  S. Harvey, Thesis, University of South Florida, 200 6.  J. M. Smith, H. C. Van Ness, and M. M. Abbott, Introduction to Chemical Engineering Thermodynamics, 6th ed. (McGraw Hill, New York, NY, 2001).  S. Gordon and B. J. McBride, NASA Reference Publica tions 1311, 1-58 (1994).  M. J. McBride and S. Gordon, Nasa Reference Publica tions 1311, 1-184 (1996).  CEAgui; 2 ed. (1996).  R. L. Myers, Y. Shishkin, O. Kordina, and S. E Saddow, Journal of Crystal Growth 285, 486-490 (2005).  K. L. Choy, Prog. in Mat. Sci. 48, 57-170 (2003).  H. O. Pierson, Handbook of Chemical Vapor Deposition: Principles, Technology and Applications (Noyes Publications, Norwich, New York, U.S.A., 1999).  . Danielsson, A. Henry, and E. Janzn, Journal of Crystal Growth 243, 170-184 (2002).  G. M. Petrov and J. L. Giuliani, Journal of Applied Physics 90, 619-636 (2001).
100  M. E. Coltrin, R. J. Kee, and J. A. Miller, Journal of Electrochemical Society 133, 1206-1213 (1986).  J. Meziere, M. Ucar, E. Blanket, M. Pons, P. Ferret and L. D. Cioccio, J.Cryst. Growth 267, 436-451 (2004).  S. P. Walch and C. E. Dateo, J. Phys.Chem A 105, 2015-2022 (2001).  Y.-p. Wu, Combustion and Flame 122, 312-326 (2000).  G. Valente, C. Cavallotti, M. Masi, and S. Carra, J ournal of Crystal Growth 230, 247-257 (2001).  A. Veneroni, F. Ormarini, and M. Masi, Cryst. Res. Technol. 40, 967-971 (2005).  S. P. Walch and C. E. Dateo, J. Phys.Chem A 106, 2931-2934 (2002).  J. M. Wittbrodt and H. B. Schlegel, Chemical Physic s Letters 265, 527531 (1997).  J. M. Wittbrodt and H. B. Schlegel, Chemical Physic s Letters 269, 391 (1997).  S. F. Nitodas and S. V. Sotirchos, Journal of the E lectrochemical Society 149, C112-C119 (2002).  Y.-p. G. Wu, Fuel 83, 2237-2248 (2004).  R. I. Masel, Chemical Kinetics and Catalysis (John Wiley and Sons, Inc, New York, NY, 2001).  A. Veneroni, F. Ormarini, D. Moscatelli, M. Masi, S Leone, M. Mauceri, G. Pistone, and G. Abbondanza, J.Cryst. Growth 275, e295-e300 (2005).  A. Veneroni and M. Masi, Chemical Vapor Deposition 12, 562-568 (2006).  R. B. Bird, W. E. Stuart, and E. N. Lightfoot, Transport phenomena, 2nd ed. (John Wiley and Sons, Inc., New York, NY, 2002)
101  COMSOL AB, COMSOL Multiphysics Reference Manual; 3 .4 ed. (2007).  COMSOL AB, COMSOL Multiphysics Reference Manual; 3 .4 ed. (2007).  H. Nagasawa and Y.-I. Yamaguchi, Journal of Crystal Growth 115, 612616 (1991).  M. J. Bozack, Physica Status Solidi B 2002, 549-580 (1997).  D. C. Montgomery, Design and Analysis of Experiments, 6 ed. (John Wiley & Sons, Inc, 2005).  The MathWorks Inc., MATLAB; 22.214.171.124 (R2006a) ed. (2006).
