DETERMINATION OF THE GAS-FLOW PATTERNS INSIDE THE HOT-

WIRE CHEMICAL VAPOR DEPOSITION SYSTEM, USING A

COMPUTATIONAL FLUID DYNAMICS SOFTWARE (FLUENT)

By

Thobeka Wittes

A thesis submitted in partial fulfillment of the requirements for the degree of Magister
Scientiae in the Department of Physics, University of the Western Cape.

Supervisor: Prof. D. Knoesen

Dept. of Physics

University of Western Cape

July 2009
 DECLARATION

I, the undersigned, hereby declare that the work contained in this thesis is my own

original work and that I have not previously in its entirety or in part submitted it at

any university for a degree

Signature: ……………………………………………………………..

Date:…………………………………………………………………..
 DETERMINATION OF THE GAS-FLOW PATTERNS INSIDE THE HOT-

WIRE CHEMICAL VAPOR DEPOSITION SYSTEM, USING A

COMPUTATIONAL FLUID DYNAMICS SOFTWARE (FLUENT)

BY

THOBEKA VICTORIA WITTES

KEYWORDS

CFD

HWCVD

Gas flow

Silane

Hydrogen

i
ABSTRACT

Computational Fluid Dynamics is the analysis of a system involving fluid flow, heat

transfer and associated phenomena such as chemical reactions by means of a

computer-based simulation. The simulations in this study are performed using (CFD)

software package FLUENT. The mixture of two gases (Silane gas (SiH4) and

Hydrogen gas (H2)) are delivered into the hot-wire chemical vapor deposition system

(HWCVD) with the two deposited substrates (glass and Silicon). This process is

performed by the solar cells group of the Physics department at the University of the

Western Cape. In this thesis, the simulation is done using a CFD software package

FLUENT, to model the gas-flow patterns inside the HWCVD system. This will show

how the gas-flow patterns are affected by the varying temperature of the heater in

each simulation performed in this study under a constant pressure of 60 Bar of the

system.

ii
ACKNOWLEDGEMENTS

Firstly I would like to thank God Almighty who kept me up until this far, guiding me

and giving me strength, knowledge and wisdom and love to succeed. Above all, I

thank God for who he is in my life.

I am truly thankful to the following people for their contribution towards making this

thesis possible:-

To my parents, for allowing me the opportunity of finishing my

studies, and their continuous support and prayers throughout this study.

Prof. Dirk Knoesen (University of the Western Cape), supervisor, for

his continuous guidance and support throughout this study.

Mr. Ian Schroeder (University of the Western Cape), for valuable

guidance and support throughout this study.

My precious family, amaHlotshane, oHala, Sophico, Yemyem,

Qholomsila, Vela-baMbhentsele, words cannot explain how I feel

about you guys. Ke Modimo fela a itseng, mara e, keya lona.

uLonwabo-lwam Wittes, my boy I thank God every time I remember

you “Philippians 1:3”, you have given me a reason to live and to look

forward to the next day of my life.

To Nombulelo Agnes Mthembu and her husband. Mmangwane

Modimo a go hlogonolofatse le lelapa la gago ka moya o montle and

for planting this seed.

iii
Nathan Zhou, Tshepo Samuel Dinoko and Stephan Smittch for

constant support and guidance throughout this study and the

background of CFD, may the good Lord do you good.

Mr. and Mrs. Maleka for your continuous guidance and support

throughout this study, and letting us into your office.

To my friends, Promise Tlhologelo Maredi, Steven Solethu Nkosi,

Takalani Madima, Nombulelo Agnes Khwaza, Thato Tshabalala, for

their moral support they provided throughout this study.

To my precious four sisters and brother, Nongetheni Rosina, Lumkile

Phillip, Nomthandazo Florence, Tombi Evelyn, Sindiswa Mildred

Wittes for their continuous prayers, moral support, patience and

guidance throughout this study.

To Khwaza’s family for your prayers and initiating my success may

the good Lord increase the gift that he has given to you.

To my home church, the Christian Apostolic Church in Zion of South

Africa (Lichtenburg District); for starting this journey of my success.

May the good Lord do you good.

I cannot stop thanking my parents, Nozizwe Elizabeth and Vuyo Petrus

Wittes for their prayers and love throughout my life. May the good

Lord bless you and keep you for me.

University of the Western Cape (UWC), iThemba LABS and the

National Research Foundation (NRF) for both financial and moral

support throughout this study and, lastly to my friends and those I have

come into contact with over this study for their constant motivation.

iv
To my family: Vuyo (Dad), Nozizwe (Mom),
Nongetheni, Lumkile, Nomthandazo,
Tombi, Sindiswa, Zolani,
Dumisani, Unathi, Mpendulo
no Lonwabo Wittes.

v
LIST OF FIGURES

Figure 1: The hot-wire chemical vapor deposition system used at the Physics department of

the University of the Western Cape. 8

Figure 2.1: Schematic cross-section of HWCVD system. 12

Figure 2.2: A domain (geometry) of HWCVD system as used in this study, designed using

Gambit. 17

Figure 2.3: A grid/mesh of the geometry of the HWCVD system. 18

Figure 3.1: A flow chart of the procedure followed in a CFD problem. 27

Figure 3.2: A domain of the gas-inlet pipes 29

Figure 3.3: A domain of the filaments, insulators and filaments stand. 30

Figure 3.4: Deposition chamber (CVD) with all the parts imported to it. 30

Figure 3.5: A grid/mesh of CVD system, with the total number of 657120 cells. 33

Figure 3.6: An overview of the segregated solution method when iterating. 38

Figure B1: Temperature contours. 60

Figure B2: Temperature contours with a plane drawn at (y0=-15, y1=15, y2=15) axis. 61

Figure B3: Pathlines colored by the temperature. 62

Figure B4: Pathlines colored by the velocity. 63

vi
LIST OF TABLES

Table 3.1: Some of the measurements of HWCVD system. 28

Table 3.2: The general data of the HWCVD system. 34

Table 3.3: Boundary conditions of HWCVD. 35

Table 3.4: Material properties used in this study. 36

Table 4.1: Mass flow-rate results for all simulations. 41

Table 4.2: Surface deposition-rate results for all simulations. 45

Table A1: A detailed CFD model used in this study. 55

LIST OF GRAPHS

Graph 4.1: A plot of the mass flow-rate on the filaments, inlets, outlet and two substrates vs.

temperature of the heater for all simulations. 42

Graph 4.2: A plot of the net mass flow-rate results around the filaments, inlet , outlet and the

two substrates vs. temperature of the heater. 43

Graph 4.3: A plot of the surface deposition-rate on the filaments, the two substrates and

substrate-holder vs. temperatures of the heater. 46

vii
NOMENCLATURE

Symbols:

Variable Description Unit

x : distance in the x-direction -

y : distance in the y-direction -

z : distance in the z-direction -

u : velocity in the x-direction m/s

v : velocity in the y-direction m/s

w : velocity in the z-direction m/s

t : time s

ρ : density kg/m3

Pr : Prandtl number -

q : heat flux Wm-2

Sccm : standard cubic centimeter per minute -

τ : stress N/m2

: dynamic molecular viscosity kgm-1s-1

δij : kronecker delta function -

λ : bulk viscosity m/s2

: kinematics viscosity m2/s

V : velocity vector m/s

T : temperature K

Sm : mass source kg/m3-s

p : pressure Pa

Ѓ : exchange coefficient (laminar + Turbulent) -

viii
Cp : specific heat at constant pressure J/kg-k

g : gravitational acceleration m/s2

K : thermal conductivity W/m-k

D/Dt : substantial derivative -

φ : dependent variable -

Abbreviations:

CFD Computational Fluid Dynamics

HWCVD Hot-wire Chemical Vapor Deposition System

ix
CONTENTS

KEYWORDS ……………………………………………………………………......i

ABSTRACT ……………………………………………………………………......ii

ACKNOWLEDGEMENT …………………………………………………………iii

LIST OF FIGURES ……………………………………………………………....vi

LIST OF TABLES ………………………………………………………………..vii

LIST OF GRAPHS ……………………………………………………………….vii

NOMENCLATURE ………………………………………………………………viii

CHAPTER 1 …… ……….…………………………………………………………...1

Introduction …………………………………………………………………………1
1.1 Background …………………………………………………………………….....1

1.2 Computational Fluid Dynamics ………………………………………………….2

1.2.1 The CAD design (Gambit)…………………………..………………4

1.2.2 Solver …………………………………………………………….......5
1.2.3 Post-processor……………………………………………………......7

1.3 Hot-wire Chemical vapor deposition system ……………………………….7

1.4 The Aim and Outline of this study …………………………………………..9

CHAPTER 2 ………………………………………………………………………...11

Techniques used in this study ………………………………………………………11

2.1 HWCVD .............................................................……………………………….11

2.1.1 Background …........................................................................................11

2.1.2 HWCVD Process ...................................................................................13
.
2.2 Background on Gambit (CAD design) ………………………………………15

