LIBRARY I

Michigan State .‘
University

 

 

 

 

.ii" .
-.|rI '

.‘1 A ‘AII

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due. ‘1
MAY BE RECALLED with earlier due date if requested.

A.
A.

 

DATE DUE DATE DUE DATE DUE
AY 2 9 2004

AA A‘

61"“

 

 

 

 

WV 1 6 2005

092209

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DaIeDue.p65-p.15

 

NUCLEATION AND GROWTH OF HETEROEPITAXIAL DIAMOND
By

Connie Rebecca Bednarski-Meinke

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of
DOCTOR OF PHILOSOPHY
Department of Physics and Astronomy

2002

ABSTRACT
NUCLEATION AND GROWTH OF HETEROEPITAXIAL DIAMOND
By

Connie Rebecca Bednarski-Meinke

This thesis describes the growth of single crystal diamond by low-pressure
microwave plasma-enhanced chemical vapor deposition (CVD). Although diamond in
the form of polycrystalline thin films has been deposited by CVD methods in the past, it
has proven difficult to devise methods that induce the diamond crystallites to align and to
coalesce into a single crystal. The method adopted here is based on heteroepitaxy, the
single crystal growth of a desired material on a chemically distinct substrate. Two
heteroepitaxial steps, carried out in sequence, were required. Epitaxial films of metallic
iridium (Ir) metal were grown on lattice-matched single crystal oxides, principally SrTiO;
and a-A1203 (sapphire). Heteroepitaxial diamond was then grown on Ir, which provides
a good match to the lattice parameter of diamond and is resistant to the high temperature
methane-hydrogen plasma.

Since diamond does not nucleate spontaneously at high densities, methods were
used to stimulate its nucleation on Ir. Low-energy ions, attracted to the substrate from
the plasma by a voltage bias, produced conditions favorable for diamond nucleation with
densities approaching 1012 cm’2 across the entire substrate. By terminating the biasing at
a series of time intervals, the sample temperature was quenched and the evolution of

diamond film formation could be followed. Scanning electron microscopy was used to

image individual diamond nuclei and crystallites, their pattern formation, and coalescence
on scales from 10 nm to 10 um.

Afier 60 minutes of growth, extremely smooth, continuous films of single crystal
diamond were obtained with dimensions the size of the substrate grth region, an area
3.5 mm in dia. By growing for extended periods, to a maximum of 48 hr, diamond plates
of thickness 35 mm were produced. Freestanding crystals exhibited (111) cleavage
surfaces, the same as natural diamond, and were transparent in visible light.
Characterization of the material by x-ray diffraction, electron backseattering diffraction,
and Raman scattering confirmed the presence of (001) oriented single crystal diamond.

By optimizing the CVD reactor geometry, this research has led to the highest
density of oriented nuclei yet reported. Coalescence therefore occurs at an early stage,
leading to greatly improved crystalline perfection and homogeneity. The discovery that
sapphire can be used as a substrate to grow (001) epitaxial Ir and diamond promises to
lead to improvements in diamond quality. Sapphire has high thermal stability coupled
with remarkably good crystalline perfection, and is available as large area substrates.

This thesis encompasses studies of substrates, epitaxial growth of electron-beam
evaporated Ir, systematic optimization of biasing procedures, exploration of the vast
parameter space of the CVD reactor, description of characterization tools, and the results
of more than 300 diamond growth experiments. Some new ideas, presented as a model of
diamond nucleation and early growth, may help explain the major improvements

obtained here.

T 0 my parents, Daniel and Reinhild Bednarski

In memory of Opa

iv

ACKNOWLEDGMENTS

I am grateful to my advisor, Brage Golding, for the opportunity to work on this
project, and learn what it takes to do real science. His patience and enthusiasm helped
me to focus on a sometimes murky road. I am indebted to our postdocs, Zhongning Dai
and An-Ping Li. Zhongning spent long hours working on the iridium deposition, key to
the success of the diamond films. An-Ping did much of the Raman and EBSD analysis.
Our group shared every fi'ustration and triumph equally.

I learned that no project of this scope is possible without the aid of knowledgeable
people. Baokang Bi, competent in many techniques, was an invaluable source of
information and encouragement. Reza Loloee helped with EBSD analysis, and
encouraged me in my writing and presentations. A. Refik Kortan is thanked for his
excellent x-ray diffraction work. Professor Jes Asmussen helped in the critical early
stages of the project. Brian Wright taught me how to run a CVD system, and the finer
points of university politics. The “machine shop guys”: Tom Palazzolo, Jim Mums and
Tom Hudson cannot be thanked enough for their patience in getting a job done just right,
on a tight schedule, with a willingness to work with whatever material we needed. Ann
Kirchmeier and Darla Conley were a great help in ordering those materials and tracking
things down, as well as giving an encouraging word.

My labmates, Ryan Kruse and Mahdokht Behravan, set excellent examples of
hard work and focus. Ryan is thanked for sitting through countless lunchtime diamond
discussions, and his help and enthusiasm with the Fourier analysis. Mahdokht, one of

few who truly understood the difficulty of this project in all aspects, is thanked for her

unending supply of encouragement and positive feedback. I also want to thank Lowell
McCann for teaching me valuable lab techniques, and useful discussions.

Finally, I never would have attempted a Ph.D. without the love and support of my
family. My sister, Melinda, never doubted I could do it, and was never shy of telling me
so. My parents have been behind me the entire way, and my brothers, Dan and Michael,
warmed a competitive streak in me early on. My husband, Jan, my rock and my

shoulder, was with me at every step.

vi

Table of Contents

LIST OF TABLES ............................................................................................................... x
LIST OF FIGURES ........................................................................................................... xi
CHAPTER 1 INTRODUCTION ....................................................................................... 1

CHAPTER 2 DIAMOND: STRUCTURAL PROPERTIES AND

CHARACTERIZATION METHODS ......................................................... 4

= 2.1 Diamond crystal structure ............................................................................ 4
2.2 Carbon bonding in diamond leads to unique properties .............................. 5
2.3 Epitaxial film structural analysis tools ......................................................... 7

' 2.3.1 Scanning electron microscopy ............................................................... 8
2.3.2 Electron backscatter diffraction (EBSD) ............................................. 1 1
2.3.3 Raman spectroscopy ............................................................................ 12
2.3.4 X-ray diffraction .................................................................................. 15
2.3.5 Atomic force microscopy (AF M) ........................................................ 19

CHAPTER 3 BASICS OF CRYSTAL NUCLEATION, EPITAXY AND PLASMA

DEPOSITION ............................................................................................ 21

' 3.1 Nucleation: qualitative ideas ...................................................................... 21
3.2 Epitaxy ....................................................................................................... 22
3.2.1 Crystal grth modes .......................................................................... 23
3.2.2 Substrates for diamond epitaxy ............................................................ 24
3.2.3 Other substrate criteria ......................................................................... 26
3.3 Thermodynamics and kinetics of diamond nucleation .............................. 27
3.3.] Chemical vapor deposition .................................................................. 27
3.3.2 Development of vapor growth methods for diamond growth .............. 28
3.3.3 Diamond chemical vapor deposition methods ..................................... 29
3.4 Plasmas and chemical vapor deposition of diamond ................................. 29
3.4.1 Properties of plasma ............................................................................. 30
3.4.2 Microwave plasma discharges ............................................................. 34
3.5 Models of diamond nucleation for vapor phase grth ............................ 34
CHAPTER 4 PRIOR STUDIES OF EPITAXIAL DIAMOND GROWTH ................... 36
4.1 Definition of diamond heteroepitaxy ......................................................... 36
4.2 Diamond epitaxy on bulk substrates .......................................................... 40
4.2.1 Cubic boron nitride .............................................................................. 40
4.2.2 Nickel ................................................................................................... 41
4.2.3 Platinum ............................................................................................... 4 1
4.2.4 Silicon .................................................................................................. 42
4.3 Diamond epitaxy on thin films .................................................................. 42

vii

4.3.1 Pt{111}/sapphire{0001}andPt{111}/SrTiO3{111} ........................... 42

4.3.2 B-SiC/Si ................................................................................................ 43
4.3.3 Iridium .................................................................................................. 44
4.4 Comments and perspectives ....................................................................... 48
CHAPTER 5 THE BIAS PROCESS ............................................................................... 49
5.1 Seeding for diamond growth ...................................................................... 49
5.2 Biasing ....................................................................................................... 50
5.3 Models and mechanisms for diamond nucleation due to BEN .................. 50
5.3.1 Ion bombardment ................................................................................. 51
5.3.2 Electron emission ................................................................................. 57
5.4 The secondary plasma ................................................................................ 59
5.5 Biased enhanced nucleation on iridium ..................................................... 64
5.6 Summary .................................................................................................... 65
CHAPTER 6 SUBSTRATE PREPARATION FOR DIAMOND HETEROEPITAXY .21
‘ 6.1 Chapter overview ....................................................................................... 66
6.2 Substrate preparation for iridium deposition ............................................. 67
6.2.1 Strontium titanate substrate preparation .............................................. 68
6.2.2 A-plane sapphire substrate preparation ................................................ 70
6.3 E-beam iridium deposition ......................................................................... 73
6.3.1 Procedure on strontium titanate and sapphire ...................................... 74
6.3.2 Description of iridium growth mechanism on strontium titanate and
sapphire ................................................................................................ 74
6.3.3 Structure and morphology of Ir films on SrTiO3 ................................ 75
6.3.4 Temperature dependence of iridium morphology on SrTiO3 .............. 79
6.3.5 Morphology and structure of iridium films on (112O)A1203 .............. 82
6.3.6 Backside coating .................................................................................. 83
., 6.4 Silicon preparation ..................................................................................... 84
6.5 Summary .................................................................................................... 85

’ CHAPTER 7 MICROWAVE PLASMA CHEMICAL VAPOR DEPOSITION OF

DIAMOND ................................................................................................ 66

7.1 Chapter overview ....................................................................................... 86
7.2 Chemical vapor deposition system ............................................................ 86
7.2.1 Reactor and vacuum chamber .............................................................. 87
7.2.2 Gas handling ........................................................................................ 90
7.2.3 Modifications for biasing ..................................................................... 91
7.2.4 Computer operation ............................................................................. 93
7.3 Operation procedure ................................................................................... 93
7.3.1 Procedure for heteroepitaxial diamond growth on iridium .................. 94
CHAPTER 8 HETEROEPITAXIAL NUCLEATION OF DIAMOND .......................... 97
8.1 Chapter overview ....................................................................................... 97
8.2 Role of bias current density ....................................................................... 98

viii

8.2.1 Arcing during biasing ........................................................................ 101

8.3 Improving bias conditions ....................................................................... 103
8.3.1 The molybdenum cap and sample holder .......................................... 107
8.3.2 Sample height relative to the plasma ................................................. 108
8.3.3 The secondary plasma ........................................................................ 110

8.4 0+ growth experiments — The effect of biasing on Ir/SrTiO3 and Ir/A1203 ,,,,,

............................................................................................................ 111
8.4.1 Bias experimental procedure (0' growth) .......................................... 111
8.4.2 20 min bias ......................................................................................... 113
8.4.3 60 min to 80 min bias ......................................................................... 115
8.4.4 120 and 180 min bias ......................................................................... 117

8.5 Quantitative analysis of 60 min bias on Ir/A1203 ..................................... 119

8.6 Discussion ................................................................................................ 124
8.6.1 Nucleation and early grth model of heteroepitaxial diamond ....... 126

8.7 Summary .................................................................................................. 127

CHAPTER 9 HETEROEPITAXIAL DIAMOND GROWTH: CHARACTERIZATION

AND ANALYSIS ...................................................................................... 86

9.1 Short growth experiments - s 3hr growth time ........................................ 129

9.2 Progression of diamond growth — Ir/SrTiO3 ............................................ 130

9.2.1 Early stages of growth — analysis using 2D-FF T ............................... 130

9.2.2 Complete coalescence — 30 to 180 minutes of growth ...................... 132

9.2.3 Comparison with other results by other groups ................................. 133

9.2.4 X-ray diffraction studies on 180 minute diamond films .................... 135

9.2.5 Effect of small angle substrate offcut - Ir/SrTiO3 ............................. 138

9.3 Evolution of diamond growth — Ir/A1203 ................................................. 140

9.4 Thick films — Ir/SrTiO; ............................................................................ 142

9.4.1 SEM cross-sections ............................................................................ 145

9.4.2 X-ray diffraction ................................................................................ 147

9.4.3 Electron backscattering diffraction .................................................... 148

9.4.4 Raman microscopy ............................................................................. 149

9.5 Summary .................................................................................................. 152
CHAPTER 10 CONCLUSIONS AND OUTLOOK ........................................................ 129
10.1 Summary of this research ......................................................................... 154

10.2 Suggestions for further work ................................................................... 156
REFERENCES ................................................................................................................ 158

ix

List of Tables

Table 1 Coefficient of thermal expansion for various substrate materials used in

diamond heteroepitaxy (at 300 K). ....................................................................... 25
Table 2. Properties of a typical microwave plasma discharge (45). ................................. 34
Table 3 Summary of diamond epitaxy experiments on bulk crystal substrates ................. 38
Table 4 Summary of epitaxy on thin film single crystal substrates ................................... 39
Table 5 Sample preparation for Ir and diamond growth system ........................................ 45
Table 6 Growth parameters for groups studying diamond heteroepitaxy on Ir ................. 46
Table 7 Diamond nucleation on Si ..................................................................................... 98

Table 8 Data from experiments carried out using the sample configuration in Figure

43. ....................................................................................................................... 100
Table 9 Comparison of diamond growth on Ir with growth on Si. .................................. 103
Table 10 Parameters for carbon saturation of the Ir surface during biasing. ................... 106
Table 11 Calculated values for a 2D hexagonal array. .................................................... 124
Table 12 Bias and growth parameters for short grth experiments .............................. 129
Table 13 Bias and 2-step growth CVD conditions for thick diamond samples ............... 142
Table 14 Cross-section thickness measured from SEM or optical micrographs. ............ 145

List of Figures

Figure 1 Diamond lattice structure. .................................................................................... 4

Figure 2 Diamond structure as two superimposed FCC lattices displaced by a0/4 along
the body diagonal .................................................................................................... 5

Figure 3 Schematic of a seaming electron microscope. ...................................................... 8

Figure 4 Penetration depth of various electrons and photons produced by electron

beam interaction with the sample. ........................................................................ 10
Figure 5 Raman scattering geometry. ................................................................................ 13
Figure 6 X-ray scattering geometry ................................................................................... 15
Figure 7 The Bragg condition for constructive interference .............................................. 16
Figure 8 Ewald construction .............................................................................................. 17
Figure 9 X-ray diffractometer geometry ............................................................................ 18
Figure 10 Schematic of an AFM in tapping mode. ............................................................ 20

Figure 11 The three growth modes as a function of surface coverage, 6, in
monolayers ............................................................................................................ 24

Figure 12 Laboratory and space plasmas on a density (II) vs. electron temperature (Tc)

log-log plot ............................................................................................................ 31
Figure 13 A DC Plasma glow discharge between two parallel plates .............................. 33
Figure 14 Nucleation and growth model of diamond grth on a substrate .................... 35
Figure 15 SEM micrograph of a 1.5 pm thick diamond film on

(111)Pt/Ir/Pt/(0001)sapphire substrate ................................................................. 43
Figure 16 Plot of nucleation density for negative bias voltages keeping the time-

integrated current constant for each data point .................................................... 55
Figure 17 Ha Balmer line (7L=656.3nm) intensity .............................................................. 60

Figure 18 I-V curve for a diamond coated silicon substrate, a silicon substrate, and a
clean graphite sample holder. ............................................................................... 61

Figure 19 I-V plot for a bare silicon substrate and a diamond covered silicon substrate ..62

xi

Figure 20 SrTi03 unit cell. a0 = 0.39050 nm ..................................................................... 68
Figure 21 SrTi03 structure with Ti02-terminated surface. ................................................ 68

Figure 22 SrTi03 surface AFM scan after ultrasonic cleaning for 10 minutes each in
acetone and methanol, followed by DI water rinse. ............................................. 69

Figure 23 AFM image and 3D profile of typical SrTi03 surface after surface

preparation. ........................................................................................................... 69
Figure 24 Representative AF M images of two different a—plane A1203 substrates. ......... 70
Figure 25 Schematic of sapphire structure, unit cell, and orientation of (1120) plane. ....71
Figure 26 Surface of (1120) a-plane of A1203. ............................................................... 72
Figure 27 Possible epitaxial relationship of (001) Ir on (1 150) A1203. ........................... 73
Figure 28 X-ray 20-0 scan of Ir on SrTi03 using Cu Ka radiation. . ............................... 75
Figure 29 The Gaussian linewidth extracted from a least-squares fit to the data: 003°

for (200) SrTiO3 and 019° for (200) Ir. ................................................................ 76
Figure 30 1x1 um2 area AFM images of Ir on SrTi03 ...................................................... 77
Figure 31 Ir roughness vs. substrate thickness measured from 1x1 um AF M scans .......... 78
Figure 32 X-ray analysis of Ir (200) vs. film thickness. ........ - ........................................... 79
Figure 33 1x1 um AFM scans of Ir on SrTi03 deposited at different substrate

temperatures .......................................................................................................... 80
Figure 34 Iridium roughness as a function of deposition temperature. ............................. 81

Figure 35 X-ray rocking curve width of Ir (200) as a function of substrate
temperature. .......................................................................................................... 81

Figure 36 X-ray diffraction rocking curves for 300 nm epitaxial Ir grown on (1120)

A1203. ................................................................................................................... 82
Figure 37 AFM scans of (1 120) sapphire substrate before (upper) and after (lower)

(OOl)Ir deposition. ................................................................................................ 83
Figure 38 Microwave plasma CVD diamond reactor system ............................................ 87
Figure 39 TM013 Electromagnetic mode of reactor cavity. ............................................... 88

xii

Figure 40 Cutaway scale drawing of CVD reactor. The base plate rests on the

vacuum chamber. .................................................................................................. 89
Figure 41Exploded view of CVD stage and sample holder setup. .................................... 91
Figure 42 Circuit for bias setup. ........................................................................................ 92
Figure 43 Sample holder, sample platform and bias stage with Si mask and 2x2 cm2

Mo sample holder. ................................................................................................ 99
Figure 44 Highly ordered diamond grown on (001)Si ..................................................... 100
Figure 45 Kt) profile for biasing on Si (22-Apr-00). ....................................................... 101
Figure 46 Illustration of the Mo sample holder and arrangement on the stainless steel

platform ............................................................................................................... 102
Figure 47 I-V of round cap geometry with Ir/oxide substrate ......................................... 103
Figure 48 Temperature vs. pressure at -150V bias and 0V bias. ..................................... 104
Figure 49 Growth temperature at 32 Torr as function of microwave input power. ......... 105
Figure 50 Temperature as a function of negative bias voltage at 32 Torr. ...................... 105
Figure 51 SEM images of sample (17-Jul-00) with conditions of Table 10 .................... 107
Figure 52 Sample 07-Nov-2000 after standard 60 min biasing and 3 hr growth. ........... 108
Figure 53 The temperature during bias vs. sample holder height; relative to the fixed

stainless steel platform in the CVD reactor. ....................................................... 109
Figure 54 Cross-section of the bias stage and the sample holder .................................... 110

Figure 55 SEM image of sample 20-Jul-02 of (001) Ir on SrTi03 after 20 minute bias .1 13

Figure 56 Center region of 20 min sample (20—Jul-02) ................................................... 114
Figure 57 20 min bias sample (20-Ju1-02) SEM image. .................................................. 114
Figure 58 High resolution SEM image of 20 min bias sample (20-Ju1-02). .................... 115
Figure 59 Central region of 60 min bias sample (18-Jul—02) ........................................... 116
Figure 60 SEM image of 80 min bias sample (22-Jul-02). .............................................. 116

Figure 61 Edge (within 200 pm of the inside cap edge) of 60 min bias sample (21-Jul-
02). ...................................................................................................................... 117

xiii

Figure 62 120 min bias sample (08-Sep—00). 30° sample tilt ........................................... 118

Figure 63 Comer of 180 min bias sample (08-Sep-00). 30° sample tilt. ......................... 118
Figure 64 SEM scan of 180 min bias sample (08-Sep-00). 30° sample tilt. ................... 119
Figure 65 High resolution SEM image of 60 min bias sample 12-Apr-02. ..................... 120

Figure 66 Equalized SEM scan (left) of sample 12-Apr-02 and its binarization (right). 121

Figure 67 2-D power spectrum of the binarized image shown in Figure 66 and its

radial cross-section ............................................................................................. 122
Figure 68 Radial distribution function plot of 60 min bias sample ................................. 123
Figure 69 Model for heteroepitaxial nucleation and growth of diamond on Ir ............... 126

Figure 70 SEM micrograph of the sample surface after 60 minutes of biasing plus 1
minute of growth (17-Jul-00) .............................................................................. 130

Figure 71 SEM micrographs of 5-20 minutes diamond growth and the corresponding
2-D power spectra. .............................................................................................. 131

Figure 72 Diamond on Ir/SrTiO3. (a) 30 min grth SEM image sample 21-Jul-00 ,
(b) 30 min growth AF M scan of 21-Jul-00 , (c) 60 min 09-Apr-01, (d) 180
min, 23-Jun-01. ................................................................................................... 132

Figure 73 (la) SEM micrograph of the center of a 173 nm 3: 2 nm thick diamond film
on (001) Ir/SrTi03 grown in this study. (2a) SEM micrograph of a 600 nm
thick diamond film on (001)1r/SrTi03 ................................................................ 134

Figure 74 (1b) SEM micrograph of the edge of the diamond film shown in Figure 73
(la), 1.5 mm from the center. (2b) Edge of the epitaxial area of the film

shown in Figure 73 (2a) ...................................................................................... 134

Figure 75 {111}—¢ X-ray scans of 180 min diamond on Ir/SrTi03. .............................. 136

Figure 76 (OKL) X-ray area scan of a 180 minute diamond film on Ir/SrTi03 ............... 137
Figure 77 Scanning electron micrographs of diamond on Ir on offcut SrTi03

substrates ............................................................................................................. 138

Figure 78 AF M scans of diamond on Ir on offcut (001) SrTi03 substrates .................... 139

Figure 79 RMS roughness of diamond on Ir on offcut (001) SrTi03 substrates. ............ 139

Figure 80 SEM micrographs of diamond on Ir/A1203 grown for 5 min (left) and 180
min (right). .......................................................................................................... 140

xiv

Figure 81 SEM scans of short growth on (111) Ir/A1203 and the corresponding 2-D
power spectra. ..................................................................................................... 141

Figure 82 (a) Typical surface of a diamond film grown for 6 hr on Ir/SrTi03. The
white speck is a piece of dust. (b) Fractured piece near the edge of the film
that shows a cleaved surface. .............................................................................. 143

Figure 83 (a) Edge (approx. 1.5 mm from center) of the 12 hr diamond film on
Ir/SrTi03. (b) Macro steps at the center of the film scanned with AF M. ........... 143

Figure 84 (a) A complete diamond film grown for 48 hrs on Ir/SrTi03. The substrate
broke in half during cooling. (b) Center of the film showing a crack in the

upper right-hand comer, and hillock-type defects. ............................................. 144
Figure 85 Optical micrograph of a single crystal slab of diamond. ................................. 145
Figure 86 Cross section of 25 um film showing cleavage angle of (1 1 1) plane ............. 146
Figure 87 Fracture surface of a 25 um diamond film revealing the (111) plane. ............ 147
Figure 88 Thickness dependence of the X-Ray rocking curve linewidth of diamond on

Ir/SrTi03. ............................................................................................................ 148
Figure 89 {111} pole figure from EBSD measurements on a 10x10um area, 120

separate points .................................................................................................... 149
Figure 90 Raman spectrum of a 25 um (001) diamond film on Ir/SrTi03 ...................... 150

Figure 91 Polarized Raman spectra for 35 um film. The arrow points to 1332 cm". ....151

Figure 92 Depth profile of Raman peak shift for 35 um film .......................................... 152

XV

Chapter 1 Introduction

Heteroepitaxial growth of advanced electronic materials depends on the
development of suitable lattice-matched substrate systems. Diamond is of particular
interest since, as a wide bandgap semiconductor, it may prove advantageous in devices.
Progress in diamond heteroepitaxy was instigated by the growth of hi ghly—oriented
crystallites of diamond on silicon, despite the existence of a large lattice parameter
mismatch (1-3). A significant advance occurred with the discovery that epitaxial iridium
films, grown as a buffer layer on magnesium oxide (MgO), could serve as a substrate for
the nucleation and growth of chemical vapor deposition (CVD) diamond (4-6). With a
lattice parameter 7.5% larger than diamond, Ir appears to have the requisite long-term
chemical and physical stability in the high-temperature environment of the CVD
hydrogen plasma. Recently, the use of strontium titanate (SrTi03) as a replacement for
Mg0 has proven useful in decreasing the mosaic spread of the epitaxial Ir and the
resultant heteroepitaxial diamond (7, 8).

In parallel with heteroepitaxial growth efforts, the conditions for achieving high
diamond nucleation densities on various substrates have been extensively examined. The
bias-enhanced nucleation process (9), in which a negative voltage applied to the substrate
results in its bombardment by low-energy positive ions extracted from the plasma, is a

key step for inducing the formation of diamond nuclei. It is important that the nuclei

adopt the underlying orientation of the substrate and that their density be maximized, so
as to lead to rapid coalescence of crystallites during the early stages of growth. Although
the conditions that lead to effective nucleation are well known in principle, there is little
agreement on the physical mechanisms that underlie the process. The process is also
system-specific to some degree, depending on details of reactor geometry and a multitude
of processing parameters.

This thesis details the results of a series of investigations of nucleation and growth

of CVD diamond grown on epitaxial Ir with a (001) oriented surface. The oxide
substrates utilized for Ir growth included (100)SrTi03 and (1 120) A1203, although

(100)Mg0 and (100)LaA103 were also studied. Prior to Ir deposition, special emphasis
was placed on the preparation and characterization of the low-index substrate surfaces. A
bias-assisted step was developed and optimized to produce high densities, of order 1012

cm'2

across the Ir surface, irrespective of the underlying oxide substrate. High
nucleation densities resulted in grth of (001) single crystal diamond films exhibiting
homogeneity on a scale of mm.

In the course of the investigation, a significant advance was made in observing the
very early stages of grth of epitaxial diamond. This will lead to an understanding of
the mechanisms behind single crystal growth and nucleation as they apply to diamond as
well as other lattice mismatched systems. This in turn will lead to the realization of
superior materials properties of diamond.

The outline of this thesis is as follows:

Chapter 2 covers the crystal structure and bonding of diamond as well as the

analytical tools used for studying iridium and diamond films. Chapter 3 provides a

background for understanding nucleation and growth of diamond. Microwave plasma
CVD is described, as well as basic properties of glow discharge plasmas. Chapter 4
reviews the current status of diamond heteroepitaxy, and places the present work in the
context of previous investigations. Chapter 5 describes the biased-enhanced nucleation
method for diamond, and the physical mechanisms that lead to nucleation. The
“secondary plasma” is discussed and its role in creating conditions for enhanced
nucleation. Chapter 6 describes our methods for substrate preparation and epitaxial Ir
with results for SrTi03 and a-plane A1203 substrates. Chapter 7 discusses the microwave
CVD reactor configuration and procedures used to grow heteroepitaxial diamond on Ir.
Chapter 8 describes our nucleation results, showing the relationship to bias current
density, and provides a rationale for our high nucleation densities. We also present a
model for the nucleation and early growth of heteroepitaxial diamond. Chapter 9 presents
results of diamond growth, characterization of thick diamond films, and a comparison
with previous reports. Chapter 10 summarizes the major findings and makes suggestions

for further work.

Chapter 2 Diamond: structural properties and

characterization methods

2. 1 Diamond crystal structure

The diamond crystal structure is face-centered cubic (FCC) with lattice parameter
ao=0.3567 nm with a (0,0,0) and (1/4,1/4,1/4) basis, Figure 1. It can be viewed as two
superimposed FCC lattices, one displaced by a0/4 along the body diagonal of the other

(Figure 2). It belongs to the space group Fd3m.l

 

lo——°0———u

Figure 1 Diamond lattice structure. The arrows show the primitive translation vectors.
Nearest neighbor bonds are shown by the solid lines with bond length 0.154
nm. ao=0.3567 nm

 

' Space groups: The combination of all available symmetry operations that generate space-filling periodic
lattices. There are 230 total space groups. Fd3m is a cubic space group. (10)

 

Figure 2 Diamond structure as two superimposed FCC lattices displaced by ao/4 along
the body diagonal. Dark lines show tetrahedral bonding.

2.2 Carbon bonding in diamond leads to unique properties

Carbon is a group IV element and in the diamond form is the prototypical and
eponymous structure for silicon, germanium and gray tin. Its four valence electrons
hybridize to form tetrahedral sp3 bonds with a C-C interatomic distance of 0. 1 54 am.
With the exception of the sp2 C-C bond, with 0.142 nm e.g., the basal plane of graphite, it
has the shortest bond of any known element, resulting in a highly compact and rigid
three-dimensional crystal.

