Pure & Appl. Chem., Vol. 66, No. 9, pp. 1789-1796, 1994.
Printed in Great Britain.
(6 1994 IUPAC
Diamond-like carbon
J Robertson
Engineering Dept, Cambridge University, Cambridge CB2 lPZ, UK
Abstract: Diamond-like carbon (DLC) is a dense, partially sp3 bonded form of
amorphous carbon prepared by ion beam or plasma deposition and frequently used as
a hard coating material. Its sp3bonding arises from C+ ions penetrating surface layers
and giving a quenched-in density increase. The formation of DLC can be viewed as
a phase transition to a denser metastable phase. The atomic structure of DLC consists
of a network of sp3 and sp2 sites. The ?r states of sp2 sites control the electronic
properties and the connectivity of sp’ sites controls the mechanical properties.
Introduction
There is great current interest in hard ceramics prepared by vapour-deposited such as diamond, diamond-like
carbon (DLC), cubic BN, boron carbides and TiN. Diamond’s unique properties make it an excellent
coating material and a potentially important high temperature semiconductor [1,2]. While the technology
of diamond deposition develops, interest turned to related materials such as DLC [1,3-51. DLC is a dense,
metastable form of amorphous carbon (a-C) or hydrogenated amorphous carbon (a-C:H) containing a
significant sp3 bonding. The sp3 bonding confers valuable ’diamond-like’ properties such as mechanical
hardness, low friction, optical transparency and chemical inertness. Although DLC films have poorer
properties than diamond films, they have some advantages, notably deposition at room temperature,
deposition onto Fe or plastic substrates and superior surface smoothness.
Deposition
Diamond itself can be deposited by various chemical vapour deposition (CVD) methods such as hot
filament, microwave plasma, etc [1,2]. Hydrocarbon source gases are used and the role of the power source
is to create an excess of atomic hydrogen which alters the thermodynamic stability of the depositing surface.
DLC was originally prepared by ion beam deposition [6] and is now prepared by many methods such as
magnetron sputtering [7], ion sputtering [8], laser plasma deposition [9], plasma deposition [4] and ion
plating [5]. The common factor in these process is deposition from a beam containing medium energy (10-
500 eV) ions [l]. This links DLC to the wider field of ion-beam modification of materials [lo].
Plasma deposition can be used to prepare CVD diamond, a-C:H and the important semiconductor a-Si:H
so it is of interest to compare the preferred conditions for each case. Diamond is typically grown on
substrates held at 700-900°C from a 99: 1 H,/methane gas mixture under moderate pressures (250 Pa) in
microwave plasmas, to maximise the atomic hydrogen flux and minimise the ion bombardment of the film.
The major growth species are radicals such as CH,’. Diamond-like a-C:H is grown on room temperature
substrates from pure hydrocarbons such as acetylene, methane or benzene, under conditions of low pressure
(3 Pa), moderate RF power, low pressure (3 Pa) and negative substrate bias to maximise the positive ion
flux at the substrate [4]. High quality a-Si:H is grown from pure silane at low RF or microwave power and
medium pressure onto substrates held at 250°C [ll]. These conditions give a growth species of SiH,’
radicals, moderate surface mobility, and minimum ion bombardment of the sample.
1789
Unauthenticated
Download Date | 8/7/18 8:31 PM
1790 J. ROBERTSON
SP3
0 ta-C:H A
...-
nn films
L L -_A-
Fig. 1. Ternary diagram showing sp', sp' and H Fig. 2. sp3 fraction vs. ion energy for ta-C
content of a-C, PD a-C:H [12] and ta-C:H [20]. deposited from filtered ion beam [16].
