USF Libraries
USF Digital Collections

Design and development of a silicon carbide chemical vapor deposition reactor

MISSING IMAGE

Material Information

Title:
Design and development of a silicon carbide chemical vapor deposition reactor
Physical Description:
Book
Language:
English
Creator:
Smith, Matthew T
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
silicon carbide
chemical vapor deposition
Dissertations, Academic -- Chemical Engineering -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The goal of this thesis is to present the design and development of a chemical vapor deposition reactor for the growth of high quality homoepitaxy silicon carbide films for electronic device applications. The work was performed in the Nanomaterials and Nanomanufacturing Research Center at the University of South Florida from 8/2001-5/2003. Chemical vapor deposition (CVD) is the technique of choice for SiC epitaxial growth. Epitaxial layers are the building blocks for use in various semiconductor device applications. This thesis reports on a SiC epitaxy process where a carrier gas (hydrogen) is saturated with reactive precursors (silane and propane) which are then delivered to a semiconductor substrate resting on a RF induction heated SiC coated graphite susceptor. Growth proceeds via a series of heterogeneous chemical reactions with several steps, including precursor adsorption, surface diffusion and desorbtion of volatile by-products. The design and development of a reactor to make this process controlled and repeatable can be accomplished using theoretical and empirical tools. Fluid flow modeling, reactor sizing, low-pressure pumping and control are engineering concepts that were explored. Work on the design and development of an atmospheric pressure cold-wall CVD (APCVD) reactor will be presented. A detailed discussion of modifications to this reactor to permit hot-wall, low-pressure CVD (LPCVD) operation will then be presented. The consequences of this process variable change will be discussed as well as the necessary design parameters. Computational fluid dynamic (CFD) calculations, which predict the flow patterns of gases in the reaction tube, will be presented. Feasible CVD reactor design that results in laminar fluid flow control is a function of the prior mentioned techniques and will be presented.
Thesis:
Thesis (M.S.Ch.E.)--University of South Florida, 2003.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Matthew T. Smith.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 86 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001441472
oclc - 53994144
notis - AJM5912
usfldc doi - E14-SFE0000145
usfldc handle - e14.145
System ID:
SFS0024841:00001


This item is only available as the following downloads:


Full Text
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 001441472
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 031203s2003 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000145
035
(OCoLC)53994144
9
AJM5912
b SE
SFE0000145
040
FHM
c FHM
090
TP145
1 100
Smith, Matthew T.
0 245
Design and development of a silicon carbide chemical vapor deposition reactor
h [electronic resource] /
by Matthew T. Smith.
260
[Tampa, Fla.] :
University of South Florida,
2003.
502
Thesis (M.S.Ch.E.)--University of South Florida, 2003.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 86 pages.
520
ABSTRACT: The goal of this thesis is to present the design and development of a chemical vapor deposition reactor for the growth of high quality homoepitaxy silicon carbide films for electronic device applications. The work was performed in the Nanomaterials and Nanomanufacturing Research Center at the University of South Florida from 8/2001-5/2003. Chemical vapor deposition (CVD) is the technique of choice for SiC epitaxial growth. Epitaxial layers are the building blocks for use in various semiconductor device applications. This thesis reports on a SiC epitaxy process where a carrier gas (hydrogen) is saturated with reactive precursors (silane and propane) which are then delivered to a semiconductor substrate resting on a RF induction heated SiC coated graphite susceptor. Growth proceeds via a series of heterogeneous chemical reactions with several steps, including precursor adsorption, surface diffusion and desorbtion of volatile by-products. The design and development of a reactor to make this process controlled and repeatable can be accomplished using theoretical and empirical tools. Fluid flow modeling, reactor sizing, low-pressure pumping and control are engineering concepts that were explored. Work on the design and development of an atmospheric pressure cold-wall CVD (APCVD) reactor will be presented. A detailed discussion of modifications to this reactor to permit hot-wall, low-pressure CVD (LPCVD) operation will then be presented. The consequences of this process variable change will be discussed as well as the necessary design parameters. Computational fluid dynamic (CFD) calculations, which predict the flow patterns of gases in the reaction tube, will be presented. Feasible CVD reactor design that results in laminar fluid flow control is a function of the prior mentioned techniques and will be presented.
590
Co-adviser: Saddow, Stephen
Co-adviser: Wolan, John
653
silicon carbide.
chemical vapor deposition.
690
Dissertations, Academic
z USF
x Chemical Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.145



PAGE 1

Design And Development Of A Silicon Carb ide Chemical Vapor Deposition Reactor by Matthew T. Smith A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Chemical Engineering Department of Chemical Engineering College of Engineering University of South Florida Co-Major Professor: John Wolan, Ph.D. Co-Major Professor: Stephen E. Saddow, Ph.D. Andrew Hoff, Ph.D. Date of Approval: November 6, 2003 Keywords: epitaxy, epitaxial, microel ectronics, vacuum, crystal growth Copyright 2003 Matthew T. Smith

PAGE 2

ACKNOWLEDGMENTS Many faculty, staff, and students have contri buted towards the successful development of this thesis work. I wish to express my sincer e gratitude to all who have contributed towards this endeavor and especially my advisors Professors Stephen Saddow and John Wolan, who gracefully and professionally have been tr ue friends to me. My special thanks go out to Professor Saddow whose dedication to the development of my professional skills was tireless and selfless. Profe ssor Saddow should also be acknowledged as the primary contributor to virtuall y all aspects of this thesis in cluding the reactor design and the guidance in bringing the process to an experimental state. Professor John Wolan inspired me to pursue graduate studies and supported this project wi th his intense knowledge of chemical engineering and focus on educational excellence. Thomas Schatner’s control system design and technical suppor t was very much appreciated during the initia l stages of this project. The support of th e NNRC staff (Robert Tufts, Ri chard Everly, and Jay Bieber) should also be acknowledged for the laboratory training and technical input they provided during the course of this wor k. I also wish to give special mention to my lovely wife, Annemarie, who has always encouraged me to pursue greater things in life. This work was supported by the DURINT program administer ed by the Office of Naval Research under Grant N00014-0110715 administered by C. E. C. Wood.

PAGE 3

i TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES iv ABSTRACT vi CHAPTER 1: INTRODUCTION 1 1.1 Silicon Carbide Overview 1 1.2 SiC Polytypism 3 1.3 Epitaxy on Off-Axis Substrates 5 1.4 Epitaxial Growth Overview 6 1.4.1 SiC CVD Review 11 1.4.2 In-Situ Doping of Epitaxi al Layers 12 1.4.3 Recent Trends in SiC CVD 13 1.5 Organization of Thesis 14 CHAPTER 2: CHEMICAL VAPOR DEPOSITION 16 2.1 CVD Theory 16 2.1.1 Fluid Mechanics 18 2.1.2 Transport Phenomena 20 2.1.3 Surface Kinetics and Equilibrium 23 2.2 SiC CVD Epitaxial Growth Chemistry 27 2.3 Summary 35 CHAPTER 3: CVD SYSTEM DEVELOPMENT 36 3.1 APCVD SiC System 36 3.1.1 APCVD Tube Design 37

PAGE 4

ii 3.1.2 Susceptor Development 38 3.1.3 Processing Hazardous Gases 42 3.2 LPCVD SiC System 46 3.2.1 Basic Design Considerations 47 3.2.2 Computational Fluid Dyna mic Simulations 50 3.2.3 Pressure Control 54 3.2.4 Line Selection 55 3.2.5 Pump Selection 56 3.2.6 System Installation 58 3.3 Hot-Wall 59 3.4 Summary 61 CHAPTER 4: RESULTS AND CONCLUSIONS 63 4.1 Experimental Results 63 4.1.1 APCVD Homoepitaxy 63 4.1.2 APCVD Characterization 64 4.2 LPCVD System Validation 67 4.3 Summary 67 4.4 Future Work 68 REFERENCES 70 APPENDICES 75 APPENDIX A: VACUUM DESIGN CALCULATIONS 76

PAGE 5

iii LIST OF TABLES Table 1.1 Properties of SiC polytypes vs. othe r common semiconductors at STP. 3 Table 2.1 Probable specious exhibited duri ng SiC CVD using silane and propane precursors in a hydrogen carrier gas. 33

PAGE 6

iv LIST OF FIGURES Figure 1.1 The stacking sequence of the thr ee most common SiC polytypes. 4 Figure 1.2 Illustration of a cross section of off-axis “vicinal” 6H-SiC substrate which exposes a high density of steps. 6 Figure 1.3 Generic process flow diag ram of a vertical CVD system. 10 Figure 2.1 Diagram of the horizontal cold -wall CVD reactor designed at USF. 17 Figure 2.2 Mass transport and diffusi on in a horizontal CVD system. 17 Figure 2.3 Control of deposition uniform ity in a horizonta l cold-wall CVD reactor with (a) the susceptor parallel to gas flow, (b) and a tilted susceptor. 22 Figure 2.4 Steps and terraces on an off-axis substrate play a larg e role in surface growth and catalysis due to kinks and surface defects. 24 Figure 2.5 Schematic of reactant adsorption on the growth surface in CVD. 25 Figure 2.6 Typical morphological problems resulting from CVD chemistry. 30 Figure 2.7 Illustration of the competing et ch and growth mechanisms on an offaxis “vicinal” 6H-SiC substrate. 32 Figure 2.8 Temperature and gas mole fraction just above a susceptor surface as determined by Lofgren et al for a hot-wall SiC LPCVD reactor. 34 Figure 3.1 Block flow diagram of the APCVD system at USF which preceded the LPCVD design. 37 Figure 3.2 Schematic representation of gra phite susceptor under active heating and cooling 39 Figure 3.3 Material is machined out of th e underside of the susceptor. 40 Figure 3.4 Diagram of the gas handling system. 45

PAGE 7

v Figure 3.4 Block diagram of the LPCVD system at USF. 46 Figure 3.6 Main parameters in sizing a vacuum system. 50 Figure 3.7 Mesh for CFD calculations as performed by Timothy Fawcett. 51 Figure 3.8 Temperature profile CFD si mulation for APCVD and LPCVD coldwall system. 52 Figure 3.9 8 slm APCVD velocity profile CFD simulation for system with new endcap. 52 Figure 3.10 30 slm and 100 mBar LPCVD velocity profile CFD simulation. 53 Figure 3.11 10 slm and 100 mBar LPCVD velocity profile CFD simulation. 53 Figure 3.12 The pressure control loop. 55 Figure 3.13 Pump performance curve fo r the RV8 BOC Edwards oil-sealed mechanical pump. 57 Figure 3.14 Pump performance curve fo r the QDP40 BOC Edwards dry pump. 58 Figure 3.15 Diagram of the plumbing exhaust system. 59 Figure 3.16 10 slm and 150 Torr hot-wall LPCVD velocity profile CFD simulation. 60 Figure 3.17 20 slm and 150 Torr hot-wall LPCVD velocity profile CFD simulation. 61 Figure 3.18 Researchers monitoring USF CVD reactor during growth. 62 Figure 4.1 SEM micrograph of an epitaxial cross-section. 65 Figure 4.2 Doping profiles determ ined by CV/IV measurements. 66 Figure 4.3 Linear regression analysis of average doping densities for various Si/C ratios. 66 Figure 4.4 Author proudly standing in front of USF CVD reactor. 69

PAGE 8

vi DESIGN AND DEVELOPMENT OF A SI C CHEMICAL VAPOR DEPOSITION REACTOR Matthew T. Smith ABSTRACT The goal of this thesis is to present the design and development of a chemical vapor deposition reactor for the growth of hi gh quality homoepitaxy silicon carbide films for electronic device applications. The work was performed in the Nanomaterials and Nanomanufacturing Research Cent er at the University of South Florida from 8/2001 – 5/2003. Chemical vapor deposition (CVD) is the technique of choice for SiC epitaxial growth. Epitaxial layers are the building blocks for use in various semiconductor device applications. This thesis reports on a SiC epitaxy process where a carrier gas (hydrogen) is saturated with reactive precursors (silane and propane) which are then delivered to a semiconductor substrate resting on a RF induction heated SiC coated graphite susceptor. Growth proceeds via a series of heterogeneou s chemical reactions with several steps, including precursor adsorption, surface diffusi on and desorbtion of volatile by-products. The design and development of a reactor to make this process controlled and repeatable can be accomplishe d using theoretical and empi rical tools. Fluid flow modeling, reactor sizing, low-pressure pumping and control are engineering concepts that

PAGE 9

vii were explored. Work on the design and deve lopment of an atmospheric pressure coldwall CVD (APCVD) reactor will be presented. A detailed discussion of modifications to this reactor to permit hot-wall, low-pre ssure CVD (LPCVD) operation will then be presented. The consequences of this process va riable change will be discussed as well as the necessary design parameters. Comput ational fluid dynamic (CFD) calculations, which predict the flow patterns of gases in th e reaction tube, will be presented. Feasible CVD reactor design that results in laminar fl uid flow control is a function of the prior mentioned techniques and will be presented.