2.3 Mesh Generation …………………………………………………...................16

2.4 Background on FLUENT …………………………………………………....18

x
 2.5 Finite Volume Methods ……………………………………………………....19

2.6 Mathematics behind CFD ……………………………………………………20

2.6.1 Navier-Stokes equations ……………………………………………20

2.6.1.1 Conservation of Mass …………………………………………….21
2.6.1.2 Momentum equations …………………………………………….22
2.6.1.3 Energy equations ………………………………………………....23

2.7 Steady state conditions and computational accuracy in FLUENT ………….24

CHAPTER 3 ………………………………………………………………………..26

SIMULATION METHODS ……………………………………………………….26

3.1 Simulation Methods ………………………………………………………...26

3.2 Designing the domain/geometry ……………………………………………27

3.3 The deposition parameters of the HWCVD system ………………………33

3.4 Starting the Simulation in FLUENT ………………………………………35

CHAPTER 4 ………………………………………………………………………..39

4. RESULTS AND DISCUSSIONS....……………………………………………39

4.1 Introduction ………………………………………………………………….39

4.2 Mass flow-rate ………………………………………………………………..40

4.3 Surface Deposition-rate ……………………………………...........................44

4.4 Discussions ……………………………………………...................................47

CHAPTER 5 ………………………………………………………………………..48

5. SUMMARY AND CONCLUSION ……………………………………………48

References ………………………………………………………………………….50

APPENDICES ……………………………………………………………………...54

xi
Appendix A ……………………………………………………………...54

A.1 Introduction ……………………………………………………..54

A.2 Calculation of the magnitude of inlet velocity …………………56
A.3 Calculations of the heat flux of the filaments ………………….56
A.4 Species Mass Fraction …………………………………………..57

Appendix B ………………………………………………………………59

B.1 Contours and Path-lines ………………………………………...59

xii
CHAPTER 1

INTRODUCTION

1.1 Background

In this study we will use the Computational Fluid Dynamics (CFD) software

FLUENT to model the gas flow patterns inside the Hot-wire chemical vapor

deposition system (HWCVD), by designing the internal geometry of the HWCVD;

using the FLUENT CAD design system GAMBIT.

The CFD software (FLUENT) is divided into three steps; namely [1, 13]:

• Pre-processor stage (which is the computer-aided design (CAD) software

package Gambit) where the modeling goals are defined (in this study to model

the gas-flow patterns inside the hot-wire chemical vapor deposition system,

using the CFD software FLUENT), the domain to be modeled is

identified/designed (the HWCVD system), finally design and create a

grid/mesh from the geometry/domain designed.

• Solver stage, where the numerical model is set, then computes and monitor

the solution.

• Post-processor stage, where the results are examined and if the results are not

satisfying then consider revision to the model.

1
1.2 Computational Fluid Dynamics

CFD is the analysis of a system involving fluid flow, heat transfer and associated

phenomena such as chemical reactions by means of computer-based simulation. The

CFD is also a numerical modeling technique that solves the Navier-Stokes equations

on a discretised domain of the geometry of interest with the appropriate flow

boundary conditions supplied. The Navier-Stokes equations are a complex non-linear

set of partial differential equations that describes the mass and momentum

conservation of a fluid [1, 13].

From the 1960s onwards the aerospace industry has integrated the CFD technique into

the design, research and development of the manufacturing of aircraft and jet engines.

More recently the methods have been applied to the design of internal combustion

engines, combustion chambers of gas turbines and furnaces. Furthermore, motor

vehicle manufactures now routinely predict drag forces, under-bonnet air flows and

the in-car environment with CFD. Increasingly CFD is becoming a vital component in

the design of industrial products and process [1, 13].

The aim of developments in the CFD field is to provide a capability comparable to

other computer-aided engineering tools such as stress analysis codes.

There are several unique advantages of CFD over experimental-based approaches to

fluid systems design which are given below [1, 13]:

• Substantial reduction of lead times and cost of new designs.

• Ability to study systems where controlled experiments are difficult or

impossible to perform (e.g. very large systems).

2
 • Ability to study systems under hazardous (dangerous) conditions at and

beyond their normal performance limits (e.g. safety studies and accident

scenarios).

• Practically unlimited level of detail of results.

The CFD code (FLUENT) has three basic components: the pre-processor, solver and

the post-processor. The solver is the heart of the code; carrying out the major

computations and providing the numerical solutions. The pre-processor and the post-

processor are the front and the end of the code, providing the user/machine interface

that allows a CFD operator to communicate with the solver: by inputting data to

define the problem to be simulated and commanding the solver to use certain models

and schemes to carry out the simulation and, finally, presenting for study the

computed results [1, 13].

Apart from these key elements, a commercial package aimed at multi-purpose

modeling will have a suite of models for various flow problems, such as various

turbulence models to cover a range of turbulence conditions and assumptions. The

CFD package also has a library of material properties for defining the fluid media and

solid boundaries in the computational domain [1].

3
1.2.1 The CAD design (GAMBIT)

Drafting in a CAD design can be done in two dimensions (2D) or three dimensions

(3D). In this study the drafting of the internal geometry of the HWCVD system was

done in 3D. With this the system is represented by three projected views at right

angles. This allows rotations in three dimensions, allowing viewing of a design from

any desired angle, even from the inside looking out [9].

The CAD design which was used in this study is GAMBIT (Geometry and Mesh

Building Intelligent Toolkit). It has a number of functions that allows the user to

define a fluid domain, known as the computational domain, and build up the physical

geometry of the zone considered, creating mesh/grid system throughout the domain

and tuning this mesh to improve computation quality, including accuracy and speed.

The user may also specify the properties of the fluid and other materials in contact

with the fluid at this stage [1, 9].

The pre-processor (GAMBIT) allows the user to define the fluid flow phenomena to

be modeled, to choose from the appropriate physical or chemical models provided by

the package, and to select the numerical parameters and initial and boundary

conditions for the computation. All these features are crucial in ensuring a quality

CFD modeling task [9].

For example, the mesh/grid refinement at this stage (GAMBIT) affects directly the

accuracy of the solution and the cost of computation, in terms of computation time

4
and hardware requirement. An optimized mesh/grid system uses less computer

memory space, requires less computing, and yet gives satisfactory accuracy [1, 9].

1.2.2 Solver

As mentioned above, the solver stage is the heart of a CFD code. It is a collection of

various algorithms and numerical techniques that performs the major computation

task described in the pre-processor above. It consists of two components that provide

a discretization for the defining equations and subsequent (resultant) solution [1, 9].

The first component uses a discretizing scheme to express the governing equations in

a discretized form for all mesh/grid elements over the whole computational domain

and discretizes the boundary equations appropriate at the boundary elements. In

summary, this section converts the partial differential equations and boundary

conditions formulae into a group of algebraic equations. The second part uses an

iterative procedure to find solutions that satisfy these boundary conditions for the

algebraic equations defining the flow domain [1, 9].

The solver in the CFD code (FLUENT), is based on the finite volume methods,

which integrates the governing equations in the finite volumes (control volumes) over

the whole computational domain. There is one generic form of equation for one flow

variable, φ which could be a velocity component, enthalpy or species concentration:

5
∂ρφ
= −∇(ρVφ ) + ∇(Γφ ∇φ ) + Sφ (1.1)
∂t

Where:

φ = dépendent variable

Γφ = exchange coefficient (laminar + turbulent)

V = flow velocity vector

ρ = flow density

Sφ = source of the variable

This integrated expression in equation (1.1) above means that the rate of change of

φ in the control volume (cells from the mesh/grid) with respect to time is equal to the

sum of the net flux of φ due to convection into the control volume, the net flux of

φ due to diffusion into the control volume and the net generation rate of φ inside the

control volume. Following integration, the solver uses various approximations, which

are based on an application of the finite difference method, to replace all the terms in

the integrated equation (the time-change rate, the convective and diffusive fluxes and

the source term) [1].

This process converts the equation into a set of algebraic equations, which are ready

to be solved by an iterative method. The iteration is the third and the last action the

solver does in the whole computation task (simulation) [1].

6
1.2.3 Post-processor

The post-processor allows the CFD operator to construct a picture of the simulated

flow problem by displaying the geometry of the problem, the computational domain

and the mesh/grid system. It may then display contour or iso-surface plots for the flow

variables, including contours plotted over specified surfaces, such as on a solid/fluid

interface or an iso-surface of a second flow variable. In many cases velocity vector

plotting is important, as is streamline presentation or particle tracking or pathlines [1].

A complete CFD simulation often requires repeating the procedure a number of times.

It is not rare to run a large number of trials before reaching a set of reasonable

solutions. The procedure includes tuning the mesh system, adjusting boundary

conditions, selecting numerical parameters, finding the correct physical model,

monitoring iteration and finally, viewing the results [1,13].

1.3 Hot-wire Chemical vapor deposition system

The technique of HWCVD is being increasingly used for the deposition of materials

such as hydrogenated amorphous or microcrystalline silicon and its alloys and as

diamond films. The main part of the deposition system consists in a vacuum chamber

evacuated by a pumping unit to some appropriate vacuum level. After achieving the

desired ultimate vacuum, the process of gas mixture is introduced via Mass flow

controllers and the pressure is kept constant by a variable conductance valve. The

process of gas is then decomposed by the heat generated by a hot filament of tungsten,

tantalum or other material into radicals which are then deposited as thin films [32].

