Sei sulla pagina 1di 32

Materials Science and Engineering, 25 (1999) 123±154

CVD diamond films: nucleation and growth


S.-Tong Leea,*, Zhangda Linb, Xin Jiangc
Center Of Super-Diamond & Advanced Films and Department of Physics & Materials Science, City University of Hong
Kong, Hong Kong, China
State Key Laboratory of Surface Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100080, China
Fraunhofer-Institut fuÈr Schicht- und Oberflaechentechnik, Bienroder Weg 54E, D-38108 Braunschweig, Germany

Received 21 April 1999; accepted 10 June 1999

Abstract
In the last decade, we have seen rapid developments in metastable diamond synthesis by means of low-
pressure chemical vapor deposition. Concurrently, a fast growing interest in diamond technology has emerged. This
review discusses the various low-pressure growth methods of diamond films. Particular attention is paid to recent
advances in the understanding of the mechanism of diamond nucleation and metastable growth. These advances are
discussed in connection with the advances in diamond heteroepitaxy, which raises hopes that single crystalline
diamond films are not far beyond reach. Modern surface science techniques applied to diamond study have played
an essential role in these achievements and their contributions are discussed. # 1999 Elsevier Science S.A.
All rights reserved.

1. Introduction

Diamond is considered an ideal material for many applications [1±6]. Its structure belongs to the
space group O7h (F4, /d 32/m) with two atoms per primitive (Bravais) cell. The structure shown in
Fig. 1 can be viewed as two interpenetrating face centered cubic lattices shifted along the body
diagonal by (1/4, 1/4, 1/4)a, where a is the dimension of the cubic (mineralogical) unit cell. Each
carbon atom has a tetrahedral configuration consisting of sp3 hybrid atomic orbitals. The {1 1 1}
crystallographic plane comprises 6-atom hexagonal rings arranged so that the adjacent atoms are
alternately dislocated upward and downward from the plane. The stacking sequence in the h111i
directions is ABC ABC ABC. . .. The lattice constant is 3.56 A Ê and the bond length is 1.54 A Ê.
Natural diamond consists of 98.9% 12C and 1.1% 13C. The characteristic Raman spectroscopic
signals for diamond are 1332 cmÿ1 for 12C and 1284 cmÿ1 for 13C. Diamond has two isomers. The
first isomer is lonsdaleite found in meteorites. The structure of lonsdaleite is derived from diamond
as shown in Fig. 2. The positioning of atoms in each plane is the same as that in the cubic structure.
However, the planes are linked in a manner which results in a stacking sequence of AB AB AB. . ..
Consequently, the atoms experience closer chemical bonding, with lattice constants in the a and c
directions of 2.52 and 4.12 A Ê respectively. The distance between adjacent atoms is 1.52 A Ê . The
ÿ1
corresponding Raman peak is in the range of 1315±1325 cm . Another isomer is graphite, the most
common form of carbon. Each carbon atom has a sp2 atomic configuration and therefore, three in-
plane sigma bonds. The remaining valence electron forms  bonds using a pz atomic orbital. Thus,

*Corresponding author.

0927-796X/99/$ ± see front matter # 1999 Elsevier Science S.A. All rights reserved.
PII: S 0 9 2 7 - 7 9 6 X ( 9 9 ) 0 0 0 0 3 - 0
124 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 1. Face-centered cubic structure of the diamond crystal.

the trigonally bonded 6-carbon rings are situated in a flat plane instead of being alternately plaited as
in diamond. The planes are layered in an ABABAB. . . sequence. The lattice constant in the basal
plane between repeating layers is 6.707 AÊ , and the in-plane, nearest neighbor spacing is 1.42 A Ê . The
ÿ1
signature Raman peak of the in-plane layers is 1580 cm .
In this review we will briefly discuss diamond properties, diamond applications, and diamond
growth by chemical vapor deposition in Sections 2 and 3. Special attention is paid to recent advances
in the understanding of the mechanism of diamond nucleation, metastable growth, as well as
heteroepitaxy in Sections 4, 5 and 6. Modern surface science techniques applied to diamond study
has played an essential role in these achievements and their contributions will be discussed in
Section 7.

Fig. 2. Crystal structure of cubic and hexagonal diamonds. The difference in stacking sequence of (1 1 1) layer pairs in two
structures has been illustrated. Differences in the atomic arrangements are highlighted by the darkened bonds.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 125

2. Properties and applications of diamond and diamond films

2.1. Properties of diamond

By virtue of its very strong chemical bonding, the structure of diamond leads to special
mechanical and elastic properties. The hardness, molar density, thermal conductivity, sound velocity,
and elastic module of diamond are the highest of all known materials while its compressibility is the
lowest of all materials (Ref. [4], pp. 8±9).
Diamond also has the largest Young's modulus among all materials. The dynamic friction
coefficient of diamond is only 0.05, as low as that of teflon and the lowest among the materials of
interest. Diamond exhibits the highest longitudinal sound velocity among all materials.
Diamond possesses the highest thermal conductivity ever known (Ref. [4], p. 4). Even so, its
thermal conductivity would increase five times more if diamond were to be fabricated using
isotopically pure carbon on account of decreased phonon scattering by differing isotopes [7,8].
The most important optical properties are the refraction index and optical absorption for a given
wavelength and temperature. Diamond has an appropriate refraction index and a small absorption
coefficient of light from infrared to ultraviolet region [4].
The Hall mobility of holes in natural diamond is 1800 cm2/v.s, and that of electrons can reach
2000 cm2/v.s. For synthetic homoepitaxial diamond, a hole mobility as high as 1400 cm2/v.s has
been obtained. For natural diamond, the hole and electron carrier drift velocities start to saturate at an
electric field strength of 104 V/cm. The saturation velocity is 107 cm/s for holes, and 2.0  107 cm/s
for electrons. Also, its electrical resistivity can reach 1015
cm.
The dielectric loss tangent is an important parameter for applications in microwave and
millimeter wave. Diamond possesses the lowest loss tangent among the compared materials (Ref.
[4], p. 13±14).
The chemical properties of diamond have been reviewed in Ref. [9]. Diamond does not react to
common acids even at elevated temperatures [3]. Treated by a hot chromic acid cleansing mixture or
a mixture of sulfuric and nitric acids, graphite slowly oxidizes while diamond is chemically inert.
However, diamond oxidizes (graphitizes) readily at high temperatures in an oxygen atmosphere and
in air. Also, molten hydroxides, the salts of oxy-acids, and some metals (Fe, Ni, Co etc.) have a
corrosive effect on diamond.
At temperatures above 870 K, diamond reacts with water vapor and CO2 [3]. The oxidation of
diamond in potassium-containing liquid salts is twice as rapid as etching by sodium-containing salts
[10]. Diamond may either chemically react with metals to form carbides, or dissolve in the metals.
Metals such as tungsten, titanium, tantalum, and zirconium react with diamond to produce carbides
at high temperatures, while iron, cobalt, nickel, manganese, and chromium dissolve diamond.
Because diamond dissolves in and/or reacts with iron or iron alloys (e.g. steels) at temperatures
above 950 K, diamond tools are unsuitable for most machining operations on ferrous metals,
including high speed and hardened steels.

2.2. Applications of diamond

By virtue of its excellent hardness and low coefficient of friction (in the range of 0.05±0.1),
diamond can be used as cutting tools. Materials most pliable to mechanical deformation by diamond
include aluminum, aluminum alloys, copper, copper alloys, chlorides, fluorides, polycarbonates,
plastics, quartz, sapphire, NaCl, Si3N4, SiC, Ti, WC, ZnS, and ZnSe.
126 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Diamond or diamond-like carbon is also used for coating magnetic disks to protect the head
crashes on magnetic disks, which require surface smoothness and hardness. In addition, fine grain
polycrystalline diamond films can be used as wire dies and water jet nozzles, since polycrystalline
artificial diamond nozzles are isotropic in hardness and lighter in weight, the latter of which is
critical for most streamlined water cutting operations.
With a thermal conductivity of 20 W/cm/8C, diamond is unparalleled as a thermal conductor.
For polycrystalline CVD diamond films, these figures are dependent upon the grain size. For a film
columnar in structure, the thermal conductivity can be lowered to 55% of the best value in the growth
direction and 25% of the best value in the lateral direction. With a high thermal conductivity,
diamond is presumed to be the ideal heat exchange material (heat sink and heat spreader). Diamond
has been utilized as an electrically insulating thermal conductor for various electronics applications.
Recently, high-power laser diodes have also been mounted on diamond in order to improve the
performance and increase the output power of the diodes.
Very large integrated circuit (VLSI) multiple chip module (MCM) compacts also use thick
diamond films as heat spreaders to increase the packaging density [11].
Diamond has the potential for both passive and active optical applications, although current
usage is only passive. Passive applications take advantage of its high thermal conductivity, corrosion
resistance, and hardness, as well as its low absorption coefficient and small coefficient of friction.
The first diamond window was used for the IR emission sensor of the Venus explorer. Diamond
windows were also used in periscopes and in missiles.
Optical matching is another passive usage of diamond. Diamond has a refractive index of 2.4,
which is lower than that of most semiconductors and higher than that of typical dielectrics. Diamond
generally has a lower refractive index than materials from which infrared detectors are made,
namely, silicon, germanium, group II±VI elements, and lead salts. Therefore, it is the preferred
material for coating applications. In addition, much progress has been achieved for longer
wavelength detectors with higher refractive indices. The efficiency of silicon solar cells has been
augmented by as much as 40% while that of germanium cells by up to 88% by diamond coatings
[12].
Recently, it has been proposed that CVD polycrystalline diamond film can be used as a very fast
optical switch ( 60 ps), due to its low dielectric constant and high breakdown voltage [13,14].
Because of its high carrier mobility, breakdown field, saturation velocity, thermal conductivity,
and wide band gap, diamond is considered an ideal material for electronic devices which function at
high temperatures, voltages, power-levels, frequencies, and radiation environments. For additional
information, readers may refer to some thorough references in this field [4,5,15±20]. The utilization
of synthetic crystals in photodetectors, light emitting diodes, nuclear radiation detectors, thermistors,
varistors, and negative resistance devices has been documented by Bazhenov et al. [16]. Meanwhile
several groups [18,20,21] have demonstrated basic field effect transistor (FET) device operation with
homoepitaxial diamond films and boron-doped layers on insulating single crystal diamond
substrates. However, wide application of diamond solid-state devices require high quality films to be
deposited on more commonly available substrates.
Device fabrication comprises a number of process steps besides the growth of a high quality
crystalline film. These steps include controlled doping of the film, selected area doping with p and n
type dopants, etching, formation of ohmic and rectifying contacts, deposition of dielectric films, and
passivation of the film surface. In the following we will review and discuss the state of the art
methods pertaining to these processes.
Heteroepitaxy of diamond has been attempted on a variety of single crystals including c-BN, b-
SiC, Si, Ni, Pt, and Ir. Although single crystalline films have not yet been obtained, heteroepitaxial
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 127

growth of diamond on silicon has proven to be a very feasible possibility and is currently being
extensively investigated. This topic will be discussed in detail in Section 7.
Doping is an essential process in device technology. Boron is the dopant that causes p-type
behavior in diamond. Homoepitaxial and polycrystalline films can be doped with boron by chemical
vapor deposition of B2H6 gas or various solid sources [22±24] as well as by ion implantation [25±
29]. While the problem of p-doping in diamond has been solved in principle, the problem of n-
doping (N, P, Li, Na, K and Ru) in diamond is still open. As summarized by Popovici and Prelas in
their review article [30], the difficulty associated with n-doping is due to distortion of the lattice
induced by large n-dopants. The lattice distortion generates a relatively shallow acceptor level in
diamond, which in turn compensates donor dopants. Thus a reasonable approach to activating n-type
diamond is to search for a method which will not induce serious lattice distortion during doping.
Geis reported that the formation of ohmic contacts is possible on highly boron-doped diamond
(about 1021 cmÿ3) via the deposition of single layers of several metals [19]. Sandhu [31] and Prins
[32] found that ohmic contacts can be formed by select-area boron implantation at room temperature
using 65 keV ions at a dose of 3  1016 cmÿ3. This high-dose ion implantation causes extensive
radiation damage which leads to graphitization of the surface layer during the subsequent annealing
treatment. The surface is then eliminated by a boiling solution of acids and contact is established
with a film of metal (Ag, Cu, or Au). At about 8008C, metallization occurs with an alloy composed
of Ag, Cu and In, which provides strong mechanical bonding. A specific contact resistance of
3.7 10ÿ3
cm2 can be obtained using an as-deposited film of Au by employing a standard TLM
model. Furthermore, a specific contact resistance with an order of magnitude of 10ÿ6
cm2 can be
achieved by Au/Ti metallization and subsequent annealing at 8508C and 10ÿ6 Torr for 30 min.
Rectifying contacts effective at both room temperature and 4008C have been fabricated by
depositing heteroepitaxial Ni films on single crystal diamond [33]. At room temperature, Ni films
deposited at 5008C, for 625 mm diameter dot, are measured to have a reverse leakage current of 2 nA
at a bias of 20 V. Also, the Ni films adhere well to the underlying diamond substrate. Another means
of producing excellent rectifying contacts on diamond is to use a composite film of co-sputtered Ta
and Si [34]. This has been demonstrated in the work by Shiomi et al. [21] where a rectifying contact
was positioned on an undoped diamond film, which was in turn placed on an already deposited
doped film. The result was a significant decrease in the reverse leakage current and an improvement
in the breakdown voltage.
Due to the difficulty in obtaining n-type diamond, designs for bipolar transistors which utilize
p±n junctions have not yet been realized. Instead, many schemes have been concentrated in metal
semiconductor field effect transistors (MESFET) or metal insulator field effect transistors
(MISFET). Gildenflat et al. [35] and Zeisse et al. [36] have fabricated these FET devices
respectively. However, the characteristics achieved thus far for this type of device are not good
enough for practical applications. The optimal designs and fabrication methods have yet to be
realized, and the quality of the deposited films also needs to be improved.