PAGE 10

1 CHAPTER 1 INTRODUCTION 1.1 Silicon Carbide Overview Silicon Carbide (SiC) has long been consid ered a material of choice for high temperature, power, voltage, and frequency appli cations. This is related to its wide band gap (2.9 eV for the 6H-SiC polytype), high saturation drift velocity (2 x 107 cm/s), and high breakdown field (2.5 x 106 V/cm). SiC shows great pot ential, in many applications, for replacing the existing semiconductor t echnologies of Si an d GaAs, which cannot tolerate high temperatures and chemically hos tile environments. In addition, SiC exhibits several impressive physical ch aracteristics. These include a high thermal conductivity (4.9 W/cm K), chemical inertness and opti cal transparency depending on the doping. The usefulness of SiC has grown beyond power el ectronics applications and is being developed for use in gas sensing [1] and other novel applications directly related to the development of supporting technologies. Although mankind has known of SiC for over 100 years, its recent expansion into the market place has made it an increasingly inte resting research material within the past 10-15 years. The development of the optoelectronics industry has created a niche for SiC substrates and consequently has kept the ma terial at the forefront of the scientific community. SiC is well lattice matche d to many III-nitride compound semiconductor materials commonly used in the optoelectronics field and thus can be used as an excellent conducting substrate material. This allows direct integration of the epitaxial layers to the device substrate unlike the more commonly used -sapphire insulating substrate.

PAGE 11

2 Recent SiC technical development has led to the success of current state-of-the-art devices. High brightness and ultra bright blue and green InGaN-based LEDs, microwave metal-semiconductor field-effect transist ors (MESFETs) on seminsulating 4H-SiC, 19kV p-i-n diodes fabricated on SiC epitaxial la yers, and thyristors are examples of existing devices that have emerged [2] as we ll as gas sensing technologies [3]. Porous SiC (PSC) is another area of recent focus for SiC device technology. Many interesting properties of this material have been discovered and reported on while many challenges lie ahead on the CVD front. Th is thesis reports on epitaxy studies performed on porous SiC which is interesti ng from an electronic device and materials stand point. The improvement of device epitaxial layers gr own on PSC substrates and a decrease in recombination centers observe d by photoluminescence has been the most compelling evidence that the a ltered structure of these epitaxial layers might result in device improvements [4]. Also, the research of epitaxy on PSC material is of obvious importance in gas sensing applications due to the catalytic effect that surface area has on gas adsorption and desorbtion [5]. Epitaxy an d characterization on type I PSC (smaller pores) and type II PSC (larger pores) will be reported. The progress of SiC is, however, limite d by the high defect density of the substrate and challenges in the material processing steps necessary for full-scale production. Bulk growth is perhaps the most pr oblematic step as micropipes plague even the best commercial substrat es. The micropipes are hexagonal voids that extend through the substrate parallel to the c-axis. The micr opipe propagates into epitaxial layers when they are grown since the terminating surface provides a growth template. Most theories of

PAGE 12

3 micropipe formation are based on Frank’s Theory [6], which suggests that a micropipe is formed via a super-screw dislocation with a large Burger’s vect or. A micropipe is essentially a conglomeration of elementary screw dislocations th at are common in SiC substrates. Despite the micr opipe problems, commercial substrates are commonly available in 2” and 3” are already in production. The status of SiC is still generally considered to be an emerging material with great potential. While there are many challenge s to overcome, epitaxy of SiC is no longer in the infancy stage. Thick, high-quality epit axial layers have been grown for use in devices and are continually bei ng improved. Epitaxy of SiC is re ported on in this thesis as well as the design and development of the pr ocessing tool, a horizontal CVD reactor. Table 1.1: Properties of SiC polytypes vs other common semiconductors at STP [7]. Parameter Si GaAs 3C-SiC 4H-SiC 6H-SiC Band Gap (eV) 1.12 1.42 2.4 3.26 3.02 Breakdown Field @ 1017 cm-3 (MV/cm) 0.6 0.6 ~1.5 3 3.2 Electron Mobility @ 1017 cm-3 (cm2/V-s) 1100 6000 800 1000 400 Saturated Drift Velocity (106 cm/s)10 8 25 20 20 Thermal Conductivity (W/cm-K) 1.5 0.5 5.0 4.9 4.9 Hole mobility @ 1016 cm-3 (cm2/V-s) 420 320 40 115 90 1.2 SiC Polytypism One of the most important properties of Si C is the wide variety of polytypes that the material exhibits (>200). Polytypism is the phenomena when a material can arrange it’s atoms in different periodic structures such that the stacking arrangement can take on

PAGE 13

4 different forms. The primary focus of this study is 4H and 6H-SiC, while 3C-SiC growth studies on Si are being performe d at USF [8]. The number (i.e – 4 and 6) represents the number of atomic layers required to rep eat the stacking sequence while the letter represents the lattice configur ation (H = hexagonal, C = c ubic). The advantage of this polytypism phenomena lies in the fact that ba nd gap energy varies from one polytype to another making it possible to s lightly tailor the electrical and optical properties of the material based on an engineered crystalline structure. Figure 1.1: The stacking se quence of the three most common SiC polytypes [9]. The stacking sequence of a substrate gu ides crystal growth since monolayers growing in the same orientation that the surf ace terminates is favorable. A crystal is made up of periodic stacks of layers where there has to be some “memory” in order to guide additional atoms during growth. Memory effect s can be explained by considering that the A B C C B C A B 3C A C A B C A B 6H C A B A C A B A 4H [ 1100 ] [ 1120 ] [ 0001 ] c-axis a-axis

PAGE 14

5 energy of dissimilar lattice sites differs. Thes e energies are super positioned ranging deep into the crystal. This superposition originat es from the elementary tetrahedral structures stacking on each other, which cau ses interactions between th e successive stacks and thus rotation occurs. 1.3 Epitaxy on Off-Axis Substrates The concept of using an off-axis substrat e to select a specific epitaxial polytype originates from the nature of step-flow epitaxy [10]. In epitaxial growth, specific stacking sequences may be accomplished by using an offaxis substrate that is cut at an angle which exposes the desired stacking (i.e., atomic ) planes. The result ing surfaces are called vicinal and lower their energies to form steps and terraces. It is the intentional introduction of misorientation, resulting in a high density of surface steps, which provides a superior growth surface [11]. Intuitively the st ep height should be that of the unit cell, however, this has been empiri cally shown not to be true. The steps are actually quite larger, which S. Tyc [12] attributed to a process known as “ste p-bunching.” During the growth process dislocations pin the steps wh ich then coalesce into large “macro” steps and become bunched. This phenomenon offers an explanation as to why step heights are actually larger in 4H-SiC material which contra dicts the fact that th e 4H-SiC unit cell is smaller along the c-axis dir ection. Step-bunching behaves quite differently for each polytype and thus the observed steps are, in fact, larger for 4H-S iC than for 6H-SiC. Takahashi et al [13] determined post-epitaxy macr o steps to be 700-2000nm wide and 10-50nm high for 4H-SiC while 6H-SiC was 600-1300nm wide and 10-30nm high. The actual mechanisms for how steps aid in epitaxy will be discussed in Chapter 2. The reader

PAGE 15

6 should observe that the substrates are cut at different angles to expose steps that are characteristic of the crystalline structure in the bulk (Fig. 1.2). Figure 1.2: Illustration of a cr oss section of off-axis “vic inal” 6H-SiC substrate which exposes a high density of steps [12]. 1.4 Epitaxial Growth Overview Many epitaxial growth methods have been used in the thin films industry including liquid phase epitaxy (LPE), molecular beam epitaxy (MBE), and chemical vapor deposition (CVD). The basic idea is to depo sit a uniform epitaxial layer that has the electronic properties required for the device. This epitaxial la yer functions as part of a circuit where it may interact with other components inte grated on the device. The deposition and characterization of epitaxial layers via CVD will be discussed in greater detail in chapters 2 and 4, respectively. The me thod of choice for a specific application is a function of film growth rate, purity, a nd uniformity. Specific growth techniques offer advantages and disadvantages over their count erparts. CVD shows great promise in the SiC industry due to the relatively low capita l, operational, and maintenance expenses, while offering a high-throughput system that de monstrates repeatabil ity and uniformity of the film. 3.5 [ 0001 ] [ 1100 ] [ 1120 ]

PAGE 16

7 LPE uses a means of contacting a liquid so lution (liquid melt) to the substrate. Supersaturation of the solid surface in contact with the liquid causes the deposition of the liquid onto the solid [14]. The basic requireme nts for LPE are a mean s of contacting and removing the solution from the substrate. Seve ral configurations of this process have been achieved including a rocker method and a linear sliding boat assembly which drags the substrate across the liquid surface and then wipes off the excess [15]. An advantage of LPE is the high growth rates reported while major disadvant ages are the lack of film uniformity and repeatability. The large number of defects associated with LPE makes it a limited processing tool and does not produce the high quality epitaxy expected in semiconductor manufacturing although applicati ons are being developed to use LPE as a means to fill micropipes [16]. It has been shown that micropipes in SiC wafers may be reduced during LPE [17]. The general observati on is that the channel of the micropipe, originating from SiC substrate, becomes sm aller during LPE. The layer required to accomplish this is generally thick. During th ick layer growth, the formation of other defects is possible and may defeat the pur pose of micropipe re duction. Step bunching, related to high growth rates, can be a source of these defects in LPE. If a thick layer is required to close micropipes, the height of th e steps can also be a disastrous problem for subsequent processing. The main idea of usi ng LPE is to fill the micropipes inside the micropipe channel first, and then use other ep itaxial growth methods to fabricate devices. Saddow et al have proven this is possible using CVD after LPE [17]. The filling process takes place inside the micropipe channel and the epitaxial growth on a flat surface may be negligible [17].

PAGE 17

8 Molecular beam epitaxy (MBE) is an ep itaxial growth method that is well established for growth of compound semic onductors such as GaN, GaAs, and AlGaAs [16]. MBE allows for very precise growth of epitaxy due to th e ultra-high-vacuum (UHV) conditions (pressure < 10-9 Torr) inside the growth chamber. The chemical components are delivered to the substrate through a physical deposition process in a UHV chamber. Source elements are heated in effusion or Knudsen cells located in the chamber. Out-gassing produced by the heat and the resulting molecular flux of the source material toward the substrate surface is ve ry well controlled and has few molecular collisions. Flux is controlled by heating elemental sources in effusion cells which then exhibit a vapor pressure and a known evapor ation rate occurs. Evaporated molecules eventually make their way to the substrate via molecular transport and adsorb onto the surface. Transport phenomenon for this is ea sily controlled due to the UHV conditions but this equipment can be capital intens ive which make it unreasonable for most SiC applications. More importantly, growth ra tes are very slow and requires very high temperatures which make MBE basically a lo w throughput research t ool for SiC growth applications. CVD is a process where one or more ga seous species reacts on a solid surface where one of the reaction products is a solid phase material. The seve ral steps that must occur in every CVD reaction include: precursor transport to the surface, adsorption or chemisorption, heterogeneous catalyzed su rface reaction(s), desorbtion of volatile byproducts, and finally transport of reaction byproducts from the surface [18]. The rate at which the process proceeds from the initial to the final state will depend on chemical kinetics and fluid dynamic transport. An analysis of the fluid dynamics will be presented

PAGE 18

9 in Chapter 3 for the USF CVD reactor and a more detailed discussi on of the reaction engineering of the system is provided in Chapter 2. The basic idea of CVD is to flow precursor s (gases that contain reactants) in a carrier gas through a heated reaction area wher e the previously mentioned steps occur. Several variations of CVD systems exist [ 19] including horizonta l and vertical [19] orientations, hot-wall and cold -wall, and numerous other va riations which are beyond the scope of this thesis. Horizontal CVD reactor s are proven to provide quality epitaxy at relatively low cost but with limited throughput Hot-wall and cold-wall refers to the temperature of the walls adjacent to where th e reacting gas stream flow. The temperature of the walls greatly influences particle nuc leation in the reaction area thus a hot-wall system is desirable where part icle nucleation is a concern. The susceptor element in high temperature systems (1000C -1800C) is gene rally heated through RF induction heating, although resistance heating elem ents are sometimes useful.

PAGE 19

10 Figure 1.3: Generic process flow diag ram of a vertical CVD system [20]. In the mid-1970’s it was realized that low-pressure CVD (LPCVD) processing could have advantages over atmospheric pre ssure systems. By reducing the pressure, it was found that the diffusion coefficient was sufficiently enhanced such that deposition became surface or reaction rate controlled [19]. The opposite is true for a higher-pressure system (atmospheric) where the effective grow th rate tends to be mass transport limited. Furthermore, a low-pressure system is c onducive to preventing gas-phase precipitation. In this event, a reaction occurs both in th e gas phase above as well as at the substrate growth surface. The gas phase reaction produces unintentional deposition (“particle rain”) of relatively large particles which cau ses non-uniformity in the epitaxial film and poor surface morphology [19]. An example of this phenomenon exists when using Si based precursors, such as silane, where Si droplet formation is problematic and known to be a device killer.