The electronic band structure of diamond is somewhat similar to silicon. The
conduction band minima occur along the six <100> directions near, but not at, the edge
of the Brillouin zone. Thus, diamond is an indirect band gap material. Whereas silicon
and germanium have bandgaps of 1.12 eV and 0.67 eV respectively, the gap for diamond
is 5.5 eV. This can be compared to other wide bandgap semiconductors, such as IV -IV

silicon carbide (2.9 eV) and III-V gallium nitride (3.4 eV). The strong sp3 bonding and

the short bond lengths in diamond lead directly to the large bandgap. Although the
indirect bandgap does not allow strong band to band luminescence, the unusual physical
properties of diamond (high thermal conductivity, hardness, high carrier mobility and
breakdown voltage) have made it extremely attractive as a potential replacement for
silicon in specific electronic applications.

Diamond has few carriers excited into the conduction band, even at relatively
high temperatures, due to its wide bandgap. At temperatures as low as 200° C, intrinsic
carrier levels in silicon, rather than dopant levels, begin to dominate the conductivity.
Furthermore, silicon devices typically fail by high temperature degradation due to electric
field induced diffusion. For these reasons, as well as its high thermal conductivity,
diamond is viewed as an attractive material to realize high temperature and high
frequency devices.

The wide bandgap and strong covalent bonding of diamond also contribute to
another unusual property: negative electron affinity (NEA) of the hydrogen-terminated
diamond surface (11). NBA implies that the energy of the lowest conduction band state
(conduction band minimum) lies above the vacuum level, i.e., the continuum of unbound
electron states outside the semiconductor. Therefore, the electrons at the surface can, in
principle, easily escape into the vacuum, an attractive property for an electron emitter.
Presently, difficulties exist in realizing this idealized picture for diamond. For example,
attempts to introduce significant amounts of donors (electrons) into diamond have been
somewhat unsuccessful, although acceptor d0ping with boron leads to a level 0.38 eV
above the valence band maximum. Finding suitable substitutional dopants is difficult in

diamond because incorporating foreign atoms into the dense diamond lattice introduces

large strain. This, in turn, may lead to dopant energy levels that lie deep in the energy

gap, far from the band extrema.

2.3 Epitaxial film structural analysis tools

The methods and equipment used to analyze the iridium and diamond films
produced in this investigation are:

1. Scanning Electron Microscopy — This method yields 2-D images of the surfaces of
films at scales from mm to 10 nm. This allows rapid visualization of the
microstructure of the films. We use SEM for initial analysis of grain coalescence,
measuring film thickness, determining cleavage planes, nucleation sites and densities,
and combined with EBSD for crystal orientation determination.

2. Raman Spectroscopy — Analysis of high-energy vibrational modes provides
“chemical” information, principally on the C-C bond type, and the effect of stress.

3. X-ray Diffraction — Yields basic structural information of the films, and the influence
of stress.

4. Atomic Force Microscopy — Produces topographic scans of surfaces with a lateral
scale of a few pm, but a vertical scale as small as 0.1 nm. Combined with SEM,

AF M can distinguish between topographic and other constrast mechanisms.

These techniques are explained in more detail below. In addition, we used optical

microscopy with Norrnarski differential interference contrast (DIC) to characterize

diamond, iridium, and oxide substrates at length scales >10 mm.

2.3.1 Scanning electron microscopy

A scanning electron microscope forms an image of a sample surface using a beam
of electrons swept across the surface of a solid sample in a raster pattern. The electron
beam is focused with a system of electromagnetic lenses and a pair of magnetic scan coils

control the position of the beam. A schematic of a SEM is shown in Figure 3.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable high voltage
Electron gun "‘t power my
! .
Electron beam \ i
\l
I
l
l 0
Magnetic I o o
condenser l o g 3
lens I g g
I 0 0
I
l
l Scanimagniticailon)
i call controls
Magnetic 0 i g
objective 3 I 3
lens 8 t 3 Computer ‘1
I O
— l W
I
l
l

 

 

To vacum 4— display

 

Sample Chamber

 

 

 

Figure 3 Schematic of a seaming electron microscope.

The Hitachi 4700 11 Field Emission microscope (FESEM) used for the imaging
work in this thesis has a cold field emission source as opposed to a hot thermionic one
found in standard SEMs. The gun is a tungsten cathode shaped to a very sharp point of
<100 nm radius. The tip is held at a high potential, so the electric field at the tip is >107
V/cm. This allows the electrons at the Fermi level to penetrate the potential barrier by
tunneling, rather than thermionic emission, giving a field emission source a beam
crossover diameter of 10 nm compared to 10 pm for thermionic guns used in
conventional SEMs. A FESEM requires less magnetoptics to demagnify the beam before
it reaches the sample, allowing for greater beam brightness, and sharper resolution. The
Hitachi 4700 II FESEM has three electromagnetic objective lenses that can focus the
beam to achieve a maximum resolution of 1.5 nm at an acceleration voltage of 15 kV
under ideal conditions.

Electron-solid interactions produce a large number of particles and photons. The
most important of these are: Secondary electrons (SE), backscattered electrons (BSE),
Auger electrons (AE) and characteristic x-rays (12). The penetration depth and origin of

these are represented in Figure 4.

  

~— Elactron range R

Figure 4 Penetration depth of various electrons and photons produced by electron
beam interaction with the sample. PE is the incident electron beam. R~102 nm,
but is dependent on the incident electron beam energy as well as the nature of
the material.

For imaging with the SEM, the most important of these are the secondary
electrons and the backscattered electrons (12, 13). With incident beam energies ~10 keV,
secondary electrons have energies <50 eV sharply peaked at 3-5 eV that are produced by
inelastic scattering from weakly bound conduction electrons in the solid. They are
detected using a positively biased collector grid that accelerates them into a scintillator
and they are emitted as photons, recorded by a photomultiplier. The device is called an
Everhart—Thomley detector. The number of secondary electrons depends on the sample
tilt angle. Enhanced emission comes from edges and small particles. This allows the
sample topography image to be formed. The signal that rasters the electron beam across

the sample also scans a beam across a CRT. The output of the E-T detector modulates

the intensity of the spot on the CRT, producing an image map. The magnification M of
the SEM image is:

M = W / w
where W is the width of the CRT and w is the width of a single x-scan across the sample.
The Hitachi 4700 II has two SE detectors, one placed above the sample near the last
objective lens, the other at a low angle with respect to the sample.

BSE are electrons that exit the sample after undergoing multiple elastic and
inelastic collisions at large angles. BSE have a broad energy spectrum between SOeV and
E=eU and may also produce secondary and Auger electrons as they propagate through
the sample. BSE move in linear trajectories that are not affected by the Everhart-
Thomley detector because of their high energies. They are instead detected with a
semiconductor detector that gives a signal proportional to the BSE energy. BSE images
have a lower resolution than SE images due to the larger volume of interaction. However
the contrast mechanism for imaging back-scattered electrons is sensitive to the atomic
number (Z) of the sample. With a wide-angle detector below the objective lens where
only BSE with high take-off angles are detected, image contrast is produced mostly by Z-

differences, and topographic contributions are suppressed.

2.3.2 Electron backscatter diffraction (EBSD)

Backseattered electron yield is also dependent on the sample crystallographic
orientation. For a crystalline surface, BSE that are scattered at low angles with respect to
the sample surface will be scattered according to the Bragg condition. Electron

backscatter patterns (EBSP) are formed when the BSE from a crystal undergo diffraction

ll

as they re-emerge from the surface (14). Instead of Bragg spots, the diffracted electrons
emerge in a pattern of lines called Kikuchi bandsz.

For EBSD analysis, we mount the samples in a CamScan 44F E SEM at a tilt
angle of 70° with respect to an incident 25 keV electron beam. The backscattered
electrons are detected on a fluorescent screen mounted in the vacuum chamber producing
the Kikuchi pattern. A high gain CamScan Ortex CCD camera transmits the image from
the screen to a computer via a video digitizing board. The diffraction image is analyzed
using the Channel4 sofiware package3 to index the pattern and obtain the orientation of
the crystal lattice. The sample stage is translated, or scanned, by computer-controlled
movement of the stage in the SEM to obtain a spatial-orientation map of the sample.

A pole figure may be obtained from the series of point-by-point orientation
measurements. A pole figure, a standard measure of the texture of a polycrystalline
sample, is a plot that shows the density distribution of orientations for a selected set of

crystal planes as a stereographic projection (14).

2.3.3 Raman spectroscopy

Raman spectroscopy is frequently used for characterizing diamond and other
carbon films (15-1 7). The Raman effect is the inelastic light scattering from internal

excitations, phonons here. Incident light of energy ha) is scattered by a polarizable

sample. This process causes a frequency shift in the scattered light of a) = 60' i STE ,

where AE corresponds to the phonon energy — the so-called Raman shifi. In 3 Raman

 

2 The Bragg condition is fulfilled for all directions of incidence where the wave vector k lies on a Kossel
cone.
3 HKL Technology, 1997.

12

spectrometer, the scattered light is dispersed by a grating, which allows peaks in the
spectrum to be resolved, and assigned to vibrational modes of the crystal (18).

In crystals, Raman selection rules allow only near zone-center optic vibrational
modes to participate in first-order Raman scattering4 (I 9). For diamond, the
characteristic first order peak that corresponds to the temperature dependent C-C
stretching mode appears at 1332 cm'1 with a linewidth of 1-2 cm'1 (15, 20—22). The
intensity of the Raman peak depends on a Raman tensor derived from the crystal
symmetry and the polarization directions of the incident and scattered light. For diamond
(point group 01.) oriented in the <001> direction, the Raman line in the (001)
backscattering geometry (Figure 5) is a function of the azimuthal angle of the incident

light polarization é, relative to the crystallographic axes.

<001>

 

 

 

 

Figure 5 Raman scattering geometry. For a (001) crystal surface in the backscattering

geometry, the incident, IE and scattered, k0“, ,

rn’

wavevectors are in the <001>

direction. life“, is polarized according to the Raman selection rules for 0;, point
group crystals.

 

° Zone-center optic: those phonon vibrations that retain the full crystal symmetry with non-zero energy at
the zone center (k =0).

13

The scattered intensities, Ill and I.L , are either parallel or perpendicular,
respectively, to the incident polarization 25,. The scattered intensity is related to the

Raman tensor (R,) and the polarization vectors (é, , 5].) by
I=c-Z:le,.-Rj-ej|2 i,j=1,2,3

The Raman selection rules for this case reduce to:

III 2 0 and

I1 = c-dz, for 5,. || [100] and

III = c - d2

[i=0, foréi||[110]
where c is a constant, d is a parameter proportional to the intensity of the incident light
and the Raman scattering cross section, and E, is the incident light polarization.

For the Raman measurements described here, a Kaiser Optical Systems
HoloProbe Raman spectrograph coupled to an Olympus BX-60 optical microscope was
used with the 532 nm line of a frequency doubled Nd:YAG laser source. The
polarization measurements used a fixed polarization of the incident beam and a M2 plate
to rotate the scattered beam polarization. A 100X objective was used on the microscope
for all measurements; this allowed the incident light to be focused to a spot
approximately 1pm in diameter. As a result, both lateral (x,y) as well as depth (2) spatial

resolution was possible. A natural type IIa diamond single crystal was used as a

reference, with a single peak at 1332.3 cm" and linewidth 2.1 cm". With a high-

14

 

resolution grating, the instrumental resolution of the Raman spectrometer used here is 1

cm".

2.3.4 X-ray diffraction
X-ray diffraction measurements are done by placing a sample in a collimated,
nearly monoenergetic x-ray beam at an angle, to, to the surface of the sample, and

detecting the diffracted x-ray intensity as a function of the scattering angle 20 (Figure 6).

 

 

 

Figure 6 X-ray scattering geometry

The Bragg condition, n2 = 2d sin 6 , corresponds to constructive interference and
a peak in the scattered intensity, where 2 is the wavelength of the incoming x-rays, d is
the lattice spacing of the crystal, and n is an integer index corresponding to the scattering

order. See Figure 7.

15

 

 

 

Figure 7 The Bragg condition for constructive interference

The Ewald construction is useful for understanding the geometry of x-ray

diffractometers. i In an Ewald construction, Figure 8, the incident x-ray If,” has an angle of

incidence 0 with respect to the hkl plane. A circle is drawn around the origin of It," with a

radius of | I? I: 9:15.

16

 

 

sample J

 

Figure 8 Ewald construction

Diffraction results along I; only if Alt. 2 If“, 4;," 2 27:5,,“ where lib“ is the

out

reciprocal lattice vector having length I/d. This can be restated as the Bragg condition:

k —k. _2$inl9

out In

._ =1 2niz',,, |=1/d:>/l=2dsinl9
27d 27:2

The Ewald circle defines the Ewald sphere of radius 21t/2t about C in 3-D.
An x-ray diffractometer is used for precise measurement of the Bragg angles of
diffraction of a crystal. A typical x-ray diffiactometer with a 4-circle Eulerian-cradle is

shown in Figure 9.

l7

 

goniomeier

 

 

 

 

Figure 9 X-ray diffractometer geometry

The goniometer (a sample mount with three mutually perpendicular axes of rotation that
allows very accurate centering of the sample in the diffractometer) is mounted on a to-
circle on a horizontal plane with a vertical rotation axis. A x-circle is perpendicular to
the (D-CII'CIC with a horizontal axis. The head of the goniometer is mounted on a third
circle -— the ¢-CII‘CIC inside of x. The detector for the diffracted x—rays is on a fourth
circle, 0, concentric with (0, but decoupled from the other three circles. The zeros of the
circles are defined such that x=0 when the (ii-axis is vertical, and (0:0 when the x-axis is

parallel to the incident x-ray beam.

2.3.4.1 The co-scan or rocking curve

For an co-scan, the x-ray source and detector remain stationary. d), x, to and 20
positions are set to a calculated reflection position (i.e. (004) for the diamond structure).

The crystal is rotated a few degrees through the reflecting position about on, and the

18

intensity of the signal is recorded as a function of the angle. In the Ewald sphere

representation, the (lo-scan rotation is perpendicular to lib“ .

2.3.4.2 The 6-263can, or co-ZBscan

For the 0-20 scan, both crystal and the detector are moved. The crystal is rotated
by A0) as for the (lo-scan, while the detector is rotated in the 20 circle at 2 times the

angular velocity of the crystal rotation, A20=2Ato at all times. In the Ewald sphere

representation, the 0-20 scan is in the direction of lib”.

2.3.5 Atomic force microscopy (AFM)

An atomic force microscope uses a very sharp tip of radius <30nm to map the
morphology of a sample surface on a scale of a few 100 nm to roughly 10 pm. The tip is
at the end of a cantilever with a spring constant of the order 1 Newton/m. We use a
Digital Instruments model 3100 in tapping mode with a Si tip to image the samples. In
tapping mode, the tip is oscillated in air near the resonant frequency of the cantilever
using a piezoelectric crystal (23). The oscillating tip is brought near enough to the
surface of the sample to touch. Changes in the topography as the tip rasters across the
surface causes damping of the oscillations. The oscillation amplitude is maintained by a
constant feedback loop in the system controller that interprets changes in signal
amplitude as changes in topography. The movement of the tip is tracked by a
photodetector that captures the light signal of a laser focused on the backside of the
cantilever as it moves across the surface. A schematic of the system is shown in Figure

10.

19

 

Am - ltude Detector I

Flgh Resolution "
! Oscillator {

‘ Nanoscopelll ‘1
System Controller

........-, Uh.

 

 

photo-
detector

 

Silicon cantflever
with Integral tlp

 

Figure 10 Schematic of an AF M in tapping mode (23).

20

 

 

Chapter 3 Basics of crystal nucleation, epitaxy and
plasma deposition

This chapter presents a brief description of crystal nucleation, the principle of
epitaxial growth and crystal growth methods that were used in this work, particularly
chemical vapor deposition (CVD). The nucleation and growth of diamond is discussed in
this context. A description of DC glow discharges is given, as it will be used in Chapter

5 to explain the phenomenon of the secondary discharge during bias.

3.] Nucleation: qualitative ideas

The process by which a substance “tries out” another phase is the essence of
nucleation. In any medium, there are thermodynamic fluctuations. In the water-ice
transition, as the temperature of water approaches 0°C, molecules slow down and
occasionally a few attach in a way resembling ice crystals. Most of these “nuclei” break
apart again into liquid phase, but as temperature decreases, attachment becomes more
favorable. Eventually a nucleus reaches a critical size and growth of the crystal is
thermodynamically favored over breaking apart again. Molecules will add to the nuclei
and the transition from water to ice proceeds.

Nucleation can be broadly divided into two regimes, homogeneous and

heterogeneous. Homogeneous nucleation describes a phase transition that is driven by

21

 

thermodynamics. Heterogeneous nucleation, on the other hand, can be compared to a
catalytic process. Formation of a crystalline film on a surface is a heterogeneous process
if the presence of the substrate activates the nucleation. Heterogeneous nucleation is
essential for epitaxial thin film growth, as the growing film uses the substrate as a

template for crystal growth.

3.2 Epitaxy

A crystalline substance (deposit) that grows on a crystalline substrate is said to be
epitaxial if one crystal plane of the deposit crystal and one of the substrate are parallel.
This can be described in terms of the Miller indices of the crystal planes and directions.
For example, (001)d<110>d||(001)s<l10>s means that a (001) plane of the deposit resides
on a (001) plane of the substrate, with the deposit <110> direction parallel to the <110>
substrate direction. In most epitaxial systems, the coincident growth planes are identical
and of low index. Common growth planes are (001) and (111) for cubic systems or
(0001) for hexagonal systems. The coincidence of high index dissimilar planes is also
feasible. If a matching of two high index planes lowers the interface free energy, then
such an arrangement will be favored in equilibrium. The lattice mismatch, f between

deposit and substrate is defined as

f = (“on - 005)

“as
where aos and 001) are the lattice parameters or atom unit spacings of the deposit and the
substrate, respectively. A smaller lattice mismatch is possible if the relative difference

becomes smaller at integer intervals of unit atom spacings, i.e. naop, maps.

22

 

Homoepitaxy or autoepitaxy describes epitaxial systems where the deposit and
substrate are the same substance. By including dopants in the epitaxial layer, one can
produce thin layers of electronically modified materials on generic, low-cost substrates,
for example, B-doped Si on Si. Growth of diamond on diamond substrates will be
referred to as diamond homoepitaxy. Heteroepitaxy describes deposit-substrate systems
where the epitaxial deposit and substrate are dissimilar and generally lattice mismatched.
Diamond heteroepitaxy will be referred to as the epitaxial growth of diamond on another

material.

3.2.1 Crystal growth modes

Following nucleation, crystal growth generally proceeds by one of three different
modes: island growth, layer by layer, and layer plus island (24). In some cases a
combination of modes may occur. In 3-D island growth, also called Volmer-Weber
growth, islands of deposit first form then coalesce after reaching a critical thickness.
Islands then form on top of this deposit, and again coalesce. Here, the atoms of the
deposit are more strongly attracted to each other than to the substrate. This is expressed
as 713 + y. > 73, where y* is the interface energy, and yo and ys are the surface energies of
deposit and substrate (25). In 2-D layer by layer, or F rank-van der Merwe growth, single
layers form initially and grow mostly laterally along the substrate surface. For layer
growth to proceed, the atoms in the deposit must be more strongly attracted to the
substrate then each other, i.e., 713 + y. < 75. The intermediate case — layer plus island, or
Stranski-Krastanov growth, occurs if the interface energy increases with increasing

deposit thickness. Usually this is the case for strained layer-growth. In SK mode, the first

23

 

2 to 3 monolayers grow in a layer type mechanism, and then islands grow on the layers.

The three modes are shown schematically in Figure 11.

9.1M]. 22%”; W W

[—
1<9<2 / '////////////7/ 7////////////

o>2 - I l \ l '
/ WW/ 7////////////
(a) (b) (c)

 

 

 

 

 

Figure 11 The three growth modes as a function of surface coverage, 0, in monolayers
(ML): (a) island, or Volmer-Weber; (b) layer-plus-island, or Stranski-
Krastanov; (c) layer-by-layer, or Frank-van der Merwe growth.

3.2.2 Substrates for diamond epitaxy

For heteroepitaxy, the choice of substrate is dictated by the following
considerations:

(1) Similar structure as the deposit. For diamond this is the diamond, face-
centered-cubic (FCC), or zinc-blende structure.

(2) One, or more, similar lattice parameters. The diamond lattice parameter is
0.3567 nm. Since the C-C sp3 bond is 0.154 nm, most substrates will have
larger lattice parameters than diamond.

(3) Stability of physical properties. The substrate should have a melting point
much greater than the deposition temperatures. For diamond CVD, the

substrate needs to be stable in a hot hydrogen plasma. Matching thermal

24

expansion coefficients (or) are preferred to minimize stress during cooling. See

 

 

 

 

 

 

 

 

 

 

Table 1.
Material or (10'6 K’)
Diamond 0.8
Silicon 2.3
Iridium 6.4
SrTiO3 9
Sapphire 7.5
Platinum 8.8
Nickel 13.4
B-SiC 2.5

 

 

 

 

Table l Coefficient of thermal expansion for various substrate materials used in
diamond heteroepitaxy (at 300 K).

(4) Surface energy match. Epitaxy is more likely if the surface energies of the
two materials are similar, as is ofien the case for materials with similar lattice
parameter and chemical bonding. Differences in surface energies lead to the
different growth modes. For epitaxy, layer by layer growth is the preferred
mode because fewer defects are introduced during growth.

In practice, it can be difficult to find substrates that satisfy the above criteria. For
diamond, the small lattice parameter is only one of a number of hurdles that must be
overcome in finding suitable substrates. CVD plasma temperatures are generally 800-
1000°C, leading to rapid etching or decomposition of many substances. Table 1 indicates

that materials used for diamond grth have thermal expansion coefficients much greater

25

than diamond. In addition, diamond has an extremely high surface energy, 5.3 J/m2 for
the (111) plane, 9.2 J/m2 for the (100) plane (26), compared to the substrate materials
used for diamond growth. Large differences in surface energy between substrate and
deposit inhibit formation of nuclei. Consequently homogeneous diamond nucleation
occurs spontaneously only on other diamond surfaces.5

In the absence of an ideal economically attractive substrate, ways to circumvent
the above requirements must be found. Diamond has already been demonstrated to grow
with highly preferred orientation on silicon despite a 52% lattice mismatch, and a large
difference in surface energy (1.5 J/m2 for Si(100)). This indicates that diamond epitaxy
on highly mismatched substrates may yet prove feasible. As will be shown in this thesis,
single crystal diamond growth is possible on substrates that provide only a reasonably

good match.

3.2.3 Other substrate criteria

Substrates should maintain phase stability over the span of temperatures
encountered in diamond growth. Furthermore for diamond growth chemical stability,
namely, carbide formation and carbon solubility need to be considered. The role of
carbide formation is unclear and will be discussed below. It is also unclear whether
carbon solubility should be a factor in substrate selection. The two above conditions are
not necessarily mutually exclusive (27). Diamond can form epitaxially on a substrate that
1) forms a flat, well-crystallized, carbide layer under diamond growth conditions, or 2)
forms no carbide, but chemically bond with carbon, and has a small carbon solubility. In

either case, the above criteria limit the choice for substrates in diamond growth.

 

5 {111} c-BN with surface energy 4.8 J/m2 and lattice parameter 0.3612 nm is an exception.

26

 

 

3.3 Thermodynamics and kinetics of diamond nucleation

In contrast to the ice-water nucleation, the multitude of condensed carbon phases
greatly complicates diamond nucleation and growth. At CVD temperatures and pressures,
graphite is the ground state of carbon. Diamond is a metastable state of carbon. At room
temperature and pressure, the free energy difference between graphite and carbon is 0.03
eV per atom (28-30). However, there is a large activation barrier between the two
phases, thus graphite is the thermodynamically favored state.

There are in general, two ways to create diamond under laboratory conditions.

 

With the high pressure, high temperature (HPHT) methods, graphite is compressed and
heated in the presence of a catalyst to thousands of atmospheres and over 2000 K.
Chemical vapor deposition methods, on the other hand, operate at sub—atmospheric

pressures (10'3 to 101 Torr) and temperatures near 1000 K.

3.3.] Chemical vapor deposition

Chemical vapor deposition (CVD) is a crystal growth technique that involves
chemical reactions at a substrate surface. CVD is related to PVD (physical vapor
deposition) and belongs to the subset of deposition techniques that includes molecular
beam epitaxy (MBE), sputtering, evaporation (including electron beam evaporation) and
laser ablation. Physical deposition processes involve only a state change at a substrate of
the depositing species. Chemical deposition processes proceed using a chemical reaction,
or set of reactions, at the surface. The rate of deposition depends on the reaction rates
and the fluid dynamics of the system (31). Generally, the exact chemistry that takes place
is usually very difficult to elucidate experimentally, as the system may be far from

equilibrium (kinetically controlled) rather than thermodynamically dominated (at

27

equilibrium) (32). CVD may be classified into two general categories: thermal and
activated. In thermal CVD, gas flows past a substrate that is heated to high temperatures
to initiate surface reactions. In activated plasma CVD, energy to ionize gas species is
provided by an external source, such as a plasma or hot filament. Surface reactions are
controlled or strongly modified by the plasma properties. This allows the substrate to
remain at a lower temperature. However, energy from the plasma also leads to heating of
the substrate, blurring the distinction between the two categories. The diamond for this

work was grown using an activated plasma CVD process.

 

3.3.2 Development of vapor growth methods for diamond growth

CVD methods have a long history of usage for semiconducting film growth (33,
34). For diamond growth, CVD methods were initially complicated by the concurrent
formation of graphite and diamond (28, 35). The realization that preferential etching of
graphite by atomic hydrogen in methane-hydrogen gas mixtures provided the basis for
interest in hydrogen plasma CVD (28, 36, 37). The report of {110} faceted
homoepitaxial diamond growth from mixtures of hydrocarbon-hydrogen gases appeared
in 1981 (30). The process involved a thermal CVD process in a closed system by carbon
transport fi'om a solid graphite source to a diamond substrate using methane, ethane,
ethylene, acetylene, and atomic hydrogen. The growth rate was 1 rim/hr at 1000° C. All
other reported methods used before this point, suffered from a slow growth rate of 0.1
um/hr or less (28). Shortly after, a grth rate of several micrometers per hour was

achieved using microwave and hot filament methods of gas decomposition (38-40).

28

3.3.3 Diamond chemical vapor deposition methods

Diamond CVD methods use gas dissociation of hydrogen and carbon into atomic
hydrogen and carbon and methyl radicals. A good review of various gas activation
methods is given in (33). Hot filament CVD (HFCVD) uses a DC filament to ionize the
gases, whereas plasma torches are powered by DC are discharges. Plasma torch
deposition achieves some of the highest deposition rates recorded — in excess of 60 rim/hr
(41-43). Combustion methods are similar, using a highly exothermic reaction, e.g.
acetylene-oxygen, to heat the gas species. ECR (electron cyclotron resonance) uses an
applied magnetic field chosen so the resulting electron gyration frequency is equal to the
microwave frequency. This allows increased plasma densities by enhanced microwave
absorption (33). RF (radio frequency) and microwave plasmas use electromagnetic
radiation to heat and dissociate the gases to form a plasma. We use the last method.
Microwaves travel from the power source via a waveguide into a reactor where they
energize and dissociate the gases in a bell jar or other confined volume. A ball-like
plasma forms a few centimeters above the substrate. The plasma is confined by the

electromagnetic fields and the physical surfaces of the reactor.

3.4 Plasmas and chemical vapor deposition of diamond

The next section contains a brief introduction to glow discharge plasmas. The
understanding of the regions of a plasma, and the current-voltage behavior is important
for describing the bias-enhanced method we used for diamond nucleation. This is put into
context of microwave plasma chemical vapor deposition — the method used for diamond

growth in this study.

29

 

3.4.1 Properties of plasma

A plasma is a collection of positive and negative charge carriers with overall
charge neutrality (44). The charge neutrality condition gives plasmas some of their
unique properties and distinguishes them from a neutral gas. For instance, electric and
magnetic fields strongly influence plasmas. Charged particles of plasma will rearrange
themselves to shield external and internal electrostatic fields to preserve charge
neutrality. Charge neutrality, however, is only true on a macroscopic scale. At distances

less than AD, the Debye length, fluctuations of local charge densities can exist. 20 (in cm)

 

l / 2
is given by: AD z 743 *[13‘ ] where T e is the electron temperature (eV), and N. is the

number density of electrons (cm'3). For a typical microwave plasma discharge, T e z 5 eV
and Ne ~1012 cm'3, gives AD z 0.02 mm. This is typically much smaller than the system

size, so the plasma is “quasineutral”. The electrons and ions arrange themselves within
the Debye length so that charge neutrality is preserved for distances greater than ’10.

Plasmas differ due to pressure, particle density and temperature — see Figure 12.

30

 

 

25

  
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  
 
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Solid Si at
room temperature
®‘/ 4 3
31:11).“ < 1
L Laser
20 plasma Focus
/ Shock
”'3” tubes
pressure Theta
/ arcs plnChes
Fusion
15 _. reactor
2
7 /////
g . gpressure / Fusion
8’ Alkali y experiments
- metal
plasmas
10 — //
4 Glow é
Flames discharges;
Earth's
ionos-
phere Age > 1 cm
5 p. Solar
corona
lnterplanetary
Solar
wind
0 l l A l l r
10'2 10'1 O 1 2 3 4 5
'0810 TJV)

Figure 12 Laboratory and space plasmas on a density (n) vs. electron temperature
(Tc) log-log plot. AD is the Debye length (45).