7
 Figure1: The hot-wire chemical vapor deposition system used at the Physics

department of the University of the Western Cape.

The HWCVD system, requires relatively simple apparatus, offer the advantages of

high deposition rate, low hydrogen content and effective decomposition of gas

molecules. HWCVD shows great potential in large area electronics application. For

the application of this technique in commercial products, the uniformity of film

thickness over a large scale is particularly important.

Usually, the thickness uniformity of film in HWCVD is mainly effected by gas

supply, catalyzer structure, diffusion of radicals and the temperature distribution of

8
substrate. A uniform and sufficient gas flow is required, because the deposition

process is a gas flow limited one [33].

The HWCVD process has been extensively used for the deposition of various

materials, including diamond, polymers, silicon thin films, boron-carbon-nitrite layers

and carbon nanotubes. The process relies on the catalytic decomposition of precursor

gases into reactive gaseous species by a resistively heated filament. A variety of

transition metals, such as tungsten (W), tantalum (Ta) and rhenium (Re), have been

utilized as the filament during the HWCVD process. This is primarily due to their

high melting points and superior mechanical stabilities at temperatures above 1500˚C.

Furthermore, different filaments have different abilities to dissociate the precursors.

For example, a Ta filament dissociates molecular hydrogen (H2) into atomic hydrogen

(Ho) twice as efficient as a tungsten filament during the deposition of microcrystalline

silicon thin films [34].

1.4 The Aim and Outline of this study

In this study, we want to determine the mixture of the gas flow patterns inside the

HWCVD system, by using the CFD software, FLUENT; using information known

from the solar cells group of the physics department in the University of the Western

Cape. This includes the flow rate of the silane gas (SiH4) at 1.2 sccm (standard cubic

centimeter) and for the hydrogen (H2) gas at 58.8 sccm, which are the two gases used

during the deposition on two substrates (silicon and glass). This information is then

9
used in this study to calculate the species mass fraction for the inlets boundary

conditions.

The two gases (silane and hydrogen) are introduced into the HWCVD system with the

two deposited substrates (silicon and glass) under a constant pressure of 60 bar and

the heat flux of the filament is about 99.21x103 W/m2, by varying the temperature of

the heater of 100 ˚C, 200 ˚C, 300 ˚C and 550 ˚C; the simulation shows how the

temperature affects the flow of the gas flow patterns inside the system.

In the second chapter we will give an overview of the HWCVD; a short historical

overview of the technique is given and key issues pertaining CFD and this study are

discussed. We will look at the background of the finite volume method, the

mathematics involved in CFD (the Navier-Stokes equations).

The third chapter will give a comprehensive simulation method for this study.

Chapter four will focus on the main results obtained from the simulation and

discussions and conclusion. References are given at the end of the chapters in these

study and appendices giving the description of the simulation method used in this

study. The contour plots and pathlines showing the movement of the gas inside the

CVD system obtained from the results in this study.

10
CHAPTER 2

TECHNIQUES USED IN THIS STUDY

2.1 HWCVD

2.1.1 Background

A HWCVD technique was introduced in 1979 by Wiesmann et al [7] as a method for

producing hydrogenated amorphous silicon thin films. However, the results of

Wiesmann were not encouraging because they used too low pressures of 5.33 x 10-4

mbar which resulted in a poor dark conductivity of the thin films as compared to the

materials prepared by plasma enhanced chemical vapor deposition (PECVD). This

technique has found its application in 1991 when Mahan et al reported on the

development of device quality a-Si:H thin films with low hydrogen content [8].

The HWCVD technique is based on the deposition of the source gases, usually silane

(SiH4) or a mixture of H2 and SiH4 gases on a heated filament. This technique is also

called the Catalytic Chemical Vapor Deposition (Cat-CVD). Figure 2.1 shows the

schematic cross-section of the HWCVD reaction chamber; in this study the gas flow

patterns inside the HWCVD are determined using a CFD software (FLUENT) as

outlined in chapter1. The source gases H2 and SiH4 are injected into the chamber and

decomposed on a heated filament. On the way from the filament to the substrates

(glass and silicon which are the two substrates), the constituent radical species

11
produced on the heated filament often react in the gas phase before deposition onto

the heated substrate [8].

Figure 2.1: Schematic cross-section of an HWCVD system, with the inlet at the top

and outlet on the side of the chamber [26].

Figure 2.1 above shows the cross-section of a HWCVD system. The deposition

materials used in the HWCVD system in this study, by the solar cells group in the

physics department of the University of the Western Cape, are based on H2 and SiH4

gases.

12
2.1.2 HWCVD Process

Hot-wire was first introduced by Wiesmann et al. [7] in 1979 when he succeeded to

deposit a Si: H by thermal decomposition of silane from a hot tungsten or carbon foil

heated to about 1600˚C. Unfortunately due to very low gas pressures (1.33 x 106 –

1.33 x 10-4mbar), he achieved too low deposition rates (0.6 – 8.7 Å/sec) and films

with low quality due to the surface filament were not understood at that stage.

Evidence of the catalytic nature of the decomposition has now been found: The 12.9

eV needed to pyrolytically cleave four Si-H bonds [27] is much higher than the

activation energy of Si radical on the surface of the main wires used (Mo, Ta and W)

of (2.6, 1.0, and 0.74 eV) respectively as reported by Tange et al [28].

Later, in 1986 Matsumura [29] obtained high quality hydro-fluorinated a-Si (a-Si: F:

H) and hydrogenated amorphous silicon (a-Si: H) by thermal catalytic reactions

between deposition-gas and a heated tungsten catalyzer. He named it “catalytic –

CVD (CTL- CVD)” method. A deposition rate larger than 20 Å/sec was achieved and

more importantly, he reported that the magnitude of the SWE of the films produced

seemed smaller than that of the glow discharge a-Si: H produced films.

J. Doyle at al [30] using what they referred to an “evaporative surface decomposition

(ESD) method” reported in 1988 a deposition of excellent a-Si: H film quality from

silane decomposition with silane pressures ranging from 4 to 30 mTorr (-5 – 40 bar).

They explained the deposition mechanism and postulate for the first time the nature of

the radicals giving growth at the substrate surface. As actually agreed upon, they

identified the 3 main phases of the deposition process:

13
 1. The deposition of the feed gas on the hot surface of the wire.

2. The evaporation of the radicals and gas-phase collision (creation of secondary

species).

3. Film growth on the substrate surface.

At the -1700 K filament temperature where they were working, they noted that the

decomposed silane gas was mostly atomized into H and Si atoms. A major fraction of

the Si and H atoms is expected to arrive at the substrate without reacting, but a

significant fraction may react to form other radicals. Hot Hydrogen and silicon atoms

which have not undergone the gas-phase collisions process were observed to dominate

the radical flux to the hydrogen-passivated a-Si: H surface. They observed that the

dominant species impinging on the substrate were SiH4 and H2. The reaction between

SiH4 and the surface dangling bond was postulated to give SH3 and a passivated

surface bond. Thus considerable Si H3, lesser SiH2 and Si2H2 (which increase with the

gas pressure) were expected to contribute to the deposition. J. Perrin et al [31] have

confirmed the nature of these radicals as exothermic gas phase reactions in the

HWCVD:

H + SiH4 → SiH3 + H2

H + SiH3 → SiH2 +H2

SiH3 + SiH3 → SiH4 + SiH2

Si + SiH4 → Si2H2 + H2

Si2H2 + SiH4 → Si3H4 + H2

14
In 1991, Mahan et al [8] demonstrated for the first time the possibility to produce

device-quality a-Si: H with reduced hydrogen content. Good quality films with less

than 1 at. % hydrogen compared to 8-12 at. % contained in good quality a-Si: H

obtained by plasma enhanced CVD, confirmed that only a small quality of hydrogen

was needed to passivate the dangling bonds. Mahan and co-authors obtained thin

films with photosensitivity (σp/ σd) – 105 and Urbach edges that, in contrast to

PECVD films, did not broaden as the hydrogen content was reduced. After the work

of Mahan, many groups worldwide took a keen interest in this Si-based new

technology. Efforts have been focused on the understanding of the deposition process

itself, the issues related to the growth and improvement of the deposition rate on the

deposition of a new class of materials which is more stable and less sensitive to the

SWE [15].

2.2 Background on Gambit (CAD design)

Gambit was released on summer 2004; its focus was to be on CAD import, geometry

conditioning and automatic meshing. This is the first step in building and analyzing a

flow model. It includes building the model within a CAD package, creating and

applying a suitable computational mesh/grid, and entering the flow boundary

conditions and fluid materials properties. These pre-processor tools used in this study,

called Gambit is part the packaged software FLUENT [1, 9].

CAD geometries are easily imported and adapted for CFD solutions in Gambit, which

is FLUENT’s own pre-processor step. The three dimensions (3D) solid modeling

option in Gambit allows for straightforward geometry constructions as well as high

15
geometry translation. This means that the geometry can be translated in three

dimensions. Among a wide range of geometry tools, Boolean operators provide a

simple way of getting from a CAD solid to a fluid domain [9].