PAGE 20

11 1.4.1 SiC CVD Review CVD is the current research and industr ial standard for SiC epitaxial growth. Some obvious device driven motivations for CVD are growth of multi-layered structure (e.g. superlattices), excellent control over dopants, integra tion of heterostructures, and the ability to produce abrupt junctions. This review will focus on SiC epitaxial growth on SiC substrates. CVD has been an integral tool in the development of the devices mentioned. Over the last 10-20 years significant work has b een done to develop the best CVD process for SiC. Perhaps the best examples of SiC CVD progress are the commercially available reactors made by Ep igress [21] among others. The design of these reactors stems from pioneering work in the Si industry which led the way for SiC researchers to adapt Si designs for a more robust system necessary for SiC growth. In the mid-1980s it was discovered that epitaxy could be performed successfully on off-axis 6H-SiC substrates. Smooth mo rphology was achieved over a wide range of growth conditions between 1450C -1500C [ 22]. Prior to this development high temperatures (>2000C) were required to produce the crystalline structures with a modest amount of polycrystalline incl usions [24]. The high density of surface steps in 6H-SiC material serve as a template for growth to proceed. This discovery made growth on offaxis cut and polished substrates standard. Work done by Olle Kordina et al [10] in the early 1990’s is perhaps the most comprehensive work on SiC CVD to date. The aim of this work was to produce SiC material suitable for power devices including the growth of thick 4H-SiC material. 4HSiC was realized to be more suited for de vice applications becau se of its superior electrical properties versus 6H-SiC (Table 1.1). A nove l hot-wall LPCVD system was

PAGE 21

12 developed and optimized which is now the mo del for most SiC CVD reactors. In this system, the susceptor is suppor ted by a carbon foam insula tion material which encases the susceptor. The insulation severely reduces radiation losses from the intensely hot susceptor (1600C – 1800C typica lly) and provides an environment that decreases the particle nucleation in the gas phase. Elemen tal Si, produced from cracking precursors due to high reaction temperatures, has the potenti al to form Si droplets. Unintentional incorporation of Si droplets into the SiC epita xial film is one of the most deleterious defects that severely inhib it device performance. These droplets can be significantly reduced by employing a low pressure system which shifts the phase equilibrium to the gas phase, thus preventing droplet formati on. The optimal pressure range for operation was determined by Kordina et al to be 100-300 mBar [10]. Growth rates in hot-wall LPCVD reactors have been reported as hi gh as 50 m/h although 10 m/h is more common for device quality material. Lower grow th rates, which are less prone to the formation of defects, consistently produ ce surfaces with good morphology and can be controlled by adjusting the partial pressure of the precursors in the reaction area [10]. 1.4.2 In-Situ Doping of Epitaxial Layers A powerful dopant control technique known as site competition epitaxy was realized to be extremely valuable in the early 1990s by D.J. Larkin et al [23]. This technique enabled a much wide r range of repeatable doping de nsities than was previously observed. Successful dopant control with th is technique is based on appropriately adjusting the flow of precursors to contro l dopant incorporation into substitutional SiC lattice sites. Since SiC is th e only compound in the Si-C equilibrium system, adjusting the Si/C mass ratio will enable the SiC to be grown while other atoms (i.e., dopants) in the

PAGE 22

13 system compete with Si and C for lattice s ites. The model for site-competition epitaxy suggests that C out competes N for C sites and Si out competes Al for Si sites [24]. This thermodynamic relationship allows the epi-gr ower to alter the Si/C mass ratio to effectively allow the incorporation of pro cess dopants or impurities into the film. Doping of SiC has proven to be challenging because the mechanical strength does not allow for diffusion as a means to selectiv ely dope an area. Epitaxial layers can be doped during growth but ion implantation is re quired to define drain and source wells, junction isolation regions, etc. Ion implanta tion techniques have been successful but ion activation results in crystal damage as th is process must be carried out at high temperatures (~1600C) [25]. Ion implanted SiC that is annealed typically results in the out-gassing of Si atoms which causes poor mo rphology to result due to so called stepbunching. Work done by Saddow et al [26] prevents the destruction of surface morphology by using a high temperat ure anneal with a Si over pressure in a CVD reactor. This process decreases the concentration gradient between the process gas and the implanted material thus the driving force for Si out-gassing is reduced. The reactor developed in this thesis has been equi pped to perform these implant annealing experiments although the focus of this work has been CVD process development. 1.4.3 Recent Trends in SiC CVD Heteroepitaxy of 3C-SiC on Si substrates is gaining momentum in the power electronics field. Although the pr operties of 3C-SiC on Si ar e currently not as favorable as 4H and 6H-SiC, this technique offers the use of Si substrates which do not have the high defect density of SiC substrates. A wi de band gap epitaxial layer on Si enables device design for high-power electronics using the well established Si substrates. In the

PAGE 23

14 late 1980s, APCVD had been used for th is application but large nonuniformity in thickness and growth protrusions were repo rted [27]. With the increasing number of reactor configurations being developed, researchers have begun to overcome these challenges. Takahashi et al [28] have reported signifi cant morphological improvements by utilizing LPCVD. The race to perfect this process and pu sh 3C-SiC into the market place is on-going with 3C-SiC growth studi es taking place at USF [8] among others. 1.5 Organization of Thesis The main thrust of this thesis is the design and development of a horizontal LPCVD reactor for the growth of SiC epitaxial layers. The development of this design was performed over several years where the APCVD design originat ed at Mississippi State University from Saddow et al [29-30] which constitutes the first point of this thesis work. An improved design of this APCVD syst em was constructed at USF in the fall of 2001. The ultimate goal of this work is to fa bricate a horizontal co ld-wall LPCVD reactor that would eventually permit a hot-wall reactor to be established. The future work will be discussed in Chapter 4. The argument for LP operation is well just ified due to potentially higher growth rates, superior film uniformity, and improve d repeatability compared to atmospheric pressure operation. In Chapter 2 it will be shown that CVD theory points toward this conclusion. Modification of an APCVD r eactor to LP operation not only requires hardware changes, but also requires process changes to create favorable fluid dynamics and growth rates. In this thesis we report on the flui d flow modeling, reactor sizing, lowpressure pumping and the control system(s ) needed to realize a cold-wall LPCVD system.

PAGE 24

15 The motivation for this thesis should be evident in the potential for SiC devices and sensors. CVD is an ideal, and indeed cr itical, tool for the deposition of uniform SiC epitaxial layers. The progress already made in SiC CVD, combined with the on-going development of new SiC devices, is sufficien t motivation for the design and development of a SiC LPCVD reactor at USF. Reaction engineering of this system including CVD theory and chemistry (Chapter 2), system design and development (Chapter 3), and finally validation (Chapter 4), will be discussed.

PAGE 25

16 CHAPTER 2 CHEMICAL VAPOR DEPOSITION 2.1 CVD Theory The focus of this thesis was the development of a horizontal cold-wall LPCVD capability at USF, with future plans for a ho t-wall upgrade. Thus the theoretical concepts required to perform this work will be pres ented in this chapter. CVD is not a new technology and pioneers in this field have set the stage for step-by-step improvement of reactor types and process conditions. Although highly complex analytical tools are now available and used to refine processes, em pirical design methods are widely used to develop CVD processes. The use of theory to provide a starting point for trial and error reactor development is a valuable tool to understand and implement. Transport phenomena and kinetics are the backbone of C VD theory and the optimal conditions for CVD reactions are based on an understanding of these concepts. Figure 2.1 shows a cross-section schematic view of the cold-wall CVD reactor at USF. As briefly discussed in Chapter 1, the events that must happen for a CVD reaction to occur can be broken down into the following : reactant gases are transported into the reactor in a carrier gas, reactant speci es diffuse through a boundary layer above the growth surface, the species are transported to the surface via diffusion, a reaction takes place on the surface where one of the products is a deposited solid, gaseous by-products are transported away from the surface, and finally are diffused away via the boundary layer [20]. Since the gas flows are continuous the film thickness w ill increase over time.

PAGE 26

17 An imperative driver for this theoretical discussion is the need for uniform fluid flow and thermal conditions in the reaction ar ea. Epitaxial layers must be uniform with respect to thickness and impurities (dopants) in order to be useful in microelectronic devices. The need for abrupt p-n junctions for these devices with a high yield per wafer is an important area for SiC processing. The main factors effecting nonuniformity in epitaxy are fluid mechanics, transport phenomena, and kinetics. Figure 2.1: Diagram of the horizontal co ld-wall CVD reactor designed at USF. Figure 2.2: Mass transport and diffu sion in a horizontal CVD system. Laminar Flow Gas Vectors gradually decrease due to resistances of layers Susceptor Velocity approaches zero near surface resulting in stagnant layer Surface Reaction Gas Inlet Exhaust Boat Susceptor RF Coil Cooling Jacket

PAGE 27

18 2.1.1 Fluid Mechanics Gas flow dynamics contribute greatly towards epitaxial layer uniformity since successful uniform mass trans port of precursors and reaction by-products is dependent on consistent flow conditions. Laminar flow is required in CVD to ensure an even cooling load and controllable transport phenomena in the reaction area. Measures must be taken to avoid turbulent flow, which can result in radical changes in precursor concentration and cooling loads with respect to time and position. Laminar flow is achieved when the fluid flows in smooth planes stacked onto each other. The planes of flow do not mix and the only mechanism for species to travel fr om one plane to another is diffusion [31]. Because of the resistance of the surrounding su rfaces, a velocity distribution of the planes is observed where vmax is generally in the center of the reactor and v = 0 at the reactor boundaries. The so called “boundary layer” is the region where the velocity changes from that of the bulk gas to zero. The boundary layer is generally defined as the locus of distances over which ninety-nine percent of the disruptive effect s occur [35] and is indicated in Fig 2.2 for reference. The orientation of the boundary layer w ith respect to the susceptor surface originates from the drag that the suscepto r exerts on the gas flow. The boundary layer would be very thin in the ab sence of obstacles and therefore a more uniform velocity profile is observed. The flow is disrupted by the drag cau sed by the hot susceptor and consequently the velocity is reduced near the susceptor surface [11]. This reduction in velocity increases along th e length of the susceptor and the boundary layer becomes gradually thicker.

PAGE 28

19 Large temperature gradients are present in the area above the susceptor in a coldwall system which may cause disturbances. Th ese disturbances result from the heating of gases which causes buoyancy [29] where th e less dense fluid tends to rise. Rough estimates of laminar conditions can be made from analyzing certain dimensionless numbers such as Reynolds, Grashof, and Ra yliegh numbers [29]. Arriving at laminar flow with suitable reactor geometry is very complex and is made easier by employing computational fluid dynamics (CFD) calculations. The CFD work reported in this thesis, which was used as a theoretical basis for optimal design of the reaction chamber geometry based on mass flow rates was performed by T. Fawcett of the USF SiC group [32]. CFD can be carried out in many different ways; the most popular, and the method used in this study, being through finite element method techniques [33]. As the na me implies, the finite element method involves dividing the fluid volume into predefined points that form a grid; then the three governing conservation equations (mass, ener gy, and momentum [34]) can be solved for each point given a set of boundary and initial conditions. The ability to easily change boundary conditions, and analyze the geometry of the reactor via simulation, vastly aids in optimizing the hardware design. CFD si mulations aide in fi nding optimum operating and geometrical reactor conditions when going from atmospheric pressure to lowpressure operation. The ultima te goal is maintaining laminar flow with minimal back mixing. Results of the CFD simulations perfor med during this thesis will be discussed later in Chapter 3.

PAGE 29

20 2.1.2 Transport Phenomena In a horizontal CVD reactor, precurso rs in a carrier gas are continuously introduced to the gas inlet where they ideally enter a laminar flow re gion. The gas state is in equilibrium far away from the heat sour ce and in a nonequilibrium state near it. The transition from equilibrium to nonequilibrium occurs over the boundary layer and is determined by viscosity, diffusion, gas-pha se kinetics, thermal conductivity, and diffusion coefficients [31]. Transport of gas sp ecies between laminar layers is the subject of intensive investigation since no reac tion can take place w ithout this action. Before considering the theoretical mech anisms for growth, it is constructive to understand the concept of the two rate limiti ng regimes for CVD reactions. These are the mass transport and the surface-reaction limited re gimes. As the pressure and/or surface reaction rate of the system d ecreases, there can be a large fl ux of reacting species at the surface. This can be thought of as an over supply condition where reactants are waiting to be consumed thus making this regime surfacereaction limited (transport controlled). The surface-reaction limited regime generally occurs when the reaction rate is low compared to the rate of species transport to the growth surface. At the other extreme, mass transport limited (surface-reaction controlled), a lack of reactants may be present at the surface and reaction kinetics would have lit tle effect on the ove rall growth rate [20] making mass transport the rate limiting step. It is favorable to be in the mass transport limited regime since the consumption of reactants can be eas ily controlled by pressure and temperature. A reduction in pressure would in crease diffusion coefficients [35] thus increasing the flux of reactants to the surface where they react quickly. Alternatively, a slight change in temperature would have little effect on the rate in the mass transport limited case thus