Stars, for instance, are in thermal equilibrium, meaning all the species are completely

ionized and Te z 7:. , where T e and T,- are the temperatures of the electrons and ions. The

electrons and ions are extremely mobile (34). In the laboratory, gas discharge plasmas

31

occur when an electric potential applied across a gap “breaks down” the gas species to
form a weakly ionized plasma (45). Glow discharges generally have particle densities of
108-1012 cm'z. These plasmas are not in thermal equilibrium because the electrons tend to

be much hotter than the ions i.e., 7;»7} , and thus the plasma temperature is normally

given by T e. The electrons are colliding with gas molecules, dissociating them at
temperatures much lower than in a thermal plasma (34). This allows plasma processes in
activated plasmas to occur at much lower temperatures than in thermal plasmas.
Ionization occurs by electron collisions in the gas, rather than by thermal means, and the
plasma is called a glow discharge.

Glow discharges can be formed at reduced pressures (P<atmosphere) by DC or
AC electric fields applied across electrodes and AC current or radio or microwave
frequencies passed through a coil around a tube. A DC plasma glow discharge and its

main features are illustrated in Figure 13.

32

 

An e glow

to) Cathode layserlsil Negative glow

 

 

 

 

 

 

 

 

 

 

     

Aston
Cathode

   

Faraday Anode Region

(b) Luminous
Intensity

Ve

(0) Potential

(d) Field

(6) Space-
charge
density

(1’) Current
density

 

Figure 13 A DC Plasma glow discharge between two parallel plates (44).

The glow discharge contains regions that are defined according to their luminosity,

potential, field strength, space charge density and current density. Starting from the

cathode the first three regions, the Aston dark space, the cathode glow and the cathode

(or Crooke Hittorl) dark space comprise the cathode region. Most of the voltage drop in

the discharge occurs in this region. The cathode glow has a relatively high ion density,

and depending on the voltage will mask the dark spaces and cover the cathode. Next,

come the negative glow region, the Faraday dark space, and the positive column. The

positive column is often considered the main discharge, because it generally expands to

occupy most of the volume. The negative glow has the brightest intensity, and electrons

carry most of the current in this region. Last the anode dark space and the anode glow

33

comprise the anode region.

3.4.2 Microwave plasma discharges

At excitation frequencies >1MHz, direct contact of the plasma with electrodes is
not needed. Electrodeless discharges can either be RF (radio frequency) or microwave
discharges. Commonly used frequencies in commercial apparatus for microwave plasma
chemical vapor deposition (MWPCVD) are 2.45 GHz and 915 MHz (33). Some

properties of typical microwave plasmas used for deposition are given in Table 2.

 

 

 

 

 

 

Power P (W) 500-5000
Pressure p (Torr) 0.1-10
Plasma density n (cm’3) 1010-10If
Electron temperature Te (eV) 2-7

Ion acceleration energy E; (V) 20-500
Fractional ionization 10'7-104

 

 

 

 

Table 2. Properties of a typical microwave plasma discharge (45 ).

A microwave plasma discharge does not typically exhibit the different regions described
for a DC discharge in the previous sections. Microwave plasma chemical vapor
deposition uses a glow discharge plasma formed at reduced pressure by microwave

excitement of the reactant gases.

3.5 Models of diamond nucleation for vapor phase growth

Most models of low pressure chemical vapor deposition of diamond consider it an

activated process of reactants: hydrogenated carbon radicals, and atomic and molecular

34

 

hydrogen. The basic ideas are shown in Figure 14 for methane and hydrogen gases. The

primary chemical reactions are summarized (32):

H, —>2H° (1)
CH, +XTI° —>CH,_,r +XH, (2)

l l l
- C—H+CH,,_,{—)-C—C—H+YH2 (3)

Equation (1) illustrates the dissociation of molecular hydrogen into atomic
hydrogen; (2) shows the reaction of the atomic hydrogen to “abstract” a hydrogen radical
from a methane molecule, forming a hydrogen molecule and a methyl radical, (3) shows
the attachment of the methyl radical to a surface hydrogen terminated carbon atom, and

subsequent detachment of a hydrogen molecule.

 

  
 

   
      
 
 

 

 

 

 

 

       

DIflelon Nucleation Growth
* * o - H
H CHx H H* CHX . I C
rib s
CH4 removal H 2 Growing
oometastable cluster . ..... C Sta
Sufaoe \’ ll ./ cl .er . ' '
Diffusion . .. . .
no buk diffusion substrate
\ \ \ \ \ \\\ \ \ \ \

 

 

 

Figure 14 Nucleation and growth model of diamond grth on a substrate (32).

35

Chapter 4 Prior studies of epitaxial diamond growth

4.] Definition of diamond heteroepitm

As a working definition of diamond heteroepitaxy, we adopt the following

criteria:

(1)

(2)

(3)

Size — Growth over areas of order 1mm2 or larger. This is necessary because
it has long been possible to grow single crystals of diamond on substrates,
including Si, that are facetted, but whose size rarely exceeds 10 pm on a
side. Diamond heteroepitaxy implies that large scale coalescence of such
oriented grains must occur over substantial regions of a substrate.
Chemical — A Raman peak at 1332 iIOCm”, the C-C stretch in diamond,
that obeys the diamond selection rules for polarization as discussed in
section 2.3.3.

Structural - X-ray, EBSD, or other diffraction confirming epitaxial
alignment of diamond with the substrate. The linewidths provide a
qualitative measure of coalescence, the distribution of orientations of
crystallites and their grain boundaries. Epitaxial diamond is expected to

have x-ray linewidths <l°.

We contrast these criteria with those describing highly ordered diamond (HOD); films

that demonstrate some degree of epitaxial alignment, but still exhibit visible grain

36

boundaries. Thus, HOD can be viewed as epitaxial growth with significantly relaxed
requirements of homogeneity.

This chapter presents the progress that has been made towards diamond
heteroepitaxy on various substrates. To date, the following non-diamond substrates have
been reported for diamond heteroepitaxial growth: Si, B-SiC, Cu, Pt(46), Ni, Mo (and
other carbide forming refractory metals), Si02, Co, Pd(46), Au(46), c-BN, sapphire,
graphite, BeO, and Ir (4 7, 48). Other than local epitaxy (on a nm scale), reports have

been published of diamond epitaxy on Si, B-SiC, c-BN, Pt, Ni, and Ir. It is useful to

 

divide the substrates into two categories: those where diamond is grown on a bulk crystal,
namely, Si, B-SiC, c-BN, Pt, and Ni, and those where diamond is grown on a epitaxially
grown thin film deposited on another substratele-SiC, Pt, and Ir. Tables 2 and 3

summarize the reports.

37

 

.mpaon 0-0 mam use New «a can 1 define 8365:8va “5:38p mm _-Eo 03; an die; a.
ease 2 _ a as so: anaemia as 3833 853358 a
.83.. cacao—om SEEM 05 hope 9 g @0326 58am acute—cm .

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

sea 3an
E: V £8: a am .592 anew
«we .3 __._aA: Vase: com a: 8 geese so: a
£5 Rambo
.5: V a: a s a as S a: Name anew
ea c .4 =52 Va: _ 3 no: a: m .08 2 _ c a
:5 we; as?
so: _-ea m <2 a: Name 2 _ c _z
5 .3 mo? .haa em: unease a: on dog “so: 2
395
$3 swam
zeane V 2883 51 832 8: ~82 so:
«5 =52 Vase: do: a: N; 85385 E8
Rambo
:2 seem
. _-aa ad Zaire V as: _ c an? a: 82x: a: some a: _ c
an a: :8 Dane 32 ___nA: Va: a : anew a: No-3 .aeaazeafi zma
A -33 953.6 953.5 8.2 5.25::
3:0 .3239 5232.: 3.59 35.3323 9:55. £8535 an
353— 553.2. Z 5E5— qu—oem 353:5 Ea..— caefiaa esteem 89:35

 

 

86593 Rambo £3 .8 3:25.630 tongue 98:86 me bean—am m 033.

38

 

8.3a 20:33-5: 35 +
.33 28:56 2: mo 538%.: o: a 82:. .ofiw 886%; 05 fl NEE vxv 53% £ $05.25 0: ”wind .5“ =3on 33 SE 2:. .

 

 

 

 

 

 

 

 

 

 

83
so .5: V583: 330 a as... .55va .9583
.m .t 92 .de 5A: Vase: 235 a: w 888A .08 __._so:
:3
. .KS V583: .883 8 83 8.5335 2.283
a» ”2.3.x 28 22 .25 5A: Vase: mom . ... 8.5% .08 __._8o:
:2
18M: .5: V583: as? .883 5.5%? assoc
a i .2 was 3: 5A: Vasc: swam a: 34 5.5% .08 =28:
.39 x:
Neeoie 8%-... a: 83 Mom...:ooe
o: no: 512 8: Q3 .08 :2 _ 9.32
x: if 82 v
7.5 2 .5 fl 3 ”852% a: 82 6.2288:
au .18 3H: cod :53: com a: S a: to .08 .5: Vs: _ c
as: m
N853 a: 83 95: _ s
at .3. com a: m .08 :5: _ s
33.5 $2
:2: swam .883 €53 8: Red 258
an .3 .933 02-85 - mom a: on -omm €53-25 __u_£§:
9535 SEE 3:. 5.25.:
85 37.5: 52.32.: 3.50 2:28:29. 235: £85.25 «85.35 .3
Lone“ EGCago—uz Ewan“ ”flaw—06m Emuaznm E—uh QEOESA— azh och—505.5% Ouahuwnflm

 

 

 

 

 

 

 

 

 

 

 

8:338 338 295 BE 56 no .0833 .3 baa—ham v 038.

39

Where multiple reports exist, the results are either combined (for work within the same
group), or the best example is chosen. Examples with thinner diamond films showing
epitaxy are given preference as being more representative. Blank fields in the above
tables indicate the value was not given or could not be deduced from the report. The

following sections explain the data in the tables in more detail.

4.2 Diamond epitaxy on bulk substrates

4.2.1 Cubic boron nitride

Cubic boron nitride was an early choice for diamond heteroepitaxy because it has
a similar lattice constant of 0.3612 nm (mismatch 1.5%), a similar structure (zinc blende),
and a similarly high surface energy (4.8 J/mz for the (111) surface). Diamond has been
grown heteroepitaxially on c-BN, with a grth process similar to diamond homoepitaxy
(49, 61, 62). Reflection high-energy electron diffraction (RHEED) showed twinned
epitaxial diamond grew on {111}c-BN with the epitaxial relation: (111)D,<110>DiH(111)c-
3N <110>c-BN using DC-plasma CVD with 0.5% CH4 in H2 at 180 Torr, and substrate
temperature of 900° C. A continuous film formed at 0.2 mm thickness (49). Polarized
Raman scattering obeyed selection rules for a 0.1-um film, but with a C-C peak at 1325
cm‘l, having a linewidth of 12.9 cm'1 that was attributed to tensile stress in the film. This
study utilized substrate areas of only 100-200 um. Although important as an early
demonstration of the feasibility of diamond heteroepitaxy, the limited substrate size of c-
BN, itself difficult to grow, means it has limited applicability. The largest c-BN single

crystals available are only 1-3 mm, and they often have many impurities (63, 64).

40

4.2.2 Nickel

Diamond (111) and (100) was reported to grow on nickel (100) and (111)
epitaxially (65). Ni (FCC) has a lattice parameter close to diamond, 0.352 nm (1.3%
mismatch). 49 hr of growth at 900-950° C in a 0.5% CH4 plasma in a HFCVD system
produced a coalesced, oriented diamond film 30 pm thick. Raman peaks appeared at
1334 cm'1 for (100) and (111) diamond on (100) and (111) Ni, with linewidths of 5 cm'1
and 8 cm’1 respectively. The processing for seeding the Ni substrates increased the
density and grth rate of the epitaxial particles, leading to some coalescence. But
roughening due to the high carbon solubility of nickel served to limit the final film
quality. The work was done on a polycrystalline Ni substrate with no mention of the
grain sizes, nor the total area of growth. Diamond crystals were reported as 3 um after 7

hrs of growth, with a density of 108 cm'z.

4.2.3 Platinum

Platinum (FCC) has a lattice parameter 0.392 nm (9% mismatch with diamond).
Diamond (111) deposited on Pt(111)bu1k substrates has been demonstrated on a 12 mm
diameter circular region by Tachibana et al (54, 66). They used microwave plasma CVD
in a quartz-tube reactor to grow diamond with 0.2-0.3% CH4 in a balance of H2 at 50-60
Torr at substrate temperatures of 850°~880° C. These conditions yielded a grth rate of
approximately 0.3 um/hr on (111) Pt. A 3 pm thick film had an X-ray Pt{ 1 1 1) peak with
linewidth of approximately 4°. The Raman spectrum showed a small peak at 1330 cm’1
with a dominant peak at 1440 cm'1 (disordered carbon). SEM showed a <111> textured

film. According to the above classification scheme, this film would be HOD.

41

 

4.2.4 Silicon

The term highly oriented diamond (HOD) was first used to describe
heteroepitaxial growth of diamond on silicon. Si (diamond-structure) has a lattice
parameter of 0.543 nm (52% mismatch). Despite this, oriented crystallites on a (100) Si
substrate have been grown using microwave plasma CVD (2, 3). X-ray pole figures
showed a (001)dimnd <110>diamond || (001)Si <110>Si relation, although, grain boundaries
are apparent between the crystals from SEM. No quantitative analysis was done to show
the degree of orientation. Raman spectroscopy revealed a diamond peak at 1335 cm",
with a broad weaker peak at 1550 cm". All of the techniques for highly oriented diamond
growth involve a multi-step bias and growth process, to exploit crystal texture evolution.
The best film on Si reported to date has a linewidth of the (004) rocking curve of 2. 10 for
a 2-in diameter 16 pm thick boron-doped film (55). Raman spectroscopy was not

reported for this film.

4.3 Diamond epitaxy on thin films

4.3.1 Pt{111}/sapphire{0001} and Pt{lll}/SrTiO3{111}

Diamond grown on Pt{111}/ SrTi03 {l l 1} gave similar Raman and X-ray results
to that grown on single crystals (67). Pt{ 1 1 1} thin films deposited epitaxially on
sapphire{0001} exhibited better structural quality than the single crystal Pt or Pt
deposited on SrTiO3 (59). Growth conditions were the same as for the bulk Pt substrates.
Higher nucleation densities were observed (>1x108 cm'z) for the same type of seeding
process as on bulk substrates. A 1.5 um thick diamond sample had a (111) x-ray rocking

curve with a linewidth of 20°. SEM showed a <111>-textured film with grain sizes of

42

 

30-50 pm. At 9 pm thickness, the film coalesced, with some remnant grain boundaries,
and (111) x-ray rocking curve linewidths of 0.4-1.0°. The Raman spectrum, with 488 nm
excitation, showed a peak at 1334 cm'1 with linewidth 15 cm], and a broad peak around
1550 cm’1 (spz-bonded carbon). The (111) x-ray rocking curve linewidth of a 1.5 pm
thick film decreased to l.l° when an interlayer of lum Ir(111) was deposited between
two 1-2 pm Pt(l 1 1) layers on sapphire(1000) (54). The film covered an area of

lelOmmz. A SEM micrograph is shown in Figure 15.

 

Figure 15 SEM micrograph of a 1.5um thick diamond film on
(1 1 1)Pt/Ir/Pt/(0001)sapphire substrate (54).

43.2 B-SiC/Si

B-SiC (zinc blende structure) with ac = 0.436 nm has a 22% mismatch to diamond.
The process for growing epitaxial diamond on SiC began by first depositing 320-500 nm
of epitaxial fi-SiC (001) on Si(OOl) substrates (68). Then a four-step process was

followed to grow diamond using MWPCVD. Particle density was reported as 0.5-1x10H

43

cm'2 afier nucleation. After the initial growth step, reflection high-energy electron
diffraction (RHEED) showed a dispersion of approximately 2°. A final growth step
yielded smooth films with some visible grain boundaries at a density of 104 cm’2 in the

best 5x5 mm2 area. Steps appeared on a 20 pm thick sample (grown for 12 hrs) with step
height 30-50 nm in <110> and (1 l0) directions. A 300 pm thick (240 hrs) sample did

not show such steps, but had a surface roughness of 100-150 nm over a 50x50 um2 area
(57). For a 20 um film, the x-ray linewidth was 1.5° representing the angular distribution
of the crystals, with an intensity due to first-order twins less than 1% of the other (111),
(220), (400) reflections. The FWHM of the x-ray rocking curve (to—scan) of the diamond
(004) was 0.92° in the best region. The 300 um film showed an improvement to 062°,
over an area larger than 1 m2. Improvement in the x-ray rocking-curve linewidth of the
thick diamond is due to the high nucleation density of oriented particles, and the effective

use of a process to select particles with the correct orientation during growth (5 7).

4.3.3 Iridium

Iridium, first proposed as a substrate for diamond heteroepitaxy in 1996, (5) is a
FCC metal with a lattice mismatch of 7 .5% to diamond. Iridium has one of the lowest
carbon solubilities (by weight) of the platinum metal group (69). It is stable with no
known phase transitions. Iridium can be grown epitaxially on other single crystal
substrates and used as a substrate for diamond growth, as now demonstrated by several
groups (4-6, 8, 70). A comparison of iridium preparation and diamond growth methods

reported by these groups is presented in Tables 5 and 6.

 

 

 

 

 

 

 

 

 

 

59:2 982 96.5 96m: 352: SC
Emonkm 539m mucosaam ”a matosam "a .Emopfi vofioE coummoaofl
EE ofixofilmxm EE 3:: Evav 88 0:2: “Emu—98%
o oomw-oo8 u go: o coca 28:8 083 u 032% 238862 82889
552.3 888-8 -- 52523 $2225 :
mcan...032.523 .moEm moEm 60:23 em: 8283 cm: 3:29: 8am
89:33”

«at at 3 Q 5 S «Q .R .213;

._a we Emhwfifiom an «o v—oohflom A“ no “flow: A“ «0 0:93am

 

 

 

 

 

 

.589? 530% vac—5% can a you 53989:.“ oifiam m oEmH

45

Gt ESE? 058 2: 5 505m :0 x83 05 E0; .

 

BEES»; 32:5 u .md

 

 

 

 

 

 

 

 

 

#80 :ofixoLm ~-Eo OS ~.Eo ”3684A ~-Eo meta—é _.o $3.33 23‘3332
mod n w .md 03:80:55. md u w .md -- @052: 2682382
U ammo U coon U coon U 094w £380
0 OSHA: n O coon O ooow U coca mflm
usSEmEEmH

88353 no 8335808
Q8 02 concave cowxxo :EEmmE 02 -- 68:8 18585
082? owhfi : =~Em :25 BE 889$

 

“coca—c mmE

 

 

 

 

 

 

 

 

 

 

 

 

 

RE: EE mm Amara . ES om 8:685 88 n I N 8635 6:: Eofi oEEmm 8 8535
Sh :3 5.330 89820 mmoifim BB 5530 59520 $22me Eofiocccg mmO
bBEeuw

:3 ”2: cos 2N top 9. Se com £305
:8. M: toe 2N toe on toe 2: mam
33.32%

as 8-8 as 8-2 as 8-2 56 2a 2:: mam
<6 3.8 LS: S -- N525 2 “guano

> cm“. > OR. > com. 2 CE. 08 >33 >03- own? :5
9% a 88 B 82 B 8: B 828 .<.z 538 32
.139 332.com .3 «o 38.5% ._a “a .6an .3 3 2.55m 9:52:93.”

 

 

 

 

 

 

 

 

n no axSEoEBon 225% meat—5m maze» e8 E83523 5380 w 03¢

46

Diamond growth on Ir was first reported by Ohtsuka et al. in 1996. They grew
epitaxial (100) Ir films by RF magnetron sputtering on cleaved (001) MgO surfaces (5 ).
Subsequent diamond deposition used dc-plasma CVD with a 5-10 minute bias at —150V
at 900° C, followed by 30 minutes of grth at 840° C. Initial studies revealed only
“patchy” dense areas of nucleation that coalesced to form single crystals. After 30
minutes of growth, pyramidal structures with distinct {111} facets appeared with density
of order 108 cm'z. Single crystal films about 1 mm2 with mean thickness of 1.5 pm
resulted (4). Cross-sectional analysis by SEM indicated residual grain boundaries. The
Raman linewidth at 1332 cm'1 was 18 cm’1 and the film had a average roughness of order
1 nm. This group has since demonstrated an 8 pm thick freestanding film of size 2x2 mm
that cleaved along the {111} planes (71). Depth profiles of the Raman spectra were
reported, but no quantitative Raman or x-ray linewidth measurements. RHEED showed a
streaky spot pattern along <110>. The epitaxial relation reported was (100)diamond
<001>diamond ”(100);r <001>1,. They observed a non-diamond carbon signal in the Raman
measurement within 1000 nm of the back surface.

Schreck et al. (8) have grown single crystal Ir films on SrTiO; substrates by
electron beam evaporation. Using microwave plasma CVD, they observed epitaxial
diamond nucleation with density 109 cm'z. Unlike the others, they added 30-50 ppm of
N2 to the gas mixture for the nucleation and growth. For an 8 urn film, they measured a
(004) rocking curve and a (311) azimuthal scan with linewidths of 034° and 065°
respectively. Raman spectra from epitaxial areas of a 600 nm film (shown in Chapter 9,
Figure 73 (2a)) had a 15 cm'1 linewidth with an upward shift of 1340 cm", which they

attributed to a biaxial stress in the (001) plane (60).

47

4.4 Comments and perspectives

Although heteroepitaxial grth of diamond has been the goal for many research
groups, few efforts have satisfied the criteria outlined in the introduction to this chapter.
Perhaps the best example is thick diamond on B-SiC; but no chemical analysis has been
reported yet. Pt does not appear to be very promising as grain boundaries persist for
large thickness on {111} surfaces. Ir seems to have the greatest potential with several
impressive results.

In the present research, we have grown iridium on SrTiO3, MgO, LaAlO3, and
A1203 substrates. We have shown that single crystal diamond is indeed possible on all of
these substrates. With nucleation densities two orders of magnitude higher than prior
reports, we have produced completely coalesced diamond films of thickness below 1 pm.
This thesis describes diamond grth on Ir surfaces that satisfy all the criteria for true

heteroepitaxy.

48

Chapter 5 The bias process

Negative biasing of the substrate in the presence of a hydrogen-hydrocarbon
plasma is an important method to achieve high nucleation densities. This chapter
describes the method and discusses some of the mechanisms that have been proposed to
explain the enhanced nucleation densities. Although of general applicability, biasing is
discussed in the context of experiments of diamond growth on silicon, the most carefully
studied system. A few remarks on the role biasing plays in nucleation on Ir are also
given. Since negative biasing of the substrate leads to low-energy ion bombardment, it is
reasonable to suspect implantation (sometimes subplantation) as the cause of the
nucleation. The appearance of a cathode sheath or secondary plasma during biasing,
however, suggests enhanced electron emission is a component of the process. We discuss

the secondary glow observed during biasing and its role on nucleation events and epitaxy.

5.1 Seeding for diamond growth

Inducing a sufficiently high nucleation density is a fundamental problem that
must be overcome in growing diamond. Diamond’s high surface energy does not allow
it to spontaneously nucleate on foreign substrates. Thus, to promote diamond growth,
non-diamond substrates must be “seeded”. Early experiments employed a scratch-
seeding method. The substrate is polished or ultrasonicated with a small grit 10-25 pm

diamond powder. This may introduce defects or implant small diamond particles in the

49

substrate allowing diamond to nucleate (77, 78). This may not promote epitaxy, as
defects or roughness in the substrate surface can propagate into the growing film, or the

nuclei will not have epitaxial registry with the substrate.

5.2 Biasing

Bias enhanced nucleation (BEN) is a process that accelerates low energy ions
towards a substrate in the presence of a dc voltage drop across the plasma. The term ion-
enhanced nucleation has also been used. In its first embodiment (9), an increased
nucleation density on Si to 10'0 cm'2 was seen when a negative bias was applied to the
substrate with respect to the plasma. Using MPCVD, bias voltages between +100 to -200
V were applied to Si (100) with resistivity of several 9 -cm. Biasing for 2-15 min
followed by 30 min of unbiased growth revealed increased nucleation densities for
voltages lower than -70 V. No oriented nuclei were observed, but the demonstration of
an alternative to scratch seeding spurred exploration of heteroepitaxial growth. Using
BEN methods, oriented nucleation of diamond on silicon carbide was next demonstrated
(79), followed shortly by heteroepitaxy attempts of diamond on silicon and other
substrates (1, 2, 80). Why this method increases nucleation density on Si, and how it

leads to epitaxy are questions under current debate.

5.3 Models and mechanisms for diamond nucleation due to BEN

Ion bombardment and electron emission have been proposed as mechanisms for
the increased nucleation densities due to BEN processing. As a negative bias to the
substrate has been shown to increase nucleation densities (9, 81), it seems logical to

assume the ion bombardment from dissociated positive ions in the plasma is responsible.

50

However, the indication that a positive bias is also effective at increasing nucleation (82,
83), that increased densities are also found during biasing with HFCVD, which does not
produce large amounts of positive ions (28, 84), and that nucleation densities can be
increased by placing diamond surfaces in the vicinity of the substrate during biasing (85,
86) led many researchers to suggest electron emission may play a prominent role in

biased enhanced nucleation.

5.3.1 Ion bombardment

Based on the experiments described in Section 5.2, Yugo et al. (9) first proposed
ion bombardment as a nucleation mechanism, claiming the acceleration of positive ions
increased the “active” species near the substrate by collisions, thereby enhancing the
diamond nucleation rate. The ion bombardment may also preferentially etch amorphous
or graphitic carbon. It was inferred that the ion bombardment mechanism is a delicate
balance between formation of sp3 bonds and etching of sp2 bonds. The following studies
clearly indicated ion bombardment plays an active role in the nucleation enhancement.

Schreck et al. (76) studied the anisotropy of ion bombardment by using
lithography to pattern a silicon wafer with evenly spaced cylindrical columns of 5 pm
diameter and 8.5 pm height. The sample was then biased with a BEN process (—200 V at
10-20 mA) that was terminated at the first sign of current increase (afier 5-20 min).
Subsequent two-hour growth showed the tops of the columns and the intervening surfaces
in between covered with a dense diamond film; the sidewalls were almost bare. The
conclusion was that anisotropy arising from the nearly ballistic ion flux explained the

spatially selective nucleation. Other groups have done similar studies with lithographic

51

patterned or etched silicon which show nucleation is increased (108-1010 cm'z) on top
surfaces, and negligible at the sidewalls (8 7, 88).

In a more recent work, similarities were noted between bias-enhanced nucleation
on silicon and nucleation on a substrate exposed to a low energy carbon ion-beam (89)
(90). Liao et al. (89) used a mass-separated ion beam deposition technique to implant Si
with 100 eV carbon ions, with reference to suggestions by Robertson et al. (91, 92) . The
substrate temperature was maintained at 800° C, and the ion dose was 2x1018 cm‘z. Upon
examination of the Si using XPS (x-ray photoelectron spectroscopy), AES (Auger
Electron Spectroscopy), Raman spectroscopy, X-ray diffraction and AF M, no Si peak
was found using XPS and the surface was firlly covered by carbon. AES showed a film
that was mostly amorphous carbon. The Raman spectrum showed peaks at 1590 cm'1
(graphite) and 1352 cm", usually attributed to microcrystalline graphite, and a small
shoulder around 1173 cm”, a signature for nanocrystalline diamond. XRD showed
characteristics of diamond and graphite grains embedded in an amorphous carbon matrix.
With AF M, they observed oriented square 80-100 nm crystallites, believed to be (0002)
graphite surrounded by amorphous carbon. Growth of these samples in a HF CVD reactor
under conditions for microcrystalline diamond resulted in diamond nanocrystalline films,
with a nucleation density comparable those produced by BEN (93).

The suggested mechanisms by which ion-bombardment leads to formation of
nucleation sites are: (1) ion subplantation, (2) growth of a carbon non-diamond layer, (3)
increased surface diffusion and (4) growth of a SiC buffer layer. These mechanisms are

discussed below.

52

5.3.1.1 Ion bombardment leads to ion subplantation

Lifshitz et al. (94, 95) proposed ion subplantation, as the main result of ion
bombardment during BEN, based on experimental evidence and molecular dynamic
simulations. Ion subplantation is the deposition of hyperthermal (1-103 eV) species via a
shallow subsurface implantation (94). To understand the role of subplantation, Robertson
et al. (91) measured the ion energy distribution (IED) using a retarding field probe
inserted in a hole in the substrate for samples biased between ——1 50 and —300V. The IED
was sharply peaked at 80 eV at the optimal conditions of 2% methane, -250V. In other
studies (96, 97) it was found that carbon ion subplantation in Si is most efficient around
100 eV. As these energies were in agreement with the Lifshitz model (94), it was
concluded that ion subplantation is the dominant mechanism of nucleation in Si, creating
“diamond-like carbon” with nanocrystalline diamond features in an ordered graphitic
deposit. Other molecular dynamics studies have also suggested an energy window
between 40-70 eV for the deposited atoms (98). A few studies have looked at the ion-
energy distribution in the plasma to determine the main species bombarding the substrate.
Katai et al. (99, 100), using ion beam mass spectrometry, found that the hydrocarbon
radicals C+, CH+, CH2”: C2H2+, and C2H3+ with average energies 50-70 eV are the
dominant species at the onset of bias. As the current increases (with the appearance of a
secondary plasma over the sample) CzHg,+ species (E=95-105 eV) and H“ and Hf
(E=160-180 eV) become dominant. Kim et al. (101) likewise investigated the plasma
Species during bias using optical emission spectroscopy (OES). Increased levels of
atomic hydrogen and hydrocarbon radicals, particularly CH+ appeared for bias voltages

between —100V and -250V. They proposed that subplantation of carbon species leads to

53

epitaxial formation of SiC and further bias leads to etching of Si atoms with replacement

by carbon atoms and forming sp3 bonded structures (102).