3. CVD processes for diamond growth

3.1. Historical aspect of CVD diamond

The cyclic process developed by Eversole [37] was the first method to demonstrate CVD
diamond growth under low pressures. In the early 1970s, Angus et al. extended this work. They grew
boron-doped diamond film on diamond grit (seed) [38]. Eversole's work was further expanded by
128 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Derjaguin et al. who performed careful physical chemistry experiments [39]. In the cyclic pyrolysis
method, diamond was used as a substrate, and diamond growth occurred homoepitaxially. However,
since cyclic, hydrocarbon pyrolysis had a very slow diamond deposition rate (1 nm/h) and required
a diamond grit substrate, its application was unrealistic.
In 1982, Matsumoto et al. made a breakthrough in CVD diamond technology [40]. They used
hot filaments (20008C) to directly activate hydrogen and hydrocarbon which were passed through
the hot filament. The diamond film was then deposited onto a non-diamond substrate located 10 mm
away from the filament. Graphite was etched simultaneously by atomic hydrogen during deposition
which rendered the cycling of deposition and etching unnecessary and therefore led to a higher
growth rate (1 mm/h).
Since then, various activating methods for diamond CVD such as DC-plasma, RF-plasma,
microwave plasma, electron cyclotron resonance-microwave plasma CVD (ECR-MPCVD), and their
modifications were developed. The role of atomic hydrogen in diamond growth has gradually been
recognized, and the growth rate approached the rate acceptable by industrial standards. In the late
1980s, synthesis of diamond under low pressures attracted the involvement of many scientists and
stimulated a `diamond fervor'. Currently, the DC-plasma jet diamond deposition method has earned
extensive attention in industry owing to its high growth rate. However, the apparatus of the DC-
plasma jet is highly expensive.
Worthy of mention is the method of fluorocarbon pyrolysis. OH radicals, O2, O, F2, and F as
graphite etchants are even better than atomic hydrogen (refer to Fig. 3). Based on these results,
Rudder et al. [41,42] predicted that the pyrolysis of fluorocarbons such as CF4 could produce
epitaxial diamond growth. In their experiment, a mixture of CF4 and F2 diluted in He was blown onto
a diamond substrate heated to 8758C. The deposited film was verified to be diamond by Raman
spectroscopy, and the deposition of graphite was not detected. The pyrolysis process happened nearly
at thermodynamic equilibrium. However, a growth rate of only 0.6 mm/h was achieved. This
technique has the potential to be energetically more efficient than the CVD methods.

Fig. 3. Carbon atom removal probabilities for the attack of an isotropic polycrystalline graphite( = 1.73 g/cm3) by O, O2,
OH, F, F2 and H (Reprint with permission J. Phys. Chem. Copyright 1973, American Chemical Society). (Ref. [4] p. 152).
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 129

Beside CVD, physical vapor deposition (PVD) methods were also attempted in parallel [43,44]
and were expected to deposit diamond at low temperatures. Recent work by S.T. Lee et al. [44]
indeed has demonstrated that diamond nanocrystals in the matrix of amorphous carbon could be
produced by direct low-energy ion bombardment using a mixture of CH4/H2/Ar ions.

3.2. Hot filament-assisted CVD

Hot filament CVD method is the earliest method used for the growth of diamond under low
pressures, and is also the most popular method.
In 1982, Matsumoto et al. [40] exploited a refractory metal filament (such as W) and heated it to
a temperature above 20008C, at which atomic hydrogen could be easily produced as H2 passed over
the hot filament. The simultaneous production of atomic hydrogen during hydrocarbon pyrolysis
could enhance the deposition of diamond. Diamond was deposited preferentially as graphite
formation was suppressed. As a result, the deposition rate of diamond increased to about 1 mm/h,
which proved valuable for industrial manufacturing. A schematic diagram of hot filament CVD
reactor was referred to [4], (p. 155). The simplicity and comparatively low capital and operating cost
of hot filament-assisted CVD have made the method popular in industry where it is imperative to
minimize the price of synthetic diamond.
The company Diamonex has used HF-CVD to grow diamond films with diameters of up to
30 cm. A wide variety of refractory materials have been used as filaments including W, Ta, and Re.
Carbide-forming refractory metals must be first carburized before starting the deposition of diamond
films. HF-CVD possesses the ability to adjust to a wide variety of carbon sources such as methane,
propane, ethane, and other hydrocarbons. Even oxygen-containing hydrocarbons including acetone,
ethanol, and methanol can be applied. The addition of oxygen-containing species may widen the
temperature range within which diamond deposition can take place.
In addition to the typical design of HF-CVD, some modifications have been developed. The
most popular is a combination of DC-plasma with HF-CVD where a bias voltage is applied to the
substrate platen and filament (or accessory electrode) [45±47]. The application of a moderately
positive voltage to the substrate platen and negative voltage to the filament (or accessory electrode)
results in electron bombardment of substrate, which induces desorption of the surface hydrogen. The
latter effect in turn increases the growth rate (up to about 10 mm/h). This technique is called
electron-assisted HF-CVD. When the bias is strong enough to establish a stable plasma discharge,
the decomposition of H2 and hydrocarbon is greatly enhanced, leading to a remarkable increase in
the growth rate (about 20 mm/h). Some laboratories even claim that the rate of diamond deposition
can exceed 30 mm/h on a 4 in. substrate. When the polarity of the bias is reversed, e.g. the substrate
platen is connected to a negative voltage, ion bombardment of the substrate surface will occur, and
enhanced nucleation of diamond on the non-diamond substrate will result. Another modification is to
replace the single hot filament with multiple filaments or a filament net for uniform film deposition
over large areas. For matching different requirements, further modifications are still under
development. The disadvantage of HF-CVD is the contamination of the diamond films with elements
from the refractory metal filaments due to the evaporation of the hot filaments.

3.3. Microwave plasma-assisted CVD

In the early 1970s, scientists found that the concentration of atomic hydrogen could be increased
by use of a DC-plasma established by an electrical discharge [39,48]. The plasma became therefore
another method to dissociate molecular hydrogen into atomic hydrogen and activate hydrocarbon
130 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 4. Schematic of a microwave plasma-enhanced CVD reactor manufactured by ASTeX.

radicals into promoting diamond formation. Besides DC plasma two other kinds of plasma with
different frequency schemes are used. The excitation frequency for microwave plasma CVD is
typically 2.45 GHz, while that for radio frequency (RF) plasma is 13.56 MHz. Microwave plasma is
unique in that microwave frequency can oscillate electrons. High ionization fractions are generated
as electrons collide with gas atoms and molecules. Microwave plasma is often said to have `hot'
electrons and `cool' ions and neutrals. A typical microwave reactor is referred to Ref. [4], p. 159.
However, the drawback of this setup is the small substrate size. A substrate about 2±3 cm in diameter
is the largest that can be introduced into a silica tube compatible with a WR284 waveguide. A
schematic representation of a microwave plasma system with a larger chamber reactor marketed by
Astex is shown in Fig. 4. Microwaves enter into the reaction chamber from a proprietary antenna
which converts a rectangular WR284 microwave signal into a circular mode. The microwave
proceeds through a silica window into the plasma enhanced CVD process chamber. The size of the
luminous plasma ball will increase with increasing microwave power. Diamond films have been
grown with the edge of the luminous plasma located about 2 cm higher than the substrate. The
substrate does not have to be in immediate contact with the luminous glow for diamond to grow via
microwave plasma. Uniform diamond films with diameters of up to 4 in. can be deposited using this
system.

3.4. RF plasma-assisted CVD

Radio frequency can be exploited to generate a plasma in two electrode configurations:


capacitively coupled parallel plates and by induction. RF plasma-assisted CVD utilizes a frequency
of 13.56 MHz. The schematic of an inductively coupled RF plasma-enhanced CVD reactor and a
parallel plate, capacitively-coupled RF reactor can be referred to Ref. [4], p. 164. RF plasma is good
in that it can be dispersed over more widespread areas than microwave plasmas. However, RF
capacitive plasma is limited in that the frequency of the plasma is optimal for sputtering, especially if
the plasma contains argon. Because ion bombardment from the plasma results in serious damages to
diamond, capacitively coupled RF plasma is not suitable for growth of high-quality diamond.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 131

Polycrystalline diamond films have been grown by the RF induction method using deposition
parameters similar to those of microwave plasma CVD [49]. Homoepitaxial diamond films have also
been deposited by RF induction PECVD [45].

3.5. DC plasma-assisted CVD

DC plasma is another method used to activate a gas source (typically a mixture of hydrogen and
hydrocarbon) for diamond growth. The schematic of a DC plasma CVD reactor is shown in Ref. [4],
p. 166. DC plasma assisted CVD possesses the ability to coat large areas, which are limited only by
the size of the electrodes and DC power supply.
As mentioned before, DC plasma has been combined with HF-CVD to obtain an increased
growth rate. Fujimori et al. [45] synthesized diamond films using a hybrid HF-CVD plus DC-plasma
method. By applying ÿ120 V to tungsten filament heated to 22008C, he achieved a deposition rate
three times as fast while maintaining the integrity of the diamond films.
Another advance in DC plasma-assisted CVD is the DC plasma jet method. Researchers in
Japan have devised a DC arc plasma-assisted CVD method which can deposit diamond films at rates
exceeding 20 mm/h. Kurihara et al. have designed a DC plasma jet facility termed DIA-JET [50,51].
DIA-JET employs a gas injection nozzle consisting of a cathode rod surrounded by an anode tube.
The typical growth rate is 80 mm/h. Norton company of USA has developed a very impressive DC-
plasma jet system which utilizes a DC power of about 100 kW. Chinese scientists at Beijing
University of Science and Technology have also built a similar system. Because various DC arc
methods can synthesize high-quality diamond on non-diamond substrates at fast growth rates, they
provide a marketable means for diamond film synthesis.

3.6. Electron cyclotron resonance microwave plasma-assisted CVD

As we discussed above, DC plasma, RF plasma, and microwave plasma all ionize and
decompose hydrogen and hydrocarbon species into hydrogen atoms and hydrocarbon radicals, and
thus promote the formation of diamond. So, we can expect electron cyclotron resonance microwave
plasma-CVD (ECR-MP-CVD) to be a method especially capable of synthesizing diamond film since
ECR-MP generates high density plasma (greater than 1  1011 cmÿ3) which is favorable for
diamond growth. In fact, Hiraki et al. [52] used ECR-MP-CVD to fabricate diamond in 1990. The
growth temperature could be reduced to 5008C. Later, Yara et al. [53] and Mantei et al. [54]
succeeded in diamond deposition using the ECR-MP-CVD technique. They obtained uniform films
at substrate temperatures as low as 3008C. The experimental arrangement for a typical ECR plasma
CVD is shown in Fig. 5.
However, due to the extremely low pressure of the ECR process (10ÿ4±10ÿ2 Torr), diamond
growth proceeds at a very low rate. Therefore, this method is only used in laboratories.

3.7. Combustion flame-assisted CVD

Combustion flame-assisted CVD method was first used by Hirose et al. [55,4]. The smoldering
tip of a welder's torch oxidizes a mixture of C2H2 and O2 gas (ratio 1 : 1). Diamond crystals form
where the tip of the bright interior section of the flame touches the substrate (temperature 800±
10508C). Some advantages of the combustion method over the conventional CVD methods include
the simplicity and cost-effectiveness of the equipment, lack of power supply, high growth rate, and
the ability to deposit diamond over large areas and on curved substrate surfaces. The disadvantages
132 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 5. Experimental arrangement for ECR plasma CVD of diamond.

of this method are also obvious. Because deposition is difficult to control, the deposited diamond
films are inhomogeneous both in microstructure and in composition. The welder's torch produces
thermal gradients over the surface of substrate, thereby causing the substrate to warp or fracture
during the process of coating large substrate areas. Hirose et al. [55] and Cappelli et al. [56] have
furthered the work and achieved great progress in enhancing the area and film quality of diamond
produced by combustion flame-assisted CVD. In the future, this technique may be applied to
fabricate diamond coatings used for tribological applications.