103 Appendix A Reactions for the gas phase model The gas phase model used in this work utilized a s et of elementary reactions describing the decomposition of the precursor gases (C3H8, SiH4, HCl and H2) and the reactions between the products of all decomposition reactions. The following sections list the reactions considered. The rate constants are wr itten in the Arrhenius form, k=ATeEa/RT. The units of A depend on the reaction order and a re given in terms of cm3, moles and seconds. A.1 Hydrogen decomposition reactions No. Bimolecular Reactions A E (K) Ref. 68 H2 + H2 => 2 H + H2 1.5e-9 0 48350 68 H2 + H => 3 H 3.7e-10 0 48350 Ref. Trimolecular Reactions A E (K) 68 H + H + H2 => H2 + H2 2.7e-31 -0.6 0 68 H + H + H => H2 + H 2.7e-30 -1.0 0 A.2 Propane decomposition reactions No. Unimolecular Reactions A E (K) Ref. 1 C3H8 => C2H5 + CH3 2.3e22 -1.8 44637 68 2 C2H5 => C2H4 + H 1.4e8 1.19 18722 68 3 C2H4 => C2H2 + H2 1.4e12 0.44 44670 68 4 C2H2 => C2H + H 1.8e15 0 62445 68 No. Bimolecular Reactions A E (K) Ref. 5 C2H5 + H => 2 CH3 6.1e-11 0 0 68 6 Reverse 1.3e-9 0 13375 68 7 C2H5 + H => C2H4 + H2 3e-12 0 0 68 8 Reverse 1.1e-11 0 34300 68 9 C2H5 + H => C2H6 6e-11 0 0 68 10 C2H5 + H2 => C2H6 + H 5.1e-24 3.6 4253 68 11 Reverse 2.4e-15 1.5 3730 68 12 C2H5 + CH3 => C2H4 + CH4 3.3e-11 0.5 0 68 13 C2H5 + CH3 => C3H8 (Reverse R1) 7.4e-11 0 0 68 14 C2H5 + CH4 => C2H6 + CH3 1.4e-22 1.14 6322 68 15 Reverse 2.5e-31 6.0 3730 68 16 C2H5 + C2H => C2H2 + C2H4 3e-12 0 0 68
104 Appendix A: (continued) No. Bimolecular Reactions A E (K) Ref. 17 C2H5 + C2H2 => C2H + C2H6 4.5e-13 0 11800 68 18 Reverse 6e-12 0 0 68 19 C2H5 + C2H4 => C2H6 + C2H3 1.1e-21 3.13 9063 68 20 Reverse 1.0e-21 3.3 5285 68 21 C2H5 + C2H3 => C2H2 + C2H6 8e-13 0 0 68 22 C2H5 + C2H3 => 2C2H4 8e-13 0 0 68 23 Reverse 8e-10 0 36000 68 24 C2H5 + C2H5 => C2H6 + C2H4 1.2e-11 0 540 68 25 C2H4 + H => C2H5 (Reverse R2) 1.4e-15 1.49 499 68 26 C2H4 + H => C2H3 + H2 2.2e-18 2.53 6160 68 27 C2H4 + CH3 => C2H3 + CH4 1.1e-23 3.7 4780 68 No. Bimolecular Reactions A E (K) Ref. 28 Reverse 2.4e-24 4.02 2754 68 29 C2H4 + C2H2 => 2C2H3 4e-11 0 34400 68 30 C2H3 + H => C2H2 + H2 1.6e-10 0 0 68 31 Reverse 4e-12 0 32700 68 32 C2H3 + H2 => C4H4 + H 5e-20 2.63 4298 68 33 C2H3 + CH2 => C2H2 + CH3 3e-11 0 0 68 34 C2H3 + CH3 => C2H2 + CH4 6.5e-13 0 0 68 35 C2H3 + C2H => 2 C2H2 1.6e-12 0 0 68 36 Reverse 1.6e-11 0 42500 68 37 C2H2 + H => C2H + H2 1e-10 0 11200 68 38 Reverse 1.9e-11 0 1450 68 39 C2H2 + H2 => C2H4 (Reverse R3) 5e-13 0 19600 68 40 C2H2 + CH3 => CH4 + C2H 3e-13 0 8700 68 41 Reverse 3e-12 0 250 68 42 C2H + H => C2H2 (Reverse R4) 3e-10 0 0 68 43 C2H + CH2 => C2H2 + CH 3e-11 0 0 68 44 CH4 + H => CH3 + H2 2.2e-20 3.0 4045 68 45 Reverse 4.8e-22 3.12 4384 68 46 CH4 + CH => C2H5 2.7e-10 0 0 68 47 CH4 + CH => C2H4 + H 5e-11 0 200 68 48 CH4 + CH2 => C2H6 1.7e-11 0 0 68 49 CH4 + CH2 => 2CH3 2.1e-11 0.5 0 68 50 CH4 + CH3 => C2H5 + H2 1.7e-11 0 11500 68 51 CH3 + H => CH2 + H2 3.3e-11 0 0 68 52 Reverse 3.3e-11 0.5 0 68 53 CH3 + H => CH4 2e-9 -0.4 0 68 54 CH3 + CH => C2H3 + H 5e-11 0 0 68 55 CH3 + CH2 => C2H4 + H 3e-11 0 0 68