5.3.1.2 Ion bombardment causes growth of a non-diamond carbon layer

McGinnis et al. (103) believed a subplantation model was not necessary. The
average energy of carbon ions at the substrate was calculated to be around 15 eV, which
would correspond to a surface process rather than subplantation (95, 104). Using SEM,
Raman microscopy, cross-sectional TEM, and x-ray diffraction, they observed evidence
of nanocrystalline diamond in a non-diamond carbon matrix. A nucleation density of
1011 cm'2 for a sample biased for 1 hr at -250 to -300 V was reported along with the
suggestion that a threshold voltage must be exceeded for the ion bombardment
mechanism to be effective ([05). In two sets of experiments, bias nucleation density
enhancements were related directly to ion bombardment. For the first, a series of
experiments were run under identical process treatments. They made nine separate runs
applying a bias from 0 to —333 V in 2% CH4. Bias duration lasted from 12 min (333V)
up to 300 min (0V). The application time for the bias was calculated to keep the time-
integrated bias current approximately equal for all samples. Higher bias voltage caused a
sharper increase in bias current, which meant a shorter bias time was necessary for the
total current flux to the substrate to remain constant. The plot of nucleation density with

applied bias is shown in Figure 16.

54

 

109

10’

10’

 

Nucleation Density (cm'z)

 

103 44.1.i.rm|.1.
050100150200250300350

Negative DC Bias Voltage (-V)

Figure 16 Plot of nucleation density for negative bias voltages keeping the time-
integrated current constant for each data point (105).

The nucleation density increased sharply above —100 V, and saturated above —210 V at a

value that is five orders of magnitude higher than for 0 V bias. Low nucleation densities
correlated with small increases in current, and this indicated the existence of a critical
voltage for the onset of nucleation. Below the critical voltage, the impinging species are
not energetic enough to overcome the nucleation barrier.

Sheldon et al. (106) used Electron Energy Loss Spectroscopy (EELS) and TEM to
determine an amorphous carbon layer is deposited on the Si substrate during a —200 V
BEN procedure. The deposit thickness, approximately 4 nm, was independent of the
substrate temperature; suggesting the carbon was deposited through a non-chemical
surface process. Ion energies of 2-20 eV were estimated. In other studies of CBS during

biasing at -150V, it was shown that atomic hydrogen increases near the substrate by more

than 20% (107). Atomic hydrogen is known to etch sp2 carbon faster than sp3 carbon

55

(108, 109), and it also has been found to stabilize sp3 clusters (1 10). The increase in
atomic hydrogen directly above the substrate may have enhanced the carbon layer that
formed due to hydrocarbon ion collisions. By stabilizing sp3 clusters, it may also serve to

increase the amount of amorphous carbon on the substrate that later nucleates diamond.

5.3.1.3 Ion bombardment increases surface diflusion

Jiang et a1. (88, 111) studied the early growth stages of diamond on Si using SEM
and AFM data, and postulated that surface diffusion of the carbon species on the
substrate is increased due to ion-bombardment. F irst-nearest-neighbor distributions of
particles imaged by AFM, for 10 minute and 12 minute bias times at a —150 V bias,
showed distributions shifted to larger distances compared to a random (Poisson)
distribution. They claimed a depletion zone develops around the nuclei and small just-
forrning islands. Therefore, the growth process induced by ion-bombardment of the
substrate can be completely explained by thermodynamics. On the other hand, Ascerelli
(112) concluded the depletion could be due to a deformation zone induced by the
diamond-Si lattice mismatch. J iang et al. however, reported that because the depletion
zone is a function of bias voltage, ion-bombardment must influence the diffusion of the
surface species (88). Other studies have been done to show atomic diffusion is enhanced

by ion bombardment during diamond BEN (113).

5.3.1.4 Ion bombardment causes growth of a silicon-carbide bufler layer

Stoner et al. (114) studied the nucleation and growth of diamond on Si using
XPS, AES, EELS (electron energy loss spectroscopy) and TEM. It was suggested that

ion-bombardment enabled the formation of an amorphous-SiC layer (>9 nm) during the

56

first hour of biasing at —250 V in a MWCVD reactor. Further biasing developed an
excess carbon concentration at the surface, which with continued ion bombardment could
have become mobile on the surface and nucleated diamond. Other studies have shown
that BEN on Si led to the formation of an epitaxial SiC layer, which nucleated oriented
diamond crystallites (1 15-1 18).

In a detailed study of bias nucleation, Stockel et al. used MPCVD to investigate
the ion-bombardment nucleation mechanism for diamond grth on Si (118, 119). Their
bias setup is nonstandard, as the (100) Si substrate is grounded, and a +250 V bias is
applied to a tungsten electrode in the plasma (120). By examining SEM and TEM images
of the carbon deposits, the following 4-step model for nucleation was proposed: (1) In the
first minutes of bias, nanocrystalline B-SiC forms on the Si substrate with a near epitaxial
relationship. Simultaneously, the Si surface is etched by the H+ flux in the plasma. (2)
The B-SiC forms a closed continuous film after about 10 min at 10 nm thickness. This
film is polycrystalline, but with an epitaxial relationship to the substrate. Excess carbon
flux leads to a supersaturation of carbon in the SiC. Diamond nuclei form in the carbon
layer, some with registry to the SiC. (3) Further biasing etches the B-SiC layer and
exposes the diamond nuclei. (4) Growth conditions allow the diamond nuclei to grow —

those with an epitaxial relationship grow as an oriented film.

5.3.2 Electron emission

Ion bombardment has been shown repeatedly to lead to enhanced diamond
nucleation. However, the following experiments indicate it may not be the only
mechanism at work, and some nucleation enhancement may be due to electron emission

of diamond.

57

Stoner et al (121) studied the effects of a diamond coated substrate holder, an
uncoated molybdenum substrate holder, and an alumina substrate holder during BEN.
For the diamond coated holder they obtained high nucleation densities and a 5-fold
increase in current during bias over the other two substrate holders. In addition, it was
observed that: (1) the diamond coating on the holder surface is etched or removed during
extended biasing (they don’t indicate how long) leading to a decrease in bias current, (2)
the plasma turned red near the surface of the sample and holder at increased voltages,
indicating increased atomic hydrogen (H°), and (3) nucleation began at the substrate edge
and moved towards the center. The enhanced nucleation may be due to (1) a material
transport mechanism, whereby diamond coated surfaces in the vicinity of the sample are
etched and carbon is transported to the substrate and deposited, or (2) electron emission
of the diamond surfaces causes increased levels of dissociated H2 and/or hydrocarbons
near the substrate. They noted that BEN experiments in a pure hydrogen plasma under
the same conditions showed no increased nucleation, so the second explanation was more
likely.

Wang et al. studied the mechanism for increased nucleation due to biasing using
HFCVD (122-124). The nucleation density was increased by one of two ways: applying
a negative bias to the substrate, or to a circular diamond-coated tungsten ring 3-5mm
above the substrate. In the second case enhanced nucleation arose from increased
electron emission from the diamond-coated tungsten filament. The substrate is positive
relative to the electrode in this case so the increased nucleation could not be due to
increased positive ion bombardment. Chen et a1. (85) also found that it is unnecessary for

the substrate to be negatively biased in HFCVD. Nucleation densities were increased if a

58

diamo

while

BEN 1
Fora

the Si

that n
the di

111C181

here.

5.4

act‘s.
prev
glO‘tt
duh;
subs:
H01

thee

diamond coated filament was placed in the vicinity of the substrate and biased negatively
while the substrate was held at a positive potential.

Perng et al. (86) studied the role of electron emission in biased MWPCVD. A
BEN process was applied to micropattemed SiO2/Si substrates and sapphire substrates.
For a pattern of silicon dots 5 and 8 um in diameter, nucleation densities of ~109 cm'2 for
the Si dots, and ~108 cm’2 for the SiO2 masked areas were measured (via SEM).
Nucleation (~108 cm'z) and grth of diamond were also observed on the edges of a 1x1
cm2 insulating sapphire substrate placed between four 1x1 cm2 silicon substrates.
Without the silicon substrates, the sapphire was only etched during bias. They proposed
that nucleation is enhanced on insulating SiO2 and sapphire due to electron emission from
the diamond that was already formed on the Si in the vicinity. Electron emission
increased the concentration of reactive species above the SiO2 and sapphire, resulting in

increased nucleation densities.

5.4 The secondary plasma

The deposition of high nucleation densities of diamond has often been reported
accompanied by a bright secondary discharge during biasing. All the reports of the
previous section, 5.3.2, discuss a bright blue, purple or red glow (85, 86, 121-123). The
glow always appeared either above the sample or the surrounding area of the holder
during biasing, and the glow is related to diamond coatings on the substrate and/or
substrate holder. Stoner et al. (121) pointed out that repeatability of bias-enhanced and
HOD diamond results was a problem in early experiments, possibly due to differences in

the coatings, and that it could be etched during biasing.

59

Biasing of the substrate affects the charge distribution in the plasma. As was
described in section 3.4. 1, mobile charged species in the plasma screen the discharge
from an externally applied field forming a “sheath” region between the main discharge
and the substrate. This will modify the conditions near the substrate and thus the

nucleation process.

Schreck et al. (76, 125) provided evidence that biasing produces a plasma state
that is a superposition of a microwave plasma and a DC glow discharge. They measured

the intensity profile for the Ho, Balmer emission 1ine° during a bias of —200 V above a

substrate (Figure 17).

 

 

1531?; ' diamond:
50 l :

£5531“ Silicon (x20)
’0? T5114; .'
3:40 12,511.11 ;
g 13.th I
9 1;} ; ; Plasma
V20 r; : : ball
E Pt 3 5
(D10 1:12.255 : :
E 113171.131 ; , WITHOUT
,_ 1,1315 - : BIAS (x20)
2 0 "£115 - i
- 13:31; 50.8 51.96

-0.6 0 0.6 1.0 1.6 2.0 2.5 3.0 3.5 4.0
Distance from surface (mm)

Figure 17 Ha Balmer line (2t=656.3nm) intensity measured above a diamond covered
sample and a bare silicon sample biased at -200V, and a sample without an

applied bias. Schreck et al. (125).

.¥

6 Ha emission/absorption line at l=656.3 nm is the first atomic transition in the Balmer series. The Balmer
Series corresponds to atomic transitions that end in the first excited state of hydrogen (E=-3.4eV).

60

The intensity above an unbiased Si sample increased monotonically from the substrate
into the main plasma discharge. With an applied bias, a peak in intensity appeared 1.95
mm above the Si substrate. The peak gained in intensity (x20) and shifted closer (0.8
mm) when the Si was covered in a diamond film. The profile of emission intensity was
very similar to that near the negative electrode in a DC glow discharge (See Figure
l3(b)). The maxima in intensities develop at voltages much lower than for a normal glow
discharge because the microwave plasma contributes to the DC glow discharge. An I-V
curve for the bias voltage on a graphite holder, a silicon substrate and diamond film
covered Si substrate (Figure 18) was also reported. The first plateau was suggested to
corresponded to positive ion saturation with the subsequent steep increase due to electron

multiplication.

 

140

 

‘3‘

§

 

 

 

CURRENT ( mA )
8

 

 

 

 

 

o 50 too 150 200 250
VOLTAGE(V)

Figure 18 I-V curve for a diamond coated silicon substrate, a silicon substrate, and a
clean graphite sample holder. Schreck et al. ( 76).

61

It was proposed that the field in the secondary plasma (or the cathode sheath) accelerates
the ions and electrons. The higher ion saturation curve for the diamond film compared to
silicon or graphite means stronger ionization of the plasma ( 76). The strong electron
multiplication above diamond covered areas on the substrate may influence the plasma
chemistry (cf. Stoner (121)) and may contribute to the nucleation on uncovered areas.
Kulisch et al. (126) reported a similar I-V characteristic for a microwave plasma
CVD system (Figure 19). Others have suggested that the increase in current over the

diamond covered substrate can be explained by enhanced secondary emission from

diamond (114, 127, 128).

 

1

 

 

 

 

 

320 .. ' T ‘
280 — 0 silicon substrate: .
240 a diamond covered substrate
g 200
1:: 160 1 1
h I ' II : III /’I
5 120 : : ,3
o g . /
so 5 X ./
I I //
| 'I‘.’ I I
40 -flazfif--rrf
0 e‘__ r' 1 r l _ L- r
o 40 -80 -120 .150 -200 .240
bias voltage [V]

Figure 19 I-V plot for a bare silicon substrate and a diamond covered silicon substrate.
Kulisch et al. (126). Region I corresponds to a mobility limited current, region
11 to the saturation of positive ions at the surface, region 111 to current
enhancement by additional ionization.

Although the shape of the I-V curve agrees with others ( 76) —- the absolute measured

62

 

current and the time scale of I(t) (for a 30 minute bias period) were poorly reproducible.
Two plausible reasons were given for this:
1. Current is strongly influenced by diamond surfaces in the substrates vicinity.
This can vary due to relative sizes of the sample and holder, or the elevation of
the sample with respect to the holder.
2. The respective volumes of the microwave plasma and the secondary discharge are
easily influenced by small changes in the sample arrangement, geometry or

quality of the diamond coating. The secondary plasma may change in volume and

 

sample coverage during biasing. If the secondary plasma is the result of a normal
glow discharge, then current changes would be apparent in changes of effective
electrode area (including diamond coverage of the holder).

Other researchers (129-I31) have carefully studied the secondary plasma emission during

biasing and found similar emission spectra and I-V curves.

5.4.1.1 The secondary plasma and epitaxy

The cathode sheath or secondary discharge may play a role in the nucleation of
heteroepitaxial diamond. Sttickel et al. observed a spatially inhomogeneous nucleation
pattern, along with a secondary plasma sheath (118, 119). The secondary plasma was
initially localized over the substrate holder and advanced radially inwards over the
substrate. This correlates with the nucleation front on the substrate. Taking advantage of
the spatial inhomogeneity of the secondary plasma, they interrupted the bias process
when the secondary plasma had advanced 1/3 of the way from the edge to the center.

The sample was examined after nucleation, and after growing out half of the sample

further. The mechanism proposed for epitaxial diamond nucleation was discussed in

63

section 5.3.1. Only over a region 1/3 to 1/2 of the way radially inward did they see
oriented diamonds. It was concluded that the epitaxial growth occurs just ahead of the
secondary plasma front and that epitaxial nucleation requires critical timing of the bias
procedure.

Thiirer et al. (132) studied the HOD nucleation and growth on Si using XRD to
examine the linewidths of azimuthal scans of diamond{220} after growth as a function of
bias voltage and bias time. For each particular voltage, azimuthal alignment of the
diamond film was only possible within a certain bias time window. The final film
alignment was determined during the biasing step. As the secondary plasma formation is

a function of time, it is likely the two are related.

5.5 Biased enhanced nucleation on iridium

In contrast to Si, few studies have investigated the role biasing plays in the
nucleation of diamond on Ir. The reports that do exist suggest the role may be quite
different.

Ohtsuka et al. (5) conducted XPS on studies on biased Ir/MgO samples and found
that no Ir-C compounds form during biasing. Instead an amorphous carbon layer
deposits. It was suggested that the effect of biasing (or ion irradiation) is to roughen the
Ir surface and deposit a thin amorphous carbon layer.

Htirmann et al. (133, 134) studied the effects of biasing on Ir/SrTiO3 substrates
with SEM and HRTEM. No evidence for an amorphous carbon interlayer was found on
thin diamond film samples. It was pointed out that defect bands extended directly from
the iridium into the diamond, which would not happen with an amorphous layer in

between (134).

64

Sawabe et al. (135) conducted cross-sectional HRTEM on diamond/Ir/MgO
samples. Misfit dislocations and the diamond/Ir interface suggested to them a strong
chemical bond between the Ir and diamond in these areas. They found no evidence for an
interlayer, but significant roughening of the Ir surface during biasing. The roughened

regions of the Ir were proposed to act as nucleation sites for diamond.

5.6 Summary

Differences in CVD reactors and experimental repeatability make it difficult to
compare results among groups, or even experiments from the same researchers.
However, it is clear that despite the disparity, ion bombardment is a vital component for
BEN as shown by Robertson, Lifshitz, J iang and others. But the observations of Stoner,
Schreck and Kulisch, etc. leave little doubt that diamond covered surfaces affect the
behavior of the plasma and thus the nucleation in a way that is poorly understood.
Biasing effects on Ir substrates are less studied. The research in this work will show that
nucleation on Ir is also strongly affected by ion bombardment and the biasing induced

secondary plasma.

65

Chapter 6 Substrate preparation for diamond

heteroepitaxy

6.1 Chapter overview

This chapter details the preparation of substrates for diamond deposition. The
substrate is a critical component in heteroepitaxial growth. Highly ordered diamond
(HOD) can be grown on (001) Si substrates; (001) Ir is used as a substrate for grth of
single crystal diamond. The discussion of epitaxial Ir deposition comprises the bulk of
this chapter. Ir is not available as a high quality, single crystal substrate, so we first

deposit it epitaxially on another substrate. We have used (001) SI‘TiO3, (001) MgO,
(001) LaAlO3, and (1 120) A1203 as underlying substrates for single crystal (001) Ir
deposition using UHV electron beam deposition. A brief review of other Ir epitaxy
studies is given, and the structure and preparation of (001) SrTi03 and (1 120) A1203 are
discussed. In both cases, the goal was to obtain smooth, terraced surfaces for the

deposition of single crystal Ir. The epitaxy of Ir on (1 120) A1203 is not yet fully

understood due to the complex structure of the (l 120) surface, and the differing

geometries of the lattices. We suggest a possible epitaxial relationship of Ir on a-plane
A1203. The UHV electron-beam evaporation system used for all Ir depositions is

described and results of studies on the Ir morphology and crystalline quality evolution

66

with

diar

6.2

rm

with temperature and thickness are presented. The preparation of Si substrates for

diamond deposition is straightforward, and is only briefly discussed.

6.2 Substrate preparation for iridium deposition

Ir has a high melting point (2454 °C), and low vapor pressure (1.47 Pa at 2443°C).
It also has a high surface energy for a metal, with a theoretical value of 3.81 J/mz for the
(001) surface (136). Ir thin films have been prepared previously by metalorganic
chemical vapor deposition (MOCVD) (137-139), radio-frequency magnetron sputtering
(140), pulsed laser deposition (141, 142) and e-beam evaporation (143). Epitaxial Ir thin

films were first demonstrated on sapphire in 1994 using MOCVD (13 7) with an epitaxy

relation of(100)1,|| (1 150)A1203 with <011>,,|| (1100) AL203, for iridium deposited at 600°

C substrate temperature. For subsequent diamond growth, Ohtsuka et al. used magnetron
sputtering to deposit epitaxial (001) Ir films onto cleaved (001) MgO surfaces. (5).
Schreck et al. demonstrated epitaxial Ir deposited via e-beam evaporation onto (001)
SrTi03 ( 7, 8). In the course of our work over the last three years, we have deposited
single crystal (001) Ir onto polished (001) SrTi03, (001) MgO, (001) LaAlO3, and a-plane
A1203 using e-beam evaporation. Each of these surfaces requires preparation before
single crystal Ir can be grown. The major portion of the work described in this thesis is
on SrTiO3 and A1203, so discussion in this chapter is limited to these two. It should be

noted that preparation of MgO and LaAlO3 substrates is similar to SrTiO3.

67

6.2.1 Strontium titanate substrate preparation
SrTiO; is a cubic perovskite (space group Pm3m), with a melting point of 2080°
C, commonly used in oxide superconductor growth. The unit cell and lattice T102

terminated structure are pictured below.

 

Figure 21 SrTiO; structure with Ti02-terminated surface.

It has lattice parameter ao=0.39050 nm, a 1.7% mismatch to fcc Ir. The two structures are

similar, so Ir deposits epitaxially on SrTiO; with the relation (100)i,|[(100)smo3 and
[110]1,|| [110]smo3- lxlx0.5 cm3 substrates were coated with photoresist and cut into

0.5x0.5 cm2 pieces using a wire saw. They were inspected for defects under an optical

68

microscope after ultrasonic cleaning in acetone and methanol for 10 mintues each,
followed by a rinse in DI water. Figure 22 shows a typical surface after cleaning.

3.0 nm

   

Figure 22 SrTi03 surface AFM scan afier ultrasonic cleaning for 10 minutes each in
acetone and methanol, followed by DI water rinse.

Substrates that did not pass inspection due to pits, scratches or defects were returned to
the supplier. Terraced, atomically flat, TiO2-terminated surfaces were prepared by
standard methods (144, 145). They were first etched for 10-60 s in a buffered pH 4.5
hydrofluoric solution (NH4F2HF = 87.5: 12.5). Higher pH resulted in rough surfaces, a
lower one in etch pits. A two hour anneal in flowing 02 followed in a tube firmace at
950° C. They were then again cleaned ultrasonically in solvents before being loaded into

the UHV system for Ir depositon. A typical (001) SrTi03 surface is shown in Figure 23.

 

Figure 23 AFM image and 3D profile of typical SrTiO; surface afier surface
preparation. Mean roughness = 0.2 nm.

69

After annealing the surface was covered in terraces across the surface, with a mean
roughness of 0.2 nm. The mean step height of the surface in Figure 23 is 0.42 i 0.03 nm,
which is within error of the lattice parameter of SrTiO3, indicating these are single atomic

steps.

6.2.2 A-plane sapphire substrate preparation

Sapphire has a hexagonal crystal structure and belongs to the space group R3 c.
It has lattice constants a=0.476 nm, c=1.299 nm, and a melting point of 2040°C. The
crystal structure of a—Al203 is a hexagonal close-packed arrangement of 02' anions, with
two-thirds of the octahedral voids occupied by Al“ cations (146). The structure is stable
and standard processing leads to terraced, defect-free surfaces. We have used both small
offcut angle (<0.1°) with <1 nm roughness (1 120) A1203 substrates for Ir deposition and
more typically supplied substrates of this orientation with offcut angles of up to 05° and
rms roughness of 5-8 nm. The substrates were coated with photoresist, cut and cleaned
following the same procedure as strontium titanate. They were also inspected under the
optical microscope for defects or pits. The samples that passed inspection were annealed,

in air, at 1450° C for 15 hours. This leads to terraced surfaces, as shown in Figure 24.

 

Figure 24 Representative AF M images of two different a-plane A1203 substrates.

7O

The above figures show two different A1203 samples. The morphology of the a-plane
surface after annealing may depend on the offcut angle of the substrate and the direction
of the miscut, which is unknown in Figure 24. The roughness of both samples is <0.2nm.

Epitaxy of Ir on a-Al203 is not straightforward because of the dissimilar lattice
structures. The a-plane of A1203 is perpendicular to the c-plane <0001> direction. It is
the lengthwise, diagonal cross-section of the unit cell (Figure 25).

Unll Ce”
./

 

 

 

 

 

 

 

 

 

 

Figure 25 Schematic of sapphire structure, unit cell, and orientation of (1120) plane.

a=b=0.476 nm and c=1.299 nm. The c-axis is aligned along the <0001>
direction.

The unit cell of A1203 has lattice parameters c=1.299 nm, a=b=0.476 nm. The

arrangement of aluminum and oxygen atoms on the (1 120) plane is shown in Figure 26.

71

 

Figure 26 Surface of (1120) a-plane of A1203. Oxygen atoms are depicted as spheres,

aluminum atoms as rods. The unit cell is outlined in black. Oxygen atoms that
lie in the same plane, are marked with an X.

The projection of the unit cell is outlined in black in the above figure. We believe, that
similar to GaN/sapphire epitaxy, the 0 terminated surface plays a role in Ir epitaxy on the

(1 120) plane (14 7, 148). 0 atoms in the unit cell which lie in the same plane are marked

with an “X” in Figure 26. Figure 27 shows the suggested (001)1r epitaxy on the 0 atoms.

72

[1T01Ir| | [0001 ]Al,03

t, [1101M |[1Toow,o,

0.275 nm

H

0.271

 

Figure 27 Possible epitaxial relationship of (001) Ir on (1 120) A1203. The dotted box
indicates the unit cell in the (1 120) A1203 plane projection.

The distance between oxygen atoms for this surface is 0.433 nm along the <0001> and

0.275 run along the (1700) directions. When the (001) Ir plane is superimposed with

<11o>1,||(1'ioo>3.203, the lattices match 3:2 in the (170) ”<0001> directions with a 11%

mismatch, and 1.2% in the <110>||<1T00> directions.

6.3 E-beam iridium deposition

Ir single crystal films are deposited using ultra high vacuum (UHV) electron beam
evaporation (75). The main chamber base pressure is 5 x 10'10 Torr, pumped with an ion

pump and a titanium sublimation pump. The load lock is evacuated after sample loading

73

using a turbo pump backed by an oil-free diaphragm roughing pump to a pressure of 10'7
Torr. The system has two water-cooled electron guns in the main vacuum chamber. The
power supply can maintain a total current of 660 mA to the guns. 99.9% Ir slugs in Cu
crucibles were evaporated with a focused electron beam with current between 300-400
mA. A 1” diameter resistive heater mounted on the sample puck heats the sample to the
deposition temperature, as measured by a thermocouple mounted to the top of the sample

holder. A crystal thickness monitor measures the deposition rate.

6.3.] Procedure on strontium titanate and sapphire

The substrates are loaded into the UHV chamber of the e—beam evaporation
system through the load-lock immediately after solvent cleaning and drying in a nitrogen
gas flow. The substrate heater is turned on and in 30-40 minutes reaches deposition
temperature of 800° C. The shutter is opened, and the Ir deposited at a rate of 0.3-0.5 A/s
to a total thickness of 150-300 nm. A 150 nm Ir film takes about 1 hr to deposit. Upon

removal, the films are examined with AF M and X-ray diffraction.

6.3.2 Description of iridium growth mechanism on strontium titanate and

sapphire

Ir growth initiates on SrTi03 and A1203 in the 3-D island growth mode (Volmer-
Weber). The films coarsen as the thickness increases. Eventually a continuous film
forms, and 2-D layer-by-layer growth can proceed. We have demonstrated this with the

followings studies on the evolution of the crystalline structure and morphology of Ir.

74

6.3.3 Structure and morphology of Ir films on SrTi03

Epitaxial Ir is deposited on SrTi03 provided the substrate temperature is above
450° C. The morphology of the film is dependent on temperature and thickness. A 150

nm film evaporated at 800° C had the X-ray diffraction pattern shown in Figure 28.

 

 

 

 

 

it
e-
5 A :3“ §-
it. ,. gs
>'< E“ i, we
-m a
L. L i l .1 l i l
20 40 6O 80 100

29 - 9 (degrees)

Figure 28 X-ray 29-9 scan of Ir on SrTi03 using Cu Ka radiation. Film was 150 nm
thick, grown at substrate temperature of 800°C.

Although the (111) Ir reflection has a relative intensity about twice that of the (200), the
20-0 scan above does not show any evidence of(111) Ir, indicating the film consisted
entirely of (100) Ir. The linewidths of the (200) SrTi03 and (200) Ir reflections are
calculated by a Gaussian fit to the experimental data, and are 003° for SrTi03 and 019°

for Ir, as shown in Figure 29.

75

 

0 Experiment
Gaussian fit

 

_ Ir(200)
FWHNI: 0.190

.593.

   
     

SrTi03(200) ,,
FWHM: 0.03°

X-ray Intensity (Arbitrary Units)

 

 

 

 

9
.
.
' D
\
_

22.5 23.0 23.5 24.0 24.5
Theta (degree)

 

Figure 29 The Gaussian linewidth extracted from a least-squares fit to the data: 0.03°
for (200) SrTi03 and O.l9° for (200) II.

By comparison, Hermann et al. have reported a rocking curve linewidth ofO. 15° for the
(200) Ir peak for a 200 nm Ir film deposited at 950° C on SrTi03 (133). Presumably the
value is somewhat better due to a thicker film deposited at a higher temperature, and

differences in instrument resolution.

Morphology of the Ir was studied using AF M for different thicknesses of Ir. Ir
was deposited in the same manner for each run, at a substrate temperature of 800° C.

Representative AF M scans are shown in Figure 30.

76

 

9nm

 

Onm

 

A. 6 mm
B. 30 nm
C. 112 nm
D. 150 nm
E. 300 nm

Figure 30 1x1 um2 area AFM images of Ir on SrTi03

Small islands emerged initially then coarsened as the film thickness increases and
eventually coalesce. At 150 nm thickness, the film is completely coalesced, only some
pinholes remain on the surface. At 300 nm the pinholes are mostly filled in, and the film
appears to grow layer-by—layer, a 2-D growth mode. The pinholes that remain have been

determined by AF M not to extend through to the substrate. The rms roughness of the

77

above Ir films is plotted against the film thickness in Figure 31.

 

 

 

12 4.0

10_ —--RMS - 32
A —o—Rmsrrhickness '
é 8~
g; . 2.4
2 e
i °‘ 5
'5 a 16 (é)
§ 4‘ n:
E
a 3 0.8
2 2~
a:

0— - *0 T 0.0

 

 

 

.

I

I l l l l
0 50 100 150 200 250 300
Film Thickness (nm)

Figure 31 Ir roughness vs. substrate thickness measured from 1x1 pm AF M scans.
Although the roughness seems to increase with increasing film thickness for thickness
<150 nm, the normalized roughness falls off with increasing thickness, then stabilizes.
Errors are estimated at 110%.