4. Nucleation mechanism of CVD diamond films

4.1. Practical significance of nucleation

Nucleation is the first and critical step of CVD diamond growth. The control of nucleation is
essential for optimizing the diamond properties such as grain size, orientation, transparency,
adhesion, and roughness that are necessary for targeted applications.
The investigation of diamond nucleation not only can lead to the controlled growth of diamond
films suitable for various applications, but also it can provide insight into the mechanism of diamond
growth. To date, the understanding of diamond nucleation is very limited. Carbon atoms can form
different types of chemical bonds via sp1, sp2, and sp3 hybridization. Diamond consists solely of sp3
bonds and is thermodynamically meta-stable compared to graphite, which is composed of sp2 bonds,
under the experimental conditions used in CVD. It is an interesting and yet intriguing problem why
meta-stable diamond can be grown on diamond or non-diamond substrates under CVD conditions.

4.2. Methods of nucleation enhancement

During the early development of CVD diamond deposition, diamond single crystals were used
as substrates [57±60]. Later, diamond seeds were used [61,62]. Most early efforts were limited to
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 133

homogeneous growth or homoepitaxy of diamond. In 1982 Matsumoto made a breakthrough in


growing diamond on non-diamond substrates without using diamond seeds. Apart from having a
very low nucleation density, a continuous film could not be formed. In 1987 Mitsuda et al. [63] found
that scratching of the substrate surface with diamond powder could greatly enhance the nucleation
density. Since then, substrate scratching has become the most common and powerful method for
achieving nucleation that can form diamond with a high nucleation density and fine uniform grain
size. For silicon substrates, which have been studied intensively, a nucleation density of 107±
108 cmÿ2 can be routinely obtained after scratching with diamond powder. In contrast, with non-
scratched substrates, the nucleation density only reaches 104 cmÿ2. Besides diamond powder other
abrasive powders such as c-BN, TaC, SiC, and even iron can be used to scratch the substrate surface
to enhance nucleation density. Nevertheless, diamond powder is regarded as the most effective
among the hard materials.
Later scientists have revealed that coating the substrate surface with graphite [64,65],
amorphous carbon [64,66], diamond-like carbon [67±70], C60, and even mechanical oil [66] can
enhance greatly nucleation density. Even so, these methods, including the scratching method
mentioned above, cannot lead to oriented nucleation or epitaxial growth on non-diamond substrates.
In 1991, Yugo et al. reported the bias-enhanced nucleation method by which they obtained a
high density of nucleation on a mirror-polished substrate (without scratching) using the MWCVD
system [71]. They applied a negative bias to the substrate during nucleation and obtained diamond
nucleation with a density as high as 109±1010 cmÿ2 on Si. Subsequent developments of bias-
enhanced nucleation by Jiang et al. [72,73] and Stoner et al. [74] have led to the heteroepitaxial
growth of diamond on silicon and silicon carbide substrates, respectively.
For the popular HF-CVD method, since the gas reactants consisting of atomic hydrogen and
hydrocarbon radicals are neutral species, a negative substrate bias cannot induce enhanced
nucleation. However, when a plasma is generated by the proper choice of biases, an enhancement of
diamond nucleation similar to that for MPCVD can be achieved for the HFCVD process [46,47,75].
Recently, other methods for enhancing diamond nucleation have been advocated. One method is the
nucleation enhancement under very low gas pressures (0.1±1 Torr), while the other is by Si+
implantation into mirror-smooth Si substrate prior to the introduction of methane into deposition
chamber. In the following various nucleation schemes will be discussed.

4.2.1. Mechanical abrasion of substrate


In 1987 Mitsuda et al. found that scratching of the surface substrate with diamond powder can
enhance greatly the nucleation density [63]. Since then, the effect of scratching has been extensively
studied. It has been demonstrated that the scratching technique can be applied to most substrates
used for diamond growth. Scratching pre-treatments with SiC [76], c-BN [77], Cu or stainless steel
[78], ZrB2 [79] and Al2O3 [80] were also shown to enhance nucleation, although their effects were
not as strong as that of diamond powder. Heavy scratching or abrading of the Si substrate surface
with diamond grit enhances the nucleation density by roughly three orders of magnitude compared
with non-scratched Si (up to 107 cmÿ2). The nucleation density is proportional to the scratching
time, and the morphology changes from large isolated crystals for short scratching times to smaller,
high number density crystals with increasing scratching time [80,81]. The grit size of the diamond
powder used for scratching also influences the nucleation density; a 0.25 m grit is the most effective
[82] for scratching by hand, and a 40±50 m powder is the best for scratching in an ultrasonic bath
using a grit suspension [83]. Why can scratching enhance diamond nucleation? One point of view is
that, during scratching with diamond, c-BN, or a-SiC powder, the residual powder or fragments are
unavoidably left in the scratched groove and act as seeds for diamond growth. Although c-BN and a-
134 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

SiC are not diamond, their structures are close enough to that of diamond. Thus, diamond grows
easily on them. Indeed, Iijima et al. [84] observed that fragments of diamond existed in the scratched
grooves of Si, upon which the growth of diamond did occur.
Another opinion is that scratching with powder merely creates a change in the surface
morphology, such as edges, steps, dislocations, and other surface defects. These kind of defects are
labeled chemical active sites, which prefer to adsorb diamond precursors together due to enhanced
bonding at high energy intersecting surfaces with a high density of unsaturated bonds and low
coordination numbers [80,85,86].
To identify this point experimentally, Dennig et al. [87] observed diamond nucleation on the
ridges lithographically etched on non-scratched Si by SEM. Singh [80] also reported enhanced
nucleation on Si etched by HF/HNO3. Recently, Si+ implantation on non-scratched Si has also been
found to enhance nucleation density [88]. In these experiments, neither diamond seeds nor a carbon
rich layer existed; only the surface morphology or surface structure was changed. However, other
attempts to enhance nucleation by creating etch pits on non-scratched Si using acids, H+ [89], or
other reactive gases [90] proved unsuccessful. In the same way, attempts to generate large numbers
of nucleating defect sites via implantation with surface saturation levels of C+ [91] or Ar+ [89]
resulted in no nucleation enhancement on the unscratched substrates. The arrays of holes of 0.1±
0.3 mm in diameter created by a focused Ga+ ion beam on a Si substrate resulted only in the
deposition of non-diamond carbon in the depressions [92]. The reason for the discrepancy among
these experiments is still unclear. To resolve the problem, the substrate surface must be characterized
thoroughly since nucleation is a surface phenomenon. Results obtained on ill-defined surfaces
invariably lead to uncertainty and controversy.

4.2.2. Biased-enhanced nucleation in microwave plasma CVD and HFCVD


It is difficult for diamond to nucleate on mirror-polished Si and silicon carbide because of their
surface free energies and lattice constants are much different from those of diamond. In 1991, Yugo
et al. [71] obtained diamond nucleation with a density of about 109±1010 cmÿ2 on a mirror-polished
Si substrate by applying a negative substrate bias voltage in a MPCVD system. With this technique,
the highest nucleation density on mirror-polished Si has reached the level of 1010±1011 cmÿ2 [93].
The nucleation mechanism has been widely studied and different models have been proposed. Yugo
et al. [94] and Gerber et al. [95] suggested a shallow ion implantation model in which the sp3 bonded
carbon clusters, formed by low energy ion implantation, function as the nucleation precursors. The
negative bias causes the positively charged ions in the growth chamber to accelerate towards and
bombard the substrate surface, thereby removing the contamination and facilitating cluster formation
on the surface. These events in turn advance diamond nucleation. Stoner et al. [96] on the other hand
indicated the critical process should be the change in plasma chemistry, such as the increase in the
concentration of atomic hydrogen caused by substrate biasing and the formation of a carbide surface
layer. Jiang et al. [97,98] found that the overall temporal evolution of the nucleation density
corresponds well with a surface kinetic model involving immobile active nucleation sites, germs, and
nuclei. They also suggested that, in addition to surface defects (point defects, steps and sp3-bonded
carbon clusters) serving as the nucleation sites, the enhanced surface diffusion and sticking
probability of carbon on silicon due to ion bombardment should be the decisive factors. The
enhancement of the surface diffusion of carbon species has been identified by investigation of the
distribution of the first nearest-neighbor distances [97].
Stubhan et al. [75] and Lin et al. [46,47] showed that, in HF-CVD system, diamond nucleation
enhancement can also be realized when a negative bias is applied to the substrate. The highest
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 135

nucleation density can reach 109±1010 cmÿ2 on mirror-polished Si, which is similar to the results
obtained using the MPCVD system.
To summarize the results obtained for the biased-enhanced nucleation, the decisive role of ion
bombardment for the nucleation of diamond has been clearly demonstrated. The nucleation sequence
can be given as following: (1) Formation of nucleation sites, (2) Formation of carbon clusters due to
the enhanced surface diffusion, and (3) Formation of stable diamond nuclei.

4.2.3. Diamond nucleation at low gas pressure


Recently, diamond nucleation has been conducted under very low pressures (0.1±1 Torr) and a
high density of diamond nucleation (109±1011 cmÿ2) has been achieved on mirror-polished Si
substrates using either HF-CVD or ECR microwave plasma CVD [99,100] without applying surface
scratching or a substrate bias. Raman spectroscopy shows both diamond and graphite lines in the
nucleated sample. Similar results have also been obtained on Ti substrates.
Using this method, diamond grains with a density >1010 cmÿ2 can be achieved, which is more
than 2 orders of magnitude higher than the highest density (107±108 cmÿ2) that can be obtained on
scratched substrates under conventional pressure (10±50 Torr). This value matches the highest
reported level to date, which was obtained using a negative substrate bias in a MPCVD system.

4.2.4. Ion implantation-enhanced nucleation


It has been demonstrated that in order to obtain a high nucleation density of diamond by CVD
methods, the substrate surface must be treated so that: (1) surface adsorption sites can be created for
the hydrocarbon radicals, and (2) the distributed region of the adsorption sites is large enough (larger
than the critical size for nucleation) for continuing growth of the diamond nuclei.
To modify the surface structure of the Si wafer, ion implantation can also be employed [88]. Si+
ions are implanted into a mirror-polished Si wafer. The implantation of Si+ ions changes only the
surface structure, and not the composition of Si. Therefore one can distinguish the surface structure
effect from others. After treatment with a Si+ energy of 25 keV (lower energy would be better) and
an implantation dose of 2  1017 cmÿ2 diamond can easily nucleate and grow on a Si wafer and
continuous diamond film can be synthesized.
Si+ implantation-enhanced nucleation is assumed to create nano-scale surface defects on the Si
substrate. These defects serve as the active sites for the adsorption of hydrocarbon radicals necessary
for initial diamond nucleation. A similar effect can also be found from the growth of diamond on
porous silicon [101], the surface of which has a rich nano-scale microstructure.

5. Growth mechanism of CVD diamond films

5.1. Overview

The mechanism of CVD diamond growth has attracted increasing attention in recent years,
mainly due to the fact that further technological advancement requires a more detailed understanding
of the fundamental phenomena responsible for diamond synthesis. Questions such as how to
grow diamond film more efficiently and economically, how to minimize the density of defects in the
films, and which sources are most effective, all require a thorough understanding of the growth
mechanism.
The first attempt to account for diamond {1 1 1} growth on the atomic scale was given by
Tsuda, Nakajima, and Oikawa [102,103], who proposed that diamond growth involved CH3+ cations
136 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

or a positively charged surface. However, this mechanism is incompatible with HF-CVD, since CH3+
are scarce and the substrate surface is uncharged under HF-CVD conditions.
Later, Chu et al. [104] proposed that the methyl radical is the dominant growth precursor under
HF-CVD conditions for all {1 1 1}, {1 0 0}, and {1 1 0} facets. They utilized isotope13 C to
distinguish the growth precursor of diamond, and concluded that the methyl radical is the main
precursor in diamond deposition. However, in a high temperature environment, CH4 and C2H2 will
decompose into various products and one cannot distinguish from which original source the products
come. Martin and coworkers [105,106] showed that methyl or methane is much more effective than
acetylene for growing diamond films.
Harris [107] proposed a growth mechanism involving only neutral CH3 and hydrogen atoms
using a nine-carbon model compound [bicyclononane (BCN)]. For the growth of diamond {1 0 0}
surfaces, which is terminated by CH2 radicals, the intermolecular H±H spacing is only about 0.77 A Ê,
nearly the same as that in the H2 molecule (0.74 AÊ ). Since the interaction is nonbonding, very strong
steric repulsion between neighboring hydrogen atoms is expected. Such a repulsion will greatly
affect the growth of the diamond {1 0 0} surface. Similar situations hold true for some other low
index surfaces as well.
Frenklach [108±111] and co-workers proposed that acetylene, which is present in greater
quantities than CH3 under typical HFCVD conditions [112±114], is the primary growth precursor on
diamond {1 1 1} and other low index surfaces.
Recently, nanocrystalline diamond attracted people's attention due to its applications and
fundamental significance. Studies on this kind of diamond revealed that they were constructed via C2
dimer [115,116] but not CH3 or C2H2.
The answers as to whether the aforementioned models are correct and to what extent they
concur with the experimental findings are unclear due to experimental difficulties in studying the
dynamic process of growth in-situ and on the atomic scale. As diamond nucleation and growth are
both surface phenomena, surface science techniques are suitable and powerful for investigating the
growth process and for understanding its mechanism. Among the surface techniques, high-resolution
electron energy loss spectroscopy (HREELS) is used to study the vibration of atoms or molecules to
provide information on the species, configurations, and adsorption sites at the surface. In the process
of diamond growth, the precursors are always hydrocarbon radicals or modifications of them, which
adsorb on the growing surface. Therefore, HREELS is well suited for the study of the types of
precursors attached on the surface and their evolution during nucleation and growth. Later we shall
discuss recent mechanistic studies of diamond growth which have utilized various surface techniques
including HREELS.