A state-of-the-art set of cleanup and conditioning tools prepares the model for

meshing. Gambit’s unique curvature and proximity based “size function” produces a

correct and smooth CFD-type mesh throughout the model. Together with a boundary

layer technology, a number of volumetric meshing schemes produce the correct mesh

for your application. Parametric variations are also inherent to the process [1, 9].

2.3 Mesh Generation

The mesh/grid generation process deals with the division of the domain (designed

geometry) under consideration into small control volumes on which the discretised

governing equations will be solved. The grid generation forms a large part in terms of

person-hour time of the CFD analysis. Gambit, is the CAD design system used in this

study to design a domain/geometry and generate a grid/mesh from the designed

domain [9,14].

Gambit uses solid modeling techniques to create a virtual model of the geometry

under consideration. There are various grid generation tools available to create a

grid/mesh in this geometry. The grid/mesh generation tools are available to create

hexahedron, tetrahedron, prism and pyramid cells [9].

16
The origin of the term mesh/grid goes back to early days of CFD when most analyses

were 2D in nature. For 2D analysis, a domain split into elements resembles a wire

mesh, hence the name. These process of obtaining an appropriate mesh/grid is termed

mesh generation (or grid generation) [9,14].

Below there are pictures of the designed domain/geometry and its mesh/grid using

the CAD design system, Gambit.

Figure 2.2: A domain (geometry) of the HWCVD system as used in this study,

designed using the CAD design system GAMBIT.

17
Figure 2.3: A grid (a mesh) of the HWCVD system as used in this study, designed

using the CAD design system GAMBIT.

2.4 Background on FLUENT

FLUENT is the computer software package of CFD which was used in this study to

model the gas flow patterns inside the HWCVD system. Fluent is the general name

for the collection of computational fluid dynamics programs sold by Fluent Inc. of

Lebanon, New Hamphire, USA [11].

18
As outlined in chapter one, every CFD code consists of three basic components, the

pre-processor (CAD design; Gambit), the solver which is the heart of the CFD code

carrying out the major computations and providing the numerical solutions, and the

post-processor for constructing pictures of plots of the simulated problem. The

FLUENT software package which was used in this study is the FLUENT version

6.3.26, where the geometry was designed and meshed in Gambit saved and then

exported the mesh to FLUENT for running the simulation.

Fluent is a general-purpose CFD code based on the finite volume method on a proper

order grid/mesh. Fluent technology offers a wide array of physical models that can be

applied to a wide array of industries. Using the dynamic and moving mesh: The user

simply sets up the initial mesh and prescribes the motion, while fluent software

automatically changes the mesh to follow the motion prescribed [11].

This is useful for modeling flow conditions in and around moving objects. There are

many spectral methods available in CFD, but in this study the one that will be used is

the Finite Volume Method [11]. The numerical solution method used in FLUENT

(Finite Volume Method), will be discussed below.

2.5 Finite Volume Methods

The finite volume method is a method for representing and evaluating partial

differential equations as algebraic equations. This is similar to the finite difference

method, where values are calculated at discrete places on a meshed geometry [11,20].

19
The phrase ‘finite volume’ refers to the small volume surrounding each node point on

a mesh. In this method, volume integrals in partial differential equations that contain a

divergence term are converted to surface integrals, using the divergence theorem.

These terms are then evaluated as fluxes at the surfaces of each finite volume.

Because the flux entering a given volume is identical to that leaving the adjacent

volume, these methods are conservative [11,23].

Another advantage of the finite volume method is that it is easily formulated to allow

for unstructured meshes. As mentioned above, this method is used in the

computational fluid dynamics package; Fluent [11].

2.6 Mathematics behind CFD

The fundamental basis of any CFD problem is the Navier-Stokes equations, which

define a single phase flow. The most fundamental consideration in CFD is how one

treats a continuous fluid in a discretized fashion on a computer. One method is to

discretize the spatial domain into small cells to form a volume mesh/grid, and then

apply a suitable algorithm to solve the equations of motion [4,5]. Below we will

discuss the conservation equations which describe the flow field.

2.6.1 Navier-Stokes Equations

The governing equations that describe the flow field are a set of non-linear partial

differential equations. The equations are derived from mass, momentum and energy

20
conservation. In chapter 1 it was outlined that the Navier-Stokes equations are a

complex non-linear set of partial differential equations that describes the mass,

momentum conservation of fluid.

In the past, engineers made further approximations and simplifications to the equation

set until they had a group of equations that they could solve. Recently, high speed

computers have been used to solve approximations to the equations using a variety of

techniques like finite difference, finite volume, finite element, and spectral methods

which are in the area of CFD [4,5,24].

2.6.1.1 Conservation of Mass

The conservation of mass in Eulerian terms in its most common form for fluids is

given as follows:

∂ρ
+ divρV = S m 2.1
∂t

Where ρ is the density, t is the time, V is the velocity vector of the fluid and Sm is a

mass source. For a constant density, equation 2.1 will reduce to the following:

divV = 0 2.2

21
2.6.1.2 Momentum equations

The Navier-Stokes equations are derived from the conservation of momentum

(Newton’s second law). The set of equations can be written in Eulerian terms as single

vector equations using the indicial notation:

DV ∂   ∂v ∂v j  
ρ = ρg − ∇ p +  µ  i +  + δ ij λdivV  2.3
Dt ∂x j   ∂x j ∂xi  

Where g is the gravitation vector, xi is the spatial co-ordinate, vi is the i-component of

the velocity vector, is the ordinary coefficient of viscosity, λ is the coefficient of

bulk viscosity, δij is the Kronecker delta function (δij = 1 if i = j and δij=0 if i≠j) and

D
is the total or substantial derivative. The relation between the ordinary viscosity
Dt

and bulk viscosity is given by the Stoke’s hypothesis [18]:

2
λ+ µ =0 2.4
3

An example of the equations 2.3 in the i-direction is written as follows:

Du ∂p ∂  ∂u 
ρ = ρg x − +  2µ + λdivV  +
Dt ∂x ∂x  ∂x 
2.5
∂   ∂u ∂v  ∂   ∂u ∂w 
 µ  +  + µ + 
∂y   ∂y ∂x  ∂z   ∂z ∂x 

22
2.6.1.3 Energy equations

The energy equation is derived from the first law of thermodynamics. The energy

relation in Eulerian terms is given by the following if it is assumed that the conduction

heat fluxes are given by Fourier’s law and that radiative effects are negligible [4,24]:

Dh Dp
ρ = + div(k∇T ) + Φ 2.6
Dt Dt

Where h is the fluid enthalpy, k is the thermal conductivity, T is the temperature of the

fluid and Ф is the viscous dissipation and is given as follows:

2 2
  ∂u  2  ∂v   ∂w   ∂v ∂u  
2

2  + 2  + 2  +  +   2
  ∂x   ∂y   ∂z   ∂x ∂y    ∂u ∂v ∂w 
Φ = µ 2  + λ  ∂x + ∂y + ∂z  2.7
2

+  ∂w ∂v   ∂u ∂w    

 ∂y ∂z +  +  +  
    ∂z ∂x  

If assuming incompressible flow as well as constant thermal conductivity and low-

velocity flow where the viscous dissipation (Ф) is zero, the equation 2.7 above will

reduce to the following equation:

DT
ρc p = k∇ 2T 2.8
Dt

The equations under 2.5.1.1 to 2.5.1.3 are the equations which are used in every CFD

simulation; as mentioned under 2.5 that the fundamental basis of any CFD problem is

23
the Navier-Stokes equations which are non-linear partial differential equations;

derived from mass, momentum and energy conservation.

In this study, these equations are used by the computer by default under the solver

stage to determine the mixture of the gas-flow patterns inside the HWCVD system;

derived from mass, momentum and energy conservation.

The solver as outlined in chapter 1 has two components in solving the computational

simulation. One of these components is to convert the partial differential equations

and the boundary conditions formulae into a group of algebraic equations and then,

the second component of the solver stage is to use the iterative procedure to find

solutions that satisfy these boundary conditions for the algebraic equations defining

the flow domain.

2.7 Steady State conditions and Computational accuracy in

FLUENT

A steady state calculation yields a solution in a shorter execution time in comparison

to a transient calculation. Nonetherless, there may be some underlying circumstances,

where convergence and/or stability cannot be guaranteed through the steady state

approach and a transient calculation may be required to march the numerical

procedure toward the steady-state condition. Other issues concerning the aspects of

computational accuracy and efficiency can also have a strong influence of the CFD

solution. Subject to the availability of computational resources, it is nearly always

inevitable that some compromise has to be reached for solving complex CFD

problems.

24
By increasing the number of cells (i.e. with decreasing the mesh spacing) in the

computational domain geometry, the accuracy of the computational solution is usually

enhanced. There is, however, a trade off that needs to be considered between

increasing computer storage running-time. One possible way that comparable

accuracies can be obtained on a courser mesh while maintaining computational

efficiency is to employ higher order dicretization schemes to solve such problems.