X-ray diffraction was performed on these films as well. The linewidth of the

Ir(200) peak is plotted as a function of thickness in Figure 32.

78

 

 

 

 

 

g 1.2
Sb
:3 1.0
E
o 0.8
E”
0.6
33
§ 04
E O
Q...
o A
E 0.2 ' fi'
0.0 l l l g l l l
0 50 100 150 200 250 300
Film thickness (nm)

Figure 32 X-ray analysis of Ir (200) vs. film thickness.

The peak width narrows with increasing film thickness, in agreement with other results

(133). It reaches a value of 0. 19° for 150 nm, and then remains constant.

6.3.4 Temperature dependence of iridium morphology on SrTi03

The evolution of surface morphology and crystalline quality of Ir:SrTi03
substrates was also studied as a function of substrate deposition temperature. 150 nm Ir
films were deposited in different runs with substrate temperatures from 450° C to 800° C.
The morphology was studied using AF M, and the crystal structure using X-ray

diffraction analysis of the (200)1r peak. The AF M images are shown in Figure 33.

79

18.5nm

      

Figure 33 1x1 um AF M scans of Ir on SrTi03 deposited at different substrate
temperatures. All films are 150 nm thick. A. 450° C, B. 600° C, C. 700° C, D.
800° C
At 450° C, the film grows in 3-D mode, indicated by the discreteness of the features.
Films deposited at higher temperatures, 600° and 700° C, show monotonically increasing
island size. At 800°C a continuous film forms. The film also becomes smoother as

deposition temprature increases. A plot of average surface roughness with depositon

temprature is shown in Figure 34.

80

2.5

 

2.0 e

1.5 -

l.0 -

Film Roughness (Ra) (nm)

0.5

 

 

0.0 P l l l
400 500 600 700 800 900

 

Substrate Temperature ('0

Figure 34 Iridium roughness as a function of deposition temperature. All films are
150 nm thick. Measurements made with AF M.

Roughness decreases with increasing substrate temperature. It increases slightly at 800°C
again — due to the agregation of pinholes on the surface. The linewidth of the Ir (200)
peak decreases with increasing substrate temperature, as shown in Figure 35. The

crystalline perfection increased with increased deposition temperature.

f‘
N

 

1.0 -

0.8 -

0.6 -

0.4 -

0.2 -

FWHM of Ir(200) Rocking Curve (degrees)

 

 

0.0 l A l A 4 4 l L l

400 500 600 700 800

 

Substrate Temperature (°C)

Figure 35 X-ray rocking curve width of Ir (200) as a function of substrate temperature.

81

6.3.5 Morphology and structure of iridium films on (1120)A1203

A 300 nm Ir film deposited at 800° C on (1 120) A1203 with e-beam evaporation

had the x-ray diffraction rocking curves shown in Figure 36.

IL
I'lrfi-ITIIA'Irl'l-I'F'

l

r 11(200)

 

L

.l

I'll-Jill.

Al2 03 (1 1 20)

n.4J.Li . l . I In...”
18.6 18.8 19.0 19.2 19.4220 22.523.023.524.024.5 25.0

0 (deg)

Figure 36 X-ray diffraction rocking curves for 300 nm epitaxial Ir grown on (1120)
A1203. Linewidths: sapphire, 0.04° ; iridium, 0.21°.

X-Iay mafiau.)

 

 

 

 

 

 

The sapphire linewidth is 004° and the Ir O.21°, comparable to that of Ir on SrTi03
(Figure 29). The morphology and crystalline quality of the iridium films on A-plane

sapphire also varied with temperature and thickness. Both (100)Ir and (111)1r epitaxy are

possible on (1120) A1203. (100)1r epitaxy occurs at substrate temperatures of 700°-800°
C during e-beam deposition with the epitaxial relation (001)1r<110>1,||(1 120),t,1203
(1l00) “203. For lower temperatures, we observed (111)l,||(1120) A1203. The epitaxy of

the Ir may also be dependent on the direction and/or the angle of the offcut on the A1203

substrates. The morphology of the cleaned, etched and annealed surface is shown in

82

Figure 37 before and afier after Ir deposition. The Ir fihn was 300 nm thick and

deposited at 800° C.

 

    

4%.

Ir average step height = 6.3mm

 

8x8 um

Figure 37 AFM scans of (1120) sapphire substrate before (upper) and afier (lower)
(001)1r deposition. Ir thickness = 300 nm. (Sample 02042002).

The surface of the A1203 is irregularly terraced afier the annealing treatment. The
average step height is 2.2 nm and the RMS roughness 0.6 nm. After Ir deposition the

surface displayed a surface with terraces and mean roughness 1.7 nm.

6.3.6 Backside coating

SrTiO3, A1203 and the other oxides used are electrically insulating. For biasing, it
was assumed necessary to have a conducting path from the Ir to the rest of the biased

stage. SrTi03 can be made conductive by heating it in vacuum at 750° C in contact with

83

the stainless steel sample holder. This may be due to oxygen reduction, which results in a
conduction carrier enhancement (133), or a chemical reaction between the oxide and the
metal. However, the method was not reliable, as the backside resistance varied for each
sample, from less than 1000 to a feka. This also did not work as well for the other
oxide substrates. Sapphire is particularly stable, and remained highly insulating after heat
treatment. To obtain reliable backside conductivity on all samples, we deposited a 100-
150 nm coating of Ir on the backside and one edge using a tilted sample holder designed

for this purpose.

6.4 Silicon preparation

We have used Si substrates early in our studies to investigate the effects of
changing sample holder geometry on current density (Chapter 8). The 3” (001) p-type Si
wafers used as substrates had well-polished single crystal surfaces with few defects.
They were coated with photoresist to protect the surface, then cut with a dicing saw into 1
or 2 cm squares. The individual samples were marked on the backside, and then cleaned
in a solvent rinse with acetone and methanol for 10 min each in an ultrasonic bath. For
bias-enhanced experiments, no other surface preparation was necessary. The samples
were placed in the CVD reactor, and subjected to a 10 min hydrogen plasma cleaning
prior to adding CH4 to remove the native oxide (1 49-151). We determined that
lengthening or shortening this step by as much as 5 min did not affect nucleation

densities.

84

6.5 Summary

Single crystal substrates of high quality are required for the deposition of single

crystal diamond. This chapter mainly discussed the preparation of Ir films on SrTi03
and A1203. Terraced, low roughness surfaces of (001) SrTi03 and (1 120) A1203

substrates were prepared using a combination of cleaning, chemical etching and
annealing at high temperatures. We deposited single crystal (001) Ir on these surfaces
using UHV electron beam evaporation. The morphology and crystalline structure of Ir on
SrTi03 was studied with AF M and X-ray diffraction as a function of substrate
temperature and film thickness. High quality smooth (001) Ir films with X-ray linewidths
of 019° and 021° for SrTi03 and A1203 substrates respectively, were achieved. The

epitaxial relationships are (001 )1;<1 10>Irll(00 l )SrTiO3<1 10>SrTi03 and
(001)1,<110>,,||(1150),.1203<1’1’00>A.203. The epitaxy relation between (001) It and

(1 120) A1203 was studied and a possible matching of the two lattices suggested.

85

Chapter 7 Microwave plasma chemical vapor deposition

of diamond

7. 1 Chapter overview

This chapter presents details of the experimental setup for CVD diamond growth.
The microwave plasma reactor is described, and modifications made to the sample holder
of the system for biasing experiments are discussed. Finally the procedure for
heteroepitaxial diamond nucleation and growth on iridium are given. Results of biasing

and growth experiments are given in the following chapters.

7.2 Chemical vapor deposition system

Diamond growth is carried out in a microwave plasma CVD reactor based on a

design by Asmussen (152). A schematic diagram of the reactor is shown in Figure 38.

86

 

       

,-
0945: Shutoff
‘ valves
'S

  

 

 

Figure 38 Microwave plasma CVD diamond reactor system

The system consists of: (1) a 2.45 GHz microwave source with a 7 kW magnetron power
supply, a rectangular wave guide, a cylindrical brass cavity and an excitation probe; (2) a
gas handling system for a stainless steel vacuum chamber outfitted with ports for
feedthroughs and gauges, a quartz dome, mass flow controllers and a pumping system;
(3) a PC control system that monitors system pressure, microwave input and reflected
power and allows timed experiments; and (4) biasing and stage temperature monitoring

circuits with a data acquisition unit.

7.2.1 Reactor and vacuum chamber

The system comprises a quartz bell jar type reactor that confines the plasma and reactant
gases to a quartz dome in the vacuum chamber. The reactor is fed with a 2 m long
rectangular waveguide from a Cober 2.45 GHz 7 kW model SF6 microwave power

source to a cylindrical brass microwave cavity. An adjustable brass excitation probe

87

extends down fi'om the top of the cavity and allows excitation of the modes. The cavity
length is also set by adjusting the position of the sliding short. The excitation mode is

TM013, shown in Figure 39.

Quarlz
dome ‘

 

 

 

 

Plasma dscharge

Figure 39 TM013 Electromagnetic mode of reactor cavity.

In TM013 mode, the electric field lines circle the inner perimeter of the cavity, and the
magnetic field lines consist of four toroids evenly spaced on the vertical axis. This gives
the plasma a hemispherical shape. The quartz dome sits inside the brass cavity as shown
in the cross-sectional schematic in Figure 40. Positioned within the dome is the stage and

sample holder.

88

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Reactor
cavity H
Slicing Short
BOOK Front
Probe
q
5 cm
<——>
Air
Cool' Quartz dome out
"‘9 __, w New port
air in / \ for pwofnefer)
sample
._, 1
Gas In -E‘. l “—— qse plate
__ =57 l
_ T _J .J
Side mew H Slage and sample holder

Figure 40 Cutaway scale drawing of CVD reactor. The base plate rests on the vacuum

chamber.

With the ionization of the reactant gases directly above the sample holder, a plasma

forms. The quartz dome and stainless steel platform that holds the Mo sample holder sit

above a vacuum chamber that has a residual gas analyzer, ion gauge, pumping system

(turbomolecular pump and roughing pump), gas handling system and electrical

feedthroughs for thermocouple, stage biasing, and heater connections. The entire sample

stage is removable, and is bolted to the baseplate of the vacuum chamber, so that the

sample sits flush with the base of the cavity.

The reactor cavity and baseplate are cooled using a closed loop cooling system

with lines running through the baseplate around the cavity base, and through the probe.

89

A Neslab chiller model CFT-300 with a 5 gallon capacity is set to 29° C when the system

is running.

7.2.2 Gas handling

The gas handling system consists of a vacuum chamber, pumped by a water
cooled Alcatel model 5150 turbomolecular pump backed by an Alcatel Model 2021 CP
roughing pump. The pump reduces the system base pressure to the mid 10'8 Torr range.
The system does not have a load lock for sample loading. After a sample is loaded, the
system is pumped for a minimum of 4 hr to reach a base pressure <5x10'7 Torr. Residual
gases then consist of water and nitrogen (<10'9 Torr) as measured by a residual gas
analyzer (RGA). Gases are mixed in the line after coming through the flow controllers
and fed into the system through a single inlet valve. The process gas feeds into a gas
distribution plate with 8 symmetric angled 0.04” holes around the base of the stage,
allowing for even gas flow into the bell jar. The gas is pumped out through eight 0.25”
symmetrically placed holes at the base of the stage plate. This is referred to as a forced
flow configuration. A MKS type 653 throttle control valve and MKS type 651C pressure
controller with 0-10 Torr and 0-100 Torr baratron pressure transducers regulate the
pressure to within 0.1 Torr of the setpoint. Exhaust gases are purged with compressed
nitrogen at the outlet of the rough pump and exhausted through the building exhaust
handling system. Each cylinder has a gas line to the inlet valve with a manual shutoff
valve after the flow controller to prevent leakage of unwanted process gases into the
system while the main gas valve is open, and also for leak checking the lines. The
system has dedicated lines for CH4 (99.999%), H2 (99.9995%), Ar (99.999%), and N2/H2

gas mixes. It is vented using 99.999% Ar to minimize nitrogen contamination of the

90

system walls. The process tanks are regulated at a line pressure of 150 psi, the small

N2/H2 bottles at 30 psi. Gas lines are flushed and leak-checked, and the vacuum system

is baked out with heat tape when a process gas tank, or the quartz dome is changed.

7.2.3 Modifications for biasing

We modified the sample holder and stage to carry out experiments on biased

nucleation and growth of diamond. A new stage was designed to include an electrically

floating sample platform and bolder, as shown in Figure 41.

BDMmg

Saxxm -~‘,

lead —
WIreS

dexxmwh
Sandehdda

 

hngmew
maHmQ
_/SMthMBk
i I
’1'- .A. d
J .5” _. .1" __ _.
r J’r I’r r ‘21”? I"
I I J"
l r an" I .
. . QuauRmQ
kxdedmxi
, Bdakm
( _, e
\

 

 

e / a.) Q
\Q‘Qg’a/ KL Siege

W‘ Plate

Figure 41 Exploded view of CVD stage and sample holder setup.

The stainless steel sample platform is electrically isolated from the stage by a 0.25”x

2.75” 0D ring of fused quartz. A K-type thermocouple is mounted on the underside of

the stainless steel platform. A 1.5” substrate heater can also be mounted underneath, but

was not used for the experiments in this thesis. The original design of the sample

platform was for a 3” Si wafer. Due to geometry constraints, the top surface area of the

platform has to remain fixed for the TM013 mode to remain stable in the reactor cavity.
We found that by electrically masking the area of the platform surrounding the 0.8 cm
Mo sample holder, we obtained a high current flux through the sample area during
biasing. Here, the sample holder and platfrom are at a negative 100-300 V potential,
while the rest of the stage remains grounded. The sample holder geometry relevant to the
nucleation of heteroepitaxial diamond is described in detail in section 8.3.3.

A Kepco DC power supply, model BOP 1000M, capable of 0-300V/0-200mA
provided the bias potential. Two 50 Q , 5W resistors were connected in series with the

power supply on both the positive and negative terminal. The schematic bias circuit

 

diagram is in Figure 42.
Isolated Sample
platform \

 

 

 

 

 

 

509% 509
l _

300V/200mA DC

Figure 42 Circuit for bias setup.

Under certain biasing conditions, we observed arcing: the plasma was unstable and sent

electrical feedback to the power supply during biasing. The resistors served two

92

purposes: to help dissipate feedback spikes from the plasma to the power supply, and to
provide a means for measuring bias current. A circular electrode immersed in the plasma,
connected by three standoff wires to the base of the stage, provided electrical contact to
the plasma. The plasma circuit had a resistance of 3-5 kQ. The return path for the bias
current is through the plasma and the bias circuit was only complete with a plasma

ignited.

7.2.4 Computer operation

The system is operated with a computer interface to control an experiment. Input
parameters are gas flow rates, system pressure, microwave input power, and duration of
experiment. The program records incident and reflected power and pressure fluctuations
at one-minute intervals. A separate computer running as a PC data recorder, with a Fluke
Hydra data acquisition unit, records the substrate temperature from the thermocouple,
pyrometer and bias current. The main system PC monitors chamber pressure, and
reflected and incident microwave power, checking they are within preset safety
parameters. It also has control of the pneumatic process (roughing pump) and main
throttle valves. The system can operate unattended for the preset experiment time.
However, adjustment of gas pressure and flow must be done manually when growing

heteroepitaxial diamond.

7.3 Operation procedure

Microwave plasma chemical vapor deposition is a complex process with many
variables. Some variables, such as pressure, microwave power, gas flow and mixture,

bias voltage and timing, can be independently controlled. Temperature and bias current,

93

however, are highly dependent on pressure, geometry, bias voltage and, to a lesser extent,
microwave power. Finding optimal parameters for diamond growth with such a large
parameter space is not trivial. After each run, the sample was examined and
characterized using optical microscopy and/or SEM and sometimes AF M, Raman
microscopy or X-ray diffraction. In most cases, experiments in which only one
parameter varied were attempted. We conducted more than 250 CVD runs on Ir/oxide
substrates and at least half as many on Si substrates. To keep track of the experiments on
Ir/oxide, we constructed a database. For each run, the input parameters of each step were
recorded (microwave power, gas pressure and flow, time and bias voltage), as well as
bias current and temperature readings of the pyrometer and thermocouple. This was
entered into a spreadsheet along with information about the substrate, pre-experiment
reactor conditions and geometry, and hyperlinks to files containing SEM images and the
data recorded by the system computers. From the spreadsheet, individual datasheets for
each experiment were made using a script that presents the relevant data in a one page
format for easy reference. The spreadsheet allowed us to quickly sort runs that were
made under specific conditions, and find the relevant data for them. We can, for
example, find all runs that had a 60 min bias at —150V followed by 60 min growth, and

compare data from different runs.

7.3.1 Procedure for heteroepitaxial diamond growth on iridium

The following is an abbreviated procedure for heteroepitaxial diamond growth on

Ir. Growth of diamond on Si followed a similar procedure, with modified parameters.

94

7. 3. 1. 1 Loading the sample

Samples are loaded into the CVD reactor in a Mo sample holder, resting on a
platform that sits on a removable stage. After a sample is placed on the Mo post, a Mo
cap is placed on top of the sample. The cap serves a three-fold purpose: (1) it keeps the
sample held in place during loading; (2) it masks any exposed oxide substrate, so that the
plasma only sees an Ir surface, and (3) it provides an electrical contact to the Ir. The
stage with the sample is loaded into the vacuum chamber and bolted into place in the

reactor. The chamber is pumped to 4 x 10'7 Torr, as measured by an ion gauge.

7.3.1.2 System startup, hydrogen cleaning, carburization and bias procedure

After pumping down, the ion gauge and RGA are shut off, and the turbo valve
closed. The pressure setpoint is set to 15 Torr. The turbo pump is turned off and the
roughing valve and gas valves opened. H2 flows into the vacuum chamber at 300 sccm.
At 10 Torr, the microwave power supply is manually turned on and set to 1 kW. The
ignition of the plasma at 10-12 Torr and 500 W of microwave power define t=0. At t=2
min, the microwave power is ramped up to 1.5 kW. It takes approximately 4-6 min for
the system to reach the pressure setpoint of 15 Torr. This stage of the procedure is called
the “hydrogen cleaning” step. At FSmin, the methane flow controller is turned on
allowing 6 seem of methane to flow into the system, and the pressure setpoint increased
to 18 Torr. The plasma is allowed to stabilize for 12 min. This step is called the
“carburization” step. At F16 min the bias power supply is turned on. At t=17 min, the
voltage control is turned on, and the bias voltage ramped up to —150 V over 20 s. After

the bias time, the voltage is ramped down to zero over 10 s. The power supply is left on,

95

with the voltage set at 0 V. This leaves the stage at a slightly positive potential with

respect to the plasma.

7.3.1.3 Growth step

After the bias is stopped, the CHa flow is set at 3 sccm. For growth times of three
hours or less, no other parameters are changed, and the microwave power is shut off after
turning off the methane channel for 1 min. For longer runs, after 1.5 hr, the pressure is
increased incrementally, and the methane flow rate further reduced. After turning off the
microwave power at the end of the run, the gas valve is closed and the throttle valve
opened fully, allowing the system to immediately pump down. This minimized
deposition of non-diamond carbon during the cool-down phase. The system is allowed to

cool, and the sample removed.

96

Chapter 8 Heteroepitaxial nucleation of diamond

8. 1 Chapter overview

This chapter describes major advances in the implementation of high density
diamond nucleation on epitaxial Ir grown on SrTiO3 and a-plane sapphire substrates. The
first part, presenting early work on nucleation on (001) silicon substrates, introduces the
CVD reactor geometry and general operating parameters. We then discuss the geometry
and CVD conditions needed to initiate nucleation on (001) Ir. To obtain a high nucleation
density on Ir, we focus on several critical issues: temperature, bias voltage/current
density, the sample holder, and overall geometry. Although the initial starting CVD
parameters for growth of diamond on iridium were similar to those for growth of
diamond on Si, we discovered that Ir degraded during processing. This led to a
systematic investigation of the effects of pressure, microwave power, bias voltage and
sample geometry on temperature. Furthermore, the environment near the growth surface
profoundly influences the nucleation. The secondary plasma that appears above the
substrate during biasing is also vital to nucleation. The second half of the chapter
describes the evolving morphology of the carbon condensate during the biasing process,
and the initial stages of diamond growth on Ir. We use a number of quantitative measures
of the initial crystallite spatial distributions: 2-D Fourier transforms and radial density

plots give us statistical information on the distribution of crystallites on Ir.

97

8.2 Role of bias current density

The first biased-enhanced nucleation experiments were carried out on 3” dia. Si
wafers. An alternative arrangement was an array of four 0.5 x 0.5” Si samples centered

on a 3.5” diameter Mo sample holder. Typical results are shown in Table 7.

 

 

 

 

Run ID Bias (Volts) Current (mA) Sample size Nucleation
Density
04-Feb-00 -150 110 0.5” square 104-109
09-Feb-00 -200 340400 0.5” square 10”
10-Feb-00 -200 280-500 3” round 10“

 

 

 

 

 

 

 

Table 7 Diamond nucleation on Si. All runs had a 15 min bias duration, 32 Torr, 3
kW microwave input power, 6 sccm methane in 300 sccm hydrogen. For 09-
Feb-00 and 10-Feb-00 runs the current rose continuously during biasing.

In some runs, the current increased with time at constant bias voltage. Although the
nucleation density approached 1011 cm'2 for high voltage and currents, we realized that
higher nucleation densities across the entire sample would require higher currents than
possible with available power supplies. We surmised that higher current densities would
allow us to maintain a secondary plasma while increasing nucleation densities.

A square Mo sample holder, 2x2 cm2 was designed to replace the 3.5” Mo holder
for 3” dia. wafers. To increase the current density, we devised a mask to shield the
stainless steel sample platform, reducing the total current while increasing the current
density at the sample. The mask was made from a 3” Si wafer diced into 4 pieces as

shown in Figure 43.

98

Molybdenum
Tungsten
Sam e holder
pl Bias ring

/ Silicon Mask - 4 pieces

/

 

 

 

 

 

 

 

Bias ring / / Stainless Steel
Stondofl \. ( ) // Sample holder
\ :1." ,1" 3 ..-‘ . _,-",,.i' if?
g . Quartz Ring
/ for electrical
\\ isolation
\ Flange
{'f— x :f \
\.‘\'.\\\ \ me
/ at) @
../' ’
>\\w/° 4715/ /
\ 6 ,/,4' j
/ \ Stage
\ / r/ / Plate

 

Figure 43 Sample holder, sample platform and bias stage with Si mask and 2x2 cm2

Mo sample holder.

The silicon mask surrounded, but did not contact, the 2 cm Mo sample holder. It was
electrically isolated from the stainless steel platform by 1 mm thick slabs of fused quartz.
The M0 holder was 1 mm above the Si mask.

Table 8 shows the results of these changes. The total current was significantly

reduced while maintaining high nucleation densities.

99

 

 

 

 

Run ID Bias (Volts) Current (mA) Sample size Nucleation
Density
12-Feb—00 -200 50—120 2 cm square 10‘0-10”
15-Feb—00 -150 35-40 2 cm square 104
18-Feb-00 -200 50-120 2 cm square 10‘0-10”

 

 

 

 

 

 

 

Table 8 Data from experiments carried out using the sample configuration in Figure

43. Sample data on silicon comparing bias voltage, current and nucleation
density. All runs had a 15 min bias duration, and were carried out at 32 Torr, 3
kW microwave input power, 2% CHit in 300 sccm H2. Current is rising during
biasing.

Assuming that all current flows through the sample, the current densities were I/A=7-l3

mA/cm2 (run 09-Feb-00) before masking, as compared to 13-30 mA/cm2 (run 18-Feb-00)

after masking. The total current decreased, but the current density increased. This

configuration allowed us to achieve current densities high enough to obtain nucleation

densities of 10'°-10ll cm'2 across the entire sample. After biasing, we then grew diamond

on Si. We successfully grew highly oriented diamond films on 2 cm square (001)Si

wafers (Figure 44).

(110)

(110)

 

2.0 um

Figure 44 Highly ordered diamond grown on (001)Si. The crystallite faces are (001)
but there is appreciable misorientation so that good coalescence does not
occur. Bias: —175 V, for 5 min at 32 Torr, 2% CH; in H2; 3 hr growth time.

100

8.2.1 Arcing during biasing

The arrangement described above was not rigid and the mask and Mo holder
could be inadvertently moved while loading into the vacuum chamber. This led to arcing
during biasing. Charge build up along the edges of the silicon mask could discharge to
the Mo holder or the stainless steel platform. Arcing also occurred due to carbon deposits
building up on the stainless steel platform, or poor contact between the molybdenum
holder and the stainless steel platform. This was reflected in large current spikes in the
I(t) profile (Figure 45).

300 -

250 -

 

200 n

 

 

.3

(II

C
l

 

100-

Current (mA)

 

50-

 

 

 

Bias On —150 V flBias Off

0.
. l. l . . , .
0.0 20.0 40.0 60.0 80.0

time (minutes)

 

Figure 45 I(t) profile for biasing on Si (22-Apr-OO). Bias: -150V for 16 min at 32 Torr,
2.2kW input MW power, 1% CH4 in H2, 150 ppm N2.

The current spikes corresponded to bright “flashes” across the sample, or around the

sample holder. Arcing, as well as low total current levels during the bias period, led to

101

inhomogeneous nucleation across the sample and grth of highly twinned
polycrystalline diamond.

A new configuration was designed to solve these problems. A Mo sample holder
to hold 0.5 x 0.5 cm2 Ir/oxide substrates (Figure 46) consisted of a round 0.8 cm post
screwed into the stainless steel platform, to insure good electrical contact. A Mo cap
rested on top of the substrate and enveloped the post. The Si mask was cut with a 1 cm
hole in the center, and the mask was supported and held in place by three alumina pegs
spaced symmetrically around the edge of the mask. Round geometry was chosen to

eliminate comers or sharp edges, and to keep with the symmetry of the rest of the system.

 

Figure 46 Illustration of the Mo sample holder and arrangement on the stainless steel
platform. A is the Mo sample cap, B, the sample post, C, the sample. Right
figure shows holder D (consisting of A,B,C) screwed into stainless steel
platform F, and Si mask E. The cap leaves a 3.5 mm region of the Ir exposed.
The alumina support pegs under the Si are not shown.

The M0 cap served to make electrical contact with the Ir, and also to shield any oxide left

exposed by incomplete iridium evaporation.

102

8.3 Improving bias conditions

For Ir/oxide substrates, the CVD conditions are compared to those used for HOD

on Si in Table 9. The current-voltage behavior is shown in the I-V curve of Figure 47.

 

 

 

 

 

 

 

 

 

 

Substrate Ir Si (for HOD)
Microwave Power 2.2 kW 2.2 kW

Pressure 32 Torr 32 Torr

Gas mix (%CH.t in H2) 5% CH4, 150ppm N2 1% CH4, 150 ppm N2
Bias -250 V -175 V

Bias time 10 min 4.5-10 min

Ring height 16 mm 16 mm

 

 

 

Table 9 Comparison of diamond growth on Ir with growth on Si.

70—

60‘

50-

40-

30-

Curent (mA)

20‘

10..

 

0

 

40

1

80

l l 1
120 160 200

I l l
240 280 320

Applied Bias (-Volts)

Figure 47 1-V of round cap geometry with Ir/oxide substrate. Line is a guide to the
eye. 2.2 kW input MW. 32 Torr, 5% C114, 16.5 mm ring height.

103

At —280 V, a bright secondary glow appeared over the iridium sample and the Mo cap

edges. Initially we tried bias and growth parameters similar to those used for HOD on Si.

A higher voltage (-250V) was used to attain current densities to sustain the secondary

plasma.

We were able to obtain nucleation densities of 109 - 10IO cm’z. However, we

noted heavily twinned diamond and Ir damage that could have resulted from the ion

bombardment or the high temperature. During biasing, the temperature rose to 870°-930°

C whereas Ir deposition took place near 800-850° C.

We then carried out experiments in which the bias voltage was varied from -250

to —150 V. Heavily twinned diamond was still observed; changing CHs concentration

also had little effect. The effects of gas pressure and microwave power and bias voltage

on temperature were then studied, with results shown in Figures 48 - 50 below.

Temperature (°C)

Figure 48 Temperature vs. pressure at -150V bias and 0V bias.

Temperature as a function of Pressure

 

 

 

 

 

 

 

 

 

 

 

 

 

950 _ 07-Jul-00 ‘ i
. I ,
900 T . I
l i a I
850 l l l .
| Tl .
800 I l _ l l
l I - .
I. l j
750 'l———r“— l ‘ - i ‘ We ~—
1 . 0P ' l
700 -
z o
' I -150V bias
650 i if 0 O bias
500 a__.__u 7__ __..__A_ 3, PMW=1.5 kW (input)
i Ring height =16mm
550 ' ' i l A] r
10 15 20 25 30 35 40

Pressure (Torr)

Temperature is

measured with an optical pyrometer focused on the Mo cap surface.

104

Growth Temperature as a function of MW power
01-Jul-00 to 06-Jul-00 lr/SrTiOa

 

 

 

 

 

 

 

 

 

 

 

 

 

950-
900-———_—. , i-.v -_.
850-_—. ,;- ,-- -.- . :-2_

A . I. ll _

g: 800 'gg

e 3 I Pyrometer

3 750 i O Thermocouple

E l p=32 torr

8, 700._2___ . _22222__ --.2..- - , s

e .