PAGE 30

21 temperature nonuniformity is not as significant as in the surface-reaction limited case. Transport limitation makes it possible for the subsequent growth and uniformity to be more selective than the surface-reaction limited regime. The transport of the gaseous reactant s, as well as the reaction by-products, through the boundary layer is analo gous to thin film diffusion wh ere each side of the film is diluted to saturated with solute. This diffusion is governed by traditional mass transfer thin film theory where the concentration gradient is the driving force [35]. Diffusion in real systems is a result of molecular vibra tion and convection due to the diffusion of mass through the system. The solute (g aseous specious other than th e carrier gas) diffuses from the fixed higher concentrated solution to th e fixed less concentrated solution. Because the reactants are consumed at the surface, the accumulation in the flow planes is zero and the process is in steady state. Steady state tr ansport means that the initial and final concentrations of the solute are independent of time. Physi cally, this suggests that the volumes of the adjacent solutions must be much greater than the film where the transport occurs [36]. As previously mentioned, the boundary layer gradually becomes thicker along the susceptor thus the transport rate is a function of position. The thicker boundary layer creates a stack of films with gradually decrea sing velocity. Transport must occur through this buffer layer making it more difficult fo r mass transport to occur downstream since there are less turbulent fluctuations. A depletion of reactants may also observed downstream since the gas mixture is initially saturated with precurs ors and becomes less saturated as the reactants are consumed. Thes e factors indicate that growth rates should be higher closer to the reactor inlet. A met hod that is widely used to manipulate the

PAGE 31

22 boundary layer is changing the geometry. Tilting the susceptor can eff ectively correct this problem by gradually reducing th e gas velocity near the center (Figure 2.3) [20]. The boundary layer becomes more uniform and an in crease in epitaxial growth is observed downstream [20]. Figure 2.3: Control of deposit ion uniformity in a horizontal cold-wall CVD reactor with (a) the susceptor parallel to gas fl ow, (b) and a tilted susceptor [20]. Another consideration must be given to the trade-off of reactant retention time (often called residence time) a nd the diffusion coefficients. As the pressure is reduced for a constant mass flow rate, the fluid expands in the tube and the velocity increases. Gas density decreases with a reducti on in pressure and thus molecu les must travel faster for the mass flow rates to remain constant. Reten tion times of the reactants will subsequently decrease but the diffusion coefficients will in crease. Shorter retention times are indicative of a system that will have a lower conversion of reactants to desired products; however, an optimal pressure range exists such that ove rall growth rate is optimized and these two competing factors cancel out [9]. (b) (a)

PAGE 32

23 2.1.3 Surface Kinetics and Equilibrium The mechanisms by which CVD reactions occur are similar to heterogeneous catalysis where gaseous reactants adsorb ont o a solid surface and then react to form a new surface. The kinetic rate law for this t ype of reaction is very complex, due to the many steps involved, making the ideal design equations difficult to implement [5]. Arriving at a ra te law is beyond the scope of this thesis but an understanding of the concepts that affect the kineti cs would be helpful. Transporta tion of the solute species to a point near the surface was discussed in the previous section but one should understand that this is not directly associated with the actual catalyzed heterogeneous surface reaction but only a necessary function for th e reaction steps to occur. When reactive species approach the surface, adsorption ta kes place which is characteristic of a heterogeneous surface reaction. Two types of adsorption that are possible are physical adsorption and chemisorption. In physical adsorption, van der Waals forces tend to play the biggest role. Van der Waals forces are weak but act over a long rang e where the energy rele ased is on the order of the enthalpy of condensati on [37]. When a molecule inte racts with the surface its energy dissipates as a result of friction and eventually settles onto the surface. This is analogous to a ball bouncing on a surface and eventu ally coming to a rest after its kinetic energy has been spent. The molecule maintain s its identity in physical adsorption since there is not sufficient energy to break bonds [37]. Physical ab sorption may also take place in CVD, such as large particle deposition, but is generally considered undesirable in microelectronics epitaxy and lengths are taken to avoid this absorption mechanism.

PAGE 33

24 The type of adsorption that takes place for catalyzed heterogeneous surface reactions, and the one that is considered to be the dominant part of the rate of the surface reaction, is chemisorption. In chemisorption, the adsorbed molecule is held onto the surface by valance forces resulting from surface defects. The electronic structure of the adsorbed molecule is perturbed, causing it to be reactive [5]. The key to SiC CVD is Si atom surface mobility since transport of atoms to the surface is required to grow a film. Off-axis substrates in SiC epitaxy play a critic al role in this process due to the irregular surface it provides. A step may have defects of its own known as kinks. When an atom or molecule comes into contact with a terrace it bounces across it under the influence of intermolecular potential and may come to a st ep or a corner formed by a kink [37] which effectively lowers the activation energy of th e growth process. Th e activation energy is the energy required to force the slowest reaction step (adsorption for SiC epitaxy [9]) that limits growth to proceed. The species has a hi gher probability of interacting with more surfaces in step-flow epitaxy and may become adsorbed. Figure 2.4: Steps and terraces on an off-axis s ubstrate play a large role in surface growth and catalysis due to kinks and surface defects. Step Terrace Kink Molecule

PAGE 34

25 Once a molecule has been adsorbed onto the surface a number of different reactions could occur to form the desired pr oduct [38]. Taylor [39] suggested that the reaction is not catalyzed over the entire surface but only at certain act ive sites resulting from surface irregularities such as chemisor bed molecules. The surface reaction could be a single-site mechanism in which only the adsorbed molecule is involved. Alternatively, the surface reaction may be a dua l-site mechanism where the adsorbed molecule interacts with another site to form the desired product. This second site may be occupied or the reactant will actually interact with both sites. Lastly, the reaction could take place between an adsorbed molecule and a molecule in the gas phase. The actual mechanism for any surface reaction is possibly a combina tion of the three possibilities although each one has a different rate law [5]. Figure 2.5: Schematic of reactant adsorpti on on the growth surface in CVD. Once a molecule has been adsorbed onto the surface (a ) a single site mechanism, (b) a dual site mechanism, or (c) a single site with a gas phase intera ction could occur. (a) (b) (c)

PAGE 35

26 The last step for a heterogeneous su rface reaction is the desorbtion of the byproducts. This is much the same as the adsorption process except in reverse. The byproduct no longer has valance forces holding it in place and become loosely bound to the surface by van der Waals forces [37]. Thes e by-products are thus highly volatile and readily release from the surface and they deso rb. The by-products are then transported, via diffusion through the boundary layer, en ter the gas stream and are then swept downstream to the reactor vent. The adsorp tion, surface reaction, and desorbtion steps are all critical steps to epitaxy and the operat or can use the knowledge of these phenomena to understand and improve the growth process. It has been argued that a low pressure system would effectiv ely increase growth rates due to the larger diffusion coefficien ts making it a reaction-rate limited growth regime. Forethought should also be given to any possible negative effects a reduction of pressure would have on the kinetics. Partial pr essure of the gases are decreased with an overall reduction in pressure making them fact ors in the ability of the gas to reach the surface and react successfully. Although the kineti cs of this system described in this thesis have yet to be worked out, surface re actions in high temperature CVD typically take place very quickly compared to mass tr ansport making it unlikely that a reduction of pressure would have a signifi cant effect on the overall re action rate. Careful weight should be given to the factor s previously mentioned when selecting operating parameters but it can not be clearly stated what the ra te limiting SiC CVD steps are until a careful kinetics and mass transport study is complete d. It has been empirically observed that growth rates are indeed optimized in a lowe r pressure SiC CVD syst em [11], most likely due to increased mass transpor t through the boundary layer.

PAGE 36

27 An important aspect of step flow epita xy is that it encourages lateral growth. Molecules tend to migrate towards surface step s and kinks where catalytic reactions take place. This means that three-dimensional grow ths are inhibited since a molecule migrates across the surface until it hits a de fect or step thus epitaxial la yers want to grow one step at a time and thus form a two-dimensional f ilm. If the ideal step height existed, the epitaxy would take place one m onolayer at a time. SiC grow th is, however, prone to defects including screw dislocat ions, particle nucleation, grow th pits, etc., particularly when growth rates are high [2], leading to three-dimensional growths which are highly undesirable. Suitable process chemistry can be used to suppress poor morphology resulting from these defects and it has been demonstrated for some time that growth on off-axis substrates suppresses three-dimensional nucleation. 2.2 SiC CVD Epitaxial Growth Chemistry SiC CVD chemistry is quite complex due to the large number of possible reactions and side reacti ons. Fortunately an analysis of gr owth is greatly aided by general knowledge even as the specifics are difficult to know for certain. SiC CVD chemistry is compounded to a great extent by the high temper atures required, the number of possible precursors, and the carrier gas which have imp lications with regards to the materials in the reactor as well as the epitaxy. This discus sion will focus on deroga tory results, related to chemistry, which have been observed dur ing the course of Si C CVD development. Chapter one detailed the use of a Si/C ratio to control doping by using the sitecompetition effect [23]. There are, however, additional chemistry issues related to this ratio which we will now discuss. A low Si/C ra tio is desirable not only to achieve low ntype doping levels, but also for preferable surface morphology. The chemistry of surface

PAGE 37

28 morphology can be classified into three types: C rich, Si rich, and moderate (Si and C flux being comparable). Although there have been many types of precursors used to grow SiC films, silane (SiH4) and propane (C3H8) are the sources of Si and C used in most systems and the system at USF. Under mo derate conditions a mo re stoichiometric deposition is observed where morphological defect s are less prone to oc cur. Si richness is perhaps the most common problematic case due to the lower decomposition ratio of C3H8. Cracking patterns reveal th e existence of Si and SiH2 due to cracked SiH4 while carbonaceous species such CH4, C2H2, C2H4, etc. result from cracked C3H8 [40-41]. This suggests that Si species are preferably absorb ed on the reactor surfaces thus encouraging a Si rich growth surface. Si rich conditions tend to produce three-dimensional particle nucleation because of the polymerization and subsequent deposition of elemental Si [9]. Si rich growth conditions can be inhibited during the growth cycle by decreasing the Si/C ratio and by introducing SiH4 only when C3H8 is in equilibrium. C rich growth has been discussed in relation to graphite decomposition in the susceptor and Si desorbtion in process heating. Typical C rich morphological defects includ e graphitization and wavy or stripe-like morphology. The effect that a low-pressure system has on chemistry relates to prevention of gas-phase precipitation. In this effect, a reaction occurs in the gas phase and subsequently at the substrate growth surface. The unint entional deposition of these relatively large particles causes nonuniformity in the epitaxial film and poor surface morphology [18]. As discussed, thermal cracking of silane leads to elemental Si in the system that has the potential to form Si clusters known as Si droplets. Gas phase nucleation takes place by a polymerization leading to the formation of particles ranging in size up to approximately

PAGE 38

29 300 [42]. Another negative effect of this phenomenon is that severe depletion of Si in the vapor phase limits the amount of Si av ailable for a surface reaction. A reduction in overall pressure would reduce th e partial pressure of Si making it less reactive in the gas phase and shifts the phase equilibrium of th e system to prevent vapor condensation [9]. Indeed it was with this anticipated benefit that this thesis research was undertaken. An excess of C from the susceptor during growth can change the Si/C ratio in the system significantly. The excess C can cause two problems; graphitization leading to morphological defects in extreme cases and l ack of dopant control via site-competition due to an excess of uncontrolled C in the ga s stream. This problem has been observed in the USF reactor where freshly exposed graphite, due to SiC-coating cracks on the susceptor, was obvious visually and resu lted in poor morphology during growth runs where this was observed. Although SiC coated susceptors are capable of producing highquality epitaxy, graphite exposure due to wear of the coating is a continuous problem in these systems and must be carefully monitored. The lifetime of the susceptors and quality of the epitaxy has been indicated to be improved by applying a tantalum carbide (TaC) coating on the graphite surface [43] since this coating is more durable than the conventional SiC coating used during these preliminary experiments. Poly-crystalline gr owth on the back-side of the substrate is observed when the SiC coatings are used as the coating transfers to the substrate during growth thus making a TaC coating even mo re desirable for ho moepitaxy application (since TaC is a dissimilar material it does not transfer to the substrate and no polycrystalline film is grown on the backside). Ta C coatings may not be used when using Si substrates for 3C-SiC epitaxy because any resi dual Ta that did not form TaC during the

PAGE 39

30 coating process forms a low temperature eutectic with Si at the temp eratures required for 3C-SiC growth [8]. The discussion in Section 2.1.3 regarding adsorption and desorbtion has implications regarding the surface chemistry. The solid-vapor equilibrium that exists is highly dependent on temperature thus precau tions must be taken when ramping the temperature to growth condi tions prior to introducing th e precursors [11]. The vapor pressure of a heated SiC surface has incongruent vapor pressures due to Si and C atoms existing in equilibrium at the surface (note that the lattice can be thought of as 2 interdependent Si and C sub-la ttices, each with their own vapor pressures). Si is the lower vapor pressure substance which will readily evaporate in th e absence of a C etching gas or a lack of Si overpressure. This inc ongruent vapor pressure may result in two disastrous effects: graphitizat ion of the surface or formation of Si droplets leading to morphological defects during the subsequent growth process. The later is a result of the C removal rate being higher than Si when a C et ching gas is used such as hydrogen. A small amount of hydrocarbon or a Si etching gas introduced durin g the heating/etching process has been shown to adequate ly prevent morphological problems associated with Si droplets [44-45]. Figure 2.6: Typical morphological pr oblems resulting from CVD chemistry. (a) (b)

PAGE 40

31 The carrier gas for most SiC CVD sy stems is hydrogen. Several factors are involved with this choice including its ability to prevent graphitization during the heating process. Hydrogen is used as a carrier because a comparatively larger stoichiometric SiC deposition area is obtained, whic h is presumably due to the ab ility of hydrogen to inhibit the formation of radical species [46]. A dditionally, hydrogen is economical and is available in ultra-high-purity (99.999% pure). There are, however, some drawbacks to using this gas. Etch rates of the graphi te susceptor increases exponentially with temperature in the presen ce of pure hydrogen [47]. It has been observed that a hydrogen carri er gas influences the etch rate of SiC during the growth process and will produce ga seous hydrocarbons and free Si [48]. Since the reaction that produces a SiC deposition is an equilibrium reaction, the reverse is possible and is governed by Le Chtelier’s pr inciple[49]. When a chemical system in a state of equilibrium is distur bed, it retains eq uilibrium by undergoing a net reaction that reduces the effect of the disturbance. Th e presence of pure hydrogen in the hot system may cause a decomposition reaction of the subs trate. This decomposition is the basis for the etching process that is common in SiC C VD to remove surface material that may be “damaged” from prior processing steps (such as polishing). Additiona lly, the equilibrium present in the system during gr owth has a net growth rate wh ich is the growth rate minus the etch rate.