Properties of DLC
The nature of the DLC depends on the deposition process used. The sp' and H content is most conveniently
displayed on a ternary phase diagram [12] as in Fig. 1. The three comers correspond to diamond, graphite
and the hydrocarbons, respectively. There is an excluded region at high H contents, where molecular solids
cannot form. The familiar forms of non-crystalline carbon such as glassy carbon and evaporated a-C lie in
the sp' corner, Sputtering methods produce hard but predominantly sp' bonded a-C [T. Ion-assisted
sputtering onto well-cooled substrates can produce a highly sp' bonded a-C [8]. Laser plasma methods can
produce rather highly sp3 bonded a-C films [9].
A particularly useful form of ion beam deposition is filtered ion beam deposition in which a magnetic filter
removes neutrals, particulates and other ions and allows deposition from a monochromatic, single species
ion beam [13-161. The resulting a-C attains an sp3content of up to 85% and is called highly tetrahedral a-C
or 'ta-C' [15]. Diffraction data suggest that its atomic structure is truly amorphous and resembles the 4-fold
coordinated random network of a-Si, except for the finite content of 3-fold sites [17]. The properties of ta-C
depend strongly on the energy of the ion beam. The sp' content, density and hardness each pass through
a maximum at an optimum ion energy of order 140 eV [16](Fig. 2). Ta-C films also possess a very high
intrinsic compressive stress arising from the deposition process which turns out to be a key signature of
DLC [15]. A linear correlation is found between sp3 fraction and density (Fig. 3) and the stress (Fig. 4).
Plasma deposition (PD) is the most popular means of producing a-C:H [4]. The substrate is attached to the
RF powered electrode which acquires a negative DC self-bias, so the substrate attracts positive ions from
the plasma. The proportion of ions in the total deposition flux is relatively low, typically 10% [18]. The
0 a C , Fallon
2.8 I rn a-C, McKenzie
A a-C:H, Tamor
0 ,,' 0
I:,' a-C
01.j " ' " ' I ' " I " " " " " ' " '
0 2 4 6 8 10 12 I
Stress, GPa
Fig. 3. Density vs. sp3 content for ta-C [16] and Fig. 4. Compressive stress vs. sp3 fraction for
ta-C:H-[20]. ta-C [15,16], PD a-C:H [19] and ta-C:H [20].
Unauthenticated
Download Date | 8/7/18 8:31 PM
Diamond-like carbon 1791
3.2
3
m
E 2.8
Y
g2.6 -
02.4
C O!
0" 2.2
2
[Link] '
100
' '
200
' '
300
' '
400
'
SdO
properties of PD a-C:H depend primarily on bias volt-
age vb and source gas. v b is a measure of the mean ion
energy and at typical operating pressures (3 Pa) Ei r:
0.4Vb. Overall, H content and sp3 fraction decline
steadily with Vb (Fig. 5 ) so that a-C:H has polymeric
character at low v b , has a maximum density and dia-
mond-like character at intermediate V, and has disorde-
red sp2 bonded character at high v b [4,19]. The overall
density variation is reminiscent of ta-C, except that the
energy scale now depends on source gas. Note that PD
Fig. 5 . sp3 fraction, H content, density, a-C:H (open symbols in Fig. 1) still has sizeable H and
band gap of PD a-C:H, after [4,19]. sp2 content which limits its diamond-like character.
Recently, a highly tetrahedral form of a-C:H, 'ta-C:H', was deposited from acetylene using a novel plasma
beam source [20]. The electrode configurations allow the powered electrode and plasma to acquire a
negative self-bias and a plasma beam exits through a grid into the substrate chamber [21]. The properties
of ta-C:H also depend strongly on ion energy, with the density and sp3 fraction reaching a peak at 200 eV
per ion (Fig. 6). The high sp3 fraction (75% of total C) attained by ta-C:H sets it well above other forms
of a-C:H in Fig. 1. The ta-C:H also has a high intrinsic compressive stress, and its stress and density each
vary linearly with sp3 fraction (Figs. 3,4). The high sp3 fraction of this ta-C:H arises from three factors:
the high ionisation of the plasma beam, the mono-energetic character of the ions and the predominance of
one ion, C2Hx+,due to the use of acetylene as a source gas with its simple ionisation pattern.