105 Appendix A: (continued) 56 CH3 + CH3 => C2H6 1.7e-9 -0.64 0 68 No. Bimolecular Reactions A E (K) Ref. 57 CH2 + H => CH + H2 5e-11 0 0 68 58 CH2 + CH => C2H2 + H 6.6e-11 0 0 68 59 CH2 + CH2 => C2H4 1.7e-12 0 0 68 60 CH2 + CH2 => C2H2 + 2 H 1.8e-10 0 0 68 61 CH2 + CH2 => C2H2 + H2 2e-11 0 400 68 62 CH + CH => C2H2 2e-10 0 0 68 No. Unimolecular Reactions A E (K) Ref. 63 CH4 => CH3 + H 8.3e13 0 52246 68 64 C2H6 => 2 CH3 1.2e22 -1.79 45834 68 A.3 Silane decomposition reactions No. Unimolecular Reactions A E(K) Ref. 1 SiH4 => SiH2 + H2 3.120e9 1.7 27550 67 No. Unimolecular Reactions A (1/s) E(K) Ref. 2 SiH4 => SiH3 + H 3.690e15 0 46830 67 3 Reverse 1.323e14 0 140 67 No. Bimolecular Reactions A E (K) Ref. 4 SiH2 + H => SiH3 3.810e13 0 1000 67 5 SiH2 + H => SiH + H2 1.204e13 0 0 67 6 SiH2 + SiH2 => Si2H2 + H2 6.5e14 0 0 67 7 SiH3 + H => SiH2 + H2 1.204e13 0 0 67 8 SiH3 + SiH2 => Si2H5 6.580e12 0 1000 67 9 SiH3 + SiH3 => SiH4 + SiH2 1.8e13 0 0 67 10 SiH + H2 => SiH3 3.45e13 0 1000 67 11 Si2H5 + SiH3 => SiH2 + Si2H6 9.033e13 0 0 67 12 Si2H6 + H => Si2H5 + H2 1.445e14 0 1250 67 13 Si2H6 + H => SiH3 + SiH4 1.445e14 0 1250 67 14 Si2H6 + SiH3 => SiH4 + Si2H5 2.409e14 0 2500 67 15 SiH4 + H => SiH3 + H2 1.686e13 0 1250 67 16 SiH4 + SiH => SiH3 + SiH2 1.38e12 0 5640 67 17 SiH4 + SiH => Si2H4 + H 3e14 0 4535 67 18 SiH4 + SiH => Si2H5 4.139e14 0 0 67 19 Si2H4 + H2 => SiH2 + SiH4 9.41e13 0 0 67 20 Reverse 9.43e10 1.1 2916 67 21 Si2H4 + SiH4 => SiH2 + Si2H6 1.73e14 0.4 0 67 22 Reverse 2.65e15 0.1 4267 67
106 Appendix A: (continued) No. Unimolecular Reactions A E (K) Ref. 23 Si2H4 => Si2H2 + H2 3.16e14 0 26690 67 24 Reverse 2.450e14 0 1000 67 25 Si2H4 => Si + SiH4 1.420e13 0.54 28980 67 26 Si2H6 => SiH2 + SiH4 1.810e10 1.7 27280 67 27 Si2H6 =>Si2H4 + H2 9.090e9 1.8 27290 67 No. Bimolecular Reactions A E(K) Ref. 28 SiH2 + Si => Si2 + H2 1.500e14 0 0 67 29 SiH2 + Si => Si2H2 7.240e12 0 1000 67 30 SiH4 + Si => 2 SiH2 9.310e12 0 1000 67 31 Si2H6 + Si => SiH2 + Si2H4 1.300e15 0 6345 67 32 Si2 + H => Si + SiH 5.150e13 0 2670 67 33 Si2 + H2 => 2 SiH 1.540e13 0 20140 67 34 Si2 + H2 => Si2H2 1.540e13 0 1000 67 35 Si2 +SiH2 => Si3 + H2 3.550e11 0 1000 67 36 Bimolecular Reactions A E(K) Ref. 37 Si2 + Si => Si3 2.060e12 0 12135 67 38 Si3 + Si => 2 Si2 2.060e12 0 12135 67 No. Bimolecular Reactions A (cm3/s) E(K) Ref. 39 Si3 + H2 => Si + Si2H2 9.790e12 0 23770 67 40 Si3 + SiH2 => Si2 + Si2H2 1.430e11 0 8160 67 No. Unimolecular Reactions A E (K) Ref. 41 Si2 => 2 Si 1e15 0 37460 67 A.4 Hydrogen chloride decomposition reactions No. Unimolecular Reactions A E (K) Ref. 1 HCl => H + Cl 4.365e13 0 41142 73 No. Bimolecular Reactions A E (K) Ref. 2 H + Cl => HCl (Reverse R1) 7.20e21 -2.0 0 72 3 Cl + H2 => HCl + H 4.786e13 0 2647.8 73 4 Cl + Cl => Cl2 2.34e14 0 -902.1 72 5 H + Cl2 => HCl + Cl 8.59e13 0 589.37 72
107 Appendix A: (continued) A.5 Organosilicon reactions No. Bimolecular Reactions A E (K) Ref. 1 SiH3 + CH3 => CH4 + SiH2 3.372e13 0 -360 67 2 SiH4 + CH3 => CH4 + SiH3 7.762e11 0 3515 67 No. Bimolecular Reactions A E (K) Ref. 3 SiH4 + C2H5 => C2H6 + SiH3 5.370e11 0 3650 67 4 Si2 + CH4 => Si2C + 2H2 3.011e15 0 10000 67 A.6 Chlorinated species reactions No. Bimolecular Reactions A E (K) Ref. 1 Si + HCl => SiCl + H 1.585e-9 0 6954.5 74 2 Si + HCl => SiHCl 4.169e11 0.5 0 74 3 SiCl + HCl => SiCl2 + H 1.072e-10 0 9814.8 74 4 SiCl + HCl => SiHCl2 4.193e-12 0 2372.6 75 5 SiCl + H2 => SiClH2 2.053e-11 0 31669 75 6 SiCl + H2 => SiHCl + H 6.681e-10 0 16406 75 7 SiCl + H2 => SiH + HCl 3.149e-11 0 15744 