5.2. Growth mechanism of diamond (1 1 1)

Many efforts have been devoted to deciphering the growth mechanism of diamond (1 1 1), and
various growth models have been presented. To judge which model is correct requires experimental
support. To this end, HREELS measurements were conducted in-situ on as-grown homoepitaxial
diamond (1 1 1) surfaces and highly oriented (1 1 1) diamond films grown on Si (0 0 1) [117]. Fig. 6
shows the HREELS spectra for different growth temperatures. These spectra are very simple. The
loss peak at 365 meV corresponds to the C±H stretching vibration, 155 meV to the C±H bending
vibration, 110 meV to the C±C stretching vibration, and 310 and 460 meV to the first and second
overtones of the C±H bending vibration. Judging from the vibration modes and comparing with the
characteristic frequencies of the molecular subgroups of CHx in the handbook [118,119], the results
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 137

Fig. 6. High-resolution electron-energy-loss spectra of grown diamond (1 1 1) facets with 0.2% CH4 remote feed and at (a)
8008C, (b) 9008C, and (c) 10008C.

show that the surface is terminated not by CH3, but by CH radicals. Thus, diamond growth on the
(1 1 1) surface proceeds in a two-by-two layers mode.
Notably, the C±H stretching vibration is perpendicular to the (1 1 1) surface, whereas the C±H
bending vibration is parallel to the (1 1 1) surface. According to the selection rule [120], the
vibration parallel to the (1 1 1) surface cannot be activated and thus should be absent in the HREELS
spectra. However, in Fig. 6, C±H bending vibration loss peak at 155 meV is indeed present. In order
to reconcile this apparent contradiction, the H on the diamond (1 1 1) surface was replaced by its
isotope D [117], and two types of adsorption sites were found to exist; one is on the (1 1 1) surface,
and the other on another facet. If we speculate that the growing diamond (1 1 1) surface consists of
(1 1 1) faces and (1 1 0) steps, it is not difficult to understand that among the two kinds of sites
mentioned above, one is located on the (1 1 1) faces and the other is on the (1 1 0) faces steps
because the bending vibration of C±H on the (1 1 0) faces steps is perpendicular to the (1 1 1)
surface. Therefore, the C±H bending vibration should be active. As Frenklach pointed out, growth on
the (1 1 1) surface is carried out in two stages. First is kernel formation and second is the propagation
stage. In the kernel formation stage, the appearance of island and the (1 1 0) faces steps are possible.
Thus, the existence of the C±H bending vibration is understandable.
The second stage of the Frenklach model, which details the connection of C2H2 in a two- layers
by two-layers manner, is supported by experimental results [117]. In addition, it has been upheld
experimentally that during kernel formation, CH3 is the possible precursor [121].
In summary, growth on the (1 1 1) surface has been proposed to be completed in two stages. In
the first stage, the activation of a surface carbon by H abstraction, adsorption, and catenation of CH3
at the active site results in kernel formation. In the second stage, as predicted by Frenklach, the
(1 1 1) surface grows along the (0 1 1) direction with acetylene only.

5.3. Growth mechanism on (1 0 0) surface

There also is a lot of debate pertaining to the growth mechanism of diamond on the (1 0 0)
surface. Harris [107] proposed a growth model involving only neutral CH3 and hydrogen atoms
using a nine-carbon model compound [bicyclononane (BCN)], and the predicted growth rate on the
(1 0 0) surface agreed well with experiment. However, the steric repulsion for the H±H site on BCN
is different from that on the diamond (1 0 0) surface. On the diamond (1 0 0) surface terminated by
CH2 radicals, a very strong steric repulsion should exist between the neighboring hydrogen atoms.
138 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 7. High-resolution electron-energy-loss spectra of (a) diamond (1 0 0) facets grown at 8008C and 1.0% CH4, (b)
(1 0 0) facets grown at 10008C, (c) the sample of (a) dosed with atomic hydrogen, and (d) dosed with oxygen.

Hamza et al. found a 1  1 LEED pattern on the diamond (1 0 0) surface at temperatures from 500 to
750 K [122]. This was attributed to the saturation of the dangling bonds of a surface carbon atom by
two hydrogen atoms, and thus surface reconstruction did not occur. This surface was a nominally
dihydrogenated surface. Upon heating to over 1300 K, the surface structure showed a 2  1
reconstruction due to the desorption of one H atom from a surface carbon atom. This surface was a
nominally monohydrogenated surface with an elongated C±C dimer bond [123]. A 2  1 reconstruction
of the (1 0 0) diamond surface grown at 10008C was also observed by scanning tunneling microscopy
(STM) and atomic force microscopy (AFM), respectively [124,125]. Using HREELS, the diamond
(1 0 0) surface grown at a temperature of 8008C was investigated [126]. The spectra are shown in Fig. 7.
Intensity losses occur at 156, 180, and 372 meV, with three smaller losses at 110, 310, and 530 meV.
Compared with the characteristic frequencies of the molecular subgroups CHx [118,119], the spectrum is
consistent with that of the CH2 radical, which has its stretching vibration at 370 meV, scissors vibration at
179 meV, wagging vibration at 157 meV, twisting vibration at 150 meV, and rocking vibration at
108 meV. The 372 meV is assigned to CH2 stretching, 180 meV to scissors, 110 meV to rocking, and
156 meV to the overlapping of wagging and twisting. The 310 meV is the overtone of the loss at
156 meV, and the loss at 530 meV is the combination of CH2 wagging and stretching. The film grown at
this temperature exhibits good crystallinity.
If the growth temperature is increased to 10008C, the film shows a `cauliflower'-like
morphology, and the Raman spectrum exhibits a broad peak at 1580 cmÿ1, which is characteristic of
graphite, while its HREELS is similar to that at 8008C. However, a prominent peak appears at
140 meV. This peak corresponds to the bending vibration of the monohydrogenated dimer. Due to
the appearance of the monohydrogenated surface, hydrogen atoms that are bonded to the (1 0 0)
surface become further separated. As a result, it releases the strong steric repulsion between the H
atoms. Nevertheless, the C±H bond of the monohydrogenated dimer possesses to some extent  bond
character. As the hydrocarbon radicals attach to the C±H bond, they also take in some  bond
character as well. Their tendency to form graphite is why we observe a broad peak at 1580 cmÿ1 in
the Raman spectrum. Noteworthy is that if the (1 0 0) surface of the as-grown sample is exposed to
atomic hydrogen (no hydrocarbon involved), the 140 meV loss peak also appears in the HREELS
spectrum. This can be explained by the abstraction of one of the two hydrogen atoms bonded to a
surface carbon by gas phase atomic hydrogen.
If the abstraction is strong enough, monohydrogenated dimers appear in some local regions
where the surface is similar to the growth surface at 10008C. If the amount of atomic hydrogen is
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 139

small, or if the abstraction proceeds at a moderate temperature (about 8008C), or if hydrocarbon is


involved in the gas source, the abstraction of H will not be strong. Thus, dihydrogenated carbon
atoms remain in the neighborhood of C with one abstracted H, and monohydrogenated dimers cannot
be formed. The carbon atom keeps then only one H and one vacant site, which may bond to the
hydrocarbon radical. As a result, diamond growth proceeds steadily.
The quantity of carbon with one hydrogen is determined by the growth temperature, the amount
of atomic hydrogen near the surface, and the concentration of activated hydrocarbon. At growth
temperatures around 10008C, the (1 0 0) surface consists of monohydrogenated H±C±C±H dimers as
well. For even higher temperatures, large amounts of hydrogen desorb from the surface and the
surface carbons become C±C dimers.
The growth rate for the CH2-terminated surface is determined by the quantity of CH3 in the gas
phase. It has been shown that atomic hydrogen can easily abstract one hydrogen atom from CH3,
therefore, CH2 radicals become the precursor of (1 0 0) diamond growth [127].

5.4. Appearance regularity of the diamond facets

The control of the appearance of diamond grain facets becomes significant not only in practical
usage, but also in the testing of established diamond growth mechanisms. The morphology of CVD
diamond films is correlated with the growth parameters. Systematic studies on the relationship
between the appearances of the (1 1 1) and (1 0 0) facets and the growth parameters such as the ratio
of CH4/H2, O2 content, and distance from the hot filament to the substrate have been performed
[128,129]. A satisfactory explanation has however not yet been presented.
It is well known that, for a kinetically controlled growth system, the crystal morphology is
determined by the appearance of facets which have the slowest growth rate in their normal direction
and by the corresponding relative growth rates [130]. Because the (1 1 0) surface is a S (stepped)
face and encounters no repulsion between adjacent hydrogen atoms, it should have the highest
growth rate and appear as diamond facets. In fact, the growth rate of the (1 1 0) surface via the CVD
method is the highest among the low-index surfaces, and either methyl-radical- or acetylene-species-
based mechanisms can be postulated. Thus in most cases, the surface of CVD diamond crystals
appear with {1 0 0} faces and {1 1 1} faces. According to the established growth model, the growth
rate of the (1 0 0) face depends on the concentration of CH2 or CH3 while the growth rate of the
(1 1 1) facet relies on both CH3 and C2H2 for kernel formation and growth.
If the CH3 concentration near the substrate is much higher than the C2H2 concentration, the
{1 0 0} growth rate will be high, and hence the {1 0 0} faces will be absent. The crystal appears as
having just {1 1 1} faces. In contrast, if the C2H2 concentration near the growing surface is
dominant, the {1 1 1} face grows fast [131], which consequently results in the appearance of {1 0 0}
facets. In most cases the concentrations of CH3 and C2H2 are comparable, so both {1 0 0) and
{1 1 1} facets are present. Harris and Weiner [132,133] have measured the dependence of the CH3
and C2H2 concentrations upon the CH4/H2 and O2/H2 ratios, as well as upon the spacing between the
hot filament and substrate. Based on their experimental results, Sun et al. [126] explained the
regularity of the appearance of the diamond facets.

5.5. The role of atomic hydrogen in CVD diamond growth

In the chemical vapor deposition of diamond crystals, the most crucial aspect is that the
hydrocarbon gas must be diluted in hydrogen to as low as about 1% and the hydrogen must be
140 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

dissociated into atomic hydrogen. The role of atomic hydrogen for CVD diamond growth has been
clearly demonstrated in the HF-CVD method [40].
It has been recognized that atomic hydrogen plays a few specific roles. First, atomic hydrogen
reacts with, or etches graphite about 20±30 times faster than diamond, so graphite and other non-
diamond phases can be removed rapidly from the substrate and only clusters with diamond structure
remain and continue to grow [126]. Second, atomic hydrogen stabilizes the diamond surface and
maintains the sp3 hybridization configuration [134]. Third, atomic hydrogen converts hydrocarbons
into radicals, a necessary precursor for diamond formation. Fourth, atomic hydrogen abstracts
hydrogen from the hydrocarbons attached on the surface [135] and thus creates active sites for
adsorption of the diamond precursor. However, too much atomic hydrogen causes unnecessarily
strong abstraction. The formation of monohydrogenated dimers will increase and the graphite phase
will readily appear, which results in the deterioration of diamond film quality.

5.6. The role of ionic hydrogen in diamond growth

It is generally accepted that diamond growth is a combined process of deposition and carbon
etching which take place concurrently. The growth of diamond occurs if the deposition rate is larger
than the etching rate. During CVD growth of diamond films, both atomic H and H+ ions in the
plasma cause etching, and H+ ions etch even faster than atomic H [136,137]. Recently it was found
that a [0 0 1]-textured top layer can be prepared on polycrystalline or [1 1 1]-textured diamond films
by the application of a negative substrate bias potential during diamond growth [137±139]. A novel
etching effect of hydrogen ions on the growth of diamond films was observed and confirmed to play
a dominant role for the [0 0 1]-textured growth. The H+ ion bombardment was performed by
applying a negative substrate bias during a microwave plasma CVD process, using only hydrogen as
the reactant gas. It was discovered that the etching efficiency of H+ ions on non-[0 0 1] oriented
grains is higher than that on grains with their (0 0 1) faces parallel to the substrate. Lateral growth of
the (0 0 1) faces can occur during the bombardment process. As a result, the size of the (0 0 1) faces
increases after H+ etching while grains oriented in other directions are etched off. This effect
provides a way to improve the orientation degree of [0 0 1] oriented diamond films and may be
helpful for obtaining very thin [0 0 1] oriented diamond films.