Deposition Mechanism
The process creating the metastable, densified phase DLC is called ion subplantation [14,22-241. At the
moderate ion energies of interest, ions they loose their energy mainly by elastic collisions with target nuclei
(nuclear stopping) and their range is only a few monolayers. In subplantation, incident ions of sufficient
energy penetrate the surface of the growing film, enter interstitial subsurface positions and increase the local
density. Penetration occurs by direct entry or knock-on displacement of a surface atom (Fig. 7). Lower
energy ions have insufficient energy to penetrate the surface and just stick to the surface to form sp? a-C.
Higher energy ions penetrate further into the film and increase the density in deeper layers. However, ions
need only exceed a certain threshold energy (just less than the displacement threshold) to penetrate the
surface, any excess energy rapidly dissipates in a 'thermal spike' [25] during which the excess density can
relax by thermally activated diffusion (Fig. 7c). An optimum ion energy occurs where the surface
penetration is maximised but relaxation is minimised. Under the energetic conditions of ion bombardment,
the local bonding hybridisation adjusts to the local density, becoming more sp3 at high density and more
sp2density at a lower density. The fractional increase in density ApIp,, for a film growing from an incident
beam containing a fraction 4 of ions of energy Ei is given by
Unauthenticated
Download Date | 8/7/18 8:31 PM
1792 J. ROBERTSON
relaxation during
ion subplantation thermal spike
direct 1 0 0 0
entry
knock-on
:1 0 100
ion energy (ev)
Fig. 7. Two stages of subplantation; ion Fig. 8. Schematic penetration probability (f) and
penetration and density relaxation. ion energy distribution (IED) for filtered ion
beam, plasma, and sputter deposition.
ApIpO = +
f/[( l/d)-f 0.0 1~P(E~/E#~] (1)
where E,, is the activation energy of relaxation, f is the penetration fraction and p is a constant of order 1.
The subplantation mechanism describes well the density variation of ta-C films, as seen in Fig. 2, with
&=3.1 eV, p0=2.2 g m . ~ m -p~ =0.1, and f represented empirically by [23] (Fig. 8)
with the penetration threshold Ep =25 eV and a spread factor E2=57eV [16].
When the depositing species is a molecular ion such as C2H2+,the ion fragments on impact at the surface.
Its energy partitions, by momentum conservation, roughly equally between two daughter C+ ions [HI,
which then undergo separate subplantations (Fig. 8). The penetration of the daughter ions is characterised
by their energy, which is 46% of the parent ion energy for C2H2+.The thermal spikes of each daughter ion
overlap (Fig. Sb), so the relaxation is characterised by the total energy - the energy of the parent ion. With
these adjustments, the subplantation model can also describe the density variation of ta-C:H [20], as shown
in Fig. 6. The penetration threshold rises markedly to 55 eV, as it now includes the energy needed to break
the C-C bond of C2H2+on impact. The sharpness of the density dependence in Fig. 6 is the key reason why
a monochromatic, single species ion beam is needed to give ta-C:H.
The deposition process is essentially the same in conventional PD a-C:H [22,23]. The lower ionisation of
the incident beam, the presence of a number of different ions and the broad ion energy spectrum (due to
collisions in the plasma sheath, Fig. 9b) convert the sharp density peak of ta-C:H (Fig. 6) into the wider,
shallower density peaks seen in Fig. 5. Note that the density of a-C:H prepared from methane, acetylene
and benzene reaches a peak at ion energies approximately in the ratio of 1:2:6 (Fig. 5), that is the same
energy per daugther C atom after energy partition [22]. The ion flux also causes dehydrogenation. The
incident species have a H/C ratio of a least 1. Ions cause hydrogen loss by preferential sputtering, creating
H2molecules which effuse out of the film. The sputtering cross-section increases with ion energy, causing
the H content to decrease with bias.