25
CHAPTER 3

SIMULATION METHODS

3.1 Simulation Methods

In this section the simulation procedure that was followed in determining the mixture

of the gas flow patterns inside the HWCVD system using the Fluent software

package, will be outlined. From the previous chapter (chapter 2), a detailed

background of the software package was given with the different tools that it is

comprised with.

The software package has a Cad design system (GAMBIT), in which the geometry of

the system to be modeled is generated and meshed using the tools in the Cad design

system. The Cad system also enables the user to set (define) the boundary conditions

for he different part of the system, which will again be set in the second stage of the

software; the solver stage. This stage of the software is known as the heart of the

software package, because this is where most of the simulation work is being done.

Under this stage, the user input the parameters to be considered in the simulation

process, the boundary conditions are also set under this stage; selecting from the tools

that are there in the package the materials to be used inside the system designed using

the Cad design system.

In this study, the materials that where selected to be used under the solver stage are

the two gases (H2 and SiH4) and the two substrates that are deposited inside the

HWCVD system (glass and silicon).The properties of silicon and glass (i.e. the

thicknesses, thermal conductivities, the pressure inside the HWCVD system) are also

defined under the solver stage. The calculation (simulation) of determining the

26
mixture of the gas flow patterns inside the HWCVD is done by the computer using the

process of iteration in the Fluent software package. The flow chart below is a step by

step illustration of the simulation procedure followed in any CFD simulation problem.

What to model
(gas flow patterns
inside HWCVD)

Design the
geometry/Domain GAMBIT

(Cad design)

Mesh the
geometry/domain

FLUENT
Flow solution (Solver and Pre-
processor stage)

Analysis

Figure 3.1: A flow chart of the procedure followed in a CFD problem (simulation).

3.2 Designing the Domain/Geometry

The first step of any simulation problem is to decide on what to model; which in this

study is to model the mixture of the gas flow patterns inside the HWCVD system

using the CFD software package FLUENT. As it was outlined in chapter 1, the pre-

processor stage of the Fluent software is comprised with a Cad design (GAMBIT)

system to create the geometry of the system to be modeled. In designing the HWCVD

27
chamber in the Cad design system, the dimensions of HWCVD system was collected

by measuring the different parts of the system. The table below shows the collected

data for the dimensions of the HWCVD system to be used to design its

domain/geometry by using the Cad design system.

Table 3.1 Some of the measurements of the HWCVD system.

Name of the part Height (length) Diameter measurements

measurement in mm in mm

Gas inlet tube 130 9.7

Outlet 50 179

Insulators 44 9.7

Filaments 11 0.1

Filaments-stand 5 121.5

HWCVD system 300 Rectangular shape

Glass (substrate) 0.8 Rectangular shape

Silicon (substrate) 0.4 Rectangular shape

Substrates-holder 122 Rectangular shape

Heater 128 Rectangular shape

The table above has only the parts of the geometry which were used in this study;

some parts of the geometry which were measured were not used due to the small

influence that they will have to the gas flow and they were increasing the number of

cells in the grid/mesh thus increasing the computing time of the simulation.

Having the measurements of the internal and external geometry of the HWCVD; the

geometry of each part was designed individually using the CAD design program

28
(Gambit) and combined to the exact space inside the system (HWCVD) for

determining the gas flow.

As outlined in chapter 2; Gambit offers many options in designing the geometry, each

of these parts were designed to its specific size and shape; by scaling them to fit in the

computer using Gambit. Figures 3.2 and 3.3 below, shows the parts of the chamber

which were designed individual, each with its own measurements; and then in figure

3.4 there is a domain of the chamber with all the parts included inside it.

Figure 3.2: a domain of gas-inlet pipes.

29
Figure 3.3: a domain of the filaments, insulators and filaments stand.

30
Figure 3.4: Deposition chamber (CVD) with all the parts imported to it by Gambit.

All the parts were designed individually and each of them saved in Gambit as xx.dbs

files (the format used in the Gambit CAD design system for storing files). These parts

were then imported to the chamber by using the measurements obtained at the

beginning of the study in placing them in their correct positions inside the chamber in

order to get the completed HWCVD chamber as given above in figure 3.4.

31
In every CFD simulation process the designed geometry of the system to be modeled

has to be meshed in order to have a complete designed system to start the simulation

process. This process of meshing was done in this study after having the geometry

given in figure 3.4. As mentioned in chapter 2, there are a number of tools which are

offered by Gambit in generating the mesh, one of the tools which was used in this

study to obtain a finer mesh; is the size functioning tool.

The mesh obtained in any CFD problem, affects directly the accuracy of the solution

and the cost of computation, in terms of computation time and hardware requirement.

An optimized mesh/grid system uses less computer memory space, requires less

computing and yet gives satisfactory and accurate results [1,2,9].

Figure 3.5 below is a mesh/grid obtained in this study for the HWCVD chamber to

model the mixture of the gas flow patterns.

32
Figure 3.5: a grid (a mesh) of CVD system, with the total number of 657120 cells.

3.3 The deposition parameters of the HWCVD system

The HWCVD system is used by the solar energy research facility at the University of

the Western Cape. The process of the HWCVD system, involves the deposition of the

two substrates by using the H2 and SiH4 gases under a constant pressure of 60 bar by

varying the temperature of the heater, these process will be performed in this study

using the CFD software package; FLUENT. The table below is the general data of the

HWCVD system with the four varying temperatures of the heater which will be used

33
 in the simulations of this study to determine the gas flow patterns inside the

deposition system.

Table 3.2: The deposition parameters used in the HWCVD system.

Pressure of HWCVD 60 bar

Temperature of the Heater 100 oC, 200 ˚C, 300 ˚C and 550 ˚C

Temperature of the filaments 1600 ˚C

Heat Flux around the filaments (Appendix A.3) 99.21x103 Wm-2

H2 mass flow rate 58.8 sccm

SiH4 mass flow rate 1.2 sccm

This data in the above table, it is some of the parameters used in the solver stage of

the software before starting the simulation (iteration process). The heat flux of the

filament is calculated using its temperature of 1600˚C, this calculation is given in

appendix A as well as the calculations for the mass flow rate of H2 and SiH4. After

obtaining the mesh/grid given in figure 3.7, the boundary conditions for the walls

(parts of the system) which were formed during mesh/rid generation where set in the

Cad design system. Below is a table of the boundary conditions specified in the Cad

design system (Gambit) in this study.

34
Table 3.3: Boundary conditions of the parts of HWCVD system.

Name of the Parts Boundary Conditions

Inlets Velocity-inlet

Outlet Pressure-outlet

Wall (Chamber) Wall

All the other parts inside the HWCVD (the two substrates (glass and silicon),

substrates-holder, filaments, insulators, and heater) which are not given in Table 3.3

were specified as wall boundary condition. After completing the process in the Cad

design system, the mesh was exported to Fluent in order to start running the

simulation. A full description of the model that is used in this study is given in the

appendix A.

The CFD software package (FLUENT) uses the Navier-Stokes equations and energy

balances over control volumes, small volumes within the geometry at a defined

location representing the HWCVD internals. The size and number of control volumes

(mesh density) is user determined and will influence the accuracy of the solution to a

degree. After boundary conditions have been introduced in the system (this is done

automatically in Gambit when generating a mesh/grid), the flow and energy balances

are solved numerically.

3.4 Starting the Simulation in FLUENT

The process of starting the simulation is done under the second stage of the Fluent

software, the solver stage. The file which was created in Gambit and exported to

35
Fluent, will be opened in this stage in order continue with the simulation problem.

The boundary conditions are being set the same as the boundary conditions which

were set in the Cad design system and setting up the parameters inside the system.

The material properties which are used in determining the mixture of the gas flow

patterns are also defined under this stage. The HWCVD system has a number of parts

with different material properties. The material properties for the HWCVD system

used in this study are given in the table below.

Table 3.4: Material properties used in this study.

Material Specific Density (ρ) Thermal conductivity

Properties Heat (Cp) (J/kg/K) (kg/m3) (K) (W/m/K)

Glass 700 2579 1.05

Alumina (Al2O3) 880 961 18

Tantalum (Ta) 140 16690 57.5

Silicon (Si) 700 2329 130

Copper (Cu) 381 8978 387.6

Steel 502.48 8030 16.27

From the table given above of the material properties, glass and silicon are the two

substrates used in the deposition process with thicknesses 0.08 cm and 0.04 cm

respectively; Alumina (Al2O3) is the material used for the two insulators of the

chamber, Tantalum (Ta) is the material for the filaments of the chamber, Copper (Cu)

is the material of the inlets-pipe of the CVD chamber and Steel is the material used

for the substrates-holder, heater, filaments-stand and Wall (chamber).

36
In this study, the segregated solution method of FLUENT was used. In this approach,

the governing equations are solved sequentially (i.e. segregated from one another).

Because the governing equations are non-linear (and coupled), several iterations of

the solution loop must be performed before a convergence is obtained. Each iteration

process consists of the steps illustrated in the figure [3.8] and outlined below [9,15]:

• Fluid properties are updated based on the current solution (if the calculation

has just begun the fluid properties will be updated based on the initialized

solution).