PAGE 41

32 Figure 2.7: Illustration of the competing etch and growth mechanisms on an off-axis “vicinal” 6H-SiC substrate. The numerous possible gas and surface chem ical specious, along with the possible surface reactions, further emphasize the state of equilibria of the system. In order for a reaction to take place the reac tants must proceed to a lower energy state to form products. This energy change is known as the fr ee-energy change of the reaction, or Gr, which varies as a function of the type of reactants, the molar ratio of the reactants, temperature, and pressure [18]. One can assess the feasib ility of a reaction o ccurring by solving the related equations [20] and obtaining a value for Gr. The reaction is said to favor the reactants if Gr is positive, favor the products if Gr is negative, and be at equilibrium when it is equal to zero. This assessment is only valid if the reaction contains the major species that exist at equilibrium. Table 2.1 lis ts the common species th at may exist in the system which was determined by Lofgren et al [50] by modeling the likely decomposition and surface reactions. 3.5 [ 0001 ] [ 1100 ] [ 1120 ] Direction of Etching Direction of Growth

PAGE 42

33 Table 2.1: Probable specious exhibited dur ing SiC CVD using silane and propane precursors in a hydrogen carrier gas [50]. Gas Phase Specious Surface and Bulk Specious C-Containing Si Containing Other Surface Bulk C Si H C C CH Si2 H2 CH Si CH2 Si3 Si CH3 SiH SiH CH4 SiH2 SiH2 C2H SiH3 HCa C2H2 SiH4 HSib C2H3 Si2H2 C2H4 Si2H3 a H atom adsorbed at a C site C2H5 H2SiSiH2 b H atom adsorbed at a Si site C2H6 H3SiSiH C3H2 Si2H5 H2CCCH Si2H6 C3H4 Si3H8 CH2CHCH2 C3H6 i-C3H7 n-C3H7 C3H8 As previously discussed, the carrier gas is initially saturated with precursors and becomes less saturated as reactants are cons umed thus depleting the driving force for transport along the length of the reaction area. The result of this is often seen as dopant nonuniformity as the impurities can out compet e the Si and C atoms when they are depleted [23]. Work done by Koshka et al [51], using CFD simulations coupled with experimental validation, conf irms a depletion of reactants along the length of the susceptor in a horizontal cold-wall CVD reactor indicating a mass transport limited regime. An analysis of the experiment s shows growth rate uniformity, surface

PAGE 43

34 morphology, and doping uniformity where highly dependent on placement of the sample with respect to the susceptor area [51] due to altered chemistry at different positions. The practical consideration which comes out of this study is the need for consistent substrate placement on the susceptor. The operator must place substrates consistently in order to perform repeatable growth studies. As the precursors enter the heated area, decomposition and consumption occurs resulting in a distribu tion that is further effected by fl uid dynamics. The mole fractions of the possible species along the susceptor lengt h are useful in refi ning growth processes and susceptor design. Since the mole fractions are indicative of the growth rate and uniformity of the system, this t ype of modeling represents some of the latest advances in CVD chemistry. Figure 2.8: Temperature and gas mole frac tion just above a susceptor surface as determined by Lofgren et al [50] for a hot-wall SiC LPCVD reactor.

PAGE 44

35 2.3 Summary CVD operation and design involves an unde rstanding of fluid dynamics, kinetics, and transport phenomena. The complexities of these concepts make CVD growth a challenging area which can be analyzed with e xperimental and analytical tools. Process refinements can be performed on a regular basis using empirical observation coupled with experimental validation. More complicat ed areas, such as fluid dynamics, are better approached through CFD simulations to design the geometry of the reactor.

PAGE 45

36 CHAPTER 3 CVD SYSTEM DEVELOPMENT 3.1 APCVD SiC System It has been stated that the manipulat ion of process parameters and reactor geometry can be used to control the growth of epitaxial layers of SiC. The essential components needed to construct a CVD reactor can be selected and/or designed and constructed with planning and careful attent ion to engineering principles. The system must be designed to be fail-safe with equipm ent that can handle the extreme environment of SiC processing, especially th e high growth temperatures (>1600 C) and highly reactive gases such as silane and hydroge n. The primary components that one must consider are the same for al l CVD reactors, namely the gas handling system, reaction vessel, susceptor, RF delivery and coil design (for inductively heated systems), automatic process control, and an exhaust package. The AP design was largely the result of previous work conducted on a nearly identical system by members of our group [30]. Distinctions exist between the previous design and the USF system but the basic compon ents of a horizontal reactor capable of epitaxy on up to a 2 inch substrate still is the same. The main ideas carried over from the other reactor are the control system design and essential instrumentation required to monitor the process. A programmable logic controller (PLC) and microprocessor were designed and constructed to allo w fail-safe safe operation of ga ses, purging of the vessel, gas metering, and temperature control.

PAGE 46

37 Figure 3.1: Block flow diagram of the APC VD System at USF [30] which preceded the LPCVD design to be discussed in Section 3.2. 3.1.1 APCVD Tube Design The APCVD tube design was performed by Burke, Schattner, and Saddow [29-30] at Mississippi State University for a CVD system similar to the one at USF. The essential element that must preface any CVD reaction tube design is th e basic diameter it must be to accommodate the substrate on which the epit axial layer is to be grown. 75mm SiC wafers are now commercially available, alt hough growth on 50mm substrates is still the commercial standard at this time. Typically th e susceptor is designed slightly larger than the substrate to ensure maximum temperatur e uniformity hence a 70mm wide susceptor, to permit growth on a 50mm substrate, was used as the basis for specifying the tube inner diameter (ID). Assuming the susceptor can be evenly heated and reactants can be uniformly delivered to it, a tube ID slightly larger than the susceptor can be selected as a first iteration step to sizing the reaction vesse l. The diameter must be large enough so that Microprocessor Control Unit Programmable Logic Controller Pyrometer Gas Supply and Deliver y Control Console Reactor Cabinet Cooling Water Exhaust to Burn Box and Scrubber Reaction Tube RF Generator

PAGE 47

38 the susceptor will have adequate spacing from the quartz tube wall. The overall length of the tube can be roughly estimated based upon pr actical heat transfer considerations and development of fluid flow regimes. This process of sizing the tube and selecting the geometry often takes multiple attempts to optimize all parameters resulting in an iterative process [29]. The iterations generally consists of choosing a carrier gas flow rate, analysis of dimensionless number (Reynolds, Grashof and Rayliegh), and finally a computer simulation to confirm laminar flow in the susceptor area. The final design selected for the USF APCVD reaction tube consists of a head tube section, a water-cooled reaction tube, a quartz boat that supports the susceptor, and an end cap. A 100 mm bore diameter was used with an overall length of 30 inches. The head tube section permanently clamps onto the reactor housing and pr ovides the quartz to metal transition. The successive sections clam p onto the head tube sections using bake-olite clamps with Viton o-rings for vacuum integrity. This design has been shown to hold adequate vacuum for CVD (< 50 mTorr) but other designs, such as differentially pumped double o-rings, are desi rable to ensure maximum vacuum integrity. The main body of the reaction tube is water cooled to reduce heat accumulation in the system and provide cooling for the o-ri ng joints, which have a maxi mum operating temperature of ~150 C. The end cap then provides the transi tion from the reaction tube to the gas exhaust system. 3.1.2 Susceptor Development The methods of designing a susceptor w ith sufficient RF coupling range from complex simulations to empirical observation coupled with practical considerations. The

PAGE 48

39 desired result is a substrate that is heated uniformly. Th e basic idea of RF inductive heating of a graphite suscepto r is to use a RF generated ma gnetic field to couple onto the graphite, which consequently produces eddy currents in th e graphite which heats the graphite through Ohmic (i.e., I2R) heating. The convective cooling effects of the process gases and radiation from the susceptor represent challenges in creating a su sceptor that will heat uniformly and at the desired temperature. Convective cooling occurs primarily at the topside of the susceptor while radiation losses occur most intensely at the edges of the susceptor. Cooling can be significantly greater at the front of the suscep tor due to the cold finger of the inlet gases extending into the hot region. With a pract ical assessment of these facts, one can conclude that the majority of the RF power s hould distribute to the areas with the greatest load (i.e. – the edges, top side, and front side). Figure 3.2: Schematic repres entation of graphite suscep tor under active heating and cooling. Convective cooling oc curs primarily at the tops ide of the susceptor while radiation losses occur most intensel y at the edges of the susceptor. Process g as flow Convective Coolin g Intense Radiation loss

PAGE 49

40 These cooling effects can be overcome by shaping the susceptor’s geometry to minimize their effects and by proper spacing of the RF coils. Increasing the amount of graphite material in the areas of most intens e cooling is a good appr oach to dealing with these issues. More graphite mass translates into more RF coupling at those areas (in addition to thermal mass) and thus the deliver y of more power. In addition increasing the thermal mass in these regions reduces the te mperature dips associated with convective cooling. It follows from Figure 3.2 that having more graphite material at the edges and topside would deliver more coupling to the a ppropriate areas. This can be accomplished by having the susceptor machined such that material is removed on the bottom side as shown in Figure 3.3. (a) (b) Figure 3.3: Material is machined out of the un derside of the suscepto r (a) to deliver less power to the areas that are subject to less cooling load and increase the thermal mass at the edges and topside (b) where the cooling load is at a maximum. To counteract the effects of the cold gas finger extending into the front area of the susceptor, the RF coil design comes into pl ay. A complete design of the coil is beyond

PAGE 50

41 the scope of this thesis but a short discussi on is necessary to understand the effects that coil spacing has on power delivery. An ideal coil would have the turns placed as close together as possible without t ouching. This is because the co il is basically a solenoid and the magnetic flux from each coil adds to yi eld the resultant inductive field at the susceptor. Thus any spacing in the coil repres ents power losses due to fringing flux lines and this fact can be used to deliver more pow er to the front of the susceptor, where the power is needed. Gradually increasing the coil spacing from front to back can increase the temperature uniformity and counteract the cold gas finger effect. Pr operly designing a coil in this manner will intentionally in troduce power losses where needed and can drastically improve uniformity. The turning and testing of several coils resulted in an optimal design with the coil spacing as close as possible at the front of the growth zone and slightly increasing spacing near the end. Professionally turned coils could further optimize heat delivery and provide better spec ifications and should be considered in the future if susceptor performan ce is found to be inadequate. A standard method of testing temperat ure uniformity and calibrating heating instrumentation in SiC CVD is the so-calle d Si melt test. Since the melting temperature of Si is close to the growth temperature, one can use this to accurately define the temperature uniformity of the susceptor near growth conditions. In this test, small pieces of Si are placed on the graphite susceptor a nd then heated until melting of the Si pieces (at 1410C) are observed. Alterna tively, the Si pieces could be placed on a SiC substrate (on top of the susceptor) as to not damage the susceptor. An IR pyrometer is aimed at the back edge of susceptor so one must de termine the temperatur e error between the measurement point (susceptor back edge) to the top side where growth takes place.

PAGE 51

42 Basically, when the Si is observed to melt the surface temperature is exactly 1410C. The pyrometer temperature reading is then recorded and the difference ( T) is noted. Additionally, the temperature acr oss the growth surface must be as uniform as possible so several pieces of Si should be placed to measure and ve rify the uniformity. This calibration is the most accurate developed but may produce some inaccuracy since process temperature can be much higher or lower and T is not constant except at 1410C. However, this is a good starting poin t for CVD processing and is the accepted standard for most SiC CVD growers. Several melt tests were performed at U SF for various susceptor arrangements and specifications. The best design was observed to have a temper ature gradient of less than 15 C along the reaction area, which is reasonable when compared to other research grade CVD reactors that utilize similar power deliv ery systems[52]. The dimensions for this design were 80mm x 70mm x 20mm with a groove of 60mm x 50mm x 5mm removed from the bottom. All corners had a 5mm radius to prevent hot-spots (radiation losses) at the corners. 3.1.3 Processing Hazardous Gases Gas handling, exhaust, and t ube sealing management are critical in CVD due to the hazardous nature of the process gases. The hazardous process gases involved, namely SiH4, make design and testing paramount to sa fe operation. Proper design must ensure purging of all gas lines duri ng start-up and shut-down to safely handle these process gases. All seals and materials were selected to ensure maximum integrity and be versatile enough for research grade use.