75 8 SiHCl + HCl => SiCl2 + H2 1.169e11 0.5 0 74 9 SiHCl + HCl => SiH2Cl2 2.455e-3 0 5137.1 77 10 SiHCl + H => SiCl + H2 3.647e-16 1.736 -609.3 75 11 SiHCl + H => SiH + HCl 1.404e-10 0 8044.3 75 12 SiH + HCl => SiH2Cl 1.106e-18 2.158 -1023 75 13 SiH + HCl => SiHCl + H 8.414e-11 0 8616 75 14 SiH + HCl => SiCl + H2 1.559e-18 1.984 859.6 75 15 SiH2Cl2 + HCl => SiHCl3 + H2 2.49e29 0 24957 78 16 SiCl2 + H => SiCl + HCl 4.068e-10 0 9498.4 75 17 SiCl2 + H2 => SiH2Cl2 2.291e-3 0 19737 77 18 SiHCl3 + H => SiCl3 + H2 2.455e12 0 1276.1 74 19 Reverse 1.4e10 0 472.8 78 20 SiHCl3 + CH3 =>CH4 + SiCl3 6.760e7 0 2162.6 78 21 SiHCl3 + Cl => HCl + SiCl3 7.230e9 0 0 78 No. Unimolecular Reactions A E (K) Ref. 22 SiH2Cl => SiH + HCl 1.869e14 0 30971 75 23 SiH2Cl => SiCl + H2 3.975e13 0 46776 75 24 SiHCl2 => SiCl + HCl 3.510e13 0 43047 75 25 SiH2Cl2 => SiCl2 + H2 7.943e13 0 37310 76 26 SiH2Cl2 => SiHCl + HCl 6.761e14 0 36564 76
108 Appendix A: (continued) No. Unimolecular Reactions A E (K) Ref. 27 SiHCl3 => SiCl2 + HCl 4.898e14 0 37106 74 No. Bimolecular Reactions A E (K) Ref. 28 CH4 + Cl => CH3 + HCl 5e13 0 1960.5 79 29 C2H2 + Cl => C2H + HCl 1.58e14 0 13531 79 30 C2H4 + Cl => C2H3 + HCl 1e14 0 29.3e3 79 31 C2H3 + Cl2 => C2H3Cl + Cl 5.25e12 0 -240.6 79 No. Bimolecular Reactions A E (K) Ref. 32 C2H5 + Cl2 => C2H5Cl + Cl 7.58e12 0 -120.3 79 33 C2H6 + Cl => C2H5 + HCl 4.64e13 0 84.2 79 34 C2H3Cl + CH3 => C2H3 + CH3Cl 3.15e11 0 12484 79 35 C2H3Cl + Cl => CHCHCl + HCl 5.00e13 0 3524.2 79 36 C2H3Cl + Cl => CH2CCl + HCl 3.00e13 0 2766.4 79 37 C2H3Cl + H => C2H3 + HCl 1.00e13 0 3271.6 79 38 C2H3Cl + H => CH2ClCH2 8.25e9 -0.1 1768.1 79 39 C2H3Cl + H =>C2H4 + Cl 2.92e13 -0.1 2970.9 79 40 CH3Cl + H => CH3 + HCl 3.72e13 0 3824.9 79 41 CH3Cl + Cl => CH2Cl + HCl 3.20e13 0 13000 79 42 CH3Cl + Cl => CH3 + Cl2 1.00e14 0 12581 79 43 CH3Cl + CH3 => CH4 + CH2Cl 3.31e11 0 4727 79 44 CH2Cl + H => CH3Cl 8.00e26 -5.05 1383.2 72 45 CH2Cl + H => CH3 + Cl 2.18e5 -0.24 108.3 72 46 CH2Cl + H2 => CH3Cl + H 1.79e12 0 6567.2 72 47 CH2Cl + CH3 => C2H5Cl 1.62e43 -9.89 3800.2 72 48 CH2Cl + CH3 => C2H5 + Cl 2.68e14 -0.57 1202.8 72 49 CH2Cl + CH3 => C2H4 + HCl 4.26e19 -2.02 1816.2 72 50 CH2Cl + CH2Cl => C2H3Cl + HCl 4.21e22 -3.02 1900.4 72 Unimolecular Reactions A E (K) Ref. 51 C2H3Cl => C2H2 + HCl 7.64e33 -6.3 36492 79 52 C2H3Cl => C2H3 + Cl 5.15e45 -10 50529 79 53 C2H5Cl => C2H4 + HCl 6.03e27 -4.5 31164 79 54 C2H5Cl => C2H5 + Cl 7.62e49 -11 46078 79 55 CH3Cl => CH3 + Cl 3.71e38 -7.61 45188 72
109 Appendix B Reactions for the surface reaction model The main surface reactions involved in the 3C-SiC deposition are listed in the following sections. Ssi and Sc stand for Si and C surface sites, while a species marked with an asterisk (*) indicates an adsorbed species. The rate constants are written in the form k = ATeEa/RT. The units of A are given in terms of [mol cm2 s1], gas and surface concentrations units are [mol cm3] and [mol cm2], respectively. Subscripts S and C are for chloride on Si or C surface sites respectivel y. B.1 Carbon species adsorption No. Surface Reactions A E(K) Ref. 1 CH4 + Ssi --> C* + 2H2 2.39 e09 0.5 0 2 CH3 + Ssi --> CH* + H2 8.51 e11 0.5 0 3 CH2 + Ssi --> C* + H2 8.91 e11 0.5 0 4 CH + Ssi --> CH* 9.12 e11 0.5 0 5 C2H5 + 2Ssi --> C* + CH* + 2H2 5.75 e20 0.5 0 6 C2H4 + 2Ssi --> 2C* + 