The deposition mechanism of sputtered and laser-deposited a-C may also involve subplantation, although
further proof is desirable. Sputtering can involve a range of ion energies, the mean is of order 10 eV [8],
lower than in ion beam deposition, and lower than the penetration threshold of C, 25 eV. The penetration
Unauthenticated
Download Date | 8/7/18 8:31 PM
Diamond-like carbon 1793
threshold equals the displacement threshold minus the cohesive energy [23]. In practice, the surface of a-C
may be described by a distribution of penetration thresholds, with local low density regions having low
thresholds, as shown schematically in Fig. 9(c). The penetration probability, f i n (l), would then arise from
the convolution of the low energy tail of the penetration distribution and the high energy tail of the ion
energy distribution. This would give the relatively low sp’ fractions (but disordered structures) found in
sputtered a-C [7,3]. A generalisation of such a model could describe many ion beam modification processes,
currently given more qualitative descriptions [10,27].
0 0
h
penetration - ions separate
relaxation - si’ngle spike
I
Fig. 9. Subplantation by a molecular ion, with
impact fragmentation, separate penetration and Fig. 10. Energy vs. volume and P-T curves for
a common thermal spike. phases. transitions between a-C phases.
The creation of DLC can also be viewed as a phase transition to a denser phase. The transition from graph-
ite to diamond occurs when the Gibbs free energy of graphite G(P,T) rises above that of diamond. The tem-
perature dependence of the transition pressure between graphite and diamond is known as the Berman-Simon
line. McKenzie [15, 271 recognised that phases present in deposited thin films could also be described by
equilibrium thermodynamics and free energy differences, despite the irreversible nature of the deposition
process. The key observation is that most deposition processes create an intrinsic stress, which can be
compressive or tensile [28], although it is rarely as large as in DLC. This stress u is biaxial and is equi-
valent to a hydrostatic pressure of 2u/3 plus a shear of d 3 . The hydrostatic component a compressive stress
can therefore stabilise a denser metastable phase. McKenzie et al [29] applied these ideas to the deposition
of cubic BN (c-BN). BN, like carbon, has a low density sp’ bonded phase hexagonal BN (h-BN) and a
denser but less stable sp3 phase, c-BN. It differs from C in that its amorphous phases are less prevalent.
McKenzie et al [29] found that c-BN was formed when the measured compressive stress exceeded the criti-
cal value expected from the known AG of c-BN.
In the case of a-C, the fully sp’ bonded a-C is expected to form when the compressive stress exceeds the
necessary threshold. Experimentally, the sp’ content of ta-C and ta-C:H varies linearly with stress, and does
not display a threshold value (Fig. 4). This suggests that a-C shows a continuous phase transition of
increasing sp’ content as shown in Fig. lo@). This is possible because the random network can accommo-
date a variable sp3 content, each with its own AG versus V curves (Fig. lOa), whereas a sharp transition
must occur between the crystalline phases h-BN and c-BN. The difference AG between sp2and sp’ a-C can
be estimated as 1 PAV from the difference in density of each form (from Fig. 3) and the hydrostatic
pressure equivalent to the total stress of 13 GPa needed to form fully sp’ a-C (Fig. 4) to give
AG = 0.10 eV (3)
This value is larger than that 0.03 eV between graphite and diamond at O’K, presumably due to the large
distortion energy of the very stiff sp’ bonded random network.
Unauthenticated
Download Date | 8/7/18 8:31 PM
. -
0 graphite --c
Energy - I 1 iL--i- -I-A L L L-
L l i
We now consider the atomic structure and electronic structure of a-C and a-C:H. The well known examples
of amorphous semiconductors and glasses such as a-Si and a-SiO, are random, a-bonded covalent networks.