0 .

r- 650 . 3
600 _ l i

l a .
550 1

 

 

 

 

 

1200 1400 1600 1800 2000 2200 2400
Microwave Input Power (Watts)

Figure 49 Growth temperature at 32 Torr as function of microwave input power.
Temperature measured with an optical pyrometer focused on the Mo cap, and a
K-type thermocouple fixed underneath the stainless steel platform.

Temperature as a function of Bias Voltage
950 - ,

925- _ . .e~ . . .-- -, -.,‘

 

900

 

875-——————L——'

850

 

 

Temperature (°C)

 

 

' l
825 p=32 ion a.
. MW input . 1500 to 2200 W

 

 

 

 

 

 

 

 

800 a l l i l .
120 140 160 180 200 220 240 260
Bias Voltage (-Volts)

Figure 50 Temperature as a function of negative bias voltage at 32 Torr. Temperature
is measured with an optical pyrometer focused on the Mo cap surface.

The results of these calibrations indicated a strong dependence of temperature on gas

105

 

pressure. In general, bias current decreased with increasing pressure at a fixed voltage.
Microwave power, on the other hand, had little effect on bias current. Furthermore,
subsequent studies showed changing the distance of the bias ring from the holder
changed the shape of the plasma, as well as the current and thus temperature. Raising the
bias ring from 16 to 32 mm elongated the plasma between the two electrodes, and the
current decreased on average 20 mA.

We lowered the temperature over a series of runs by decreasing pressure from 32
to 18 Torr, decreasing the microwave power fi'om 2.2 to 1.5 kW, lowering the bias
voltage from —250 to —150V, and increasing the ring height from 16 to 32 m. We also
eliminated N2 from the feed gas. Under these conditions, we found the temperature was
710 i 10° C during biasing as measured by an optical pyrometer focused on the Mo cap.

For a 30 min bias, the Ir appeared undamaged with locally high nucleation
densities, 1010-1012 cm'z. We next sought to improve nucleation uniformity. Increasing
the bias voltage only caused Ir damage, and fast growth due to high temperatures. We
extended the bias time with the intent to saturate the surface with carbon. The near-

optimal biasing parameters are shown in Table 10.

 

 

 

 

 

 

Microwave power 1.5 kW
CH2/H2 ratio 2%

Pressure 18 Torr
Bias Voltage -150 V
Bias time 60 min
Ring height 32 mm

 

 

 

 

Table 10 Parameters for carbon saturation of the Ir surface during biasing.

106

 

SEM images showed a substrate covered in a bright electron—emissive dense carbon

deposit (Figure 51).

   

Figure 51 SEM images of sample (17-Jul-00) with conditions of Table 10. The
circular bright area in (a) corresponds to the biased region of the 5 mm square
Ir/oxide sample, exposed to the plasma through the round opening of the Mo
sample cap. The surrounding dark area is bare Ir, masked by the sample cap.
The narrow band around the bright area is due to shadowing by the sample cap.
The objects in the lower right and left of the image are copper sample clips,
part of the SEM sample holder. (b) shows a high resolution scan of the central
region of the sample, covered in a uniform dense carbon deposit. The sample
was in the plasma for 1 min beyond biasing, so some growth and coalescence
of nuclei has already occurred.

Short growth after biasing under somewhat modified CVD conditions (Table 12, Section

9.1) showed that this carbon layer was necessary for the nucleation and growth of

diamond single crystal films.

8.3.1 The molybdenum cap and sample holder

After many CVD runs, the Mo cap became thickly coated with diamond which
began to flake off. The uncoated surface of a new cap and sample holder did not yield a
uniform coverage of the sample, but rather inhomogeneous nucleation as shown in Figure

52.

107

 

Figure 52 Sample 07-Nov-2000 after standard 60 min biasing and 3 hr growth. Mo
cap was polished and uncoated. Bright region has a very high density of
highly twinned crystals

8.3.2 Sample height relative to the plasma

We observed bias and growth temperature were sensitive to the vertical position
of the sample. Figure 53 shows a plot of bias temperature vs. position of the top of the
sample holder (relative to the fixed stainless steel platform). Only bias temperature is

plotted here for clarity, the grth temperature was 40-50° C lower for each run.

108

 

800 -

760 -

720 -

 

Temperature (°C)
0

640 _ 0 Round holder
I Square holder

 

 

 

 

 

l

1.5 2.0 2.5 3.0 3.5 4.0 4.5
Sample height relative to platform (mm)

Figure 53 The temperature during bias vs. sample holder height; relative to the fixed
stainless steel platform in the CVD reactor. A round and square sample holder
was used. Pressure, 18 Torr, except for data at 4.0 mm, p=16 Torr.
Temperature was measured using an optical pyrometer focused on the Mo cap
surface.

The strong temperature dependence is surprising, and is essentially independent of
sample holder shape. A temperature change of 50° C occurred for a 0.5 mm change in
height. Since there is a narrow temperature window in which diamond “nucleation” can
occur, the precise positioning of the sample with respect to the plasma is crucial. We
found the optimal condtions for diamond nucleation and growth occurred at a 2.4 mm
height relative to the platform, for which the sample temperature was 710° C for bias and
650° C for growth. The height maximized the nucleation density. Qualitatively, it
appears that variations in height may also expose the grth surface to different plasma

species at different concentrations.

109

 

 

8.3.3 The secondary plasma

During biasing we observed a bright, secondary, plasma that occurred directly
above the sample, separated from the main plasma discharge (Figures 57 and 58). We
made the following observations: (1) The presence of a secondary plasma during biasing
is a necessary, but not sufficient, condition for efficient diamond nucleation on Ir; (2)
Instability of the secondary plasma during biasing is fatal to good nucleation; (3) Bias

voltage, bias current, geometry and the cap coating influence the secondary plasma glow.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   

@—
-H

L-

 

 

 

 

 

:P

Figure 54 Cross-section of the bias stage and the sample holder. The bias ring
confines the main plasma discharge. A secondary discharge appears above the
sample area. A. Quartz dome. B. Main plasma discharge, C. Secondary plasma
discharge. D. Stainless steel platform, E. Quartz isolation ring, F. Stainless
steel stage, G. Alumina centering posts for the Si mask, H. Alumina washers
for mask isolation, 1. Si mask, J. Bias ring and supports, K. Mo sample holder.

The small secondary discharge appears pale, almost colorless, and is much brighter in

intensity than the main discharge. There is also faint blue glow over the diamond coated

110

cap. Whereas the secondary plasma extinguishes immediately when the bias is turned
off, the blue glow over the cap remains. Other groups have reported on a blue, purple or
red secondary plasma above the sample, or the diamond covered sample holder (85, 86,
121-123), but they have not noted that the secondary discharge was confined at the
sample. We believe that the characteristics of the secondary plasma result from a dc
glow discharge induced by the biasing process in the specific reactor geometry used.
From the discussion of glow discharges in Section 3.4.1, the secondary discharge could
correspond to the negative glow. Other groups ( 76, 125) have reported the secondary
discharge corresponded to the cathode sheath. If the secondary plasma is a negative glow
of a dc discharge, then electrons are carrying the current in this region. The pale, slightly
blueish color of the secondary plasma is indicative of the negative glow region of a
hydrogen glow discharge (44). The cathode glow in a hydrogen plasma should have a

reddish/brown color.

8.4 0+ growth experiments — The effect of biasing on Ir/SrTi03 and Ir/Al203

8.4.1 Bias experimental procedure (0+ growth)

The experiments described in the remainder of this chapter consist of a CVD
biasing procedure with no subsequent growth. These investigations were concerned with
the time development of the bias process. However, some short time growth is inevitable
even if the plasma is extinguished after bias. This is the reason for adopting the

terminology “0”

growth. To reliably compare a series of runs, we made every effort to
insure a reproducible process for every experiment. 0+ growth experiments (or biasing

only) were carried out with the following conditions: 1800 W microwave power, -150 V

111

 

DC bias, 2 sccm CH4 in 300 sccm H2 at 18 Torr. Timing began with ignition of the

plasma at 1000 W and 10-12 Torr. There is some variability in the exact pressure when

the plasma is ignited, because the pressure rises very quickly at this stage. The process

included the following steps:

t=0 (1) Hydrogen cleaning: 300 sccm hydrogen is flowed into the bell jar to
raise the pressure to 15 Torr.

t=5 min (2) “Carburization”: the methane flow is turned on to 2 seem. The
pressure setpoint is increased to 18 Torr.

t= 15 min (3) The bias power supply is switched on. The voltage control on the
power supply is switched on at t=l6min.

t=17 min (4) The bias voltage is increased smoothly from O to —150V over 20 s.

t=20-180 min (5) The biasing period. At the end of the biasing period, the microwave
power is turned off. This extinguishes the plasma and the biasing
current drops abruptly to zero. The throttle and gas valves are closed
immediately to stop the gas flow into the system; the bias power
supply is switched off. The system is pumped out using the backing
pump.

The important part of the process is the rapid termination of the biasing step, leading to a

“quenching” of the sample temperature.

Examination of the Ir surface with SEM has proven useful in visualizing the
carbon distribution, and the early stages of diamond growth. When examining the
following images, it should be noted that there are two effective means of interpreting

contrast in a SEM image. Higher topographical features are brighter because they present

112

 

 

 

a larger volume to the electron beam; so more scattered electrons are produced.
Diamond, in contrast to other forms of carbon, produces a bright SEM image because of
its high electron emissivity. This is critical, as extremely small volumes of diamond can
be detected.

The following sections describe how carbon condenses on Ir/SrTi03 samples as

the biasing period varies from 20 to 180 min.

8.4.2 20 min bias

Although the Ir surface appeared unchanged under visual inspection after 20 min

bias, SEM revealed striking changes to the surface (Figure 55).

 

1mm

Figure 55 SEM image of sample 20-Ju1-02 of (001) It on SrTiO3 after 20 minute bias.

The above SEM scan shows bright contrast in the circular region at the center of the 5
mm substrate. The exposed area is circular, a result of the 3.5 mm aperture of the Mo
cap. The edge of the exposed area is brighter due to a buildup of carbon at the inside

edges of the cap. The area labeled A in Figure 55, is expanded in Figure 56 revealing

bright contrast that originates from an inhomogeneous diamond, or diamond-like, deposit

113

ill

that covers 75 :t 4% of the surface.

 

10 um
Figure 56 Center region of 20 min sample (20-Jul-02)
The percentage of carbon coverage is calculated by binarizing the image, and measuring
the ratio of white to black areas. Variation in the threshold limits determines the error.
We interpret the bright condensate as diamond that nucleated from the carbon deposited
on the Ir over the course of the 20 min bias. The dark regions are believed to be Ir, not
yet covered by diamond. Figure 57, at 10 times greater magnification, shows that the

condensate is oriented with respect to the in-plane Ir crystallographic axes.

 

Figure 57 20 min bias sample (20-Jul-02) SEM image. 1 pm

Block-like areas, 100-200 nm, are oriented along Ir [110] directions. The dark regions

114

 

 

are either bare Ir, or areas covered with non-diamond carbon where diamond has not yet
nucleated. A higher magnification image, Figure 58, shows the structures in the dark
regions. These lines of diamond appear to decorate surface features that are aligned with

Ir [110] directions.

 

300 nm
Figure 58 High resolution SEM image of 20 min bias sample (20-Jul-02).

The “whorls" may correspond to diamond that preferentially nucleates on topographic
features of the Ir, e. g., step edges. The thickness of the deposits has not been measured

directly with AF M, but are expected to be only a few nrn.

8.4.3 60 min to 80 min bias

Runs of 60 and 80 min bias led to similar results for coverage: 82 i 4% for 60
min, 89 :h 4% for 80 min. An SEM image of the central region of a 60 min bias sample

(the same scale as Figure 57) shows an increased area of diamond deposition (Figure 59).

115

 

Figure 59 Central region of 60 min bias sample (18-Jul-02).

Here, dark regions correspond to areas devoid of diamond that lie along [110] directions.
The 80 min bias sample was nearly indistinguishable from these results at this
magnification.

At higher magnification, Figure 60, remarkable features emerge. An SEM scan

of the bright area on an 80 min bias sample is shown in Figure 60.

 

100 nm
Figure 60 SEM image of 80 min bias sample (22-Jul-02).

The image shows small bright objects on a dark background with a mean diameter of 10-
15 nm distributed across the Ir surface. In some cases, as in Figure 60, the particle edges

appear aligned with It [110] directions. A quantitative analysis of the distribution of

116

 

particles is given in Section 8.5. If the outermost edge of this sample is examined, Figure

61, it shows a pattern similar to that seen at the shortest bias times (20 min).

 

500 nm

Figure 61 Edge (within 200 pm of the inside cap edge) of 60 min bias sample (21-Jul-
02).

The above SEM scan was taken only about 200 pm from the inside cap edge. With the
exception of this small, circumferential region, the density was uniformly high over the

rest of the substrate.

8.4.4 120 and 180 min bias

Biasing for yet longer times (120 and 180 minutes) proved detrimental to the Ir
surfaces. An SEM image of a 120 min bias sample shows new features that did not

appear at shorter times (Figure 62).

117

 

Figure 62 120 min bias sample (08-Sep-00). 30° sample tilt

Small white asperities on a scale of 200-300 nm appeared on the surface. The white

 

spots appeared as “cone-like” growths in the carbon deposit. This indicated that extended
biasing caused instabilities to develop in the carbon deposit and the occurrence of growth
during biasing. The instabilities may initiate at defects in the iridium surface. The above
scan was taken at a 30° tilt of the substrate in the SEM.

180 min of biasing resulted in significant damage to the iridium surface (Figure

63).

 

Figure 63 Comer of 180 min bias sample (08-Sep-00). 30° sample tilt.

The Ir buckled and detached from the substrate surface. This SEM image is of the lower

118

right comer of the biased sample area. A ridge of carbon formed along a circular ring
defined by the inner edge of the Mo cap. An image of a spot, such as at the region

labeled A, is shown in Figure 64 at higher magnification.

 

Figure 64 SEM scan of 180 min bias sample (08-Sep-00). 30° sample tilt.

The spots show a “forest” of the cone-like growths that self-assemble into lam-diameter
clusters that appeared also on the 120 min bias sample. It may be that the cones stimulate
growth of more neighboring cones. The range in cone sizes suggests that they did not

appear simultaneously, but emerge over time.

8.5 Quantitative analysis of 60 min bias on Ir/Al203

Our standard process for single crystal diamond involves a 60 min bias, therefore
this bias time is of particular interest. The sensitivity of the SEM to diamond enabled us
to distinguish individual crystallites that emerge at a very early grth stage of diamond
on Ir. (See Figure 60 and Figure 65). The clarity of the images has allowed us to
characterize the size and spatial distributions using several statistical measures: (1) direct
real-space particle analysis to obtain the areal density and size distribution; (2) 2-

dimensional Fourier transforms of binarized images to reveal spatial symmetry, and (3)

119

radial density fimctions that show interparticle correlations. The particles are of order 10
nm in size. The particle density (n) was calculated by binarizing the image, so that the
particles became black blobs on a white background. They were then labeled and
counted over the entire SEM image area using the NIH ImageJ software package (153)
This was done for SEM images on different samples of the same bias time. The particle
density is calculated directly by dividing the number of particles by the image area. For
60 min samples, n = 3.2 :t 0.7 x 1011 cm'z. The nucleation and early grth was similar
on Ir/SrTiO3 and Ir/A1203 samples. The analysis contained in the rest of this chapter was
done on 60 min biased Ir/Al203 samples.

The particle distribution at first appears random for a 60 min biased sample, as

can be seen in Figure 65.

 

100 nm
Figure 65 High resolution SEM image of 60 min bias sample 12-Apr-02.

Closer inspection, however, showed evidence for spatial correlations. In order to test for

this possibility, we calculated 2-D Fourier transforms as follows:

120

A grayscale SEM image was first cropped to 512x512 pixels, and then filtered to
equalize the level of contrast, since the SEM scans often had a grayscale gradient across
the digitized image. The filtered image was then transformed to a 1-bit black/white
image. The threshold level was set by inspection. Since this setting is somewhat
subjective, it was varied to determine the effect of different levels and chosen to maintain
a reasonable representation of the original image. An example of the filtered SEM

image, and its binarized version are shown in Figure 66 for a 60 min bias sample.

7' .; ‘ :4 me n -~...«.- v-a- -:.

ol-w' ?-"' ;<;,.'.‘v .1

 

4—>
100 nm

Figure 66 Equalized SEM scan (left) of sample 12-Apr-02 and its binarization (right).
Nuclei correspond to light spots in the SEM image, dark spots in the binarized
image.

Although the sizes of the particles are somewhat dependent on the binarization threshold,
there is little effect on the number of particles in a 100 nm2 area, or on their relative
positions. The 2-D FFT of the binarized image was calculated using the Image/J

software program. The power spectrum of the 2-D image from the SEM image of Figure

66 is shown in Figure 67, with intensity on a logarithmic scale.

121

0
U.
.00.
n
e
to

 

Y

R (nm'l)
Integrated Intensity (log)
0
0)

 

 

-O.1 r 1* I '
0.00 0.04 0.00 0.12 0.16
.020 .010 0 0.10 0.20 kn ("m")
k‘(nm'1)

 

Figure 67 2-D power spectrum of the binarized image shown in Figure 66 and its
radial cross—section. The power spectrum has been smoothed by applying a
filter that takes the average of the 3x3 neighborhood of each pixel. This
process was applied twice, so small features are broadened.

The Fourier transform is useful for revealing averaged symmetries of spatial features in
an image. Figure 67 shows an intense central peak, an artifact of the finite area of the
transform, and a prominent ring at kn: 0.068 nm", with the relation R=kR",
corresponding to a real-space distance of R=l4.7 nm. A smaller peak at kR=0.028 nm'l
emerged from the cross-section at R=35.7 nm.

The radial distribution function can give information about the distribution of
nearest and next nearest neighbors. It is often used to describe the spatial correlations
between molecules in a liquid ([54, 155), particularly in analyzing X-ray scattering data

of liquids (156). It is here modified for 2D. A density, p(r), is defined for the number of
particles (n) per unit area of an annulus of width, dr, as p(r)dr = Zd—n. The radial

717‘
distribution function, g(r) is then defined as the density function at a radius r divided by

, where p0 is N/(rtRz). For large r, g(r) —> 1.

 

the total areal density, i.e. g(r): p (r)
p

0

122

The radial distribution of small crystallites was found by assigning a (x,y)
position to every particle in the binarized image of the sample pictured in Figure 66. For
this analysis, the entire SEM image was used (1280x960 pixels) in order to keep the
largest data set possible. A cutoff value of 10 px was assigned, so blobs smaller than 10
px in the above image were ignored in the analysis. The (x,y) positions were found
using the Image/J software measurement tool, which picked a center of mass for each
blob in the image, and assigned it a coordinate relative to the upper left comer of the
image. The output was saved as a text file. The distances between all pairs of
coordinates were calculated using a script, which binned the measurements, and saved the

results to a text file. The normalized RDF for the image of Figure 66 is shown in Figure

68.

2.0 11)

(2X3) (4)15)

 

 

 

1.5-

90)

0.5 d

 

0.0

I
I
I
I
I
I
l
I
I
I
I
I
l
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
l

 

 

 

20

40

R (nm)

l ' T

60 80

Figure 68 Radial distribution function plot of 60 min bias sample. Dropped lines
show predicted peak positions for a 2D hexagonal array. A properly
normalized RDF should asymptotically approach 1 as R —) 00. The fall off at
large R is due to finite system size (Rm. z240 nm).

123

The RDF shows a first nearest neighbor peak at 13.8 nm followed by two broader peaks.
The position of the first peak agrees well with the measurement from the 2-D Fourier
transform and indicates strong correlations in radial distances of adjacent crystallites.
The presence of higher order peaks indicates that correlations persist to much longer
distances. The peak positions for a hexagonal array are tabulated in Table 11 for a first

nearest neighbor distance, a = 13.8 nm.

 

 

 

 

 

 

Neighbor Distance Calculated for
a=13.8 nm

1st a 13.8

2“(1 ,5, 23.9

3rd 2a 27.6

4th fig 36.5

5th 3a 41.4

 

 

 

 

 

Table 11 Calculated values for a 2D hexagonal array.

The predicted and measured values up to 4th neighbor distances agree well. The second
peak in the RDF maybe be significantly broadened because it involves overlap of the 2"°
and the 3'° neighbor peaks. The third peak in the RDF falls in between the calculated

values of the 4th and 5th neighbors.

8. 6 Discussion

The time evolution experiments of biasing on Ir indicate there is a window for the
bias time. A bias period that is too short results in incomplete sample coverage, and a
bias period that is too long causes damage to the Ir. However, close examination of the
short bias time (20 min) was useful for identifying the island aggregation of the carbon
on the iridium. Aggregation of depositing material on a substrate is known for other

systems of epitaxial deposition, and has been studied using STM (Scanning Tunneling

124

Microscopy) and computer simulations (157-160). It is possible that certain areas of the
Ir, such as step edges, have different sticking coefficients for carbon, and are preferential
for the carbon. This has been shown to be the case in Si/(001)Si epitaxy, (161), for the
transition between 2-D layer by layer and island growth.

Inspection of SEM images of 60 min bias samples, such as in Figure 65 and
Figure 66, suggested to us a hexagonal close-packed array of particles. Most crystallites
appear to have an average of six nearest neighbors. The RDF peak positions agree well
with a hexagonal array, and indicated the correlation goes further then can be seen by
inspection. This suggests that the distribution results from “interactions” among nuclei.
We infer that nucleation does not need to occur at special topographic features of the
substrate. Careful examination by SEM also reveals that the crystallite shapes, even at
this early stage, are faceted with [110] lateral faces. Furthermore, the narrow distribution
of sizes of crystallites across the entire surface of the Ir strongly suggests that nucleation
begins synchronously with the termination of bias current. The areal density measured, n
= 3.2 :1: 0.7 x 10H cm'z, is a lower limit on the nucleation density. As some coalescence
has already occurred, it is likely closer to 1012 cm“. Our measurements for nucleation
densities are unambiguous, based on a count of particles per unit area. This is significant
because nucleation densities in other reports have usually been estimated from outgrown
crystals. In Chapter 4, Table 6 and the accompanying discussion, we pointed out that the
highest nucleation densities previously reported on Ir are 10°-109 cm'2 (4-8). Other work
that has examined the initial distribution of crystallites, looked at the distribution of
nearest neighbor distances for diamond crystallites on silicon (88, 111). It was found that

the measured nearest neighbor distances follow a random nucleation model, with no

125

spatial correlation. We believe the correlated distribution of crystallites is directly related
to the high nucleation densities observed on the Ir after biasing. This arrangement of
nuclei could not happen unless the nucleation occurred as the sample was quenched with

the cessation of the bias. This leads to the formation of single crystal diamond after short

growth times.

8.6.] Nucleation and early growth model of heteroepitaxial diamond

Based on the evidence presented in this chapter, we propose the following model

for the nucleation and growth of heteroepitaxial diamond (Figure 69).

‘11.1." r21." ‘r‘ «‘1‘.‘¢‘.‘;t‘_“." a" t" 0‘? 01".‘_‘.~“'.“.-“r i
.......
‘III'JII Ill Iii-H," ‘.l|’-Il III' II' | ' Ill-l’lt-I' '1'. If ' ' """
. . .~ ..............
1‘ r . r : .‘ r" .fi u'_ r ...............
...................................
-‘ I n I‘- l f' r' r" I‘ n ‘‘‘‘‘‘‘‘‘‘

'. '.' ,-I, .' r .
I l ‘ 1 1
.rlij-lI‘II I l I I’ll 11111
1‘ I" I.

i 0
1,11,... ,. . x. l.lll. ‘

I. 0'" l' f I '\ ‘ r' .“ y' - ‘ I' ‘

'\ ' ' 11111111 ‘I 1
:rllltlilllf..I:-rl'llf> . ,»II'-|I-I|l‘liv“":l . . .

I c n. r l' r I n‘ v 1‘ .' r‘ r' l r‘ s l’ r 1 l‘ I
L" ‘ “a...“m " u -a-.u'.w AR!“ 25.x; .t-eanean
' I

_ a

0

_ ”Substrate _ _

,1't '
lt lijil'

--..- 9:5. ...-u..._...‘.;- 11...”! .5'

Ir _
Substrate _

.3 '_"lL."' a»

     

 

.....

Fla-'5,” mu, n TF9! EFF“!

 

    

_ Substrate ., l i: Substrate _ E

 

Figure 69 Model for heteroepitaxial nucleation and growth of diamond on Ir. 1) Ion
bombardment and/or enhanced electron emission in the sample vicinity leads
to deposition of a carbon layer on the iridium. 2) Cessation of the bias current
leads to immediate nuclei formation on the surface. 3) Early grth involves
depletion of local carbon, and coalescence of nearby nuclei into crystallites. 4)

Growth from vapor phase proceeds on a surface already almost entirely
covered by diamond.

126

1. Biasing deposits carbon on the surface of the Ir. In the presence of a bias current,

the carbon condensate is in a highly excited, nonequilibrium state, and nucleation

 

does not take place. It is also possible that long bias times (60-80 minutes) are
necessary for the carbon to diffuse into the iridium and saturate the surface.

2. Turning off the bias voltage causes the temperature to drop rapidly with the
extinguishing of the secondary plasma over the sample. The carbon condensate
cools, and nucleation is initiated.

3. Early growth is dominated by depletion of the carbon deposited during biasing by

 

the initial nuclei. It is possible that there are much higher densities of nucleation
sites initially (>10'2), but competition for surface carbon amplifies growth of
larger nuclei.

4. Growth proceeds from the vapor phase as the surface carbon is depleted.
Coalescence occurs via lateral overgrowth methods while the crystallites are still

very small, because they are so densely packed.

8. 7 Summary

In summary, experiments initially on Si and later on Ir/oxide substrates allowed
us to increase the current density to the sample leading to increased nucleation densities.
Bias geometry and conditions which are conducive to the appearance of a secondary
plasma over the sample during biasing are crucial to the success of the diamond
nucleation. We developed a precise processing procedure that leads to rapid quenching of
our samples after bias. This allowed us to examine previously unseen early stages of
diamond growth on Ir. The initial arrangement and density of crystallites on the iridium

were investigated quantitatively. The Fourier transform clearly shows a circularly

127

 

symmetric ring corresponding to R=l4.7 nm in real space, a nearest-neighbor correlation
length. Radial correlations persist to large distances. The RDF plot confirmed this, and
distances of nearest neighbors are consistent with a hexagonal close packed 2D array for
to 3'° and 4th neighbors. We ended by presenting a new model for the nucleation and

early growth of diamond on Ir.

 

128

Chapter 9 Heteroepitaxial diamond growth:
characterization and analysis

This chapter examines the growth of diamond on iridium on strontium titanate
and sapphire substrates. The first section discusses experiments with growth times of s 3

hr on Ir/SrTiO3, and on Ir/Al2O3. The other section describes thick diamond films grown

on Ir/SrTiO3 substrates.

9.1 Short growth experiments - 53hr growth time

Short growth experiments in this section are defined as experiments where we
biased for one hour, under conditions conducive to high nucleation density, then changed

CVD parameters to growth conditions that lasted as long as 3 hr. Bias and growth

conditions are given in Table 12.

 

 

 

 

 

 

Bias Conditions Growth Conditions
Bias Voltage -150 V --
CH4/H2 ratio 2% 1%
Pressure 18 Torr 18 Torr
MW power 1500 W 1500 W
Time 60 min 1 min to 3hr

 

 

 

 

 

Table 12 Bias and growth parameters for short growth experiments

129

 

9.2 Progression of diamond growth — Ir/SrT i0 3

9.2.1 Early stages of growth — analysis using 2D-FFT
We studied the early growth and coalescence of diamond on Ir/SrTi03 with
interrupted growth experiments at intervals of l, 5, 10, and 20 minutes. An SEM image

of the surface after approximately 1 min growth is shown in Figure 70.

50

 

Figure 70 SEM micrograph of the sample surface after 60 min of biasing plus 1 min of
growth (17-Jul-00).

The surface is covered in small oriented diamond particles that are already starting to
coalesce. SEM micrographs were taken under the same conditions at identical resolution

for 5, 10 and 20 min growth. 2D-FFTs were performed on the SEM micrographs to

 

obtain an objective understanding of the early grain growth and coalescence. Even for
short growth times, coalescence has proceeded enough to make individual grains
indistinguishable. Therefore, in contrast to the Fourier analysis of the previous chapter,
the transforms here reflect the voids between incompletely coalesced adjacent crystallites
rather than individual particles. In the binarization procedure the voids appear as dark

regions and the diamond as white. The real space SEM micrographs are shown at the left

130

and the corresponding reciprocal space power spectra on the right in Figure 71.

   

           

«J
9“,.”
~.,ut . -'
‘1‘! V1.7"; :9

'.’
,.

-0,15 -0.10 -0.05 0 0.05 0.10 0.15
K(nm“)

10 mln

K (nm")

 

-0.15 -o.10 -o.05 0 0.05 0.10 0.15
Kl(nm')

20 min

1

K (nm'l)

   

-0.15 -0.10 -0.05 o 0.05 0.10 0.15
K‘(nm")

Figure 71 SEM micrographs of 5-20 minutes diamond growth and the corresponding
2-D power spectra.