PAGE 52

43 The manifold for gas distributi on is capable of handing Ar, H2, C3H8 (97% H2), SiH4 (97% H2), 3% N2 (97% H2), and an auxiliary gas for future work. The manifold provides 3 basic functions: pro cess gas to the reaction tube purge gas to the reaction tube, and process gas to th e vent. The PLC and PC cont rol these functions [30] by outputting user input and ensuri ng fail-safe operation. The 2 ba sic modes of operation are “process monitor” and “purge” mode. Venting of gases can be done in any mode but the other functions are restricted. Process gases can only be active when the control system is in “process mode” to avoid dumping flowing gas when the system is not properly configured. The process and vent function me ters, via mass flow controllers (MFC’s) under PC control, the supply gas and directs it to the appropriate line via an isolation valve that allows flow to ei ther the vent or process mani fold (or both). The isolation valves are controlled by the user via the c ontrol system front pane l which sends a signal to the corresponding solenoid base d on user input and internal safe operation checks. Gas flow to the process manifold is also controlled by the PLC to ensure the reactor is in the correct configuration, as st ated above. The purging function can deliver Ar or H2 from the gas manifold via a small section of stai nless-steel line prec eding the tube and subsequently through the tube and exhaust sy stem. Purging of Ar occurs automatically when any interlock is compromised. The mani fold was constructed of stainless steel using VCR connections so that the syst em was modular (for maintenance and modification purposes). Figure 3.4 is a diagram of the gas handling system, including the manifold and should provide the reader w ith a more complete understanding of the system.

PAGE 53

44 The seals of all process lines should be selected and tested as rigorously as possible. Even the tightest seals are subject to small leak rates due to pressure gradients. A standard vacuum check should be performed whenever a seal is compromised such as tube changing and routine maintenance. Altern atively, some lines are more appropriately checked by holding high pressure for a prol onged time period (12-24 hours). The most effective test used during the development was helium leak dete ction. This tool uses mass spectroscopy, tuned for He, to analyze the molecular flow inside the vacuumed lines being tested. The seals can be saturated with He by applying a small amount (~1 ml/min) around the perimeter of the seal which, if a leak exists at that point, is then drawn into the vacuum system and subsequently into the analyzer. Using this technique, a 10-8 Torr partial pressure He leak was found. The only dr awback to this test is that the vacuum conditions required (<50 mTorr) do not necessa rily represent the processing conditions. Essentially this means that leak rates ca n radically change over different pressure gradients and may not be found with the dete ctor. However, since 50mTorr is below the LPCVD process pressure, this represents a “wor st case” scenario and likely is the best means for securing many leaks in the system. A combination of the various methods can be applied to ensure the integrity of the gas handling system.

PAGE 54

45 Figure 3.4: Diagram of the gas handling syst em. The manifold is outlined by the dotted line as shown. SiH4 is pyroflouric even in dilute concen trations thus an explosive risk is possible during all phases of processing with SiH4. Additionally, SiH4 tends to produce SiOx particulates even with small amounts of O2 present, which may accumulate without proper design and maintenance procedur es. This residue can be from SiH4 reactivity with oxygen that out-gasses from the qua rtz reaction tube or from small leaks in the process lines. SiOx is an explosive risk in itself and must be eliminated whenever possible. For these reasons, purging is absolute nece ssity in all lines that contact SiH4 before and after use. A purge panel was salvaged and modifi ed for the purpose of purging the main gas

PAGE 55

46 line that goes from the SiH4 bottle to the isolation valve at the manifold. A standard Ar purge is used to completely purge down the lines after use. 3.2 LPCVD SiC System The design of the low-pressure system was implemented from June 2002 to May 2003 by the author. The need to stay with the current trends in CVD as discussed in Chapters 1 and 2 led to a horizontal low pressure reactor design. There are several elements to this design such as tube modi fication, plumbing design, pressure control, pump selection, and operation. Operation parame ters are subject to frequent change as new processes develop so the system was designed with maximum flexibility. Figure 3.5: Block diagram of the LPCVD sy stem developed during this thesis work. The new components are the low-pressure exhausting system shown on right. Microprocessor Control Unit Programmable Logic Controller Temperature Sensor Gas Supply and Delivery Control Console Reactor Cabinet Pumping Package Throttle Valve Pressure Controller Cooling Water Pressure Transducer Particulate Trap Pump Exhaust to Burn Box and Scrubber Dry Pump Reaction Tube Pressure Sensor RF Generator

PAGE 56

47 3.2.1 Basic Design Considerations The design of a low-pressure pumping syst em may be arrived at by attention to classic fluid flow design rules. Resistance of the fluid flowin g against the pipe walls and fittings is important in viscous flow regimes which are the operation parameters in this version of LPCVD. Molecular flow is typica lly recognized to occu r when pressures are less than 50 mTorr, which is not the case for the USF LPCVD reactor except during vessel purging. SiC LPCVD is conducted around 100 Torr. Therefore viscous fluid flow design equations dominate most considerations here. However, the system is subject to the molecular flow regime during initial pump downs to evacuate the vessel but this is not a condition that occurs during LPCVD gr owth. The vacuum design must take into consideration this most important function by estimating system pump down so that proper pump selection can be made. During a pumpdown from atmospheric conditions, the system passes through different types of flow regimes. As pumpdown proceeds, flow begins as very turbulent and transitions to laminar flow as the averag e velocity of molecules leaving the system decreases. These types of regimes are viscous with molecules dragging each other from a short mean free path. As the pressure is reduced further, the mean fr ee path lengthens and the molecule collisions decrea se brings the flow regime in to molecular flow. The term molecular flow is used to describe molecu les transferring back and forth in a random motion. The continuous force applied by the pump causes a pressure gradient and thus a density gradient. This density gradient changes the volumetric flow throughout the process lines but the mass flow rate w ill stay the same during steady state.

PAGE 57

48 The mass flow rate is ca lled the throughput (Q ) and is a measure of the number of molecules passing a given plane per uni t time. The units Torr-liters per second conventionally represent mass flow rate in vacuum systems which comes from the ideal gas law (pressure times volume is a measure of a quantity of gas). Considering a general system where gas is leaked in and subsequent ly pumped out one can see that the pressure is different at different parts of the system. At steady state the pressure decreases from the gas inlet to the reactor to the pump which is obviously the low pr essure point of the system. In this case the throughput is the sa me throughout the vacuum system since there is no accumulation of the molecules along the vacuum path. The volumetric flow rate at any point in the system is called the pumping speed (S). The units of pumping speed are thus liters per second. The throughput is the same throughout the system but the pumping speed is not since the gas is continuously expanding as it flows towards the pump. The vol umetric flow rate then increases as the fluid moves closer to the pump. The pumping speed and throughput are related by Q = SxP. Selecting vacuum component and si zing lines involves analyzing fluid resistance or the inverse (conductance) and can be determined by vacuum sizing equations [53]. The pipe walls and various geometries the fluid may have to travel through causes friction and thus resistance. With the obvious el ectrical analogy, the resistance Z is defined as the proportionality constant between pressure drop and mass flow and the inverse is the conductance C. 2 1P P Q C [1]

PAGE 58

49 In viscous flow regimes conducta nce can be further defined as L PD Cv 43000, [2] where P is the average pressure, in Torr, D is the tube diameter, in inches, and L is the length, in inches. Assuming the fluid is dry air at 20C the constant 3000 applies giving C the units of liters per second. All of this is, of course, assuming the ideal gas law and using Bernoulli’s equation to re late pressure difference. The conductance and pumping speed have the same units but they are not synonymous. Conductance is a property that ex presses the amount of three dimensional spaces while pumping speed is a property of position. The conductance of a fitting describes the resistance to the fluid from the volumetric and surface area while a pumping speed describes the mass flow rate for a given pressure. Calculating conductance is essential to designing a vacuum system as a larger conductance is more desirable for higher vacuum conditions. To arrive at a large conductance, large diameters and minimal lengths should be incorporat ed. More collisions with the walls occur for more wall surface area. In molecular flow, more collisi ons give the opportunity for a molecule to go the wrong way; hence pumpdowns occur mu ch more rapidly for higher conductance systems. The volumetric flow rate through the inlet of the pump is referred to as the pumping speed of the pump, Sp. If a passage from the vacuum chamber to the pump is known to be Ct, then the volumetric flow rate at the chamber, S1, is of interest for CVD. Using S1 to be the desirable volumetric flow ra te based on CFD simulations, one can size and select lines and fitting to obtain Sp that is reasonable to select a pump.

PAGE 59

50 Figure 3.6: Main parameters in sizing a vacuum system. S1 is the volumetric flow rate exiting the chamber and Sp is the pumping speed at the pump. Ct is the total conductance estimated through the lines a nd fittings that connect the chamber and the pump. In SiC CVD deposition of SiOx is a reality that must be considered in all components. A pleated paper particulate trap capable of capturing par ticles greater than 2 m was evaluated to be the best option to remove as much SiOx as possible. The only other option may be to heat the lines to re duce particulate accumu lation. This filter was placed about 20 inches downstream of the reac tion tube to promote particulate trapping upstream immediately from sensitive downstream instruments such as the pressure transducer and throttle valve. Equilibrium will shift as the flowing fluid cools and SiOx contamination is always a risk so one must select robust components that will tolerate harsh service. 3.2.2 Computational Fluid Dynamics Simulations The CFD simulations were performe d by Timothy Fawcett of the USF SiC group and were presented at the 2002 AIChE annual conference [54], which described some of this thesis work. The theoretical di scussion in section 2.1.1. was applied to this project via simulations using FIDAP 6.0 [55]. The mesh used to solve for the input and Chamber Pump Ct S1 Sp

PAGE 60

51 output of finite cells is displayed in Figure 3.7. A fine grid was used in the reaction area to maximize the accuracy of the analytical results shown in Figures 3.8-3.11. A Temperature profile for the APCVD and LPCVD cold-wall systems indicates that the gas is heated to desirable reacti on temperatures only directly near the susceptor. An 8 slm APCVD gas velocity profile with newly desi gned endcap for LPCVD is given in Figure 3.9 where laminar flow is evident in th e reaction area. 30 slm and 100 mBar LPCVD velocity profiles predicted an unstable region in the reactor in addition to the non-laminar flow across the susceptor which indicates undesirable operating conditions as shown in Figure 3.10. 10 slm and 100 mBar LPCVD veloc ity profile (see Figure 3.11) display some back mixing, however, laminar flow in the reaction area makes these process conditions feasible as a starti ng point to develop a process. Figure 3.7: Mesh for CFD calculations as performed by Timothy Fawcett [54]. Dimensions are as follows: Tube bore – 100m m, tube length – 640 mm, gas inlet – ”, gas outlet 40mm. Simulations performed using FIDAP 6.0. Gas Inlet Boat Susceptor Cold Wall Outlet

PAGE 61

52 Figure 3.8: Temperature profile CFD si mulation for APCVD and LPCVD cold-wall system as performed by Tim Fawcett [54]. Figure 3.9: Eight slm APCVD Velocity profile CFD simulation for system with new endcap as performed by Tim Fa wcett [54]. Laminar flow is evident in the reaction area. P = 1 atm Q = 8 SLM Wall Temp = 60C Inlet Temp = 25C Susceptor Temp = 1600C Re = 365

PAGE 62

53 Figure 3.10: Thirty slm and 100 mBar LPC VD velocity profile CFD simulation as performed by Tim Fawcett [54] An unstable region, in additi on to the non-laminar flow across the susceptor, indicates und esirable operation conditions. Figure 3.11: Ten slm and 100 mBar LPCVD velocity profile CFD simulation as performed by Tim Fawcett [54] Although some back mixing is evident, laminar flow in the reaction area make thes e process conditions feasib le as a starting point. P = 100 mbar Q = 10 SLM Wall Temp = 60C Inlet Temp = 25C Susceptor Temp = 1600C Re = 15 A little bit of back mixing P = 100 mbar Q = 30 SLM Wall Temp = 60C Inlet Temp = 25C Susceptor Temp = 1600C Re = 46

PAGE 63

54 3.2.3 Pressure Control The first step in design is recognizi ng the instrumentation that must be employed to provide desired functions. The ta rget for this system is vessel evacuation followed by pressure control in the 75-225 Torr range. Vessel evacuation was designed in the APCVD system so that was kept func tional by including a vent bypass to permit APCVD operation after upgrade to LPCVD. Howe ver, pump selection has to address this issue by providing a low ultimate vacuum for the reactor pump and purge processing step. The control loop consists of a pressure transducer, co ntrol valve, and a pressure controller which has been left out of the data acquisition loop for si mplicity sake although this option is left open for future work. By far the easiest and cheapest way to accomplish pressure control is purchasing a commercially available pressure controller. BOC Edwards model Barocel 600 was selected to provide maximum vers atility in the range of 0-1000 Torr. This model is corrosive resistant all stainless steel constructi on with a calibrated pressure measurement when exhaust system temperatur e is between 15-80 C. Cooling the piping with various muffin fans was used to prev ent the transducer el ement from accumulating heat is excess of this temperature spec ification. The correspo nding controller, BOC Edwards Downstream Pressure Controller Model 1800, is capable of proportional control of valve position base d on comparison of user inpu t and the signal from the transducer. The controller also will lock a desired valve pos ition, provide a “soft start”, and default to the last valve position if interrupted. The control valve selected, BOC Edwards Butterfly Valve Model 1850, is capable of controlling 15-37 slm which gives versatility over the wide range of future process parameters.