5.7. The role of atomic oxygen in diamond growth

The addition of oxygen to the reaction gases, as either CO, O2 or alcohol, has a beneficial effect
on the growth rate and quality of the CVD diamond films and allows for diamond growth to occur at
low temperatures [140]. It was found [141] that although hydrocarbon species are somewhat reduced
by oxygen addition, oxygen has only a relatively small effect on the mole fractions of radical species
such as H and CH3. Also, OH is formed at concentrations sufficient for the removal of non-diamond
carbon at a rate comparable to that of diamond growth. It was found that [142], at low temperatures
(<8008C), the addition of oxygen not only enhances the growth rate but also extends the region of
diamond formation. At temperatures lower than 5008C, oxygen was explained to have stronger
preferential etching behaviors than hydrogen. Another possible explanation about the role of atomic
oxygen is the effective abstraction of surface hydrogen [143], which balances strongly the
abstraction of surface hydrogen. These occurrences are helpful for diamond growth. On the other
hand, the addition of too much oxygen will cause too strong of a hydrogen abstraction and even
oxidation of the surface, which in turn will deteriorate the diamond film.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 141

6. Heteroepitaxy of diamond

6.1. A great challenge

As mentioned in Section 2.2 pertaining to the electronic properties of diamond, diamond is an


important material for high temperature, high speed, high power, and high radiation-tolerant
electronic devices. However, to manufacture such devices, single crystals or epitaxial mono-
crystalline films are required. Diamond can be grown epitaxially on either natural or high-pressure-
synthesized diamond [144±152]. However, natural diamond is rare and expensive, while high-
pressure-synthesized diamond crystals have only been obtained in small sizes (<0.5 mm). Thus, it
appears the only practical way to grow diamond films is heteroepitaxially.
To date, c-BN is the best substrate on which diamond heteroepitaxy can be easily achieved
[153±158], because of its close similarity to diamond in lattice parameter (1.3% misfit) and surface
energy (4. J/m2 for the (1 1 1) surface of c-BN [159], 6 J/m2 for low-index planes of diamond
[160]). Unfortunately, large c-BN single crystals (>2 mm) cannot be grown at present, although
polycrystalline c-BN films can be grown with an extremely small grain size (several nanometers).
The second best candidate for diamond heteroepitaxy is b-SiC. In spite of a large lattice misfit of
about 22%, Stoner and Glass [74] obtained highly oriented diamond crystallites using the bias-
enhanced nucleation method. They obtained diamond crystallites with the following crystallographic
relationships: diamond (1 0 0)//b-SiC (1 0 0), diamond [011]//b-SiC [011].
In 1992, Jiang and Klages reported [72,73] that [0 0 1]-oriented diamond films can be
epitaxially grown on the (0 0 1) plane of single crystal silicon, also by applying a negative electrical
potential to the substrate without the intentional formation of an intermediate layer. They found by
high-resolution electron microscopy that diamond (0 0 1) grew directly on silicon (0 0 1) [161,162]
(Fig. 8). They stated that b-SiC is not a necessary interfacial layer required for diamond
heteroepitaxy on Si. Later, Lin et al. [163±166] and Stuhban at al. [75] also realized heteroepitaxial
growth of diamond (0 0 1) on silicon (0 0 1) in a HF-CVD system.
The structural quality of diamond films has been gradually improved during the recent years. It
has been possible to reduce the orientation deviation on Si (1 0 0) substrates from a best value of
about 98 in 1992 [167], as determined from the full width at half maximum (FWHM) of X-ray

Fig. 8. High resolution lattice image along the [1 1 0] direction of the diamond±silicon interface (Ref. [161]).
142 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

rocking curves, to a best value of about 28 today [168]. The misorientation of diamond crystallite
grown on b-SiC (0 0 1) substrates has also been strongly reduced. Diamond films with FWHM
values smaller than 18 have been prepared on b-SiC (0 0 1) [169]. The application of a two-step
process (a [0 0 1] fast growth and a [1 1 1] fast growth) produced diamond particles misoriented by 2
to 38 at the initial stage of deposition. However, as the diamond deposition proceeded, the
misoriented particles disappeared and eventually smooth diamond films were formed.
Heteroepitaxial growth of diamond was also attempted on other types of substrates. Using HF-
CVD, oriented diamond crystallites were deposited on a BeO single crystal [170]. Diamond films
were also grown heteroepitaxially on Ni [171], Pt [172] and Co [173] substrates by treating the
substrate surface with diamond powder. However, due to the complicated processing or the toxicity
of the substrate material (BeO), it is questionable whether these approaches can have practical
applications.
Recently, superior-quality heteroepitaxial growth of diamond on iridium layers has been
reported. The Ir layers have been deposited on the (0 0 1) cleavage planes [174±176] of MgO or on
the mechanically polished (1 0 0) planes of SrTiO3 [177]. Iridium has a lattice constant of 3.840 A Ê,
which is close to that of diamond 3.567 A Ê , and it (not a catalyst) does not form carbide under the
conditions for diamond deposition. Thus Ir would seem to be an excellent substrate for diamond
heteroepitaxy. By using bias-enhanced nucleation and subsequent growth by a MPCVD process,
highly oriented diamond crystallites have been grown on Ir. The size of the crystallites obtained is
around 1 mm, while the XRD polar and azimuthal spread for the crystal orientation are less than 18.
Therefore, this approach may offer a promising route for realization of large-area diamond
heteroepitaxy.
In terms of practicality, the realization of diamond heteroepitaxy directly on silicon is
particularly attractive, because Si wafers are easily available and extensively used in electronic
industry. The epitaxial growth of diamond on Si is highly desirable, particularly in view of its
convenient integration of diamond electronics with Si technology. Therefore the realization of
diamond heteroepitaxy on Si has become a great challenge of immense technological and scientific
interest. In the following we will concentrate on this issue.

6.2. Oriented heterogeneous nucleation of diamond

Generally, diamond heteroepitaxy on silicon proceeds in two stages, namely, nucleation and
growth. It is well known that it is difficult for diamond to nucleate on mirror-polished silicon
substrates. Under conventional nucleation without substrate pretreatment, the nucleation density is
only about 104 cmÿ2, and only the growth of isolated individual diamond crystals is obtained.
Because substrate scratching causes serious damage to the arrangement and periodic structures of the
surface atoms, epitaxial growth cannot occur via this technique. Bias-enhanced nucleation was
applied to a mirror-polished Si substrate in a MPCVD system to avoid the substrate scratching with
diamond powder. Fig. 9 shows a schematic diagram of the process parameters. Prior to the bias-
enhanced nucleation, in-situ hydrogen plasma etching was performed in order to remove the native
surface oxide layer. By carefully controlling the nucleation process, diamond nucleation with more
than 90% [0 0 1]-oriented nuclei was achieved [98]. Later, high-density nucleation and oriented
diamond films were also achieved by applying a negative bias to the substrate in a HF-CVD system
[75,163].
The investigation of the nucleation process reveals a narrow parameter window for epitaxial
nucleation. The crucial parameters are the substrate temperature, methane concentration, applied
bias voltage to the substrate during nucleation, and the nucleation time. A critical bias voltage exists,
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 143

Fig. 9. Schematic diagram of the microwave plasma CVD process for preparing heteroepitaxial diamond films (Ref. [98]).

although experimental results show different values for various reaction chambers. Typically, the
critical bias voltage for a process pressure of 20±40 mbar is approximately ÿ80 to ÿ120 V (Fig. 10)
[178].
Atomic force microscopy (AFM) and reflection-high-energy-electron diffraction (RHEED)
were employed to investigate the early stage of diamond nucleation [179,97]. The RHEED patterns
(Fig. 11) obtained for different nucleation times contained spots due to diffraction by diamond after
several minutes of deposition, which suggested a certain period of time is necessary for the initiation
of nucleation. It is demonstrated that the diamond nuclei formed at the start of the nucleation are
oriented. As the nucleation time increases, the nuclei increases dramatically in density to
5  1010 cmÿ2 (Fig. 12) [97] and are also randomly oriented. These results confirm the importance
of controlling the nucleation process for epitaxial growth [98]. Recently, Schreck et al. studied this

Fig. 10. Diamond nucleation density as a function of substrate bias voltage (Ref. [178]).
144 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 11. RHEED patterns obtained for different nucleation time (a) 7.5 min, (b) 10 min, (c) 12.5 min, (d) 15 min (Ref.
[97]).

processing space for heteroepitaxial nucleation of diamond on Si (0 0 1) using the bias process by X-
ray diffraction texture measurements [180,181]. The temporal development of the azimuthal pole
density distribution is shown in Fig. 13. It can be found that the bias time lies within a distinct time

Fig. 12. Nucleation density (islands/cm2) vs. deposition time. In the abscissa an induction time of 6.5 min is subtracted.
The curve is obtained by computer modeling. The open and full circles in the plots represent data calculated from crystal
size distribution and obtained by direct particle counting, respectively (Ref. [97]).
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 145

Fig. 13. Azimuthal {2 2 0} pole density distribution for diamond films. Results were obtained from a variation of the
biasing time. (Ref. [186]).

interval. The width of the time window and the bias time for optimal crystal alignment decrease
sharply as the absolute value of the bias voltage decreases (Fig. 14).
A possible mechanism of the loss of diamond epitaxy was suggested by Jiang et al. [182] and
Schreck et al. [180] after investigating diamond growth under bias conditions. The authors reported
on slightly misoriented crystallites grown homoepitaxially on {1 0 0} facets under bias conditions.
According to their results, the crystal misorientation can be traced back to either the formation of
defects during the homoepitaxial growth of the crystallites induced by ion bombardment or the
increased re-nucleation of strongly misoriented grains.

6.3. Heteroepitaxal growth of diamond films

To date, heteroepitaxial growth on Si is limited to microwave plasma CVD [73,98] hot-filament


CVD [75,163]. Significant progresses have been achieved in obtaining heterogeneous nucleation of
oriented diamond crystallites via bias-enhanced nucleation.
It has been noted that if the growth temperature is lower and the nucleation period (incubation
time required before the appearance of nuclei) is shorter, diamond can nucleate directly on a
Si substrate without forming a SiC intermediate layer at the interface. The maximum area
of the epitaxial layer (single crystalline layer) of diamond grown directly on Si can reach
20  20 mm2.
146 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

Fig. 14. Variation of the process time window for oriented nucleation with the bias voltage (Ref. [180]).

So far, the epitaxial crystallite of diamond grown on Si (1 0 0) has the following


crystallographic relationships to the substrate: diamond(1 0 0)//silicon(1 0 0) and diamond[1 1 0]//
silicon[1 1 0].
Textured growth utilizes an `evolutionary' selection (Van der Drift growth) if the growth process
proceeds to a large thickness, and only permits crystallites which are oriented in the fastest growth
direction (approximately parallel to the general growth direction) to grow further. The fastest growth
direction can be varied by changing the process parameters, the most important of which are the
substrate temperature and methane concentration [183]). Recently, it was found that the application
of a proper bias potential to the substrate also leads to a strong selection of the growth direction
[182]. The selection effect is attributed to the direction-dependent etching of the diamond crystal by
H+ ions [184]. Therefore, a textured film texture can be achieved for thin films. Both of the
aforementioned approaches have been used to prepare epitaxially oriented films [183,139].
Heteroepitaxial [0 0 1]-oriented diamond films with considerably increased lateral grain size
and strongly improved orientational perfection could be prepared by microwave plasma-assisted
chemical vapor deposition using a [0 0 1]-textured growth process on Si (0 0 1) substrates followed
by a [1 1 0] step-flow growth process [168,185]. The diamond films were characterized by atomic
force microscopy, scanning electron microscopy, and transmission electron microscopy. The results
indicate that diamond crystals increase their lateral dimensions at the (0 0 1) film surface either by
coalescence of grains combined with a termination of the propagation of grain boundaries (Fig. 15)
or by changing the grain boundary plane orientations from preferentially vertical to preferentially
parallel directions with respect to the (0 0 1) growth faces. In the second case, the grains with
relatively large angle deviation from the ideal epitaxial orientation are overgrown by those with
relatively small angle deviation. As a result, the degree of orientational perfection of the films
improves considerably in comparison to that of films prepared by the established process of [0 0 1]-
textured growth. The presence of boron in the gas phase was found to strongly enhance the step-flow
lateral grain growth. It was possible to achieve deposition of a thin boron-doped diamond films
characterized by a full width at half maximum value of the measured tilt angle distribution of only
2.18.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 147

Fig. 15. High-resolution lattice image of a grain boundary region, showing disappearance of a 28 low-angle grain boundary
(Ref. [161]).