2H2 9.33 e17 0.5 0 7 C2H3 + 2Ssi --> C* + CH* + H2 5.88 e20 0.5 0 8 C2H2 + 2Ssi --> C* + H2 1.20 e19 0.5 0 82 B.2 Silicon species adsorption No. Surface Reactions A E(K) Ref. 9 SiH2 + Sc --> SiH2* 6.02 e11 0.5 0 10 SiH4 + Sc --> SiH2* + H2 3.16 e10 0.5 9399.657 11 SiH3 + Sc --> SiH* + H2 6.02 e11 0.5 0 12 SiH + Sc --> SiH* 6.16 e11 0.5 0 13 Si + Sc --> Si* 6.30 e11 0.5 0 14 Si2H5 + 2Sc --> SiH* + Si* + 2H2 3.89 e20 0.5 0 15 Si2 + 2Sc --> 2Si* 4.07 e20 0.5 0 16 Si2H6 + 2Sc --> 2Si* + 3H2 2.08 e20 0.5 9399.657 17 Si2H4 + 2Sc --> 2SiH2* 3.98 e20 0.5 0 18 Si2H2 + 2Sc --> 2SiH* 4.07 e20 0.5 0 19 Si3 + 3Sc --> 3Si* 2.29 e29 0.5 0 20 H2 + 2Sc --> 2H* 2.29 e11 0.5 0 82
110 Appendix B: (continued) B.3 Mixed species adsorption No. Surface Reactions A E(K) Ref. 21 SiC2 + Sc + 2Ssi --> Si* + 2C* 8.70 e20 0 22 Si2C + 2Sc + Ssi --> 2Si* + C* 1.14 e21 0.5 0 23 SiC2H2 + 2Ssi + Sc --> C* + CH* + SiH* 4.36 e20 0.5 0 82 B.4 Chlorinated species adsorption No. Surface Reactions A E(K) Ref. 24 SiHCl3 + 2Ssi + 2Sc --> SiCl* + H* + 2Cls* 2.63 e16 0.5 0 25 SiH2Cl2 + Ssi + 3Sc --> SiCl* + 2H* + Cls* 3.80 e08 0.5 0 26 SiCl4 + 2Ssi + 2Sc --> SiCl* + Clc* + 2Cls* 2.34 e16 0.5 0 27 SiCl2 + 2Sc --> SiCl* + Clc* 3.09 e19 0.5 0 28 SiCl2 + Sc + Ssi --> SiCl* + Cls* 3.09 e19 0.5 0 29 HCl + Ssi + Sc --> CH* + Cls* 3.54 e12 0.5 0 30 SiCl + Sc --> SiCl* 4.16 e11 0.5 0 31 SiHCl + 2Sc --> SiCl* + H* 3.31 e20 0.5 0 32 SiHCl + Sc --> Si* + HCl 4.16 e12 0.5 0 82 B.5 Reaction between surface and gaseous species No. Surface Reactions A E(K) Ref. 33 H + CH* --> C* + H2 3.54 e12 0.5 0 34 H + C* --> CH* 3.54 e12 0.5 0 35 CH* --> 0.5H2 + C* 1.00 e23 0.5 28735.43 36 CH* + H2 --> Ssi + CH3 2.29 e11 0.5 44262.63 37 CH* + 0.5H2 --> C* + H2 2.29 e12 0.5 0 38 C* + 0.5H2 --> CH* 2.29 e12 0.5 0 39 SiCl* + 0.5H2 --> Si* + HCl 2.34 e15 0.5 30194.85 40 SiCl* + H --> Si* + HCl 3.31 e15 0.5 30194.85 41 SiCl* + HCl --> SiCl2 + H + Sc 3.54 e10 0.5 0 42 2Clc* + H2 --> 2HCl + 2Sc 2.34 e12 0.5 39545.19 82
111 Appendix B: (continued) No. Surface Reactions A E(K) Ref. 43 Clc* + H --> HCl + 2Sc 3.16 e12 0.5 0 44 2Clc* + SiCl2 --> SiCl4 + 2Sc 3.23 e11 0.5 12581.19 45 2Cls* + H2 --> 2HCl + 2Ssi 2.34 e12 0.5 45194.65 46 Cls* + H --> HCl + Ssi 3.16 e12 0.5 0 47 Cls* + Clc* + H2 --> 2HCl + Sc + Ssi 2.34 e12 0.5 42368.41 48 Cls* + Clc* + SiCl2 --> SiCl4 + Sc + Ssi 3.23 e10 0.5 12581.19 82 B.6 Reaction between surface species No. Surface Reactions A E(K) Ref. 49 2SiH* --> Si* + H2 1.00 e25 0 30698.1 50 SiH2* --> Si* + H2 1.00 e19 0 30698.1 51 SiCl* + Clc* --> SiCl2 + 2Sc 1.00 e19 0 10112.76 52 SiCl* + Cls* --> SiCl2 + Ssi + Sc 1.00 e19 0 45194.65 53 2SiCl* --> SiCl2 + Si* + 2Sc 1.00 e19 0 45194.65 54 H* + SiCl* --> HCl + Si* + Sc 1.00 e23 0 25165.4 55 CH* + Cls* --> HCl + C* + Ssi 1.00 e19 0 45194.65 56 Si* + Cls* --> SiCl* + Ssi 1.00 e17 0 0 57 H* + Cls* --> HCl + Sc + Ssi 1.00 e23 0 45194.65 58 CH* + Clc* --> HCl + C* + Sc 1.00 e18 0 42368.41 59 Si* + Clc* --> SiCl* + Sc 1.00 e17 0 0 60 H* + Clc* --> HCl + 2Sc 1.00 e23 0 42368.41 61 H* + H* --> H2 + 2Sc 1.00 e24 0 30698.1 62 CH* + CH* --> 2C* + H2 1.00 e23 0 30698.1 63 CH* + H* --> H2 + C* + Sc 1.00 e23 0 30698.1 64 CH* + CH* --> C2H2 + 2Ssi 1.00 e23 0 44262.63 82 B.7 Desorption reactions No. Surface Reactions A E(K) Ref. 65 Si* --> S2 + Si 1.00 e13 0 20381.53 66 2Si* + C* --> Si2C + 2Sc + Ssi 1.00 e24 0 0 67 Si* + 2C* --> SiC2 + Sc + 2Ssi 1.00 e24 0 0 82