• The u, v and w momentum equations are each solved in turn using current

values for pressure and face mass fluxes, in order to update the velocity filed.

• Since the velocities obtained in the previous step may not satisfy the

continuity equation locally, a “Poisson-type” equation for the pressure

correction is derived from the continuity equation and the linearised

momentum equations. This pressure correction equation is then solved to

obtain the necessary corrections to the pressure and velocity fields and face

mass fluxes such that continuity is satisfied [15].

• Equations for scalars such as energy and species are solved using the

previously updated values of the other variables.

• When all these steps are met in a simulation, then a check for convergence of

the equation set is made [15].

37
 Update properties

Solve momentum equations

Solve pressure-correction (continuity) equations.

Update pressure, face mass flow rate

Solve scalar equations (energy and species)

Convergence Stop

Figure 3.6: An overview of the segregated solution method when iterating [15].

An iterative process decreases the error in the solution until a satisfactory solution has

been reached (i.e. the simulation is converged). The iterative process is the final step

of the solver stage which is the process used to start the simulation. Each simulation

in this study was calculated for 10000 iterations. The total number of simulations

performed in this study is about four simulations; each simulation was performed

using the same procedure outlined above, but with a different temperature of the

heater (i.e. 100 ˚C, 200 ˚C, 300 ˚C and 550 ˚C). The results obtained from the four

simulations after all this procedure is done (when the simulation is converged) will be

given and discussed in the next chapter.

38
CHAPTER 4:

RESULTS AND DISCUSSIONS

4.1 Introduction

This section focuses on the results obtained in determining the gas flow patterns

inside the hot-wire CVD system. There are a number of simulations which were

performed in this study for one to come into having these results. As mentioned in

chapter one under post-processor, a complete simulation often requires repeating the

procedure a number of times. This includes tuning the mesh system, adjusting

boundary conditions, selecting numerical parameters, finding the right physical

model, monitoring the iteration and viewing the results.

The results in this chapter are for the simulations which were done under constant

pressure of 60 µbar, constant heat flux of the filaments of the HWCVD of about

99.21x103 Wm-2 (Appendix A.3)

; by varying the temperature of the heater from 100 ˚C,

200 ˚C, 300 ˚C and 550 ˚C, thus the simulations show how the temperature affects the

gas-flow patterns inside the HWCVD system. The results will give an indication of

what influence the mass flow-rate and the surface deposition-rate of the mixture of

gas particles inside the HWCVD system.

The results for the four simulations performed will be given and discussed in this

chapter. There are two parts of the results which we will concentrate on in this study,

that is, the results for the mass flow rate of the gas around the following important

parts of the system; the filaments, the two substrates, inlet and outlet of the HWCVD

39
system. The second part of the results is the results for the surface deposition rate

around the filaments, the two substrates and the substrates-holder inside the HWCVD

system.

4.2 Mass flow-rate

This part of this study is about the mass flow-rate results. Table 4.1 is the results for

the mass flow-rate obtained from the four simulations performed in this study,

followed by the plot (graph of the mass flow-rate around the filaments, the two

substrates, the inlet and outlet versus the temperature of the heater). The results in this

table (table 4.1) show the values for the mass flow-rate into the system from the first

to the fourth column of the results to be positive. The positive value means the mass

flow-rate into the system through the inlets, and the negative mass flow-rate values

indicates the total mass flow-rate out of the system through the outlet.

The results in table 4.1 also shows that the mass flow-rate out of the system is greater

than the mass flow-rate flowing into the system, this is because of the small value of

the velocity magnitude for the inlets which was calculated in chapter 3 to be 0.18(A.2)

using the flow-rate of the two gases (H2 = 58.8 sccm and SiH4 = 1.2 sccm) flowing

into the HWCVD system. Thus this value (velocity magnitude) makes the constants

value of the flow into the system through the inlet to be small (5.74 x10-12).

In table 4.2 below, are the results for the surface deposition-rate obtained from the

four simulations performed in this study, followed by a graph of the surface

40
deposition-rate (around the filaments, the two substrates and the substrates-holder)

versus the temperature of the heater for all simulations.

Table 4.1 Mass flow-rate results for all simulations.

Name Temperature of the Heater (˚C)

Of the Parts 100 200 300 550

Filaments -2.12x10-17 -4.17x10-17 -8.61x10-17 -6.32x10-16

Inlets 5.74x10-12 5.74x10-12 5.74x10-12 5.74x10-12

Outlet -5.99x10-12 -5.96x10-12 -5.82x10-12 -5.66x10-12

Substrate1 -9.35x10-20 -2.15x10-19 -4.81x10-19 -5.01x10-18

Substrate2 -2.52x10-19 -4.51x10-19 -8.32x10-19 -6.49x10-18

Net Mass flow -2.45x10-13 -2.17x10-13 -7.57x10-14 7.62x10-14

From the Table 4.1, we can see from the four simulations that as the temperature of

the heater are increased from 100 ˚C to 200 ˚C, 300 ˚C and 550 ˚C the mass flow rate

flowing respectively around the filaments and the two substrates is decreasing. But the

flow-rate of mass flowing out of the system through the outlet is increasing as the

temperature of the heater is increased inside the chamber.

The temperature around the filaments of the HWCVD system at 1500 ˚C is slightly

lower than the filament at 1600 oC. It is suggested that although chemical energy is

released during the catalytic process, this is balanced by the removal of heat by the

higher flow rate at the slightly lower temperature of the filament.

This is in agreement with practice where at higher flow rates and substrate

temperatures the power to the filament had to be increased slightly to keep the

41
temperature at 1600 ˚C. The decreasing flow-rate at the outlet at increasing

temperatures is suggested to be due to the faster removal rate of materials from the

gas due to the higher deposition rate at the substrates. The results also show an

increasing negative flow-rate at the filaments and substrate, this could be explained by

the amount of material deposition on the filament and the substrates effective gas

removal from the system.

Mass flow-rate (kg/s) versus

Temperature (K) of the heater
6.00E-012
Inlets

4.00E-012

2.00E-012
Mass flow-rate (kg/s)

Substrate1&2
0.00E+000
Filaments

-2.00E-012

-4.00E-012

-6.00E-012 Outlet

300 400 500 600 700 800 900

Temperature of the Heater (K)

Graph 4.1: A plot of the mass flow-rate on the filaments ,inlets, outlet and two

substrates vs temperature of the heater for all four simulations.

The net flow-rate it is just the indication of the excess flow-rate flowing out of the

system. This is only true at the lower temperatures at higher temperature it decreases

42
and at 550 oC it actually is less than coming in, meaning some of the gas remains

behind. This would be due to the gas-atoms deposited at a high enough rates on the

substrates to reduce the flow at the outlet.

1.00E-013

5.00E-014

0.00E+000
Mass flow-rate (Kg/s)

-5.00E-014

-1.00E-013

-1.50E-013

-2.00E-013

-2.50E-013

300 400 500 600 700 800 900

Temperature of the Heater (K)

Graph 4.2: A plot of the net mass flow-rate results around the filaments, inlet, outlet

and the two substrates versus the temperature of the heater.

As we can see that the mass flow-rate flowing into the system for each simulation is

about 5.74 x 10-12 kg/s and the one flowing out of the system for about three

simulations, 100, 200 and 300 oC the outlet mass flow-rate is greater than the inlets

flow rate. This is a small difference of about -2.46 x 10-13 which is about 4.3%. The

plot in graph 4.1 shows the plot of the temperature of the heater all four simulations

versus the mass flow-rate around the filaments, inlets and outlet. The net mass flow

43
rate results are also plotted in graph 4.2, this graph shows that the mass flow rate

increases with the increasing temperature inside the HWCVD system. This is in

agreement with the known results of the influence that the temperature inside the

HWCVD system, as the temperature inside the system is increased the mass flow rate

of the inside the deposition system also increases.

4.3 Surface Deposition-rate

The substrate heater temperature and the filament temperature inside the HWCVD

system is very a very important process parameter affecting both the deposition rate

and the properties of the deposited films.

One way of increasing the deposition rate is to increase the breaking of the SiH4 and

H2 molecules by increasing the temperature inside the HWCVD system. However, the

use of higher temperatures in the HWCVD system increases the evaporation of

tungsten from the filament, which, may lead to contamination of the growing film.

The use of lower temperatures leads to the formation of tungsten silicate on the hot

filament [38]. In section, the results of the surface deposition-rate are given.

44
Table 4.2: Surface deposition-rate results for all four simulations.

Name of the Temperatures of the heater

Parts 100 ˚C 200 ˚C 300 ˚C 550 ˚C

Filaments 8.16x10-15 1.61x10-14 3.32x10-14 2.43x10-13

Substrate1 1.04x10-16 2.39x10-16 5.34x10-16 5.57x10-15

Substrate2 2.81x10-16 5.02x10-16 9.24x10-16 7.21x10-15

Sub-Holder 2.75x10-18 5.18x10-18 9.97x10-18 8.38x10-17

Net Deposition 4.37x10-17 8.61x10-17 1.78x10-16 1.31x10-15

Table 4.2 above, shows the results for the surface deposition of the four simulations

performed in this study. From the above results we can see that the surface deposition

around the filaments, the two substrates and the substrates-holder increases with the

increasing temperature of the heater for each simulation performed in this study. Thus

this can be concluded that the deposition of the mixture of gases into the HWCVD

system is influenced by the temperature around the system.