Their chemical bonding is generally the same as in the crystal, with the same coordination number and bond
length. Disorder is primarily topological, giving 5- and 7-membered rings in a-Si in addition to the 6-
membered rings, ’chairs’, of crystalline Si. The disorder energy of a-Si roughly equals the elastic energy
due to bond length and bond angle distortions. Electronic structure is primarily a function of short range
order. Thus, the electronic density of states (DOS) resembles a smoothed version of the crystalline DOS.
a-C and a-C:H differ in that they possess both a and a bonding. This situation is complex (a) because sp”
and sp3 sites have similar energies and (b) because a bonding allows more possibilities such as resonance
and longer range bonding. Robertson and O’Reilly [30,31] proposed a model for the case of small disorder
energy, based on analogies to organic molecules. They noted that the a and a states are largely separable,
as in the Huckel approximation. The a bonds of sp’ and sp’ sites again form the network’s skeleton, and
their energetics depend on only their bond lengths and bond angles. The a states of sp2sites in general form
a half-filled band. They will gain energy if their arrangement opens up a gap at the Fermi level &, as this
lowers the energy of the occupied states. Many arrangements of sp’ sites will do this. The simplest is a
pairing of sp’ sites, as in ethylene. Further energy is gained if sp’ sites form planar 6-fold rings (aromatic
stabilisation), and yet further energy if these rings fuse together into aromatic clusters, preferably compact
clusters. In this manner, the binding energy of ?r states grows with sp” cluster size. However, the a binding
energy is independent of clustering. Thus, the total binding energy is maximised if sp’ sites and sp’ sites
segregate into a series of a-bonded clusters embedded in a sp’ bonded matrix.
The a states form the band edges and so these states determine the band gap (Fig 11). The band gap of flat
aromatic clusters can be calculated in the Huckel approximation to be [30]
E = 2R/M1/’ = 6IM’” eV (4)
where M is the number of rings in the cluster. This allows M to be deduced from the observed optical gap.
A major factor in favour of aromatic clusters is that they are the smallest configurations to account for the
relatively narrow observed optical gaps of 1 - 2 eV [30-321.
It is now clear that this cluster model is a reasonable model of glassy C but over-estimates the degree of
clustering in interesting forms of a-C such as DLC. Diffraction spectra [33] and molecular dynamics studies
[34-361 suggest a much lower degree of ordering of the sp’ sites. The problem is that the disorder potential
in interesting forms of a-C and a-C:H is quite large, and in particular it is larger than the ordering energies
of sp’ sites. Much of our understanding of a-bonded networks is gained from a-Si:H. Here, high quality
a-Si:H has a disorder potential (deduced from the Urbach slope) of only 0.05 eV, much less than a bond
energy 2.3 eV or even a ’weak bond’ energy of 1 eV. In contrast, DLC is purposefully prepared under
conditions of ion bombardment, to promote sp’ bonding, which gives a high disorder of typically 0.2-0.3
Unauthenticated
Download Date | 8/7/18 8:31 PM
Diamond-like carbon 1795
eV. This is greater than typical ordering energies, which are shown as a function of number of atoms in
the cluster in Fig. 12. For example, the aromatic stabilisation of 6-fold rings is estimated from hydrogena-
tion energies to be 23 kcals or 0.22 eV per site [37] which is similar to the disorder potential. The further
3 per site is unlikely to be recovered as it depends so
energy gain due to fusing of 6-fold rings of ~ 0 . eV
slowly on cluster size. The disorder energy AE is seen to be similar to the energy gained by clustering. This
suggests that clustering will be inhibited by ion bombardment.
A modified model is therefore required to describe the actual case of relatively high disorder. The sp’ sites
are likely to form only pairs, single aromatic rings or very small aromatic clusters. However, we must then
account for the small size of the gap. Tight-binding calculations on 6-fold sp’ rings have found that strong
out-of-plane ’chair’ distortions of B symmetry will narrow the gap to about 1.2 eV. This would account the
optical gaps of most samples of a-C:H but is still too large for sputtered a-C. Odd-membered rings were
also considered. Neutral odd-membered rings are not favoured energetically because they have half-filled
states near Ep. However, a combination of a negatively charged 5-fold ring and a positive 7-fold ring,
denoted R;-R,+, has states at -1.25 eV and 0.05 eV, giving a gap of 1.3 eV. Out-of-plane chair distortions
of these rings narrows the gap to 0.8 eV. It is therefore possible to account for the observed optical gaps
in terms of distorted, single sp2 ring structures.