PAGE 64

55 Figure 3.12: The pressure c ontrol loop. Actual model num bers selected as shown. 3.2.4 Line Selection The detailed calculation performed during this thesis are listed is Appendix A. In this section the design pr ocedure used to select the vacuum system components is discussed and the reader is referred to the appe ndix for details. Line selection begins with basic space considerations as well as selec tion of fittings and instrumentation required. The reactor at USF had an area below th e cabinet specifically to house vacuum components thus line lengths were basically fixe d within this volume. This leaves the line diameter and the pumping speed as the only r eal variables to play with in the design equations presented. The goal of weighing th ese variables is to minimize the pumping speed while still have a practic al line diameter to install. High conductance, with a small line diameter compared to the line length (~8f t), results from the vi scous fluid flow being Pressure Transducer (BOC Edwards 600 ) Control Valve (BOC Edwards 1850 ) Pressure Controller (BOC Edwards 1800 ) Controller generates a signal that is proportional to the error found by the pressure sensor The control valve makes an adjustment in conductance Pressure signal is sent to controller where it is compared to setpoint User Input

PAGE 65

56 more significantly influenced by friction with flowing molecules than the walls. Appropriate engineering for this situati on would be to oversize slightly for any unforeseen increased demand and 40mm hardwa re is well above a minimal conductance range as per the calculations in Appendix A. NW40 hardware specifies 40mm diameter so it was selected. Furthermore, valve se lection for process requirements (15-37slm) would permit NW40 connection to be used. The pump selection can then be made based on the pumping speed calculated in Appendix A and considerations to the gases to be processed. 3.2.5 Pump Selection In the pressure ranges required for SiC LPCVD (75-225 Torr), there are two types of pumps that may be used; Oil-sealed mechanical pumps, or the more state-of-theart dry pumps. Pumping of hot corrosive gases such as SiH4 can lead to precipitation of hazardous SiOx on the contacted parts. Oil-seal ed mechanical pumps have been developed with purging capabilities as well as Fomblin Oil (only needed if O2 is present in system at high concentration or corrosive fluid) and filtration units to deal with accumulation of SiOx. Dry pumps can more effectively deal with SiOx by maintaining a high internal temperature and using interstage purging. Costs for each pump can be carefully evaluate d with attention not only to initial cost but maintenance cost that will be incurred during its life. The initial cost for mechanical pumps is usually lower than dry pumps for the same capacity but oil and maintenance costs over the pump’s life can quickly accumulate. The initial plan was to purchase a mechanical pump w ith minimal required capacity to bring the process up, which was 76 l/m (2.6 CFM) per Appendix A. The RV8

PAGE 66

57 BOC Edwards Mechanical Pump had adequate pumping speed specifications according to the pump curve presented in Figure 3.13. Eventually, a QDP40 BOC Edwards Dry Pump was available during the course of this project and was installe d to very adequately service this process. A pump curve fo r the QDP40 is presented in Figure 3.14. As with any research project, numer ous challenges were encountered which forced changes to the initial design made. Initial sizing of the pump indicated that the RV8 would be sufficient for initial proce ssing and a larger capacity pump would be purchased if needed to scale up the process. The flaw in this was the high operating pressure (~100 Torr) caused viscous flow through the pump oil and the process gases absorbed the oil subsequently draining the pump. While systems do exist to deal with this problem, pump maintenance and oil costs quickly add to near the cost of a high capacity dry pump so a dry pump was procured (BOC Edwards QDP40) and operation found to be satisfactory for the LPCVD operation. 3.13: Pump performance curve for the BOC Edwards model RV8 oil-sealed mechanical pump [56]. Note that P = 100Torr operati on permits ~ 5CFM pumping speed which is adequate for this LPCVD operation.

PAGE 67

58 Figure 3.14: Pump performance curve for the BOC Edwards model QDP40 dry pump [56]. Note P = 133 mBar = 100 Torr shoul d result in ~20 CFM pumping capacity. 3.2.6 System Installation The system installation was carried out from December 2002 to May 2003. Exhaust and pressure control hardware were mounted under the reactor cabinet as the APCVD design allowed for. NW40 flanges and fittings where used wherever possible to permit ease of installation and maintenance. The pleated paper filter element has NW25 flanges but extremely high conductance at lo w vacuum as specified by the manufacturer make the cost savings over the larger size worthy.

PAGE 68

59 Figure 3.15: Diagram of the LPCVD plum bing exhaust system designed during the course of this thesis. The rough and the ve nt1 lines are bypasses to allow AP operation and pumpdowns as provided by AP design [30]. 3.3 Hot-Wall One of the planned system upgrades was to develop a hot-wall epi process based on the work of Olle Kordina [43]. Dr Kordina’s academic work [9] contributed significantly toward the design of our eventual hot-wall hardware. The basic concept has been discussed in Chapter 2 and components fo r the hot-wall system were procured near the end of this thesis work. A hot-wall insert was designed using graphite foam as an insulator and to suspend the susceptor elem ent. Isothermal design of the susceptor Pressure Controller To Vent P P Filter P Reactor Exhaust AP Vent Valve AP Rough Valve VAC/Vent Valve N2 Purge 110V Inlet Shaft Seal Pump 208V Thermister Interlocks Motor Starter On/Off PWR

PAGE 69

60 assembly with minimal RF delivery is critical to this technique. A first iteration design was made with the help of Dr. Kordina and simulations by T. Faw cett (see Figures 3.14 and 3.15). RF delivery has to be considered to mi nimize coupling to the graphite foam and maximize coupling to the susceptor in order to achieve isothermal h eating of the reaction area at high temperatures (up to 2000C). A rectangular configura tion for the reaction area was not only detrimental to be optim um for fluid dynamics per Figures 3.16 and 3.17, but also for the susceptor design as well. A susceptor was procured to achieve an isothermal design. The outside corners have a significant chamfer on them to reduce the mass being heated and hotspots occurring close to the reaction tube walls. The two piece design makes machining easier and has been in dicated to increase coupling [9]. Testing and further development of hot-wall is left for future work by Shailaja Rao and S.E. Saddow of the USF SiC group. Figure 3.16: 10 slm and 150 Torr hot-wall LPCVD velocity profile CFD simulation performed by Tim Fawcett [54]. Laminar flow is predicted in th e reaction ar ea although slight turbulence near the head of tube is also apparent.

PAGE 70

61 Figure 3.17: 20slm and 150 Torr hot-wall LPC VD velocity profile CFD simulation as performed by Tim Fawcett [54]. Laminar flow is still predicted in the reaction area with more severe turbulence n ear the head of tube. 3.4 Summary The system design and installation was a compilation of work performed at MSU and USF. Extensive teamwork was used during the course of this project to utilize all skills available to the team. The product of this work is a totally integrated control system with extensive vacuum hardware and process tr ansport. Analytical techniques were used wherever feasible along with experienced input to develop a system capable of performing SiC LPCVD. Experimental LPCVD growth trials are the next step in evaluating this reactor design. APCVD runs we re conducted while this reactor was being further developed to have LP capabilities. T hose results will be pr esented in Chapter 4 and LPCVD future work will be suggested in Chapter 4.

PAGE 71

62 Figure 3.18: Researchers monitoring the USF CVD reactor during growth.

PAGE 72

63 CHAPTER 4 RESULTS AND CONCLUSIONS 4.1 Experimental Results The focus of this thesis was the design and development of a reactor capable of SiC epitaxy thus some experimental valid ation should be reported. The initial experiments performed in APCVD will be repo rted on and should serve as validation for the overall system design. The discussion in Chapter 2 indicates that LPCVD validation can be achieved through varying pressure, temperature, precu rsor and carrier gas flow rates. Modifications of the reactor vessel and susceptor may also be necessary to achieve fluid dynamics and/or a temper ature profile conducive to uniform growth. Although these items are directly linked to this thesis it is in the spirit of the team to pass this torch to other students. 4.1.1 APCVD Homoepitaxy Growth studies performed early during the course of development of this reactor will be reported on in this section. Namely gr owth run numbers: USF-02-010, USF-02-011, USF-02-012, USF-02-014, U SF-02-015, USF-02-016, and USF-02-017. Gas flow rates were determined based on past experience with this reactor design [17]. The growth temperature was set to 1600C a nd this was calibrated using a Si melt test and by setting the backside susceptor temp erature to 1390C. The carrier gas, 100% H2, flow rate in these runs was 8 slm. 3% SiH4 (balance H2) was always 80 sccm which was determined previously to have the appropria te saturation [17]. The flow rate of 3% C3H8

PAGE 73

64 (balance H2) was varied in accordance with the site competition doping method explained in Chapter 1 [23]. The Si/C ratio in the aforementioned runs was 0.13, 0.237, 0.11, 0.15, 0.18, 0.15 and 0.18, respectively. USF-02-016 and USF-02-017 were intentionally doped with 20 sccm 1% N2 (balance H2) to confirm site competition doping was taking place within the epitaxial layer. Growth time wa s 90 minutes in all cases. The samples were taken from the same wafer which was 6H, ntype, off-axis, production grade from Cree. The growth took place on the polished Si f ace. The samples were cut from this wafer using a wafer saw into approximately 1cm x 1cm pieces. The sample prep consisted of a 15 minute ultrasonic clean in Acetone, immediately followed by a 15 minute ultrasoni c clean in Isopropanol A deionized water rinse was carried out after the solvent cl eans and followed by a 10 minute hot bath (~75C) in sulfuric acid. Fina lly, a buffered oxide etch was performed prior to growth for 2-4 minutes. Rinsing with DI water took place before and after the oxide etch and the sample was finally dried with a N2 so as not to leave water residue. Two SiC-coated susceptors were used during the course of these runs that measured 70mm x 80 x 20mm. They were test ed to have approximately 5C variation about the sample area. Each su sceptor was conditioned for 90 mi nutes with a Si/C ratio of 0.15 to help ensure equilibria of the depositions during the runs. 4.1.2 APCVD Characterization Three specific techniques were used to confirm that epitaxy took place. The first method was optical inspection with a micr oscope at various magnifications. Image capture was not available at the time of thes e experiments thus no data can be presented here but the surface appeared specular whic h is typical of uniform epitaxy. Secondly, the

PAGE 74

65 samples were cleaved and the cross-sectio ns were analyzed in a scanning electron microscope (SEM). Due to the inherent di fficulty in cleaving SiC, only one sample displayed an epitaxial layer th at was clearly visible (USF02-010) and is shown in Figure 4.1. Lastly, CV and IV measurements were ta ken to determine doping profiles and results are shown in Figures 4.2-4.3. Figure 4.1: Cross-section SEM micrograph of a post-growth cleaved sample, USF-02010. The epitaxial layer is cleary visible to be ~3 m indicating a grow th rate of ~2 m/hr for this 90 minute growth run.

PAGE 75

66 Doping Density for various Si/C Mass Ratios 1E+15 1E+16 1E+17 1E+18 00.20.40.60.81 Depth (um)Density (cm^-3 ) 0.18 (doped) USF-02-017 0.18 USF-02-015 0.24 USF-02-011 0.15 USF-02-014 0.15 (doped) USF-02-016 0.11 USF-02-010 0.13 USF-02-010 Figure 4.2: Doping profiles determined by CV/IV measurements for the APCVD runs performed during this thesis work to validate the process. The Si/C for each run is indicated. Doping Density vs Si/C Mass Ratio 1E+15 1E+16 1E+17 1E+18 1E+19 00.10.20.30.40.50.60.7 Si/C Mass RatioDoping Density (cm^-3 unintentionally doped intentionally doped Figure 4.3: Linear regression an alysis of average doping densiti es for various Si/C ratios. This proves that the APCVD process developed during this thesis was consistent with the site competition model.