The 2-D power spectra reveal the anisotropic orientations of the voids appearing in the

SEM images. As coalescence proceeds, the average distance between voids increases;

131

thus the power spectrum contracts in reciprocal space. Alignment of the voids emerges
as an increased sharpening of the four-fold symmetry in the power spectrum. In each
case, the voids are aligned with the <011> directions of the Ir. The Fourier transform
demonstrated that after 20 minutes of growth, voids were almost exclusively aligned

along these directions.

9.2.2 Complete coalescence — 30 to 180 minutes of growth

SEM and AFM studies demonstrated the evolution of coalescence of films grown

for 30 min, 60 min and 180 min (Figure 72).

 

Figure 72 Diamond on Ir/SrTiO3, (a) 30 min growth SEM image sample 21-Jul-00 ,
(b) 30 min growth AFM scan of 21-Jul-00 , (c) 60 min 09-Apr-01, (d) 180 min,
23-Jun-01.

A high magnification image of the central region of the 30 min sample, (a), showed few
residual voids. An AF M scan, (b), of the central region had a mean roughness of 2 nm.

The remnant voids are the darker regions. 60 to 180 min of grth after biasing under

132

the conditions of Table 12 resulted in completely coalesced continuous films with the
occasional void over a few um2 area, (0) and (d). The thickness of a 180 min film (19-
Jun-01) was 0.54 i 0.02 pm obtained from SEM images of a fracture surface. Optical
microscopy showed a uniform, transparent film, with features of the substrate visible
through the diamond film. The bright spots in the SEM images of Figure 72 are small
asperities on the surface; thus, they are related to surface roughness. The AP M scan (b)
in Figure 72 with its vertical scale, indicates that these features are a few nm high.
Charge builds up on these areas as the electron beam scans across the surface and
electron emission is enhanced. Their presence is most noticeable on films of small
thickness with few voids. Voids, which may contain graphitic or other carbon, may shunt
the SEM current to the conducting Ir substrate, reducing the enhanced emission from the

“tip”. This reduces the charge buildup in this region.

9.2.3 Comparison with other results by other groups

Our 60 min films, which exhibit complete coalescence and coverage, may be
contrasted with a 600 nm film on Ir: SrTiO3 reported by Schreck et al. shown in Figure 73

and Figure 74 (8).

133

 

Figure 73 (1a) SEM micrograph of the center of a 173 nm :t 2 nm thick diamond film
on (001) Ir/SrTiO3 grown in this study. Thickness was measured by SEM
from cross-sections (2a) SEM micrograph of a 600 nm thick diamond fihn on
(001)1r/SrTi03, Schreck et al. (8).

 

Figure 74 (1b) SEM micrograph of the edge of the diamond film shown in Figure 73
(1a), 1.5 mm from the center. (2b) Edge of the epitaxial area of the film shown
in Figure 73 (2a) grown by Schreck et al. (8).

Aside from an occasional void, our films show complete coalescence at the center,
whereas thicker films of other groups still show evidence of individual crystallites. At
the edge, we have a few voids, but the scale of features is much smaller than comparable
studies. We believe that the early coalescence of our films is a direct result of the high
nucleation densities reported in the previous chapter. We observed such film growth on

both Ir/SrTi03 and Ir/A12O3 substrates. Subota et al. (6) also reported on a thin diamond

134

 

film grown for 2.5 hr total after 60 min biasing. They did not give the film thickness, and
SEM images show some holes and residual voids. Thinner films (1hr bias and 30 min
growth) show morphology similar to those of Schreck et al. in Figure 73. Ohtsuka et al.
report an average roughness of 1 nm for most of the epitaxial area from AF M for a 1.5
pm thick film (4), as compared to 2 run we measured for a 30 min film. For growth of 30
minutes after biasing, they have a nucleation density of ~10° cm“, and isolated epitaxial

particles are observed (5).

9.2.4 X-ray diffraction studies on 180 minute diamond films
The crystallographic relationship of the diamond, Ir and SrTiO3 was determined

to be (100)dta<1 10>dia||(100)1,<1 10>11r ||(100)s,1~i03<1 10>smo3 by a {111}-<|> scan of a 180

min diamond sample (Figure 75). The crystal was rotated by an angle 0 about the normal
to the (001), and reflections appear at angles corresponding to the expected {111} pole

positions.

135

 

 

 

1o - (111) (ill) (fir) (lit)
105 - i f 3 i -
10‘ -(a) '1 ‘f’

:3 SrTiO3 '3; .

 

 

 

         

 

 

E 106 - l
'E

=1 105 L(b) “
.e ' l

e 10‘ - r i
(U 3

i—l 1 '- v
.2 °,

2 10 -

2 t

5 10

 

 

 

 

 

 

0 A 90 A 180 1 270 360
Phi [degrees]

Figure 75 {111}—11> X-ray scans of 180 min diamond on Ir/SrTiO3. Data taken by
AR. Kortan.

The four {111} reflections of the diamond, It and SrTiO3 fall at the same angles,
indicating in-plane as well as rotational alignment of all three crystals. The equal
intensity of the peaks along 0 requires perfect alignment of the sample in the x-ray beam.

An x-ray reciprocal-space map is shown in Figure 76 for a (OKL) area scan taken using a

136

 

position sensitive detector.

L [STO units]

 

1.8 1.9 2.0 2.1 2.2 2.3
K [STO units]

Figure 76 (OKL) X-ray area scan of a 180 min diamond film on Ir/SrTi03. The central
dark spot is SrTiO3 at (0,2,2), and all others are relative. The dark spot at
(0,205,205) is Ir. Diamond is at (0,2.2,2.2). An unknown phase lies between
SrTiO3 and Ir, aligned with SrTiO3 in the K—direction, and Ir in the L-
direction. Data taken by AR. Kortan.

The intensity contours in reciprocal-space are given relative to SrTiO3, at (0,2,2). The Ir
reflection is at (0,205,205) and diamond (0,2.2,2.2), in the upper right-hand side of the
map. The small spot with the K of SrTiO3 and the L of Ir indicates a stressed phase
between the two that likely forms due to the lattice mismatch. Both the Ir and the
diamond are strained with respect to the SrTiO3 substrate. The asymmetry of the scan
with the spots stretched along a line of constant K=2.03, and along L=constant*K,

indicates strain in these directions.

137

9.2.5 Effect of small angle substrate offcut - Ir/SrTiO3

Diamond was grown on SrTiO3 substrates, offcut with respect to the (100)
direction by 7, 5, 1, and 0.1 deg. Growth is influenced by surface topography; as offcut
angles increase, the step density increases. AFM scans showed mean roughness of 2-3
nm after 150 nm Ir deposition for all samples. Diamond grown on these substrates,
however, exhibited different morphology and roughness, which depended on the SrTiO3
offcut angle. All samples were grown in successive runs under identical CVD reactor
conditions with a 1 hr bias and 1 hr growth. SEM micrographs are shown in Figure 77

and AFM scans in Figure 78.

   

Figure 77 Scanning electron micrographs of diamond on Ir on offcut SrTiO3
substrates. Offcut angle of SrTiO3 (in deg): (a) 0.1, (b) 5, (c) 7. CVD run: (a)
25-Mar-02, (b) 23-Mar-02, (0) 22-Mar-02.

138

 

21.11.111.111
A

 

Figure 78 AFM scans of diamond on Ir on ofi‘cut (001) SrTiO3 substrates. All scans
are 1x1 umz. SrTiO3 offcut angle (in deg): A. 0.1, B. l, C. 5, D. 7. CVD run:
A. 25-Mar-02, B. 24-Mar-02, C. 23-Mar-02, D. 22-Mar-02.

Figure 79 shows the mean roughness of the diamond as a function of the offcut angle of

SrTiO3.
O
I
5 _

a

5

a 4 -

E . '

a, O

8 .

n: 3 -

U) I

2

(I

2 _ I 4x4 um AFM scan
0 1x1 um AFM scan

 

 

1 I I l l I l I I I

-1 0 1 2 3 4 5 6 7 8
Offcut angle of (100) surface (degrees)

Figure 79 RMS roughness of diamond on Ir on offcut (001) SrTiO3 substrates.

139

Roughness increased as the offcut angle increased. 0.1 deg offcut substrates gave

diamond surfaces with the smoothest surfaces.

9.3 Evolution of diamond growth — Ir/Al203

We studied the very early stages of diamond growth on Ir/Al203, as well as
growth times of 5 min to 180 min. As discussed in Chapter 6, both (001)- and (111)-
oriented Ir films can grow on a-plane A1203 substrates, depending on deposition
temperature and possibly substrate offcut angle.

For (100) Ir on A1203, diamond grown on 5 to 180 min samples looked similar to

diamond with those same growth times on Ir/SrTi03 (Figure 80).

    

Figure 80 SEM micrographs of diamond on Ir/Al2O3 grown for 5 min (left) and 180
min (right).

X-ray diffraction rocking curve linewidths of diamond grown for 60 min on Ir/Al2O3
were also indistinguishable from diamond grown on Ir/SrTi03. We studied the initial
growth of diamond on Ir/Al2O3 by rapid quenching after 20 and 90 seconds of grth as
described in Section 8.4.1. SEM micrographs of our observations and the accompanying

2-D power spectra are shown in Figure 81.

140

 

 

-0.20 010 01 0.10 0.20
100nm k,(nm')

Figure 81 SEM scans of short growth on (111) Ir/Al203 and the corresponding 2-D
power spectra.

The SEM micrographs and power spectra revealed a 6—fold symmetry we had not seen on
any previous samples. Post-CVD X-ray diffiaction on these surfaces revealed a (111) Ir
surface, indicating that (111) Ir nucleated on (1 120) A1203. This confirmed the symmetry
seen in the Fourier transforms, and verified the transforms could be a valuable tool in
visualizing angular alignment of crystallites after very short growth periods. The bright
features in SEM micrographs may be evidence for (1 11) diamond deposition on (111) Ir.
If the diamond were further grown under conditions conducive to (111) growth, the fihns

could grow and coalesce similar to those on (001) Ir. We have not investigated further

141

the growth of (1 1 1) diamond in this work, but it is worth noting that these results strongly
suggest it is possible to grow (111) single crystal diamond on Ir.
9.4 T hick films - Ir/SrTi03

Diamond films were grown on Ir/SrTiO3 for periods of 6, 12, 36 and 48 hours.

For thicker films the growth consists of a 2-stage process as outlined in Table 13.

 

 

 

 

 

 

Bias Conditions Growth I Growth II
Bias Voltage -150 V -- --
CH4/H2 ratio 2% 1% 0.5-.75%
Pressure 18 Torr 18 Torr 28 Torr
MW power 1500 W 1500 W 1500 W
Time 60 min 90 min 4.5 - 46.5 hr

 

 

 

 

 

 

 

Table 13 Bias and 2-step growth CVD conditions for thick diamond samples.

After biasing, the samples were grown for 90 min after changing the CH4 flow rate from
6 seem to 3 sccm (2% to 1% CH4zH2). Then, CH4 flow was reduced to 2.25 sccm
(0.75%) and the pressure raised from 18 to 28 Torr by increasing the pressure setpoint by
1 Torr every 3 min.

Subsequent growth occurred under these conditions. SEM and AF M scans of the
resulting films are shown in Figure 82 to Figure 84. The diamond films grown for 6

hours are extremely smooth (Figure 82).

142

’— _

 

If?

 

Figure 82 (a) Typical surface of a diamond film grown for 6 hr on Ir/SrTiO3. The
white speck is a piece of dust. (b) Fractured piece near the edge of the film that
shows a cleaved surface.

6 hr films crack and break into pieces upon cooling in the CVD system due to stress
from the mismatch of thermal expansion coefficients. The above images were made from

remaining pieces of the film.

The 12 hr film cracked in half upon cooling, but otherwise remained intact

(Figure 83).

113nm

 

Figure 83 (a) Edge (approx. 1.5 mm from center) of the 12 hr diamond film on
Ir/SrTiO3. (b) Macro steps at the center of the film scanned with AFM.

AF M and SEM scans revealed that although most of the fihn is smooth and featureless, in
some areas it had macro-steps of average height 20 nm on the surface. Similar macro-step

bunching has been reported in homoepitaxial diamond growth (162, 163). However, it is

143

 

unclear if this is indicative of a 2-D layer-by-layer growth mechanism.

A 48 hr film is shown in Figure 84.

   

Figure 84 (a) A complete diamond film grown for 48 hrs on Ir/SrTiO3. The substrate
broke in half during cooling. (b) Center of the film showing a crack in the
upper right-hand corner, and hillock-type defects.

Films grown for 36 or 48 hours remained intact through the cool-down period, although
sometimes the substrate cracked, as shown above. The films were smooth and optically
transparent. They also had denumerable defects on the surface (~100 cm'z). These could
be due to particulates deposited on the sample during growth, since the defects were not
observed on samples grown for short times. Defects at this density can often be ascribed
to particulates (I 64). They may be similar to defects that form in homoepitaxial growth
under similar methane concentrations (0.5-2%) (163, 165). They should be suppressed
with appropriately modified growth conditions.

An optical micrograph of a cleaved 36 hr freestanding sample is shown in Figure
85. Films grown for longer than 12 hr detached from the SrTiO3 during cooling, with Ir
adhering to the diamond surface. It was removed with an electrochemical etch

comprising 15: 10:1 CaCl2:H2O:HCl (166).

 

Figure 85 Optical micrograph of a single crystal slab of diamond. Cleaved from a 3
mm dia sample, it was originally about 0.01 carat.

The 25 pm thick sample cleaved along the crystal axes and was optically transparent.
The surfaces are as grown. No polishing or thinning was carried out. The piece in Figure

85 is a portion of the original 3 mm sample - approximately 0.01 ct weight.

9.4.1 SEM cross-sections
The thicknesses of the 6 to 48 hr films were measured from SEM images or by
optical microscope analysis of cross-sections. The thickness was measured at 3 to 4

points along a cleaved or fractured edge. The average value is reported in Table 14.

 

 

Total growth time (hours) Thickness (fl!)
6

 

 

 

3.7 :l: 0.1
12 8.5 :l: 0.2
36 25 i 1
48 35 d: 1

 

 

 

 

Table 14 Cross-section thickness measured from SEM or optical micrographs.

145

The 6 hr film thickness measurement was taken from sample pieces closer to the edge of
the film than the center, therefore this measurement is likely to be lower than average.

The average growth rate is estimated at 0.7 urn/hr for these films.

9.4.1.1 Cleavage angle measurement

Single crystals cleave at angles that correspond to crystallographic planes. For
diamond (001), (111) is the preferred cleavage plane. The angle between (111) and (001)
planes is 54.74 deg. Shown in Figure 86 is a 25 um diamond sample with a (111)

cleavage plane.

 

Figure 86 Cross section of 25 um film showing cleavage angle of (111) plane. The
<001> direction is normal to the top surface.

The measured clevage angle for this sample is ISO-126.1 = 53.9 deg. The SEM

micrograph in Figure 87 shows the cross-section (1 11) plane of the same diamond film.

146

 

Figure 87 Fracture surface of a 25 um diamond film revealing the (111) plane.

Although the SEM image looks unremarkable, it should be noted that columnar grth
structure typical of polycrystalline diamond films is not present. Instead, horizontal

“layers” are apparent, particularly near the growth surface of the diamond.

9.4.2 X-ray diffraction

X-ray rocking curves were measured for films grown from 1 to 36 hr on
Ir/SrTiO3. The diamond (004) linewidth as a function of the film thickness is shown in

Figure 88.

147

1.6-

1.4-

1.2-

1.0-

0.8 -
0.6 - \
0.4 - .

0.2

Linewidth (degrees)

 

 

‘

I I I I

o 5 10 15 20 25
Diamond thickness (um)

Figure 88 Thickness dependence of the X-Ray rocking curve linewidth of diamond on
Ir/SrTiO3.

The linewidth of the diamond (004) peak decreases, from 1.5 deg for the 173 nm thick
film, to 0.34 deg for the 25 pm thick film. Schreck et al. reported similar thickness
dependence of the diamond (004) linewidth on Ir/SrTi03, with a value of 0. 1 7° for a 34
um film (7). However, it is not known whether this value was measured for a sample

with the substrate side polished off.

9.4.3 Electron backscattering diffraction

We used electron backscattering diffraction (EBSD) to obtain a {111} pole figure
from a 25 um diamond sample. This was done by scanning the electron beam at 120
separate points in a 10 x 10 um2 area. At each point an EBSD pattern was recorded, and

its coordinates indexed. The orientation at each point is represented by plotting the

positions of its {111} poles in a stereographic projection, which forms the {111} pole

148

 

figure shown in Figure 89. Each point measured in the EBSD has three {111} poles, so

360 poles are plotted in the projection.

 

 

 

Figure 89 {111} pole figure from EBSD measurements on a 10x10um area, 120
separate points. <100> lies at the center of the projection, and the horizontal
and vertical axes are along [110] directions.

If the surface had no texture, the measurements would be uniformly destributed in the
projection. Instead, the measurements are clustered around the {100} poles indicating

the characteristic pattern of a cubic crystal with (100), [001] texture.

9.4.4 Raman microscopy

For a 25 um diamond film, removed from the iridium substrate, we observed the

Raman spectrum in Figure 90.

149

 

 

 

 

 

-l
A 1331cm
.3
c
3
2‘
S
a:
.o
b
33.
a?
in
r:
a)
*0
E
500 Y 1000 ' 1560 ' 2000

Wavenumber shift (cm'1)

 

Figure 90 Ralman spectrum of a 25 pm (001) diamond film on Ir/SrTiO3. Linewidth is
8 cm' .

The diamond peak at 1331 cm'1 had a linewidth of 8 cm". The background that leads to
a shallow maximum near 1500 cm'1 indicates the presence of some disordered carbon
(15, 16). The peak shift from the expected 1332 cm‘1 position is an indication of stress in

the film. To investigate this further, we performed polarized Raman spectroscopy in the

 

(001) backscattering configuration on a 35 pm film. We first measured a reference
sample of type Ila natural diamond. It had a single peak at 1332.3 cm'1 with linewidth

2.1 cm‘1 in the geometry ei||e3||[110]. The spectrum of the sample is shown in Figure 91.

150

800

 

 

 

  
    

 

 

 

 

 

 

 

 

(a) Elllmor. E,ll[1001
700 - (b)Ei||E‘||[010]
k,llk,||[001]
600 -
’1}?
g 500 L ‘.
. (3,1 ‘1 '\ L f' .
,e , ‘1, ommu mmm
s 400 r t" xfl=13307 cm",wl=6.0 cm"
%‘ 300 _ xa=1336.0 cm",w2=4.9 cm"
c
9.
E 200 .-
100 -
o " __..__‘_I

1310 1320 1330 1340 1350 1360
Raman shift (cm")

Figure 91 Polarized Raman spectra for 35 pm film. The arrow points to 1332 cm".

The specturrn obeys the Raman selection rules: for if; and It; parallel to <001>: the

intensity of the scattered beam polarized in the <100> direction, (a), is proportional to the
intensity of the incident light. The intensity polarized along <010>, (b), is due only to
polarization leak, as seen also for the natural diamond. However, the Raman peak of (a)

is split. Lorentz fitting shows two peaks shifted 1.6 cm'1 and 3.7 cm'1 away from 1332.3

 

cm“. The peak shifts are attributed to internal stresses — a compressive biaxial stress in
the (001) plane and a tensile stress along the growth direction <001> (167). The

spectrum was measured for different focus depths through the thickness of the sample

(Figure 92).

151

 

 

 

 

I
s s l s s i r
2 i-
1 EIIOIOI. E,l|l1001
is r
0
E 0 r
.C
In
.2
3 .1
a t"
t r
2" r 9 1 s r
. r
3 l n l n l n l 4 l a l
0 5 10 15 20 25

dstance to surface (110)

Figure 92 Depth profile of Raman peak shift for 35 um film.

In changing the focus depth, only the intensity of the peaks changed, their positions
remained the same, which indicated the stress is uniformly distributed along the <001>

growth direction.

9.5 Summary

We have examined the growth evolution of diamond films on Ir/SrTiO3 and
Ir/Al2O3. The progression of coalescence for thin films was studied using 2-D Fourier
transforms of SEM micrographs with interrupted grth experiments. After 20 minutes
of growth, the remaining grain boundaries showed alignment predominantly along [011]
directions. High nucleation densities initiated by the biasing procedure led directly to the

early coalescence of crystallites on the substrate. The films show complete coalescence

at S 200 nm thickness. Our results may be compared to other reports that show

152

 

 

persistence of individual crystallites for films 2 600 nm thickness. X-ray diffraction, in a

 

{111}-d) scan demonstrated full epitaxial alignment of diamond, Ir and SrTiO3 with the

relation: (100)dia<1 10>dia| I( 100)1,<1 10>], ”(100)3{n03<l 10>SrTiO3-

Single crystal diamond grown heteroepitaxially on Ir/Al203 substrates was

demonstrated for the first time. This is a significant advance towards realization of wafer

 

scale diamond films, since sapphire is more widely available in large area defect-free

form than SrTiO3. We have also demonstrated that (111) diamond epitaxy may be

 

possible on (111) It by investigating the early stages of diamond growth on this surface.
The thick diamond films are single crystals, as evidenced by cleavage, X-ray
diffraction, EBSD and Raman measurements. The thickness dependence of the X-ray
diffraction linewidths agrees qualitatively with an earlier study. We have shown, for the
first time, heteroepitaxial diamond that obeys the Raman selection rules for single crystal
diamond. The peak splitting in the Raman peak indicated the films are under internal
stress. We have demonstrated a uniform film over the entire 3.5 mm diameter area

exposed to the plasma.

 

153

Chapter 10 Conclusions and outlook

In the course of this research, we have demonstrated or observed a number of
unique phenomena while investigating nucleation and growth of heteroepitaxial diamond.
We have demonstrated that epitaxial iridium thin films provide a suitable surface for the
study of diamond nucleation. Our study of the early growth of diamond just after the
onset of nucleation has enabled us to develop a new model of the nucleation of

heteroepitaxial diamond.

10.1 Summary of this research

The following results on the nucleation of heteroepitaxial diamond were
presented:
1. The highest nucleation density reported (3.2 x 10H cm'z) for heteroepitaxial
diamond growth. This number comes from an unambiguous count of particles on
the surface. The evidence of particle coalescence at a very early stage suggested

the actual nucleation density is more likely ~10” cm'2

. High nucleation densities
required optimization of the bias assisted procedure and geometry in a microwave
plasma CVD system.

2. High-resolution SEM images showed that the diamond crystallites in an early

stage of growth display a remarkably narrow distribution of sizes, indicating a

synchronous onset of nucleation. Quantitative analysis using 2-D Fourier

154

transforms and radial distribution plots of particles after a “quenching” process
revealed radial spatial correlations. This implies that there are “effective”
interactions among crystals that destroy randomness.

The evidence of the first two points led us to outline a new model for the
nucleation and early growth of diamond on Ir. The biasing process provides the
necessary conditions for diamond nucleation, possibly as carbon saturation of the
surface. The nucleation event, however, is a result of the quenching process with
the termination of the biasing. High nucleation densities then lead directly to the
early coalescence of the films.

Growth of single crystal diamond on iridium/oxide substrates was also studied in

detail. The films were characterized using SEM, AF M, optical microscopy, Raman

spectroscopy, X-ray diffraction and EBSD. The following aspects of the growth of

heteroepitaxial diamond were presented in this work:

1.

The early stages of diamond heteroepitaxial growth and coalescence. In order to
observe the earliest growth stages we were required to formulate a precise
procedure to ensure repeatability. We performed 2-D Fourier transform analysis
on SEM micrographs taken after short growth periods. The power spectra of the
transforms clearly showed the coarsening and alignment of the diamond films
over relatively short growth times.

By growing for extended periods, to a maximum of 48 hr, diamond plates of
thickness 35 um were produced. Freestanding crystals exhibited (111) cleavage
surfaces, the same as natural diamond, and were transparent in visible light.

Characterization of the diamond by x-ray diffraction, electron backscattering

155

 

 

 

 

 

diffraction, and Raman scattering confirmed the existence of (001) oriented single
crystal diamond. To our knowledge, these are the highest quality thick diamond
films yet produced.

3. The discovery of iridium/sapphire as a new substrate for heteroepitaxial diamond
growth. Our finding that a-plane sapphire can be used to grow (001) diamond

promises to lead to improvements in diamond heteroepitaxy.

10.2 Suggestions for further work

Semiconducting diamond is attractive for applications of high temperature/high
frequency devices because of its superior physical properties and its wide bandgap. This
will require efficient n- and p- type d0ping. Acceptor (p-type) doping of diamond with
boron leads to a level 0.38 eV above the valence band maximum, reasonable for device
purposes. Donor (n-type) doping of diamond at a similar level has not yet been realized,
mostly due to the strain that incorporating large atoms into the diamond lattice
introduces. However, the method of nucleation presented in this thesis, may allow
dopants to be introduced so that they are incorporated readily in the lattice. Electrical
mobility measurements could then be carried out on heteroepitaxial grown films.
Without doping, polycrystalline diamond films have already been demonstrated as
nuclear particle detectors (168-1 70). Radiation hardness and high temperature operation
of such diamond devices is superior to similar silicon detectors. Unfortunately grain
boundaries in polycrystalline diamond lead to carrier recombination and lower efficiency
(169). Single crystal diamond would remove this problem. Photo-excitation of the
diamond films would allow a measurement of the lifetime of free carriers, an important

parameter for photodetectors.

156

 

 

 

An obvious extension of this work would be the reduction of defect densities in
the final thick films, and an understanding of their formation and origins. The defects we
found in our single crystal heteroepitaxial diamond look qualitatively similar to common
defects found in homoepitaxial diamond. This suggests that the grth of the diamond
after a certain thickness proceeds identically to diamond grown on diamond. Related to
this issue is the scalability of the sample size. We suspect that scaling the entire process
will involve careful consideration of the secondary plasma during biasing. Optical
spectroscopy of the secondary plasma during bias may provide some clues to
understanding the specific roles that biasing accomplishes. Study of the secondary
plasma and the scaling may indirectly lead to a better understanding of the nucleation
mechanism.

Finally, the peak splitting of the polarized Raman spectra remains to be studied
and understood. It is clear that the peak splitting must be related to the stress in the films,

but such peak splitting has not been observed before in single crystal diamond films.

157

 

 

References

1. S. Wolter, B. Stoner, J. G. P. Ellis, et al., Textured growth of diamond on silicon via in

situ carburization and bias-enhanced nucleation, Appl. Phys. Lett. 62, 1215
(1993)

N

. X. Jiang and C.-P. Klages, Heteroepitaxial Diamond Growth On (100) Silicon, Diam.
Relat. Mat. 2, 1112 (1993).

3. C. Wild, R. Kohl, N. Herres, et al., Oriented C VD Diamond Films - Twin Formation,
Structure and Morphology, Diam. Relat. Mat. 3, 373 (1994).

4. K. Ohtsuka, H. Fukuda, K. Suzuki, et al., Fabrication of epitaxial Diamond Thin Film
on Iridium, Jpn. J. Appl. Phys. Pt. 2 36, L1214 (1997).

M

. K. Ohtsuka, K. Suzuki, A. Sawabe, et al., Epitaxial growth of diamond on Iridium,
Jpn. J. Appl. Phys. Pt. 2 35, L1072 (1996).

6. T. Tsubota, M. Ohta, K. Kusakabe, et al., Heteroepitaxial growth of diamond on an

iridium (100) substrate using microwave plasma-assisted chemical vapor
deposition, Diam. Relat. Mat. 9, 1380 (2000).

7. M. Schreck, F. Horrnann, H. Roll, et al., Diamond nucleation on iridium bufler layers
and subsequent textured growth: A route for the realization of single-crystal
diamond films, Appl. Phys. Lett. 78, 192 (2001).

8. M. Schreck, H. Roll and B. Stritzker, Diamond/Ir/SrT 1'03: A material combination for
improved heteroepitaxial diamond films, Appl. Phys. Lett. 74, 650 (1999).

9. S. Yugo, T. Kanai, T. Kimura, et al., Generation of diamond nuclei by electric field in
plasma chemical vapor depostion, Appl. Phys. Lett. 58, 1036 (1991).

10. W. Massa, Crystal Structure Determination (Springer-Verlag, Berlin, 2000).

11. B. B. Pate, in Diamond: Electronic Properties and Applications L. S. Pan and D. R.
Kania, Eds. (Kluwer Academic Publishers, Boston/Dordrecht/London, 1996).

12. L. Reimer, Scanning Electron Microscopy. J. Enoch, et al., Eds, Springer Series in
Optical Sciences (Springer-Verlag, Berlin, 1985).

13. S. F legler, J. Heckman and K. Klomparens, Scanning and Transmission Electron

158

 

 

 

 

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

Microscopy: An Introduction (W.H. Freeman and Company, New York, 1993).

B. D. Cullity, Elements of X-ray Difi‘raction. M. Cohen, Ed., Addison-Wesley
Metallurgy Series (Addison-Wesley Publishing Co, Inc, Reading, MA, 1956).

D. S. Knight and W. B. White, Characterization of Diamond Films By Raman-
Spectroscopy, J. Mater. Res. 4, 385 (1989).

J. R. Dennison, M. Holtz and G. Swain, Raman spectroscopy of carbon materials,
Spectroscopy 11, 38 (1996).

R. J. Nemanich, J. T. Glass, G. Lucovsky, et al., Ramon-Scattering Characterization
of Carbon Bonding in Diamond and Diamondlike Thin-Films, J. Vac. Sci.
Technol. A-Vac. Surf. Films 6, 1783 (1988).