The relative importance of aromatic and olefinic groups still needs more consideration. The original basis
for aromatic groups was to account for the observed gap. The presence of aromatic groups is supported by
the presence of some shoulder or peak in the Raman spectra at the 1350 cm-’ D mode of disorder graphitic
units in a-C:H [4]. However, high resolution neutron diffraction data of Burke [38] find a C-C sp2 bond
length of < 1.40A, suggesting more olefinic than aromatic bonding. The early infra-red spectra of Dischler
[39] also showed more olefinic =CH groups in as deposited a-C:H. Molecular dynamics simulations have
not found large aromatic clusters. Galli et al [34] found that the sp2 sites of a-C of density 2.2 g m . ~ m - ~
formed thick layers with numerous odd-membered rings. Frauenheim et a1 [35] find that the sp2 sites of a-
C:H tend to form short chains rather than rings. In contrast, Wang et al [40,41] find mainly aromatic
ordering of sp’ sites in simulations of low density a-C and higher density ta-C.
The electronic DOS of a-C(:H) is probed experimentally by a combination of photoemission for the valence
band [42,43] and electron energy loss for the conduction band [16].The ?r states form a knee at top of the
valence band and the relative weight of this feature can be used to deduce the sp2 fraction. The sharpness
of valence band features shows the transition from polymeric to diamond-like character [42]. The empty
?r* states form a pronounced peak in the C K-edge energy loss spectra at 285 eV below the u* edge at 288
eV. The area of the ?r* peak is a popular, accurate means of deriving the sp2 fraction [16] of a-C(:H) films
(infra-red spectra are frequently used, but are inaccurate as they ignore unhydrogenated C sites [12]).
Mechanical Properties
The valuable mechanical properties of diamond and DLC such as high Youngs modulus and hardness arise
from its strong, directional sp3 bonds. In contrast, the u bonds of graphitic bonding or the C-H bonds of
hydrocarbon polymers contribute little to the modulus because they do not form a three dimensional net-
work. Thus, the elastic modulus E of the DLC network can be shown to depend only on the mean C-C
coordination of the sp3 sites Z,, [44-481. Indeed E varies with &, as [46]
The hardness H is a measure of the yield strength. Yield in covalent solids occurs by bond breaking, so H
can be shown to vary as [47]
with E given by (5). Equations (5) and (6) were found to give a good description of the modulus and
hardness of a-C and a-C:H in terms of the underlying bonding where this is known [47,48].
Acknowledgements. The author is very grateful to M Weiler and H Ehrhardt for discussions on their data.
Unauthenticated
Download Date | 8/7/18 8:31 PM
1796 J. ROBERTSON
References
1 J C Angus, C C Hayman, Science 241 877 (1988)
2 K E Spear, this volume
3 J. Robertson, Prog Solid State Chem 21 199 (1991); Surface Coatings Technol 50 185 (1992)
4 P Koidl, C Wild, B Dischler, J Wagner, M Ramsteiner, Mat Sci Forum 52 41 (1990)
5 Y Catherine, in ’Diamond and Diamond-like Carbon’, NATO AS1 266B (Plenum, 1991) p193