112 Appendix B: (continued) B.8 HCl etching reaction No. Surface Reactions A E(K) Ref. 68 HCl + SiC(b) --> SiCl* + CH* 3.54 e10 0.5 0 82 B.9 Growth reactions No. Surface Reactions A E(K) Ref. 69 Si* + C* --> SiC(b) + Ssi + Sc 1.00 e17 0 0 70 2Si* + C* --> Si2C(b) + 2Sc + Ssi 1.00 e23 0 0 71 Si* + 2C* --> SiC2(b) + Sc + 2Ssi 1.00 e23 0 0 72 SiCl* + C* --> SiC(b) + Sc + Ssi + Cl 1.00 e17 0 0 73 Si* + CH* --> SiC(b) + Ssi + Sc + H 1.00 e17 0 0 74 SiH* + CH* --> SiC(b) + Ssi + Sc + 2H 1.00 e17 0 0 75 SiH* + C* --> SiC(b) + Ssi + Sc + H 1.00 e17 0 0 76 SiCl* + CH* --> SiC(b) + HCl + Ssi + Sc 1.00 e17 0 0 82
113 Appendix C Simulation procedure In order to perform the simulations the COMSOL Rea ction Engineering Lab and COMSOL Multiphysics have to be used iteratively. Th e following sections summarize the main steps necessary to develop the model and o btain a solution. C.1 Modeling using COMSOL Multiphysics Draw the desired reactor geometry Open Reaction Engineering Lab from COMSOL Multiphys ics C.2 Modeling using COMSOL Reaction Engineering Lab Gas phase reactions Choose model/model Select the calculate thermodynamic properties and c alculate transport properties Type process temperature and pressure Obtain gas phase reaction model Input gas phase reactions Input Arrhenius parameters: A, n, Ea Input species molecular weight Input species initial concentration Input species transport properties: \ and r Input species thermo parameters: NASA polynomial co efficients Select H2 as solvent and choose lock concentration/activity Compute the solution C.3 Exporting Model to COMSOL Multiphysics File/Export/Model to COMSOL Multiphysics Choose Geom1(2D) Select export mass balance
114 Appendix C: (continued) Choose application mode: Convection and Diffusion:N ew Type gas_mass in group name Select export energy balance Choose application mode: Convection and Conduction: New Type gas_energy in group name Select export momentum balances Choose application mode: Incompressible Navier-Stok es: New Type gas_momentum in group name Click export C.4 Modeling using COMSOL Reaction Engineering Lab Adsorption/Desorption Type T_surf in T edit field Type p_0 in p edit field Obtain surface reaction model Input surface phase reactions Input Arrhenius parameters: A, n, Ea Input species molecular weight Input species initial concentration C.5 Export Model to COMSOL Multiphysics File/Export/Model to COMSOL Multiphysics From domain level list choose: Interior boundary Clear export energy balance check box Clear export energy balance check box Select export mass balance Choose application mode: Convection and diffusion(c hcd)
115 Appendix C: (continued) Type boundary_mass in group name edit field Click export C.6 Modeling using COMSOL Multiphysics Verify that the model contains: Convection and Diff usion(chcd), Convection and Conduction(chcc) and Incompressible Navier-Stokes(c hns) Type model global constants Choose: Convection and Diffusion(chcd) Select appropriate subdomain settings and boundary conditions Choose: Convection and Conduction(chcc) Choose: Incompressible Navier-Stokes(chns) Choose: Incompressible Navier-Stokes(chns) Select appropriate subdomain settings and boundary conditions Click ok C.7 Mesh Generation Choose Mesh/Free Mesh parameters On global page, select Fine On Boundary page, select the substrate, then type 2 e-3 in maximum element size Click ok C.8 Computing the solution Click solver parameters button From solver list, select stationanry Click the advance tab and clear the Stop if error d ue to undefined operation checkbox Click Ok