Below is the plot of the temperature of the heater for all four simulations versus the

surface deposition-rate around the filaments, the two substrates and the substrates-

holder.

45
 2
Surface Deposition-rate (kg/m -s) vs
Temperature (K) of the Heater
2.50E-013
Filaments
2.00E-013
Surface deposition-rate (Kg/m2/s)

1.50E-013

1.00E-013

5.00E-014

0.00E+000
Substrate1&2
-5.00E-014

-1.00E-013 Substrates-holder
-1.50E-013

-2.00E-013

-2.50E-013
300 400 500 600 700 800 900
Temperature of the heater (K)

Graph 4.3: A plot of the surface deposition-rate on the filaments, the two substrates

and a substrates-holder vs temperatures of the heater.

The plot above shows the surface deposition-rate around the filaments, the two

substrates and the substrates-holder for all four simulations by varying the

temperature of the heater. From this plot, it can be seen that the surface deposition-

rate increases with the increasing temperature of the heater in the HWCVD system.

This would mean that the deposition is influenced by the temperature of the system,

with other parameters like the pressure inside the HWCVD system which is in

agreement with known results [27].

46
4.4 Discussions

CFD is the analysis of a system involving fluid flow, heat transfer and associated

phenomena such as chemical reactions by means of a computer-based simulation. In

this study the mixture of the gas flow patterns inside the HWCVD system was

modeled using CFD software package, FLUENT. The geometry of the HWCVD

system was designed and meshed using the CAD design system called Gambit which

is the first stage of the software package. In the second stage of the software (the

Solver stage), most of the work was done by the computer itself; whereby using the

discretizing scheme to express the governing equations in a discretized form for all

mesh elements and discretizes the boundary equations appropriate at the boundary

elements. In the solve stage the partial differential equations (Navier-Stokes

equations) and the boundary conditions formulae are converted into a group of

algebraic equation and the iteration procedure takes place. The simulations in this

study converted after 6000-8000 iterations which took about 12 hours for the each

simulation to converge and, this was due to the many parts that the HWCVD system

contain and the finer mesh that was obtained.

The results given in 4.2 and 4.3 above are an indication of what influence or impact

the mass flow-rate and the substrate flow-rate of the mixture of gases inside the

HWCVD system. For one to come into a conclusion of having simulation results

there are various procedure to be followed i.e. creating the geometry, changing the

sixe of the mesh/grid of the geometry created, changing values for the parameters

used in this study, setting-up boundary conditions, and re-running the simulation by

performing the iteration. The results in table 4.2 shows that the parts which are closer

to the heater and filaments have an increasing mass flow-rate as the temperature of the

heater is increased and, this is in agreement with the known result [27, 38].

47
Chapter 5

SUMMARY AND CONCLUSION

A model based on computational fluid dynamics is, in principle, more precise and can

include many features present in the fluid flow analysis; but requires much more

computer time to run and needs a lot of information to model the boundary conditions.

In performing a computational simulation using FLUENT, quality meshes should be

generated to optimize the accuracy of the results calculated in the simulation. The

mesh obtained in this study had Equisize skew (indication of the quality of the mesh)

of 0.793051; which can be brought down to a smaller number leading to having very

fine and much optimized mesh giving accurate results.

As outline under the table 4.2 and graph 4.2 in chapter 4, it can be concluded that the

flow of the gas-flow patterns inside the hot-wire chemical vapor deposition system is

influenced by the temperature, with other parameters like the pressure of the system

and the rate in which the gas flows inside the system. In this study, we experienced a

problem of having a simulation running for many hours due to the complex system we

were modeling having parts with different shapes. It is again important to mention

that in this study, there were no experimental results to compare the results obtained

using the FLUENT software with. Most importantly, the simulations performed in this

study gave an indication of how much the flow-rate and the deposition rate is affected

by the temperature inside the HWCVD system.

48
In the appendix B there are plots of the path-lines obtained from the post-processor of

FLUENT software, which shows how the gas flow patterns are distributed and flows

throughout the system. These lines are an excellent tool in the FLUENT software to

show how the mixture of the gas travels inside the HWCVD system. There are also

contours of the temperature given in the appendix B, this is to show how the

temperature of the heater which was varied in this study is distributed throughout the

HWCVD system. From this tool it is shown how the parts which are closer to the

heater and the filaments also had a higher temperature as compared to the parts which

are far from the heated filament and heater.

49
REFERENCES

[1] J.F. Douglas, J.M. Gasiorek, J.A. Swaffield, L.B. Jack, Fluid

mechanics, fifth edition, (2005) 1-102.

[2] H.K. Versteeg and W. Malalasekera, An introduction to Computational

Fluid Dynamics, The Finite Volume Methods, (1995) 4-92.

[3] C.J. Arendse, Hydrogenated amorphous silicon; optical properties and

hydrogen concentration, MSc thesis, University of the Western Cape, (1998).

[4] J. Lorenz, H.O. Kreiss, Initial-Boundary value problems and the

Navier-Stokes equations, (1989) 1-18, 325-337.

[5] B. Khanbabaei, A. Ghasemizad and H. Farajollahi, CFD-calculation of

Fluid Flow in VVER-1000 Reactor, Journal of Applied Sciences 8 (5), (2008)

780-787.

[6] K. Muralidah, T. Sundarajan, Computational Fluid Flow and Heat

Transfer, Kanpur Series of Advances Texts, Narosa Publishing House, New

Delhi, (1995).

[7] H. Wiesmann, A.K. Gosh, T. McMahon, M. Strongin, J. Appl.Phys.,

50 (5), (1979) 3752-3754.

[8] A.H. Mahan, J. Carapella, B.P. Nelson, R.S. Crandall, I. Balberg, J.

Appl.Phys., 69, (1991) 6728.

[9] Fluent Inc, Fluent Version 6.3 Manual, Centerra Resource Park, 10

Cavendish Court, Lebanon, New Hamphire, USA, (2006).

[10] C. Hirsch, Numerical Computation of Internal and External Flows,

Second Edition, (2007) 595-624.

[11] R. Eymard, T. Gallovet, R. Habin, Finite Volume Methods, (2006) 4-

93.

50
[12] [WWW1] http://www.grc.nasa.gov/WWW/k-12/airplane/nseqs.html

[Accessed 12 April 2009].

[13] A. Riddle, D. Carruthers, A. Sharpe, C. Mchugh, J. Stocker,

Comparison between FLUENT and ADMS for atmospheric dispersion

modeling, (2003).

[14] Fluent Inc, Gambit Version 2 Manuals, Centerra Resource Park, 10

Cavendish Court, Lebanon, New Hamphire, USA, (2001).

[15] S. Halindintwali, A study of Hydrogenated nanocrystalline silicon thin

films Deposited by hot-wire chemical vapour deposition (HWCVD), PhD

thesis, University of the Western Cape, (2005).

[16] S. Gu, C.N. Eastwick, K. Simmons, G. McCartney, Computational

Fluid Dynamics modeling of gas flow characterictics in a high-velocity oxy-

fuel thermal spray system ,Journal of thermal spray technology, 10, (3), (2001)

461-469.

[17] D.J. deKock, Optimal Turnish Design Methodology in a Continuous

Casting Process, PhD thesis, University of Pretoria, (2005).

[18] G.G Stokes, On the Theories of Internal friction of fluid in Motion,

Trans. Cambridge Phil.soc.vol8, (1845) 287-305.

[19] T. Frankiewicz, C.M. Lee, Using CFD simulation to model fluid

motion in Process vessels on fixed and floating platforms, Natco Group,

Houston, TX USA.

[20] T.J. Jasinski, E.P. Childs, Numerical Modeling Tools for chemical

vapor deposition, NASA Contractor Report 4480, Creare Inc. Hanover, New

Hampshire, (1992) 39-67.

51
[21] C.J. Brouckaert, T. Huang, C.A. Buckely, Applications of

Computational Fluid Dynamics Modeling in water treatment, WRC Report

No. 1075/1/05, (2005) 4-30.

[22] C.J. Brouckaert, C.A. Buckley, The application of Computational

Fluid Dynamics to wastewater treatment plants, WRC Report No. 648/1/02,

(2005) 4-14.

[23] G.F.J. Temple, An Introduction to fluid dynamics, Oxfords at the

Clarendon Press, (1958) 10-28.

[24] G.H. Cottet, P.D. Koumoutsakos, Vortex Methods, Theory and

Practice, (2000) 124-130.

[25] G.L. Kaiser, T. Meersmann, J.W. Logan, A. Pines, Visualization of gas

flow and diffusion in porous media, CA 94720, University of California,

(2000).

[26] K.K.S. Lau, H.G. Pryce Lewis, S.J. Limb, M.C. Kwan, K.K. Gleason,

Hot-wire chemical vapor deposition (HWCVD) of fluorocarbon and organo-

silicon thin films, Department of chemical engineering, Massachusetts Institute

of Technology, Cambridge, MA, USA, volume 395, Issue 1-2, (2001) 288-

291.