PAGE 76

67 4.2 LPCVD System Validation System validation was made by performing heat and pressure control tests at desirable carrier gas flow rate s. Desirable flow rates can be defined as 15-37slm in the pressure range of 75-225 Torr at growth temperature (~1600C). The system was designed with these parameters, as discusse d in Chapters 2 and 3, to provide maximum versatility over the lifetime of the reactor. Modifications to this system would likely be necessary to provide functiona lity outside of these parameters. The first LPCVD growth run, which proved the system performance to be within the design specifications, was conducted under the following process condi tions: Run No. USF-03-072, P = 35 Torr, T = 1580C, H2 = 10 slm, 3% SiH4 = 80 sccm, and 3% C3H8 = 80 sccm (Si/C = 0.30). The result was a very nice single crystal epi la yer also observed under the optical microscope. Further growth studies are required to fully validate the system [8]. 4.3 Summary LPCVD system design and development was intended to produce repeatable SiC epitaxy with doping control on substrates ranging in size of up to 2 inches while having fail-safe operation and adequate pro cess control. Growths rates are important factors as well which should produce uniformity of epitaxial layers and doping densities. A system capable of achieving such standa rds was developed for APCVD operation and further developed for LPCVD. The LPCVD a ddition was designed and implemented with the intention of refining epitaxy unifor mity and increase growth rates. The APCVD system was a reproduction of other systems designed by Dr. Saddow [30] and required intens ive work by this author, alon g with a supporting team, to procure, install, and debug all hardwa re. Doping and growth rate control was

PAGE 77

68 demonstrated by the 7 growth runs reviewed in this chapter as well as the numerous hours of growth runs performed by the author which included hydrogen etch runs as well as APCVD growth runs. The growth runs performed subsequent to the 7 growth run results illustrated in this chapter all produced similar results and is unnecessary to write an extensive review of this work. LPCVD validation consisted of demons trating low-pressure control during growth conditions. CFD simulations, performed by Tim Fawcett [54], give an indication of desirable flow rates to achieve these improvements with a new end cap design resulting from the authors design input. Grow th studies are further required to fully validate the reactor geometry. 4.4 Future Work To fully validate the LPCVD system a full matrix of experiments should be performed and subsequent characterization und ertaken. It would be optimal to perform these experiments on 2 inch wafers which are no t a suitable option due to the high cost of substrates. The shift in equilibria described in Chapter 3 should make it possible to perform these studies over a wider range of Si/C rati os and achieve lower doping densities and higher growth rates compared to APCVD. LPCVD reactor validation and further development was performed after the author left this section of the research group and moved on to other research areas within the university. The initial fo cus of growth studies after LPCVD system completion shifted to heteroepitaxy of 3C SiC on Si substrates [8]. With this change in focus also comes a change in focus for future works and reactor validation. The complete validation

PAGE 78

69 of the LPCVD system is detailed in the th esis of another member of the CVD growth team [8]. The reader is referred to this work for fu rther discussion along these lines. Figure 4.4: Author proudly standi ng in front of USF CVD reactor.

PAGE 79

70 REFERENCES [1] T. Fawcett, J.T. Wolan, R.L. Myers, J. Walker, and S.E. Saddow, "Hydrogen gas sensors using 3C-SiC/Si epitaxial layers," Late News Paper, International Conference on SiC and Related Materials (ICSCRM'03 ) 2003, Lyon, France, Oct. 6-10 2003. [2] A.R. Powell, L.B. Rowland, “SiC Mate rials – Progress, St atus, and Potential Roadblocks,” Proceedings of the IEEE, Vol. 90, No. 6, June 2002. [3] J.G. Pope, “Solid State Hydrogen Gas Se nsors Based on SiC,” MS Thesis, University of South Florida, July. 2003. [4] Stephen E. Saddow, Marina Mynbae va and Mike MacMillan, "Porous SiC Technology," Chapter 8 in Silicon Carbide: materials, devices and applications Editors: Zhe Chuan FENG and Jian H. ZHAO, as a vol ume of the book Series: Optoelectronic Properties of Semiconductors and Superlattices Editor in chief: M. O. Manasreh, Publisher: Taylor and Fr ancis Engineering, Jan. 2003. [5] H.S. Fogler, Elements of Chemical Reaction Engineering Prentice Hall, Upper Saddle River, NJ, 1999. [6] F.C. Frank, “Capillary Equilibrium of Disl ocated Crystals,” Acta Crystal, Vol. 4, p. 497, 1950. [7] S.J. Pearton, Processing of Wide Band Gap Semiconductors, Norwich: Noyes Publications, 2000. [8] R. Myers, “Growth of 3C-SiC on Novel S ubstrates,” MS Thesis, University of South Florida, Dec. 2003. [9] O. Kordina. “Growth and Characterization of Silicon Carbide Power Device Material.” Dissertation. De partment of Physics and Measurement Technology at Linkoping University, 1994. [10] H. Matsunami, K. Shibahara, N. Ku roda, W. Yoo, S. Nishino, “Amorphous and Crystalline Silicon Carbide,” Springer Pr oceedings in Physics, Vol. 34, pg. 34-39, 1989. [11] A. Ellison, “Silicon Carbide Growth by High Temperature CVD Techniques,” Dissertation No. 599, Dept. of Physic s and Measurement Technology, Linkoping University, Sweden, 1999.

PAGE 80

71 [12] S. Tyc, J. Physique I 4, 617, 1994. [13] K. Takahashi, M. Uchida, T. Yokogawa O. Kusumoto, K. Yamashita, R. Miyanaga, M. Kitabattake, “Surface Mo rphology of SiC Epitaxial La yers Grown by Vertical HotWall Type CVD,” Materials Scie nce Forum, Vols. 389-393, pg. 243-246, 2002. [14] M. Syvajarvi, R.Yakimova, I.G. Ivanov, E. Janzen, “Growth of 4H-SiC from liquid phase Epitaxy,” Materials Science & Engineer ing B: Solid-State Materials for Advanced Technology, Vol. B46, No. 1-3, pg 329332, Apr 1997. [15] S. Mahajan, K.S. Sree Harsa, Prin ciples of Growth and Processing of Semiconductors McGraw Hill Inc. 1999. [16] V.A. Dimitriev, S.V. Rendakova, V.A. Ivantsov, and C.H. Carter, “Method for Reducing Micropipe Formation in the Epitaxi al Growth of Silicon Carbide and the Resulting Structures,” U.S. Patent 5679153, Oct. 21, 1997. [17]T. E. Schattner, J. B. Casady, M. S. Mazzola, V. A. Dimitriev, S. V. Rentakova, and S. E. Saddow, "4H-SiC Device Scaling Devel opment on Repaired Micropipe Substrates" Materials Science Forum Vols. 338-342 ( 2000) pp. 1203-1206, Trans Tech Publications, Switzerland. [18] A. Sherman, Chemical Vapor Depos ition for Microelectronics, Principles, Technology, and Applications, Noyes Publications Westwood, New Jersey, U.S.A. Copyright 1987. [19] M.R. Leys, “The Chimney Reactor, A Co mpilation of Results,” Philips Research Laboratories, Eindoven, the Netherlands, 1985. [20] H. Pierson. Handbook of Chemical Vapor Deposition Norwich: Noye s Publications, 1999. [21] http://www.epigress.com/rammeny_ejjava.html [22] H.S. Kong, J.T. Glass, R.F. Davis, “Chemical Vapor Deposition and Characterization of 6H-SiC thin films on off-ax is 6H-SiC Substrates,” Journal of Applied Physics, Vol. 64, No. 5, pg. 2672-2679, Sept., 1988. [23] D.J. Larkin, P.J. Neudeck, J.A. Powell, L.G. Matus, “Site-competition Epitaxy for Superior Silicon Carbide Electronics,” A pplied Physics Letter, Vol. 65, No. 3, pg. 1659, Sept. 1994.

PAGE 81

72 [24] W.J. Choyke, “The Physics and Chemis try of Carbides, Nitrides, and Borides,” NATO ASI Series E: Applied Sciences, edit ed by R. Freer, Vol. 85, pg. 1-111, Kluwer, Dordrecht, 1990. [25] M. Capano, S. Ryu, M.R. Meeoch, J.A. Cooper, and M.R. Buss, Journal of Electronic Materials, Vol. 27, No. 4, 1998. [26] S.E. Saddow, J. Williams, T. Isaac s-Smith, M.A. Capano, J.A. Cooper, M.S. Mazzola, A.J. Hsieh, J.B. Casady, “High Temperature Implant Activation in 4H and 6HSiC in a Silane Ambient to Reduce Step B unching,” Materials Science Forum, Vols. 338342, pg. 901-904, 2000. [27] M. Shirohara et al, Journal of Applied Physics, Vol. 27, pg. L434, 1988. [28] T. Takahashi et al, “Surface Morphology of 3C-SiC Heteroepitaxial Layers Grown by LPCVD on Si Substrates,” Material s Science Forum, Vols. 264-268, pg. 207-210, 1998. [29] M.T. Burke, “Design and Simulation of a CVD Reaction Tube for Silicon Carbide Epitaxial Growth,” Department of Electri cal and Computer E ngineering, Mississippi State University, EE-4012, 1997. [30] T. Schattner, “Homoepitaxial Growth of 4H and 6H-SiC in a 75mm Reactor,” Masters Thesis, Mississippi State University, Mississippi, May 2000. [31] W.J. Thompson, Introduc tion to Transport Phenomena Prentice Hall, Upper Saddle River, NJ, 2000. [32] T. Fawcett, Research Experience fo r Undergraduates Symposium, University of South Florida, December 2002. [33] R, Lohner, “Finite Elements Method in CFD,” International Journal for Numerical Methods in Engineering, Vol. 24, No. 9, pg. 1741-1756, Sep 1987. [34] J.R. Elliot, C.T. Lira, Introducto ry Chemical Engineering Thermodynamics Prentice-Hall, Upper Sadle River, NJ, 1999. [35] A.L. Hines, R.N. Maddox, Mass Transfer: Fundamentals and Applications PrenticeHall, Upper Sadle River, NJ, 1985. [36] E.L. Cussler, Diffusion: Ma ss transfer in Fluid Systems Cambridge University Press, Cambridge, UK, 1997. [37] P. Atkins. Physical Chemistry W.H. Freeman and Company, New York, 1998.

PAGE 82

73 [38] O. Levenspiel, Chemical Reaction Engineering John Wiley and Sons, United States of America, 1999. [39] H.S. Taylor, Proc. R. Soc. London, A108, 105, 1928. [40] C.D. Stiriespring, J.C. Wormhoudt, J ournal of Crystal Growth, Vol. 87, pg 481, 1988. [41] M.D. Allendorf, R.j. Kee, Journal of the Electrochemistry Society, Vol. 138, pg 841, 1991. [42] T.U.M.S. Murthy, N. Miyamoto, M. Sh imbo, J. Nishizawa, Journal of Crystal Growth, Vol. 33, pg. 1-7, 1976. [43] O. Kordina, Personal consultation with Olle Kordina. [44] A.A. Burk, L.B. Rowland, Journa l of Crystal Growth, Vol. 167, pg. 586-595, 1996. [45] H. Muller-Frumbhaar, T.W. Burkhardt, D. M. Knoll, Journal of Crystal Growth, Vol. 38, pg. 13-22, 1977. [46] M.D. Allendorf, Journal of the Electrochemical Society, Vol. 140, pg. 747-753, 1993. [47] T.L. Chu, R.B. Cambell, Journal of the Electrochemical Society, Vol. 112, pg. 955956, 1965. [48] T.L. Chu, R.B. Cambell, Journal of the Electrochemical Society, Vol. 112, pg. 955, 1965. [49] M.S. Silberberg, Chemistry: The Molecular Nature of Matter McGraw Hill, 2000. [50] P.M. Lofgren, W. Ji, C. Hallin, C.Y. G u, “Modeling of Silicon Carbide Growth in a Hot-Wall CVD Process,” Journal of the El ectrochemical Society, Vol. 147, pg. 164-175, 2000. [51] G. Melnychuk, Y. Koshka, S. Yingquan, M. Mazzola, C. U. Pittman, “Computational Modeling for the Devel opment of CVD SiC Epitaxial Growth Processes,” European Conference on SiC and Related Materials ( ECSCRM'02 ), Linkping, Sweden, September 1 5, 2002. [52] D.J. Larkin, private communication.

PAGE 83

74 [53] T. Gessert, R. Outlaw, H. Patton, “Fundamentals of Vacuum Technology,” AVS Short Course Program, Burlington, MA, 2002. [54] M.T. Smith, T. Fawcett, J. Wolan, S. Saddow, “Design and Development of a SiC LPCVD Reactor,” AIChE Annual Conference, Nov 2002. [55] http://www.fluent.com/software/req/nt.htm [56] http://www.bocedwards.com

PAGE 84

75APPENDICES

PAGE 85

76] / _[ _ ] / _[ int int_ _ ] / _[ tan ] _[ ] _[ ] _[ s l pump at speed pumping S s l erest of po at speed pumping S s l ce conduc C inches lenght L inches Diameter D Torr pressure average Pp APPENDIX A: VACUUM DESIGN CALCULATIONS Molecular Flow Vacuum Design Calculations Assuming cool dry air is the fluid L D P C380 16 0 96 57 1 10 50 804 3 hardwareC hardware tube totalC C C 1 1 1 24 6totalC Note: Conductance not of important for low vacuum in this system. Only Pumpdown times… P P S V tiln l mm mm mm V V Vhardware tube10 2400 ) 20 ( 900 ) 50 (2 2 With QDP40 Pump with speed approximated to 2.7 l/s (will change radically during pumpdown) 310 50 760 ln 7 2 10 t onds t sec 34 24 36 6 10 50 803 3 tubeC

PAGE 86

77] / _[ _ ] / _[ int int_ _ ] / _[ tan ] _[ ] _[ ] _[ s l pump at speed pumping S s l erest of po at speed pumping S s l ce conduc C inches lenght L inches Diameter D Torr pressure average Pp Appendix A (continued) Viscous Vacuum Design Calculations Assuming cool dry air is the fluid L D P C43000 4 410 9 1 96 57 1 100 3000 hardwareC hardware tube totalC C C 1 1 1 410 9 1 totalC 2 2 1 1V P V P 2* 100 10 760 V slm Torr s l m l V / 26 1 / 762 total pC S S 1 1 1 410 9 1 1 1 25 1 1 pS m l s l Sp/ 76 / 26 1 7 410 1 36 6 100 3000 tubeC