6.4. Interface between diamond and silicon

Early studies revealed that the growth of diamond on Si occurred through a b-SiC transfer layer
[186,187]. This perhaps is expected considering that the lattice misfit between diamond and b-SiC
(22%) is much smaller than that between diamond and silicon (52%). However, recently, a direct
lattice observation of the interface between silicon and epitaxially grown diamond crystal has been
successfully performed [161,162]. It is clearly demonstrated that the diamond nuclei are grown in
direct contact with the silicon substrate (Fig. 8). Because the ratio of the lattice constants of silicon
and diamond is close to 1.5, a nearly perfect 3 to 2 correspondence (1.5% mismatch) of lattice
spacing is seen at the interface, e.g. every three Si (1 1 1) fringes are matched well with four
diamond (1 1 1) fringes (the mismatch is about 1.5%). Individual 608 interface misfit dislocations for
every third (1 1 1) atomic plane, with a spacing between two such dislocations of about 7 A Ê , can be
clearly identified (Fig. 8). The presented results are reproducible for many of the epitaxially oriented
diamond grains observed and they provide strong evidence that diamond crystal can be epitaxially
grown directly on Si. From the experimental results, it can be concluded that if the growth
temperature is not too high and the incubation time is sufficiently short, diamond can grow directly
on the Si substrate, avoiding the formation of the b-SiC interlayer. A thin epitaxial SiC intermediate
layer is unnecessary for diamond heteroepitaxy. Recent theoretical investigations of the interface
structure between diamond and silicon have also provided support for the direct growth of diamond
on silicon [188,189].

6.5. Possible improvements of the heteroepitaxy of diamond on silicon

Although great progresses have been made in the heteroepitaxy of diamond on silicon, large
area, single crystalline diamond layers have still not been obtained. In order to realize large-area and
high-quality epitaxy, further efforts are needed. For the heteroepitaxy of diamond directly on silicon,
the following approaches may be taken into consideration.
148 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

6.5.1. Improvement of the surface treatment


This is the first and essential requirement. The substrate surface must be clean, without surface
contamination and oxidation, and the dangling bonds of the Si surface atoms must be saturated by
hydrogen. During the process of bias-enhanced nucleation, surface damage due to ion bombardment
and the formation of amorphous carbon must be avoided.

6.5.2. Control of nucleation must be as exact as possible


The bias voltage, biasing time, and pressure strongly influence crystal orientations. The
bombardment of energetic ions above a critical energy is necessary for the formation of nuclei. A
negative influence by ion bombardment on the alignment of the diamond grains has also been
demonstrated. An improved epitaxy requires a compromise of the positive and negative influences.
On the other hand, at present the vacuum for CVD diamond nucleation and growth is low and the
residual gas in the chamber would contaminate or even oxidize the Si surface and change the surface
status. Therefore improvement of the base vacuum in the growth chamber and the purity of gas
source would be helpful in achieving better epitaxy.

7. Surface science studies of CVD diamond growth

The nucleation, growth, and epitaxy of CVD diamond all proceed on the surface. These
essential phenomena belong to the fields of surface science, and thus surface techniques are useful
and powerful tools in studying CVD diamond growth. Especially, the growth of diamond under low
pressure reveals a very complicated process, which to a certain extent is similar to the growth process
of some functional organic materials, i.e. the growth process is mainly completed by the tailoring
and interconnecting of functional radicals. Diamond growth under low pressure is not a simple
deposition of carbon atoms, for in this way only graphite is obtained; diamond growth is completed
by the surface adsorption, migration, and interconnection of hydrocarbon radicals as well as the
reconstruction (reformation) of carbon atoms in the lattice. Such a complicated process can be
clarified only by surface techniques.
The relationship between CVD diamond studies and surface techniques has been extensively
discussed in the review articles entitled Characterization of diamond films written by Zhu et al. [190]
and Hydrogen-terminated diamond surface and interface by Kawarada [191]. Zhu's paper also
introduced the principle of various surface techniques and their application in diamond surface
characterization, which we will not repeat in this paper. Here, we will just highlight the usefulness of
surface techniques in the studies of diamond nucleation, growth, and epitaxy.

7.1. In preparation of substrate for nucleation and growth

Prior to diamond nucleation and growth, the substrate must first be cleaned completely in order
to obtain an uncontaminated surface; second, the clean surface then undergoes surface pre-treatment
in some cases. Of the many surface cleaning and pre-treatment methods that are available, which one
is effective? How can a surface be cleaned and still maintain the integrity of the surface structure? To
evaluate the results from the different cleaning methods, surface techniques such as Auger electron
spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), low energy electron diffraction
(LEED), reflected high-energy electron diffraction (RHEED), high resolution electron energy loss
spectroscopy (HREELS), scanning tunneling microscope (STM), and atomic force microscope
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 149

(AFM) are often used. They identify surface composition, surface structure, surface bonding, as well
as surface electronic structure after the cleaning and pre-treatment stages.

7.2. In nucleation studies

CVD diamond hardly grows on the surface of other materials, however, under certain conditions
it can be grown on some materials. Why? In order to answer this question, it is essential to study the
properties of the substrate surface including its composition, structure, reconstruction, morphology,
and electronic structure by using surface techniques. The reason for this is that nucleation depends
heavily on the status of the substrate surface. Why are the data of diamond nucleation obtained from
different laboratories frequently in disagreement? The main explanation is that the substrate surfaces
used for nucleation and growth are not well-defined, therefore, due to differences in the surface
status, it is not surprising that the nucleation results vary and are even controversial. In order to
obtain `clean results' of nucleation process, a well-defined surface of the substrate must be used as a
starting point. Thus, surface techniques are needed to identify the substrate surface. Furthermore,
surface techniques are also imperative for investigating factors pertaining to the nucleation process,
such as the adsorption sites, configuration of adsorbates, and their evolution.

7.3. In diamond growth studies

Many debates still remain over the mechanism of CVD diamond growth. Which species (CH,
CH2, CH3, C2H2, etc.) is the precursor for diamond growth? Where are the adsorption sites of these
precursors? How do they interconnect with each other to form a diamond crystal? The answers to
these questions are still unclear. Although many models have been proposed, for the lack of
experiment data at the atomic level, a satisfactory mechanism has not yet been developed. Here,
surface techniques can be utilized to study the issues mentioned above and provide needy data for
understanding diamond growth.
LEED (RHEED) and STM (AFM) offer information relating to the surface and reconstruction
before and after adsorption of hydrocarbon radicals and hydrogen atoms. Both techniques
complement each other. Here, we just introduce high-resolution-electron energy loss spectroscopy
(HREELS), which, similar to infrared (IR) spectroscopy, has become a powerful tool for studying
surface vibration. Given the information of the vibration mode (number of loss peak) and vibration
frequency (position of loss peaks) obtained from the HREELS spectra, we can deduce the
configuration of the adsorbates (CH, CH2, CH3, etc.) and the adsorption sites on the surface as well
as the evolution of the adsorbates during growth. Therefore, HREELS provides the data necessary
for understanding the diamond growth mechanism. Details of HREELS can be found in Refs. [191±
194]. When HREELS is combined with other techniques, like AES, LEED/STM in an UHV system,
more complete and reliable information about diamond growth can be obtained.

8. Summary

Diamond possesses numerous unmatchable properties in mechanical, thermal, optical, and


electronic aspects, therefore it demonstrates extensive application prospects. Unfortunately, natural
diamond is rare and expensive, which prohibits extensive usage. At the same time, high-pressure-
high-temperature-diamond can only be obtained in small crystallite amounts, which also cause many
limitations to their applications. The deposition of diamond films by chemical vapor deposition has
150 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

served as a great breakthrough in the last decade of the 20th century and has made many applications
of diamond possible. In order to increase the growth rate and improve the quality of diamond films,
many methods have been developed in succession, including HF-CVD, MP-CVD, RF-P-CVD, DC-
P-CVD, ECR-MP-CVD, and combustion flame-CVD. Due to the intensive efforts expended on the
part of research scientists and industrialists worldwide, CVD methods have reached a rather high
level at present. Diamond is opening the door to the fields of coating, cutting tool, optical, and
thermal management applications. Currently, several products with respectable quantities of
diamond have entered the market. This is a great advance in diamond development. However, despite
the many successes achieved in CVD diamond, serious difficulties still lie ahead of us. So far, we
still have not succeeded in growing mono-crystalline diamond film with a large area, which limits
greatly the important application of diamond in high-temperature electronic devices. Naturally, to
solve the problem of diamond heteroepitaxy is a big challenge, and it has become a hot topic in
diamond research and attracted many researchers to this study.
In order to solve such complicated and difficult problems, first of all, we must identify the
mechanism of diamond nucleation, growth, and epitaxy, because only on the basis of mechanistic
understanding can we succeed in large-area heteroepitaxial growth of diamond.
In the studies of diamond nucleation, growth, and epitaxy, surface techniques serve as very
useful and powerful tools, and of course, the combination of surface techniques with other analytical
methods will be even more powerful. At present, it is clear that diamond facets with different
orientations have their own growth modes, atomic hydrogen and oxygen have essential roles in the
growth process, and the appearance of facets with different orientation can be regulated, etc. In the
nucleation aspect, the effect of substrate surface conditions on nucleation has been studied, the role
of atomic hydrogen and the concentration of hydrocarbon radicals on the nucleation has been
emphasized, and several methods for high density nucleation of diamond on mirror-polished Si or
SiC have been developed. As for heteroepitaxy, excellent and reproducible results have been
obtained mainly on Si and SiC. Highly oriented diamond films grown epitaxially on Si or SiC can be
achieved in many laboratories without difficulty. The diameter of the crystallites can reach 20 mm.
Although many achievements have been made in diamond nucleation, growth, and epitaxy, they still
fall short of a thorough understanding of the mechanism. Therefore, a lot of studies are still required.
Nevertheless, we will anticipate success in large-area diamond heteroepitaxy in the near future. This
is the main line of the paper.

References

[1] J.C. Angus, C.C. Hayman, Science 214 (1988) 913.


[2] W. Yarbrough, R. Messier, Science 247 (1990) 688.
[3] J.E. Field, The Properties of Diamond, Academic Press, Oxford, 1979.
[4] R.F. Davis (Ed.), Diamond Films and Coatings, Noyes Publications, New Jersey, 1992
[5] L.S. Pan, D.R. Kania (Eds.), Diamond: Electronic Properties and Applications, Kluwer Academic Publishers,
Boston, 1995.
[6] J. Wilks, E. Wilks (Eds.), Properties and Application of Diamond, Butterworth Heinemann Ltd.
[7] Max N. Yoder, Diamond properties and application, in: Robert F. Davis (Ed.), Diamond Films and Coatings,
Development, Properties, and Applications, Noyes Publications, Park Ridge, New Jersey, p. 5.
[8] R. Seitz, The end of the age of sand, presented at the Gorham Advanced Materials Institute, Marco Island, FL 15±17
Oct., 1989.
[9] A. Backon, A. Szymanski, Practical uses of Diamond, in polish, translated into English by P. Daniel, pp. 29±31.
[10] A.P. Rudenko et al., DAN SSSR, Ser, mat. Fiz. 163(5) (1965) 1169.
[11] R.C. Eden, Diamond Rel. Mater. 25±7 (1993) 1051.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 151