[27] H.L. Duan, G.A. Zaharias, S.F. Bent, MRS Symp. Proc. 715, (2002)

A26.3.1.

[28] S. Tange, K. Inouie, K. Konokura, M. Koski, Thin Solid Films 395,

(2001) 42.

[29] H. Matsumura, Japanese Journal of Applied Physics, 25 (12), (1986)

L949-L951.

52
[30] J. Doyle, R. Robertson, G.H. Lin, M.Z. He, A. Gallagher, J. Appl.

Phys. 64 (6), (1988) 3215-3223.

[31] J. Perrin, O. Leroy, M.C. Bordage, Contrib. Plasma Phys. 36, (1996) 3.

[32] J. Thangala, S. Vaddiraju, S. Malhota, V. Chakrapani, M.K. Sunkara, A

hot-wire chemical vapor deposition (HWCVD) method for metal oxide and

their alloy nanowire arrays, Thin Solid Film 517 (12), (2009) 3600-3605.

[33] Q. Zhang, M. Zhu, L. Wang, F. Liu, Influence of heated catalyzer on

thermal distribution of substrate in HWCVD system, Thin Solid Film 430,

(2003) 50-53.

[34] C.J. Oliphant, C.J. Arendse, G.F. Malgas, D.E. Motaung, T.F.G.

Muller, B.A. Julies, D. Knoesen, Filament poisoning at typical carbon

nanotube deposition condition by hot wire CVD, Journal of Material Science

44 (10), 2610-2616.

[35] T. Hadgu, S. W. Webb, M.T. Itamura, Comparison of CFD Natural

Convection and Conduction-only Models for Heat Transfer in the Yucca

Mountain Project Drifts, Albuquerque, New Mexico: Sandia National

Laboratories (2004).

[36] J. Tu, G.H. Yeon, C. Liu, Computational Fluid dynamics: a practical

approach, First edition, (2008) 192-271.

[37] J.H. Song, B.D. Ahn, S.D. Shin, N.M. Hwang, H.J. Kim, Effect of wire

temperature on the microstructural evolution of Si films in hot-wire chemical

vapor deposition for digital display and photovoltaic devices, Solar energy

Materials & Solar cells 93, (2009) 324-328.

[38] B.P. Swain, R.O. Dusane, Effect of filament temperature on HWCVD

deposited a-SiC: H, Material Letters 60, (2006) 2915-2919.

53
APPENDICES

APPENDIX A

A.1 Introduction

The steady state segregated solver in FLUENT is used in this study. The segregated

solver approach results in the governing equations being solved sequentially. An

implicit linearization technique is applied in the segregated solution of the modeled

equations previously described (in Chapter 2). Thus this results in a linear system of

equations at each computational cell. The equations are coupled and non-linear;

therefore, several iterations of the equation set are required to obtain a converged

solution.

FLUENT uses a control-volume method to solve the governing equations. The

equations are discrete for each computational cell. In applying this solution method

the CFD simulation stores flow properties (e.g. dependent variables) at the cell

centers. However, face values are required for the convection terms in the discretized

equations. Face values are obtained by interpolation form the cell centers using a

second-order upwind scheme for the momentum and energy equations and a first-

order upwind scheme for the turbulence equations [35].

In this section a details CFD model (procedure followed in the FLUENT model at

the solver stage) which was used in this study is given; and the calculation for the

magnitude of the velocity of the gas flowing into the system, the heat flux of the

filament and the species mass fraction which were used in the defining the boundary

conditions in FLUENT.

54
Table A1: A detailed CFD model as used in this study.

Description SETTING
Model Pressure based, 3D steady state, laminar
viscous model
Discretisation scheme Second-order for pressure
Second-order upwind for all other
equations (momentum, energy etc.)
Glass properties:
Specific Heat 700 J/kg/K
Density 2579 kg/m3
Thermal conductivity 1.05 W/m/K

Silicon properties:
Specific Heat 700 J/kg/K
Density 2329 kg/m3
Thermal conductivity 130 W/m/K
Alumina properties:
Specific Heat 800 J/kg/K
Density 961 kg/m3
Thermal conductivity 18 W/m/K
Tantalum properties:
Specific Heat 140 J/kg/K
Density 16690 kg/m3
Thermal conductivity 57.5 W/m/K
Copper properties:
Specific Heat 381 J/kg/K
Density 8978 kg/m3
Thermal conductivity 387.6 W/m/K
Inlet boundary condition:
Velocity specification method Components
Reference frame Absolute
y-velocity inlet -0.18
Inlet temperature 300 K

Outlet boundary condition: Zero gradient for all equations normal to

boundary.
Filaments boundary conditions:
Material name Aluminum
Thermal condition Heat flux
Heater boundary conditions:
Material name Steel
Wall boundary condition:
Free stream temperature 300 K
Other boundary walls Filaments-stand, Inlet pipe, Insulators,
substrate 1, substrate 2 and substrate-
holder.

55
A.2 Calculations of the magnitude of Inlet velocity

The total flow rate in the inlet is 60 sccm (this means that there will be a flow

of 1cm3 per second). The number of holes in the inlet-pipe is 7 (inlets).

• The radius of the inlets = 0.5 mm = 0.5 x 10-3 m per hole.

• The volume of the inlets is 1 cm3 = 1 x 10-6 m3, which then gives the

flow-rate of the gases per second into the system to be: 1 x 10-6 m3/sec.

Therefore the velocity of the gas flowing into the HWCVD system is:

flow − rate 1 × 10 −6 m 3 / sec

Velocity = = = 0.18m / s
Area [ ]
π (0.5 × 10 −3 m) 2 × 7

This velocity was used in setting the boundary conditions of the inlets in Fluent. This

velocity of the inlet is on the negative y-direction (-y) in the geometry of the CVD

system. Thus the boundary condition was set-up for the y-velocity to be -0.18 m/s.

A.3 Calculations for the heat flux of the filament

The total number of the filaments in the HWCVD system is seven (7) having the

length of 11 mm and the radius of 0.05 mm. The temperature of the filaments of the

HWCVD system is 1600˚C. Having this information, the heat flux of the filaments

was calculated as shown below and used in the simulation when setting the boundary

conditions of the filaments in solver stage of fluent.

56
 Area filaments = 7 × (2πrL ) = 2.419 × 10 −3 m 2

The current flowing around the filament = 3.8 A ≈ 4.0 A

The voltage applied over the filaments = 56 V ≈ 60 V

Therefore the power flowing around the filaments is calculated as follows:-

Power = VI = 60V × 4.0 A = 240W

Then the Heat-flux of the filaments is given as:-

Power 240W
Heat − Flux = = −3
= 99.21 × 10 3 W / m 2
Area 2.419 × 10 m 2

The calculations in (3.2.2) and (3.2.3) below were used in setting up the boundary

conditions in FLUENT for the inlet magnitude velocity and the species mass fraction

of the inlets.

A.4 Species Mass Fraction

The total flow-rate in the inlet of the HWCVD system used in this study is 60 sccm, in

which is the total flow-rate of the two gases added together. As given in the

nomenclature of this study, the sccm is the standard cubic centimeter which is the

units for the flow rate of the gases inside the HWCVD system.

57
1sccm = 7.44 × 10 −7 mol / sec

The flow-rate of Hydrogen(H 2 ) = 58.8sccm

Therefore the mass fraction for Hydrogen is:

H 2 = 2 g / mol × 58.8sccm × 7.44 × 10 −7 mol / sec

8.75 × 10 −5 g / sec
= × 1kg = 8.75 × 10 −8 kg / sec
1000 g

And the flow-rate for SiH 4 = 1.2sccm

Therefore, mass fraction of SiH4 is calculated as follows:

1.2 sccm × [(28.09 g / mol ) + (4 g / mol )]× 7.44 × 10 −7 mol / sec

SiH 4 =
1sccm
−5
2.86 × 10 g / sec
= × 1kg = 2.86 × 10 −8 kg / sec
1000 g

Then the percentage of the flow rate:

SiH 4 2.86 × 10 −8 kg / sec

Flow − Ratepercentage = = = 0.25
H 2 + SiH 4 1.16 × 10 −7 kg / sec

58
APPENDIX B

This appendix gives a detail flow results of the gas-flow patterns inside the HWCVD

system. The appendix shows the temperature, velocity and the gas-flow distribution

inside the system.

B.1 Contours and Path-lines

In figure B 1 and B2 shows the contour plots of the temperature around the walls of

the deposition system. From the plots, it can be seen that the regions near to the

filaments and the heater are hotter than the other parts. It can also be seen that above

the heater and below the filaments stand are the cold dead zones of the deposition

system from the plots.

In figure B3 and B4 is the pathlines colored by the temperature and the velocity

respectively, giving a full overview of the flow of the gases inside the HWCVD

system.

59
Figure B1: Temperature contours

60
Figure B2: Temperature contours with a plane drawn at (y0=-15, y1=15, y2=-15) axis.

61
Figure B3: Path-lines colored by the Temperature.

62
Figure B4: Path-lines colored by the velocity.

63