116 Appendix C: (continued) Choose Solve/Solver manager On the Solve For page, select Convection and Conduc tion and Non-Isothermal Flow On the script page, select Automatically add comman ds when solving checkbox Click Apply, then click Solve. This step solves the momentum and energy balances to get the good initial value. On the Initial Value page, select Current Solution in the Initial Value area On the Solve For page, select only Convection and D iffusion Click Apply and then Click Solve to find the mass b alance Finally, select all the application modes by clicki ng the Solve For tab and selecting Geom 1 folder. Click Aplly then Click Solve Now that the final solution have been obtained, go to Script page and clear the Automatically add commands when solving check box. This settings allow you to used the same solution technique if you which to al ter parameters in the model and solve again. Save your document
117 Appendix D Statistical Design of Experiments (DOE) D.1 25-2 fractional factorial DOE construction A fractional factorial design of experiments is a variation of the full factorial DOE in which only a subset of the experimental matrix i s performed. A factorial design is an experimental strategy in which factors are varied t ogether instead of one at a time. In this task; two factors levels will be considered. This m eans that a full factorial will explore all possible combinations of the factors levels; in thi s case 2k possible combinations, where k is the number of factors under consideration and th e number 2 represents the two levels for each factor. The following discussion is intended to explain ho w fractional DOEs are constructed such that the experimenter can determin e which fraction of the full factorial DOE has to be conducted. The following sample will be based on the construction of a one quarter fractional factorial design with 5 fact ors, each one studied at 2 levels (25-2) for a total of 8 experimental runs (henceforth referred to as simply runs) plus the center points instead of 32 runs if the full factorial des ign was applied. To construct the design, what is called a defining relation (I) must be specified. I=ABD and I=ACE were chosen as design generators. T hese generators produce a design of experiments of resolution III which is the highe st resolution for the quarter fractional factorial DOE in this case. By resolution III DOE i s meant that no main effects are aliased with any other main effects but main effect s are aliased with two-factor interactions and two factor interactions may be ali ased to each other. For more details the reader is encourage to refer to external sources.88 In order to determine the runs to be performed the 25 full factorial DOE matrix is specified first. Then the quarter of the full facto rial to be considered is determined based on the already defined generators I=ABD and I=ACE. This is shown in Table A.1. In the combination column represents the total number of c ombinations for the 25 full factorial DOE, equivalent to 32 runs. The columns A to E, rep resent the factors under