[12] A. Altshuler, Diamond Technology in the USSR, Delphic Associates Press, 1989.
[13] K. Baba, Y. Aikawa, N. Shohata, H. Yoneda, K.-I. Ueda, NEC Research Development 363 (1995) 369.
[14] H. Yoneda, K.-I. Ueda, Y. Aikwaa, K. Baba, N. Shohata, Appl. Phys. Lett. 664 (1995) 460.
[15] A.T. Collins, E.C. Lightowlers, in: J.E. Field (Ed.), The Properties of Diamond, Academic Press, San Diego, 1979.
[16] V.K. Bazhenov, I.M. Vikulin, A.G. Gontar, Sov. Phys. Semicond. 19 (1985) 829.
[17] A.T. Collins, Semiconductor Sci. Technol. 4 (1989) 605.
[18] K. Shenai, R.S. Scott, B.J. Baliga, IEEE Trans. Electron Dev. 36 (1989) 1811.
[19] M.W. Geis, Proc. IEEE 79 (1991) 669.
[20] G.S. Gildenblat, S.A. Grot, C.W. Hatfield, A.R. Badzian, IEEE Electron Dev. Lett. 12 (1991) 37.
[21] H. Shiomi, Y. Nishibayashi, N. Fujimori, in: R. Messier, J.T. Glass, J.E. Butter, R. Roy (Eds.), Proc. 2nd Intl. Conf.
New Diamond Sci. Technol. Mat. Res. Soc., Pittsburgh, 1991, p. 975.
[22] K. Nishimura, K. Das, J.T. Glass, J. Appl. Phys. 69 (1991) 3142 .
[23] N. Fujimori, T. Imai, A. Doi, Vacuum 36 (1986) 99.
[24] M.W. Geis, in J.T. Glass, R. Messier, N. Fujimori (Eds.), Diamond, Silicon Carbide and Related Wide Bandgap
Semiconductors, 162, Mat. Res. Soc. Symp. Proc., 1990, p. 15.
[25] V.S. Vavilov, M.I. Guseva, E.A. Konorova, V.F. Sergienko, Sov. Phys. Semicond. 4 (1970) 6.
[26] G. Braunstein, R. Kalish, J. Appl. Phys. 54 (1983) 2106.
[27] J.F. Prins, Phys. Rev. B. 38 (1988) 5576.
[28] G.S. Sandhu, M.L. Swanson, W.K. Chu, Appl. Phys. Lett. 55 (1989) 1397.
[29] P.R. de la Houssaye, C.M. Penchina, C.A. Hewett, R.G. Wilson, J.R. Zeidlers, in: R. Messier, J.T. Glass, J.E. Butler,
R. Roy (Eds.), Proc. 2nd Intl. Conf. New Diamond Sci. Techn., Mat. Res. Soc., Pittsburgh, 1991, p. 113.
[30] G. Popovici, M.A. Prelas, Diamond Rel. Mater. 4 (1995) 1305.
[31] G.S. Sandhu, Ph.D. Thesis, Univ. of North Carolina at Chapel Hill, 1989.
[32] J.F. Prins, J. Phys. D: Appl. Phys. 22 (1989) 1562.
[33] T.P. Humphreys, J.V. LaBrasca, R.J. Nemanich, K. Das, J.B. Posthill, Electron. Lett. 27 (1991) 1515.
[34] S.R. Sahaida, D.G. Thompson, in: T.D. Moustakos, J.I. Pankove, Y. Hamakawa (Eds.), Mater. Res. Soc. Symp. Proc.
1992.
[35] G.S. Gildenblat, S.A. Grot, C.W. Hatfield, A.R. Badzian, IEEE Electron Dev. Lett. 12 (1991) 37.
[36] C.R. Zeisse, C.A. Hewett, R. Nguyen, J.R. Zeidler, R.G. Wilson, IEEE Electron Dev. Lett. 12 (1991) 602.
[37] W.G. Eversole, U.S. Patent 3,030,188, April 17, 1962.
[38] D.J. Poferl, N.C. Gardner, J.C. Angus, J. Appl. Phys. 444 (1973) 1428.
[39] B.V. Spitsyn, L.L. Bovilov, B.V. Derjaguin, J. Crys. Growth 52 (1981) 219.
[40] S. Matsumoto, Y. Sato, M. Tsutsimi, N. Setaka, J. Mat. Sci. 17 (1982) 3106.
[41] R.A. Rudder, J.B. Posthill, R.J. Markunas, Semiannual Report on Semiconductor Diamond Technology, SDIO/ONR
contract N00014-86-C-0460, November 1989
[42] R.A. Rudder, J. B Posthill, R.J. Markunas, Electron. Lett. 25 (1989) 1220.
[43] W.M. Lau, S.-T. Lee, F. Qin, Surface Modification Technologies V, 1992, p. 263.
[44] X.S. Sun, N. Wang, H.K. Woo, W.J. Zhang, G. Yu, I. Bello, C.S. Lee, S.T. Lee, Diamond & Relat. Mater., in press.
[45] N. Fujimori, A. Ikegaya, T. Imai, K. Fukushima, N. Ota, in: J. Dismukes, et al. (Eds.), Diamond and Diamond-like
Films, Electrochem. Soc. Proc., Vol PV 89-12, Pennington, NJ, 1989, p. 465.
[46] Qijin Chen, Zhangda Lin, Appl. Phys. Lett. 68(17) (1996) 2450.
[47] Qijin Chen, Jie Yang, Zhangda Lin, Appl. Phys. Lett. 67(13) (1995) 1853.
[48] A.R. Badzian, R.C. De Vries, Mater. Res. Bull. 23 (1988) 385.
[49] D.E. Meyer, T.B. Kustka, R.O, Dillon, J.A. Woollam, in: A. Feldman, S. Holly (Eds.), Diamond Optics II, SPIE
Proc., 1146:21, Bellingham, WA, 1990.
[50] K. Kurihara, K. Sasaki, M. Kawarada, N. Koshino, Appl. Phys. Lett. 526 (1988) 437.
[51] M. Kawarada, K. Kurihara, K. Sasaki, A. Teshima, K. Koshino, in: A. Feldman, S. Holly (Eds.), Diamond Optics II,
SPIE Proc. 1146, Bellingham, WA, 1990, p. 28.
[52] Akio Hiraki, Hiroshi Kawarada, Jin Wei, Jun-Ichi Suzuki, Surface and Coating Technology 43/44 (1990) 10.
[53] Takuya Yara, Motokazu Yuasa, Manabu Shimizu, Hiroshi Makita, Akimitsu Hatta, Jun-ichi Suzuki, Toshimichi Ito,
Akio Hiraki, Jpn J. Appl. Phys. 33(1) (1994) 4405.
[54] D. Mantei, Zoltan Ring, Mark Schweizer, Spirit Tlali, Howard E. Jackson, Jpn. J of Appl. Phys. 35 (1996) 2516.
[55] Y. Hirose, N. Kondo, Program and Book of Abstracts, Japan Applied Physics 1988 spring meeting, March 29, 1988,
p. 434.
[56] M.A. Cappelli, P.H. Paul, J. Appl. Phys. 675 (1990) 2596.
[57] J.J. Lander, J. Morrison, Surf. Sci. 2 (1964) 553.
[58] B.V. Spitsyn, L.L. Bouilov, B.V. Deryagin, J. Cryst. Growth 52 (1981) 219.
[59] S. Matsumoto, Y. Sato, M. Kamo, J. Tanaka, N. Setaka, in Proceedings of the 7th International Conference on
Vacuum Metallurgy, Tokyo, Japan, 1982. Iron and Steel Institute of Japan, Tokyo, Japan, 1982, pp. 386±391.
[60] N. Fujimori, T. Imai, H. Nakahata, Symposium N Plasma Assisted Deposition of New Materials, MRS Fall 1987
Meeting. Boston, MA. Nov. 30±Dec. 5, 1987.
152 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

[61] J.C. Angus, Y. Wang, M. Sunkara, Ann. Rev. Mater. Sci. 21 (1991) 221.
[62] W.G. Eversole, 1962, U S Patent No. 3, 030, 187; 3, 030, 188.
[63] K. Mitsuda, Y. Kojima, T. Yoshida, K. Akashi, J. Mater. Sci. 22 (1987) 1557.
[64] J.C. Angus, Z. Li, M. Sunkara, R. Gat, A.B. Anderson, S.P. Mehandru, M.W. Geis, in: A.J. Purdes, J.C. Angus, R.F.
Davis, B.M. Meyerson, K.E. Spear, M. Yoder (Eds.), Proceedings of the 2nd International Symposium on Diamond
Materials, The Electrochemical Society, Pennington, N J, 1991, p. 125.
[65] Z. Feng, K. Komvopoulos, I.G. Brown, D.C. Bogy, J. Appl. Phys. 744 (1993) 2841.
[66] A.A. Morrish, P.E. Pehesson, Appl. Phys. Lett. 59 (1991) 417.
[67] J. Singh, M. Vellaikal, J. Appl. Phys. 73 (1993) 2831.
[68] K.V. Ravi, M. Peters, L. Plano, S. Yokota, M. Pinneco, in First International Conference on the New Diamond
Science and Technology, Program and Abstr. 24±26, Oct. 1988. (JNDF, Tokyo, 1988), pp. 96±97.
[69] K.V. Ravi, C.A. Koch, Appl. Phys. Lett. 57 (1992) 348.
[70] K.V. Ravi, C.A. Koch, H.S. Hu, A. Joshi, J. Mater. Res. 5 (1990) 2356.
[71] S. Yugo, T. Kanai, T. Kimura, T. Muto, Appl. Phys. Lett. 58 (1991) 1036.
[72] X. Jiang, C.-P Klages, Diamond Relat. Mater. 2 (1993) 1112.
[73] X. Jiang, C.-P. Klages, R. Zachai, M. Hartweg, H.-J. FuÈsser, Appl. Phys. Lett. 62 (1993) 3438.
[74] B.R. Stoner, J.T. Glass, Appl. Phys. Lett. 60 (1992) 698.
[75] F. Stubhan, M. Ferguson, H.-J. FuÈsser, R.J. Behm, Appl. Phys. Lett. 66 (1995) 1900.
[76] A. Sawabe, T. Inuzuka, Thin Solid Films 137 (1986) 89.
[77] K. Suzuki, A. Swabe, H. Yasuda, T. Inuzuka, Appl. Phys. Lett. 50 (1987) 728.
[78] C.P. Chang, D.L. Flamm, D.E. Ibbotson, J.A. Mucha, J. Appl. Phys. 63 (1988) 1744.
[79] P.A. Denning, D.A. Stevenson, Proc. of the 2nd Int. Conf. on New Diamond Sci. and Techn. Pittsburg, PA, MRS,
1991, pp. 403±408.
[80] B. Singh, O. Mesker, A.W. Levine, Y. Arie, Growth of Polycrystalline Diamond Particles and Films by Hot-
Filament Chemical Vapor Deposition, Report CN-5300, David Sarnoff Research Center, SRI International,
Princeton, NJ, 1988.
[81] B. Singh, Y. Arie, A.W. Levine, O.R. Mesker, Appl. Phys. Lett. 526 (1988) 451.
[82] C.P. Beetz Jr, T.A. Perry, Hot and Cold Wall Filament-Assisted Growth of Diamond Films, General Motors
Research Publication GMR-6093, November 24, 1987, pp. 1±21.
[83] S. Yugo, T. Kimura, H. Kanai, in: S. Saito, O. Fukunaga, M. Yoshihara (Eds.), Science and Technology of New
Diamond, KTK Scientific Pub., Tokyo, 1990, pp. 119±128.
[84] S. Iijima, Y. Aikawa, K. Baba, J. Mater. Res. 6 (1991) 1491.
[85] S.P. Murarka, M.C. Peckerar, Electronic Materials Science and Technology, Academic Press Inc. San Diego, CA,
1989.
[86] S. Yugo, A. Izumi, T. Kanai, T. Muto, T. Kimura, Proc. of the 2nd Int. Conf. on New Diamond Science and
Technology, Pittsburg, PA, MRS, 1991, pp. 385±390.
[87] P.A. Dennig, D.A. Stevenson, Proc. of the 2nd Int. Conf. on New Diamond Science and Technology, Pittsburg, PA,
MRS, 1991, pp. 403±408.
[88] Jie Yang, Xiaowei Su, Qijin Chen, Zhangda Lin, Appl. Phys. Lett. 66(24) (1995) 3284.
[89] P.K. Bachmann, R. Weimer, W. Drawl, Y. Liou, R. Messier, Diamond Technology Initiative Symposium, Paper T20,
SDIC/IST ONR, Crystal City, VA, July 12±14, 1988.
[90] C.P. Chang, D.L. Flamm, D.E. Ibbotson, J.A. Mucha, J. Appl. Phys. 63 (1988) 1744.
[91] M. Marchywka, P.E. Pehrsson, D.J. Vestyck, D. Moses, Appl. Phys Lett. 63 (1993) 3521.
[92] A.R. Kirkpatrick, B.W. Ward, N.P. Sconomou, J. Vac. Sci. Tech. B7 (1989) 19471949.
[93] B.R. Stoner, G.-H.M. Ma, S.D. Wolter, J.T. Glass, Phys. Rev. B45 (1992) 11067.
[94] S. Yugo, T. Kimura, T. Kanai, Diamond and Relat. Mater. 2 (1993) 328.
[95] J. Gerber, S. Sattel, K. Jung, H. Ehrhard, J. Robertson, Diamond and Relat. Mater. 4 (1995) 559.
[96] B.R. Stoner, G.H. Ma, S.D. Wolter, W. Zhou, Y.-C. Wang, R.F. Davis, J.T. Glass, Diamond and Relat. Mater. 2
(1993) 142.
[97] X. Jiang, K. Schiffmann, C.-P. Klages, Phys. Rev. B50 (1994) 8402.
[98] X. Jiang, C.-P. Klages, Phys. Stat. Sol. 154 (1996) 175.
[99] S.T. Lee, Y.W. Lam, Zhangda Lin, Yan Chen, Qijin Chen, Phys. Rev. B. 5524 (1997) 15937.
[100] C. Sun, W.J. Zhang, J. Yang, I. Bello, C.S. Lee, S.T. Lee, to be published.
[101] Zhaohui Liu, B.Q. Zong, Zhangda Lin, Thin Solid Films 254 (1995) 3.
[102] M. Tsuda, M. Nakajima, S. Oikawa, J. Am. Chem. Soc. 108 (1986) 5780.
[103] M. Tsuda, M. Nakajima, S. Oikawa, Jpn. J. Appl. Phys. 26 (1987) L527.
[104] C.J. Chu, M.P. D'Evelyn, R.H. Hauge, J.L. Margrave, J. Appl. Phys. 70 (1991) 1659.
[105] L.R. Martin, M.W. Hill, J. Mater. Sci. Lett. 9 (1990) 621.
[106] S.J. Harris, L.R. Martin, J. Mater. Res. 5 (1990) 2313.
[107] S.J. Harris, Appl. Phys. Lett. 56 (1990) 2298.
[108] M. Frenklach, K.E. Spear, J. Mater. Res. 3 (1988) 133.
S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154 153

[109] D. Huang, M. Frenklach, M. Maroncelli, J. Phys. Chem. 92 (1988) 6379.