M. C. Tobin, Laser Raman Spectroscopy. P. J. Elving and I. M. Kolthoff, Eds,
Chemical Analysis (Wiley-Interscience, New York, 1971), vol. 35.

M. Cardona, in Light Scattering in Solids II M. Cardona and G. Giintherodt, Eds.
(Springer-Verlag, Berlin, 1982), vol. 50, p. 19.

S. A. Solin and A. K. Ramdas, Raman Spectrum of Diamond, Phys. Rev. B 1, 1687
(1970)

R. S. Krishnan, paper presented at the Proc Indian Acad Sci Sect A 1946.

M. A. Washington and H. Z. Cummins, Linewidth of the sharp two-phonon peak in
diamond, Phys. Rev. B 15, 5840 (1977).

C. B. Prater, P. G. Maivald, K. J. Kjoller, et al., “Tapping Mode Imaging:
Applications and Technology” Tech. Report No. AN04 (Digital Instruments,
http://www.di.com).

E. Bauer, Z. Kristallogr. 110, 372 (1958).

J. A. Venables, Introduction to Surface and Thin Film Processes (Cambridge
University Press, Cambridge, 2000).

J. E. Field, in The Properties of Natural and Synthetic Diamond J. E. Field, Ed.
(Academic Press, London, 1992) p. 667.

A. Sawabe and H. Fukuda, Heteroepitaxial growth of diamond, Elec. & Comm.

159

 

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

Japan Pt 2 81, 162 (1998).

J. C. Angus and C. C. Hayman, Low-Pressure, Metastable Growth of Diamond and
Diamondlike Phases, Low-Pressure, Metastable Growth of Diamond and
Diamondlike Phases 241, 913 (1988).

P. May, C VD Diamond - a new technology for the future?, Endeavor Mag. 19, 101
(1995).

B. Spitzyn, L. Bouilov and B. Derjaguin, Vapor growth of diamond on diamond and
other surfaces, J. Cryst. Growth 52, 219 (1981).

A. Sherman, Chemical Vapor Deposition for Microelectronics; Principles
Technology, and Applications. R F. Bunshah and G. E. McGuire, Eds., Materials
Science and Process Technology Series (Noyes Publications, Park Ridge, NJ,
1987).

B. Lux and R. Haubner, in Diamond and diamond-like films and coatings R. E. e. a.
Clausing, Ed. (Plenum Press, New York, 1991) p. 579.

L. S. G. Plano, in Diamond: Electronic Properties and Applications L. S. Pan and D.
R. Kania, Eds. (Kluwer Academic Publishers, Boston/Dordrecht/London, 1996).

P. K. Bachmann, G. Gartner and H. Lydtin, Plasma-assisted Chemical Vapor
Deposition Processes, MRS Bull. 13, 51 (1988).

W. G. Eversole, US Patent No. 3030187, 3030188 (1962).

D. J. Poferl, N. C. Gardner and J. C. Angus, Growth of Boron-Doped Diamond Seed
Crystals by Vapor-Deposition, J. Appl. Phys. 44, 1428 (1973).

J. C. Angus, H. A. Will and W. S. Stanko, Growth of Diamond Seed Crystals by
Vapor Deposition, J. Appl. Phys. 39, 2915 (1968).

M. Kamo, Y. Sato, S. Matsumoto, et al., Diamond Synthesis fiom Gas-Phase in
Microwave Plasma, J. Cryst. Growth 62, 642 (1983).

S. Matsumoto, Y. Sato, M. Tsutsumi, et al., Growth of Diamond Particles from
Methane-Hydrogen Gas, J. Mater. Sci. 17, 3106 (1982).

S. Matsumoto, Y. Sato, M. Kamo, et al., Vapor-Deposition of Diamond Particles
from Methane, Jpn. J. Appl. Phys. Part 2 - Lett. 21, L183 (1982).

160

 

 

 

41

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

N. Ohtake and M. Yoshikawa, Eflects of Oxygen Addition on Growth of Diamond
Film by Arc- Discharge Plasma-Jet Chemical— Vapor-Deposition, Jpn. J. App.
Phys. Pt. 1 32, 2067 (1993).

S. Matsumoto, Y. Manabe and Y. Hibino, Diamond Deposition Using an X-Y Stage
in a Dc Plasma-Jet Chemical Vapor-Deposition, J. Mater. Sci. 27, 5905 (1992).

S. Matsumoto, M. Hino and T. Kobayashi, Synthesis of Diamond Films in a Rf
Induction Thermal Plasma, Appl. Phys. Lett. 51, 737 (1987).

H. V. Boenig, Plasma Science and Technology (Cornell University Press, London,
1982).

M. A. Lieberman and A. J. Lichtenberg, Principles of Plasma Discharges and
Materials Processing (John Wiley & Sons, Inc., New York, 1994).

W. Kalss, R. Haubner and B. Lux, Diamond deposition on noble metals, Diam.
Relat. Mat. 6, 240 (1997).

T. Suzuki and A. Argoitia, Current status of heteroepitaxy of C VD diamond, Phys.
Status Solidi A 154, 239 (1996).

S. T. Lee, Z. D. Lin and X. Jiang, C VD diamond films: nucleation and growth,
Mater. Sci. Eng. R-Rep. 25, 123 (1999).

S. Koizumi, T. Murakami, T. Inuzuka, et al., Epitaxial-Growth of Diamond Thin-
Films 0n Cubic Boron Nitride(111) Surfaces By DC Plasma Chemical Vapor-
Deposition, Appl. Phys. Lett. 57, 563 (1990).

M. Yoshikawa, H. Ishida, A. Ishitani, et al., Study of Crystallographic Orientations

in the Diamond Film 0n Cubic Boron-Nitride Using Raman Microprobe, Appl.
Phys. Lett. 57, 428 (1990).

T. Inuzuka, S. Koizumi and K. Suzuki, Epitaxial growth of diamond thin films on
foreign substrates, Diam. Relat. Mat. 1, 175 (1992).

P. C. Yang, C. A. Wolden, W. Liu, et al., Coalesced oriented diamond films on
nickel, J. Mater. Res. 13, 1120 (1998).

Z. Sitar, W. Liu, P. C. Yang, et al., Heteroepitaxial nucleation of diamond on nickel,
Diam. Relat. Mat. 7, 276 (1998).

T. Tachibana, Y. Yokota, K. Nishimura, et al., Heteroepitaxial diamond growth on

161

 

 

 

 

platinum(111) by the Shintani process, Diam. Relat. Mat. 5, 197 (1996).

55. X. Jiang, K. Schiffmann, C. P. Klages, et al., Coalescence and overgrowth of
diamond grains for improved heteroepitaxy on silicon(001), J. Appl. Phys. 83,
2511(1998)

56. H. Kawarada, T. Suesada and H. Nagasawa, Heteroepitaxial Growth of Smooth and
Continuous Diamond Thin Films 0n Silicon Substrates Via High-Quality Silicon-
Carbide Bufler Layers, Appl. Phys. Lett. 66, 583 (1995).

57. H. Kawarada, C. Wild, N. Herres, et al., Heteroepitaxial growth of highly oriented
diamond on cubic silicon carbide, J. Appl. Phys. 81, 3490 (1997).

58. Y. Yokota, T. Tachibana, K. Miyata, et al., Heteroepitaxial diamond growth on
platinum(111) by Shintani process, Diam. Films Technol. 6, 165 (1996).

59. T. Tachibana, Y. Yokota, K. Kobashi, et al., Heteroepitaxial growth of {1 I 1}-
oriented diamond films on platinum {I I 1}/sapphire{0001} substrates, J. Cryst.
Growth 205, 163 (1999).

60. M. Schreck, H. Roll, J. Michler, et al., Stress distribution in thin heteroepitaxial

diamond films on Ir/SrTi03 studied by x—ray dr'ffi‘action, Raman spectroscopy, and
finite element simulations, J. Appl. Phys. 88, 2456 (2000).

61. R. 0. Jones and O. Gunnarsson, The density fitnctional formalism, its applications
and prospects, Rev. Mod. Phys. 61, 689 (1989).

62. S. Koizumi, Heteroepitaxy of diamond on cubic boron nitride, Isr. J. Chem. 38, 33
(1998)

63. W. Yarbrough and R. Messier, Current Issues and Problems in the chemical vapor
deposition of diamond, Science 247, 688 (1990).

64. T. Taniguchi and S. Yamaoka, Spontaneous nucleation of cubic boron nitride single
crystal by temperature gradient method under high pressure, J. Cryst. Growth
222, 549 (2001).

65. P. C. Yang, R. Schlesser, C. A. Wolden, et al., Control of diamond heteroepitaxy on
nickel by optical reflectance, Appl. Phys. Lett. 70, 2960 (1997).

66. Y. Shintani, Growth of highly (111)-oriented, highly coalesced diamond films on
platinum(111) surface: A possibility of heteroepitaxy, J. Mater. Res. 11, 2955
(1996)

162

 

 

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

79.

T. Tachibana, Y. Yokota, K. Miyata, et al., Diamond films heteroepitaxially grown
on platinum (111), Phys. Rev. B 56, 15967 (1997).

H. Nagasawa, Y. Yamaguchi, T. Izumi, et al., Heteroepitaxial growth and ESR
evaluation of 3C-SiC, Appl. Surf. Sci. 70/71, 542 (1993).

M. Hansen, Constitution of Binary Alloys. R. Mehl, Ed., Metallurgy and

Metallurgical Engineering Series (McGraw-Hill Book Co., New York, Toronto,
London, ed. Second, 1958).

C. Bednarski, Z. Dai, A.-P. Li, et al., Studies of Heteroepitaxial Growth of Diamond
paper presented at the 13th European Conference on Diamond, Diamond-like
Materials, Carbon Nanotubes, Nitrides and Silicon Carbide, Granada, Spain 2002.

K. Suzuki, H. Fukuda, T. Yamada, et al., Epitaxially grown free-standing diamond
platelet, Diam. Relat. Mat. 10, 2153 (2001).

K. Suzuki, H. Fukuda and A. Sawabe, Epitaxial growth of diamond thin films on
iridium surfaces paper presented at the The 5th NTRIM International Symposium
on Advanced Materials (ISAM '98), Tsukuba, Japan 1998.

T. Saito, S. Tsuruga, N. Ohya, et al., Epitaxial nucleation of diamond on iridium
substrate by bias treatment, for microwave plasma-assisted chemical vapor
deposition, Diam. Relat. Mat. 7, 1381 (1998).

F. Htirmann, M. Schreck and B. Stritzker, First stages of diamond nucleation on
iridium bufler layers, Diam. Relat. Mat. 10, 1617 (2001).

Z. Dai, A.-P. Li, C. Bednarski, et al., Epitaxial Iridium Growth on Strontium
Titanate, Mater. Res. Soc. Symp. Proc. 648, 1135.1 (2001).

M. Schreck and B. Stritzker, Nucleation and growth of heteroepitaxial diamond films
on silicon, Phys. Status Solidi A 154, 197 (1996).

P. A. Dennig and D. A. Stevenson, Influence of Substrate Topography 0n the
Nucleation of Diamond Thin-Films, Appl. Phys. Lett. 59, 1562 (1991).

P. Karve, S. R. Sainkar and S. T. Kshirsagar, Role of surface features in C VD
diamond nucleation on surface- pretreated substrates, Mater. Lett. 34, 387
(1998).

B. Stoner and J. Glass, Textured diamond growth on (100) ,B-SiC via microwave
plasma chemical vapor deposition, Appl. Phys. Lett. 60, 698 (1992).

163

 

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

90.

91.

92.

X. J iang, C. P. Klages, Z. R., et al., Epitaxial Diamond T hin-F ilms On (001) Silicon
Substrates, Appl. Phys. Lett. 62, 3438 (1993).

Y. Ma, T. Tsurumi, N. Shinoda, et al., Effect of bias enhanced nucleation on the

nucleation density of diamond in microwave plasma C VD, Diam. Relat. Mat. 4,
1325 (1995).

M. Katoh, M. Aoki and H. Kawarada, Plasma-Enhanced Diamond Nucleation on Si,
Jpn. J. Appl. Phys. Part 2 - Lett. 33, L194 (1994).

M. J. Chiang and M. H. Hon, Enhanced nucleation of diamond films assisted by
positive dc bias, Thin Solid Films 389, 68 (2001).

R. 1. Cherry and T. Whitrnore, Experimental Apparatus for Determining Charge-
Exchange between Hydrogen and Hydrocarbon Groups in the Hot-Filament
Deposition of Diamond, Appl. Phys. Lett. 67, 3040 (1995).

Q. Chen and Z. Lin, Experimental approach to the mechanism of the negative bias
enhanced nucleation of diamond on Si via hot filament chemical vapor deposition,
J. Appl. Phys. 80, 797 (1996).

K. Pemg, K. S. Liu and I. N. Lin, Evidence of electron-emission-enhanced

nucleation of diamonds in microwave plasma-enhanced chemical vapor
deposition, Appl. Phys. Lett. 79, 3257 (2001).

A. F liiter, G. Schaarschmidt, B. Mainz, et al., Deposition of Diamond on Patterned
Silicon Substrates, Diam. Relat. Mat. 4, 930 (1995).

X. J iang, Eflects of ion bombardment on the nucleation and growth of diamond films,
Phys. Rev. B 58, 7064 (1998).

M. Y. Liao, F. G. Qin, J. H. Zhang, et al., Ion bombardment as the initial stage of
diamond film growth, J. Appl. Phys. 89, 1983 (2001).

J. Yang, X. W. Su, Q. J. Chen, et al., Si + Implantation - a Pretreatment Method For
Diamond Nucleation On a Si Wafer, Appl. Phys. Lett. 66, 3284 (1995).

J. Robertson, J. Gerber, S. Sattel, et al., Mechanism of bias-enhanced nucleation of
diamond on Si, Appl. Phys. Lett. 66, 3287 (1995).

J. Gerber, S. Sattel, H. Ehrhardt, et al., Investigation of bias enhanced nucleation of
diamond on silicon, J. Appl. Phys. 79, 4388 (1996).

164

 

 

 

93.

94.

95.

96.

97.

98.

99.

100

101

102

103

104

105

M. Y. Liao, X. M. Meng, X. T. Zhou, et al., Nanodiamond formation by hot-filament
chemical vapor deposition on carbon ions bombarded Si, J. Cryst. Growth 236,
85 (2002).

Y. Lifshitz, S. R. Kasi and J. W. Rabalais, Subplantation model for film growth from
hyperthermal species: Application to diamond, Phys. Rev. Lett. 62, 1290 (1989).

S. Uhlmann, T. Frauenheim and Y. Lifshitz, Molecular—dynamics study of the
fimdamental processes involved in subplantation of diamondlike carbon, Phys.
Rev. Lett. 81, 641 (1998).

J. Robertson, The Deposition Mechanism of Diamond-Like a-C and a-C-H, Diam.
Relat. Mat. 3, 361 (1994).

J. Robertson, Deposition Mechanisms for Promoting Sp(3) Bonding in Diamond-
Like Carbon, Diam. Relat. Mat. 2, 984 (1993).

H. P. Kaukonen and R. M. Nieminen, Molecular-Dynamics Simulation of the Growth
of Diamond-Like Films by Energetic Carbon-Atom Beams, Phys. Rev. Lett. 68,
620 (1992).

S. Katai, Z. Tass, G. Hars, et al., Measurement of ion energy distributions in the bias
enhanced nucleation of chemical vapor deposited diamond, J. Appl. Phys. 86,
5549 (1999).

. S. Katai, A. Kovats, I. Maros, et al., Ion energy distributions and their evolution
during bias- enhanced nucleation of chemical vapor deposition of diamond,
Diam. Relat. Mat. 9, 317 (2000).

. Y. Kim and J. Lee, Formation of highly oriented diamond film on (100) silicon,
Formation of highly oriented diamond film on (100) silicon 81, 3660 (1997).

. Y. Kim, Y. Han and J. Lee, The effects of a negative bias on the nucleation of
oriented diamond on Si, Diam. Rel. Mater. 7, 96 (1998).

. S. P. McGinnis, M. A. Kelly, S. B. Hagstrom, et al., Observation of diamond
nanocrystals in carbon films deposited during ion-assisted microwave plasma
nucleation pretreatments, J. Appl. Phys. 79, 170 (1996).

. L. Hanley and S. B. Sinnott, The growth and modification of materials via ion-
surface processing, Surf. Sci. 500, 500 (2002).

. S. McGinnis, M. Kelly and S. Hagstroem, Evidence of an energetic ion

165

 

 

 

106.

107.

108.

109.

110.

111.

112.

113.

114.

115.

116.

117.

bombardment mechanism for bias enhanced nucleation of diamond, Appl. Phys.
Lett. 66, 3117 (1995).

B. Sheldon, R. Csencsits, J. Rankin, et al., Bias-enhanced nucleation of diamond

during microwave-assisted chemical vapor deposition, J Appl. Phys. 75, 5001
(1994).

Y. Shigesato, R. E. Boekenhauer and B. W. Sheldon, Emission-Spectroscopy
During Direct-Current-Biased, Microwave- Plasma Chemical- Vapor-Deposition
of Diamond, Appl. Phys. Lett. 63, 314 (1993).

 

J. C. Angus, H. A. Will and W. S. Stranko, J. Appl. Phys. 39, 2915 (1968).

N. Setaka, Diamond Synthesis from Vapor-Phase and Its Growth-Process, J. Mater.
Res. 4, 664 (1989).

 

J. C. Angus, A. Argoitia, R. Gat, et al., Chemical Vapor-Deposition of Diamond,
Philos. Trans. R. Soc. Lond. Ser. A-Math. Phys. Eng. Sci. 342, 195 (1993).

X. Jiang, K. Schiffmann and C. Klages, Nucleation and initial growth phase of
diamond thin films on (100) silicon, Phys. Rev. B 50, 8402 (1994).

P. Ascarelli, E. Cappelli and F. Pinzari, Nearest-neighbor distance distribution of
diamond nuclei on substrate surfaces, Appl. Phys. Lett. 70, 1697 (1997).

K. Pemg, K. S. Liu and I. N. Lin, Study on Bias-Enhanced Nucleation of Diamonds
by Simulating the Time Dependence of Bias Current, J. Appl. Phys. 91, 3934
(2002).

B. Stoner, G. Ma, S. Wolter, et al., Characterization of bias-enhanced nucleation of

diamond on silicon by the in vacuo surface analysis and transmission electron
microscopy, Phys. Rev. B 45, 11067 (1992).

 

T. Suesada, N. Nakamura, H. Nagasawa, et al., Initial Growth of Heteroepitaxial

Diamond on Si(001) Substrates Via ,B-SiC Bufi'er Layer, Jpn. J. Appl. Phys. Pt. 1
34, 4898 (1995).

E. Maillard-Schaller, O. M. Kuttel, P. Groning. et al., Local heteroepitaxy of
diamond on silicon (100): A study of the interface structure, Phys. Rev. B 55,
15895(1997)

M. Stammler, R. Stockel, L. Ley, et al., Diamond nucleation on silicon during bias
treatment in chemical vapour deposition as analysed by electron microscopy,

166

Diam. Relat. Mat. 6, 747 (1997).

118. R. Stockel, M. Stammler, K. Janischowsky, et al., Diamond Nucleation under Bias
Conditions, J. Appl. Phys. 83, 531 (1998).

119. R. Stockel, K. Janischowsky, S. Rohmfeld, et al., Growth of diamond on silicon
during the bias pretreatment in chemical vapor deposition of polycrystalline
diamond films, J. Appl. Phys. 79, 768 (1996).

120. R. Stockel, K. Janischowsky, s. Rohmfeld, et al., Diamond growth during bias pre- ‘
treatment in the microwave CVD of diamond, Diam. Relat. Mat. 5, 321 (1996).

121. B. R. Stoner, G. H. Ma, S. D. Wolter, et al., Epitaxial Nucleation of Diamond 0n ,6»
SiC Via Bias-Enhanced Microwave Plasma Chemical- Vapor-Deposition, Diam.
Relat. Mat. 2, 142 (1993). 4

 

122. W. L. Wang, G. Sanchez, M. C. Polo, et al., Nucleation and initial growth of

diamond by biased hot filament chemical vapour deposition, Appl. Phys. A. 65,
241 (1997).

123. W. L. Wang, K. J. Liao, J. Wang. et al., Mechanism of diamond nucleation

enhancement by electron emission via hot filament chemical vapor deposition,
Diam. Relat. Mat. 8, 123 (1999).

124. G. Sanchez, W. L. Wang, M. C. Polo, et al., Nucleation of diamond on silicon by
biased HF C VD: a comparative study, Diam. Relat. Mat. 7, 200 (1998).

125. M. Schreck, T. Bauer and B. Stritzker, Optical characterization of the cathode

plasma sheath during the biasing step for diamond nucleation on silicon, Diam.
Relat. Mat. 4, 553 (1995).

126. W. Kulisch, L. Ackermann and B. Sobisch, 0n the mechanisms of bias enhanced
nucleation of diamond, Phys. Status Solidi A 154, 155 (1996).

127. D. K. Milne, P. G. Roberts, P. John, et al., Epitaxy of Diamond on Silicon, Diam.
Relat. Mat. 4, 394 (1995).

128. R. Beckmann, B. Sobisch, W. Kulisch, et al., Investigation of the Bias Nucleation
Process in Microwave Plasma-Enhanced Chemical- Vapor-Deposition of
Diamond, Diam. Relat. Mat. 3, 555 (1994).

129. M. Whitfield, R. Jackman and J. Foord, Spatially resolved optical emission
spectroscopy of the secondary glow observed during biasing of a microwave

167

 

plasma, Vacuum S6, 15 (2000).

130. M. D. Whitfield, J. S. Foord, J. A. Savage, et al., Characterisation of the secondary
glow region of a biased microwave plasma by optical emission spectroscopy,
Diam. Relat. Mat. 9, 305 (2000).

131. S. Barrat, S. Saada, I. Dieguez, et al., Diamond deposition by chemical vapor
deposition process: Study of the bias enhanced nucleation step, J. Appl. Phys. 84,
1870 (1998).

132. K. H. Thfirer, M. Schreck and B. Stritzker, Limiting processes for diamond '
epitaxial alignment on silicon, Phys. Rev. B 57, 15454 (1998).

133. F. Hormann, H. R011, M. Schreck, et al., Epitaxial Ir layers on SrTiO3 as substrates
for diamond nucleation: deposition of the films and modification in the C VD
environment, Diam. Relat. Mat. 9, 256 (2000).

 

134. F. Hormann, H. Y. Peng, T. Bauer, et al., Flat epitaxial diamond/Ir(001) interface
visualized by high resolution transmission electron microscopy, Surf. Sci. 513,
525(2002)

135. A. Sawabe, H. Fukuda, T. Suzuki, et al., Interface between C VD diamond and
iridium films, Surf. Sci. 467, L845 (2000).

136. H. L. Shriver and N. M. Rosengaard, Surface energy and work firnction of
elemental metals, Phys. Rev. B 46, 7157 (1992).

137. R. Vargas, T. Goto, W. Zhang. et al., Epitaxial-Growth of Iridium and Platinum

Films 0n Sapphire By Metalorganic Chemical— Vapor-Deposition, Appl. Phys.
Lett. 65, 1094 (1994).

138. Y. M. Sun, J. P. Endle, K. Smith, et al., Iridium film growth with iridium tris-
acetylacetonate: oxygen and substrate eflects, Thin Solid Films 346, 100 (1999).

139. S. K. Dey, J. Goswami, C. G. Wang, et al., Preparation of iridium films by liquid
source metalorganic chemical vapor deposition, Jpn. J. Appl. Phys. Part 2 - Lett.
38, L1052 (1999).

140. M. A. El Khakani, M. Chaker and B. Le Drogoff, Iridium thin films deposited by
radio-frequency magnetron sputtering, J. Vac. Sci. Technol. A-Vac. Surf. Films
16, 885 (1998).

141. M. A. El Khakani, B. Le Drogoff and M. Chaker, Efl'ect of the deposition

168

142

143

144

145

146

147

148

149

150

151

152

153

temperature on the properties of iridium thin films grown by means of pulsed
laser deposition, J. Mater. Res. 14, 3241 (1999).

. B. Le Drogoff, M. A. El Khakani, P. R. M. Silva, et al., Surface properties of
pulsed laser deposited Ir, Rh, and Iratha 1 thin films for use as microelectrode

arrays in electroanalytical heavy metal trace sensors, Appl. Surf. Sci. 152, 77
(1999)

. J. Heiderg, R. Daghighi-Ruhi, H. V. Weyssenhoff, et al., Laser-stimulated Noble
Metal Deposition from Gaseous and Condensed Acetylacetonates, Mater. Res.
Soc. Symp. Proc. 101, 221 (1998).

. M. Kawasaki, A. Ohtomo, T. Arakane, et al., Atomic control of SrT i03 surface for
perfect epitaxy of perovskite oxides, Appl. Surf. Sci. 107, 102 (1996).

. G. Koster, B. L. Kropman, G. Rijnders, et al., Quasi-ideal strontium titanate crystal
surfaces through formation of strontium hydroxide, Appl. Phys. Lett. 73, 2920
(1998).

. M. L. Kronberg, Acta Metall. 5, 507 (1957).

. D. Doppalapudi, E. Iliopoulos, S. N. Basu, et al., Epitaxial growth of gallium
nitride thin films on A-plane sapphire by molecular beam epitaxy, J. Appl. Phys.
85, 3582 ( 1999).

. J. Bai, T. Wang, H. D. Li, et al., (0001 ) oriented GaN epilayer grown on (1 1-20)
sapphire by MOC VD, J. Cryst. Growth 231, 41 (2001).

. Y. Kim, K. Lee and J. Lee, Deposition of heteroepitaxial diamond film on (100)
silicon in the dense plasma, Appl. Phys. Lett. 68, 756 (1996).

. X. Jiang and C. P. Klages, Recent developments in heteroepitaxial nucleation and
growth of diamond on silicon, Phys. Status Solidi A 154, 175 (1996).

. S. Wolter, T. Borst, A. Vescan, et al., The nucleation of highly oriented diamond on
silicon via an alternating current substrate bias, Appl. Phys. Lett. 68, 3558
(1996)

. K. P. Kuo and J. Asmussen, An experimental study of high pressure synthesis of
diamond films using a microwave cavity plasma reactor, Diam. Relat. Mat. 6,
1097 (1997).

. W. Rasband, Image/J, ver '1.28', National Institutes of Health, (available for

169

154

155

156

157

158

159

160

161

162

163

164

165

166

download at http://rsb.info.nih.gov/ij)

. D. H. Trevena, The Liquid Phase. N. Mott and G. R. Noakes, Eds., Wykeham
Science Series (Wykeham Publications Ltd., London, 1975).

. J. P. Hansen and I. R. McDonald, Theory of Simple Liquids (Academic Press,
London, ed. 2nd, 1986).

. G. H. A. Cole, An Introduction to the Statistical Theory of Classical Simple Dense
Fluids (Pergamon Press, Oxford, 1967).

. C. Ratsch, A. Zangwill, P. Smilauer, et al., Saturation and Scaling of Epitaxial
Island Densities, Phys. Rev. Lett. 72, 3194 (1994).

. C. Pearson, M. Krueger and E. Ganz, Direct Tests of Microscopic Growth Models

using Hot Scanning Tunneling Microscopy Movies, Phys. Rev. Lett. 76, 2306
(1996).

. J. A. Stroscio and D. T. Pierce, Scaling of diflussion-mediated island growth in
iron-on-iron homoepitaxy, Phys. Rev. B 49, 8522 (1994).

. G. S. Bales and D. C. Chrzan, Dynamics of Irreversible Island Growth during
Submonolayer Epitaxy, Phys. Rev. B 50, 6057 (1994).

. B. Voigtlander, T. Weber, P. Smilauer, et al., Transition fiom Island Growth to
Step-F low Growth for Si/Si(100) Epitaxy, Phys. Rev. Lett. 78, 2164 (1997).

. K. Hayashi, S. Yamanaka, H. Okushi, et al., Homoepitaxial diamond films with
large terraces, Appl. Phys. Lett. 68, 1220 (1996).

. K. Hayashi, S. Yamanaka, H. Watanabe, et al., Diamond films epitaxially grown by
step-flow mode, J. Cryst. Growth 183, 338 (1998).

. R. Hull and J. C. Bean, in Strained-Layer Superlattices: Materials Science and
Technology T. P. Pearsall, Ed. (Academic Press, Inc., Boston, 1991), vol. 33, p. 1.

. H. Watanabe, D. Takeuchi, S. Yamanaka, et al., Homoepitaxial diamond film with
an atomicallyflat surface over a large area, Diam. Relat. Mat. 8, 1272 (1999).

. J. Lindahl, T. Takanen and L. Montelius, Easy and reproducible method for making
sharp tips of Pt/Ir, J. Vac. Sci. Technol. B 16, 3077 (1998).

170

167. A.-P. Li, Unpublished.

168. L. S. Pan, S. Han, D. R. Kania, et al., Particle - and photoinduced conductivity in
type-Ila diamonds, J. Appl. Phys. 74, 1086 (1993).

169. M. A. Plano, S. Zhao, C. F. Gardinier, et al., Thickness Dependence of the
Electrical Characteristics of Chemical- Vapor-Deposited Diamond Films, Appl.
Phys. Lett. 64, 193 (1994).

170. M. Marinelli, E. Milani, A. Paoletti, et al., High-quality diamond grown by

chemical-vapor deposition: Improved collection efficiency in alpha-particle
detection, Appl. Phys. Lett. 75, 3216 (1999).

171

1

Hi“

1!
02

111111311

3