6 S Aisenberg, R Chabot, J Appl Phys 42 2953 (1971)
7 N Cho, K M Krishnan, D K Vries, M D Rubin, C B Hopper, B Brushan, D B Bogy, J Mater Res
5 2543 (1990)
8 J J Cuomo, J P Doyle, J Bruley, J C Liu, App Phys Letts 58 466 (1991)
9 D L Pappas, K L Saenger, J Bruley, W Krakow, J J Cuomo, J Appl Phys 71 5675 (1992)
10 J E Sundgren, in ’Diamond and Diamond-like Carbon’, NATO AS1 226B (Plenum, 1991) p47
11 R A Street, ’Hydrogenated Amorphous Silicon’ (Cambridge University Press, 1991)
12 W Jacob, W Moller, App Phys Lett 63 1771 (1993)
13 J. Koskinen, J App Phys 63 2094 (1988)
14 [Link], S. R. Kasi, J. W. Rabalais, Phys Rev Lett 68 620 (1989); Phys Rev B 41 10468 (1990)
15 D R McKenzie, D Muller, B A Pailthrope, Phys Rev Lett 67 773 (1991)
16 P J Fallon, V S Veerasamy, C A Davis, J Robertson, G Amaratunga, W I Milne, J Koskinen, Phys
Rev B 48 4777 (1993)
17 P H Gaskell, A Saeed, P C Chieux, D R McKenzie, Phys Rev Lett 67 1286 (1991)
18 W Moller, Appl Phys A 56 527 (1993)
19 M A Tamor, W C Vassell, K R Carduner, App Phys Lett 58 592 (1991)
20 M Weiler, S Sattel, K Jung, H Ehrhardt, V S Veerasamy, J Robertson, Appl Phys Lett, to be
published (1994)
21 H Oechsner, B Tomcik, Surface Coating Technol 47 162 (1991)
22 J Robertson, Phil Trans Roy SOCA 342 277 (1993)
23 J Robertson, Diamond Related Mats 2 (1993) 984; ibid 3 361 (1994)
24 C A Davis, Thin Solid Films 226 (1993) 30
25 F Seitz, J S Koehler, in ’Solid State Physics’ ed. F Seitz and D Turnbull, vol 2 p305 (Academic
Press, 1956)
26 I Petrov, F Adibi, J E Greene, L Hultman, J E Sundgren, App Phys Lett 63 36 (1993)
27 D R McKenzie, J Vac Sci Technol B 11 1928 (1993)
28 H. Windischmann, J App Phys 62 1800 (1987)
29 D R McKenzie, W McFall, W Sainty, C A Davis, R Collins, Diamond Related Mats 2 970 (1993)
30 J Robertson, E P O’Reilly, Phys Rev B 35 2946 (1987)
31 J Robertson, Adv Phys 35 317 (1986)
32 J L Bredas, G B Street, J Phys C 18 L651 (1985)
33 F Li, J S Lannin, Phys Rev Lett 65 1905 (1991)
34 G Galli, R M Martin, R Car, M Parinello, Phys Rev Lett 62 555 (1989)
35 Th Frauenheim, P Blaudeck, U Stephan, G Jungnickel, Phys Rev B 48 4823 (1993)
36 C H Lee, W R L Lambrecht, B Segall, P C Kelires, Phys Rev B (to be published 1993)
37 A Streitweiser, C H Heathcock, ’Introduction to Organic Chemistry’ (MacMillan, 1985) p565
38 T Burke, R Newport, W S Howells, K Gilkes, P Gaskell, J Non-Cryst Solids 164 1139 (1993)
39 B Dischler, A Bubenzer, P Koidl, Solid State Commun 48 105 (1983)
40 C Z Wang, K Ho, C T Chan, Phys Rev Lett 70 611 (1993)
41 C Z Wang, K Ho, Phys Rev Lett 71 1184 (1993)
42 P Oelhafen, D Ugolini, S Schelz, J Eitle, in ’Diamond and Diamond-like Carbon’, NATO ASI, vol
226B (Plenum 1991) p377
43 J Schafer, J Ristein, L Ley, J Non-Cryst Solids 164 1123 (1993)
44 J C Phillips J Non-Cryst Solids 34 153 (1979)
45 J C Angus, F Jansen, J Vac Sci Technol A 6 (1988) 1778
46 H He, M F Thorpe, Phys Rev Lett 54 2107 (1985)
47 J Robertson, Phys Rev Lett 68 220 (1992)
48 J Robertson, Diamond Related Mats 1 397 (1992)
Unauthenticated
Download Date | 8/7/18 8:31 PM