[110] M. Frenklach, H. Wang, Phys. Rev. B 43 (1991) 1520.
[111] D. Huang, M. Frenklach, J. Phys. Chem. 95 (1991) 3692.
[112] F.G. Celii, J.E. Butler, Appl. Phys. Lett. 54 (1989) 1031.
[113] M. Frenklach, J. Appl. Phys 65 (1989) 5142.
[114] D.G. Goodwin, G.G. Gavillet, J. Appl. Phys. 68 (1990) 6393.
[115] D.M. Gruen, MRS Bull. 239 (1998) 32.
[116] P.C. Redfern, D.A. Horner, L.A. Curtiss, D.M. Green, J. Phys. Chem. 100 (1996) 11654.
[117] Biwu Sun, Xiaopin Zhang, Qinzhe Zhang, Zhangda Lin, Appl. Phys. Lett. 62(1) (1993) 31.
[118] T. Shimanouchi, Table of Vibrational Frequencies, Natl. Bur. Stand. Ref. Data Ser., Natl. Bur. Stand. (U.S.) Circ.
No. 39, U.S. GPO, Washington, DC, 1968, Consolidated Vols. I and II
[119] J. Phys. Chem. Ref. Data 6 (1977) 993.
[120] H. Ibach, D.L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations, Academic New York, 1982.
[121] L.B. Zhao, K.A. Feng, Z. D Lin, Diamond Rel. Mater. 3 (1993) 155.
[122] A.V. Hamza, G.D. Kubiak, R.H. Stulen, Surf. Sci. 237 (1990) 35.
[123] S.P. Mehandru, A.B. Anderson, Surf. Sci. 248 (1991) 369.
[124] T. Tsuno, T. Imai, Y. Nishibayasi, K. Hamada, N. Fujimori, Jpn. J. Appl. Phys. 30 (1991) 1063.
[125] L.F. Sutcu, Appl. Phys. Lett. 60 (1992) 1685.
[126] Biwu Sun, Xiaopin Zhang, Zhangda Lin, Phys. Rev. B. 47(15) (1993) 9816.
[127] K.A. Feng, Z.D. Lin, J. Phys Chem. Solids 556 (1994) 525.
[128] K. Kobashi, K. Nishimura, Y. Kawate, T. Horiuchi, Phys. Rev. B. 38 (1988) 4067.
[129] Y.J. Baik, K.Y. Eun, Applications of Diamond Films and Related Materials, Elsevier, New York, 1991, p. 521.
[130] Ch. Wild, N. Herres, P. Koidl, J. Appl. Phys. 68 (1990) 973.
[131] Zhangda Lin, Biwu Sun, J. Vacuum Society Japan 37(7) (1994) 579.
[132] S.J. Harris, A.M. Weiner, J. Appl. Phys. 67 (1990) 6520.
[133] S.J. Harris, D.N. Betton, A.M. Weiner, S.J. Schmieg, J. Appl. Phys. 66 (1989) 5353.
[134] B.J. Waclawski, D.T. Pierce, N. Swanson, R.J. Celotta, J. Vac. Sci. Technol. 21 (1982) 368.
[135] T.R. Anthony, Vacuum 41 (1990) 1356.
[136] K. Ishibashi, S. Furukawa, Jpn. J. Appl. Phys. 24 (1985) 912.
[137] X. Jiang, W.J. Zhang, M. Paul, C.-P. Klages, Appl. Phys. Lett. 68 (1996) 1927.
[138] W.J. Zhang, X. Jiang, Appl. Phys. Lett. 68 (1996) 2195.
[139] W.J. Zhang, X. Jiang, Y.B. Xia, J. Appl. Phys. 82 (1997) 1896.
[140] Y. Muranaka, H. Yamashita, H. Miyadera, J. Vac. Sci. Technol. A. 9 (1991) 76.
[141] S.J. Harris, A.M. Weiner, Appl. Phys. Lett. 55 (1989) 2179.
[142] Y. Liou, A. Inspektor, R. Weimer, D. Knight, R. Messier, J. Mater. Res. 5 (1990) 2305.
[143] Biwu Sun, Xiaopin Zhang, Qinzhe Zhang, Zhangda Lin, J. Appl. Phys. 73(9) (1993) 4614.
[144] S.-T. Lee, Y.W. Lam, Zhangda Lin, Yan Chen, Zhiqiang Gao, Phys. Rev. B. 54 (1996) 14185.
[145] A. Badzian, T. Badzian, X.H. Wang, Carbon 28 (1990) 804.
[146] M. Kamo, H. Yurimoto, Y. Sato, Appl. Surface Sci. 33/34 (1988) 553.
[147] B.V. Spitsyn, L.L. Bouilov, B.V. Derjaguin, J. Crystal. Growth 42 (1981) 219.
[148] B.V. Derjaguin, B.V. Spitsyn, A.E. Gorodetsky, A.P. Zakharov, L.I. Bouilov, A.E. Aleksenko, J. Crystal. Growth 31
(1975) 44.
[149] H. Nakazawa, Y. Kanayama, M. Kamo, K. Osumi, Thin Solid Films 151 (1987) 199.
[150] K. Kobashi, K. Nishizawa, Y. Kawate, T. Horiuchi, Phys. Rev. B. 38 (1988) 4067.
[151] N. Fujimori, T. Imai, H. Nakahata, H. Shiomi, Y. Nishibayashi, Mater. Res. Soc. Symp. Proc. 162 (1990) 109.
[152] M. Yoshikawa, H. Ishida, A. Ishitani, T. Murakami, S. Koizumi, T. Inuzuka, Appl. Phys. Lett. 57 (1990) 428.
[153] S. Koizumi, T. Murakami, K. Suzuki, T. Inuzuka, Appl. Phys. Lett. 57 (1990) 563.
[154] S. Koizumi, T. Inuzuka, Jpn. J. Appl. Phys. 32 (1993) 3920.
[155] A. Argoitia, J.S. Ma, J.C. Angus, Appl. Phys. Lett. 63 (1993) 1336.
[156] A. Argoitia, J.C. Angus, J.S. Ma, L. Wang, P. Pirouz, W. Lambrecht, J. Mater. Res. 9 (1994) 1849.
[157] R.O. Jones, O. Gunnarsson, Rev. Mod. Phys. 61 (1989) 689.
[158] T. Tomizuka, S. Shikata, Jpn. J. Appl. Phys. 32 (1993) 3938.
[159] R.C. De Vries, Cubic Boron Nitride: Handbook of Properties, G.E. Representative. No. 72 CRD, 1972, p. 178.
[160] T. Suzuki, A. Argoitia, Phys. Stat. Sol. 154 (1996) 239.
[161] X. Jiang, C.L. Jia, Appl. Phys. Lett. 679 (1995) 1197.
[162] C.L. Jia, K. Urban, X. Jiang, Phys. Rev. B 52 (1995) 5164.
[163] Jie Yang, Zhangda Lin, Li-Xin Wang, Sing Jin, Ze Zhang, Appl. Phys. Lett. 65(25) (1994) 3203.
[164] Ke-an Feng, Jie Yang, Zhangda Lin, Phys. Rev. B. 51(4) (1995) 2264.
[165] Jie Yang, Zhangda Lin, Lixin Wang, Sing Jin, Ze Zhang, J. Phys. D: Appl. Phys. 28 (1995) 1153.
[166] Qijin Chen, Li-Xin Wang, Ze Zhang, Jie Yang, Zhangda Lin, Appl. Phys. Lett. 68(2) (1996) 176.
[167] X. Jiang, C.-P. Klages, M. RoÈsler, R. Zachai, M. Hartweg, H.-J. FuÈsser, Appl. Phys. A 57 (1993) 483.
154 S.-T. Lee et al. / Materials Science and Engineering 25 (1999) 123±154

[168] X. Jiang, K. Schiffmann, C.-P. Klages, D. Wittorf, C.L. Jia, K. Urban, W. Jaeger, J. Appl. Phys. 83 (1998) 2511.
[169] H. Kawarada, Ch. Wild, N. Herres, R. Locher, P. Koidl, J. Appl. Phys. 81 (1997) 3490.
[170] A. Argoitia, J.C. Angus, L. Wang, X.I. Ning, P. Pirouz, J. Appl. Phys. 73 (1993) 4305.
[171] W. Zhu, P.C. Yang, J.T. Glass, Appl. Phys. Lett. 63 (1993) 1640.
[172] T. Tachibana, Y. Yokota, K. Miyata, K. Kobashi, Y. Shintani, Diamond Rel. Mater. 5 (1996) 197.
[173] W. Liu, D.A. Tucker, P. Yang, J.T. Glass, J. Appl. Phys. 78 (1995) 1291.
[174] K. Ohtsuka, K. Suzuki, A. Sawabe, T. Inuzuka, Jpn. J. Appl. Phys. Part 2 35 (1996) L1072.
[175] K. Ohtsuka, H. Fukuda, K. Suzuki, A. Sawabe, Jpn. J. Appl. Phys. Part 2. 36 (1997) L1214.
[176] T. Saito, S. Tsuruga, N. Ohya, K. Kusakabe, S. Morooka, H. Maeda, A. Sawabe, K. Suzuki, Diamond Rel. Mater. 7
(1998) 1381.
[177] M. Schreck, H. Roll, B. Stritzker, Appl. Phys. Lett. 74 (1999) 650.
[178] X. Jiang, C.-P. Klages, R. Zachai, M. Hartweg, H.-J. Fuesser, Diamond Rel. Mater. 2 (1993) 407.
[179] X. Jiang, K. Schiffimann, A. Westphal, C.-P. Klages, Appl. Phys. Lett. 639 (1993) 1203.
[180] M. Schreck, K.H. Thuerer, R. Klarmann, B. Stritzker, J. Appl. Phys. 81 (1997) 3096.
[181] M. Schreck, K.H. Thuerer, B. Stritzker, J. Appl. Phys. 81 (1997) 3092.
[182] X. Jiang, W.J. Zhang, C.-P. Klages, Phys. Rev. B58 (1998) 7064.
[183] C. Wild, P. Koidl, W. Mueller-Sebert, H. Walcher, R. Kohl, N. Herres, R. Locher, R. Samlenski, R. Bren, Diamond
Rel. Mater. 2 (1993) 158.
[184] W.J. Zhang, X. Jiang, C.-P. Klages, J. Cryst. Growth 171 (1997) 485.
[185] X. Jiang, C.L. Jia, Appl. Phys. Lett. 69 (1996) 3902.
[186] S.D. Wolter, B.R. Stoner, J.T. Glass, P.J. Ellis, D.S. Buhaenko, C.E. Jenkins, P. Southworth, Appl. Phys. Lett. 6211
(1993) 1215.
[187] W. Zhu, X.H. Wang, B.R. Stoner, G.H.M. Ma, H.S. Kong, M.W.H. Braun, J.T. Glass, Phys. Rev. B. 47 (1993) 6529.
[188] R.Q. Zhang, W.L. Wang, J. Esteva, E. Bertran, Thin Solid Films 3171-2 (1998) 6.
[189] X. Jiang, R.Q. Zhang, G. Yu, S.T. Lee, Phys. Rev. B, to be published.
[190] Wei Zhu, Hua-Shuangn Kong, Jeffrey T. Glass, Characterization of diamond films, in: Robert F. Davis (Ed.),
Diamond Films and Coatings, Noyes Publications, Park Ridge, NJ, USA, pp. 244±338.
[191] Hiroshi Kawarada, Surface Science Reports 26 (1996) 205.
[192] B.J. Waclawski, D.T. Pierce, N. Swanson, R.J. Celotta, J. Vac. Sci. Technol. 21 (1982) 368.
[193] S.-T. Lee, G. Apai, Phys. Rev. B. 48 (1993) 2684.
[194] T. Aizawa, T. Ando, M. Kamo, Y. Sato, Phys. Rev. B 48 (1993) 18348.

Potrebbero piacerti anche