118 Appendix D: (continued) consideration. The numbers +1 and -1 represent the low and the high level value of each factor. The table is then filled by assigning a val ue of +1 to the corresponding factor specified on each row of the combination column and a -1 to the remaining factors. Finally, the design generators are used to determin e the runs to be considered, in this case the runs where ABD, ACE are positives (highlighted) Since this is a quarter of the design then the other three quarters can be identif ied by using the following relationship combinations I=ABD with I= -ACE, I= -ABD with I=ACE and I= -ABD with I= -ACE. As far as the design selection, which quarter shoul d be run should not affect the final analysis of the experimental matrix or the conclusi ons that are drawn from it. Finally, once the design is constructed and execut ed the results are then analyzed via statistical techniques. Typically the analysis of variance (ANOVA) test is used. ANOVA refers to a collection of statistical model w hich compare means by splitting the overall observed variance into different parts due to different factors which are estimated and/or tested. Further details on how the ANOVA ana lysis is performed can be found elsewhere.88 Table D.1 Construction of the 25-2 fractional factorial design (highlighted) from the 25 full factorial DOE when generators are I=ABD and I=ACE. Run Combination A B C D E ABD ACE 1 -1 -1 -1 -1 -1 -1 -1 -1 2 a 1 -1 -1 -1 -1 1 1 3 b -1 1 -1 -1 -1 1 -1 4 c -1 -1 1 -1 -1 -1 1 5 d -1 -1 -1 1 -1 1 -1 6 e -1 -1 -1 -1 1 -1 1 7 ab 1 1 -1 -1 -1 -1 1 8 ac 1 -1 1 -1 -1 1 -1 9 ad 1 -1 -1 1 -1 -1 1
119 Appendix D: (continued) Run Combination A B C D E ABD ACE 10 ae 1 -1 -1 -1 1 1 -1 11 bc -1 1 1 -1 -1 1 1 12 bd -1 1 -1 1 -1 -1 -1 13 be -1 1 -1 -1 1 1 1 14 cd -1 -1 1 1 -1 1 1 15 ce -1 -1 1 -1 1 -1 -1 16 de -1 -1 -1 1 1 1 1 17 abc 1 1 1 -1 -1 -1 -1 18 abd 1 1 -1 1 -1 1 1 19 abe 1 1 -1 -1 1 -1 -1 20 acd 1 -1 1 1 -1 -1 -1 21 ace 1 -1 1 -1 1 1 1 22 ade 1 -1 -1 1 1 -1 -1 23 bcd -1 1 1 1 -1 -1 1 24 bce -1 1 1 -1 1 1 -1 25 bde -1 1 -1 1 1 -1 1 26 cde -1 -1 1 1 1 1 -1 27 abcd 1 1 1 1 -1 1 -1 28 abce 1 1 1 -1 1 -1 1 29 abde 1 1 -1 1 1 1 -1 30 acde 1 -1 1 1 1 -1 1 31 bcde -1 1 1 1 1 -1 -1 32 = 25 abcde 1 1 1 1 1 1 1
120 About the Author Meralys Reyes-Natal received a Bachelors Degree i n Chemical Engineering from the University of Puerto Rico in 2003. Since that t ime she has been pursuing her Doctor of Philosophy degree in Chemical Engineering at the University of South Florida. Her research focus is on the heteroepitaxial growth of 3C-SiC layers by CVD using the SiH4C3H8-HCl-H2 chemistry. This work was performed in the departme nt of Electrical Engineering and has been supported by federal and i ndustrial grants. Her dissertation work has been presented at the 2005 International C onference on Silicon Carbide and Related Materials (ICSCRM) (Pittsburg, PA), 2006 Sp ring Materials Research Society (MRS) meeting (San Francisco, CA), 2006 European Co nference on Silicon Carbide and Related Materials (ECSCRM) (New Castle, UK), and la stly at the 2007 Electronic Materials Conference (EMC) (Notre Dame, IN).