PHYSICAL REVIEW B 72, 085401 共2005兲

Raman spectroscopy of hydrogenated amorphous carbons

C. Casiraghi, A. C. Ferrari,* and J. Robertson

Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, United Kingdom
共Received 10 March 2005; revised manuscript received 16 May 2005; published 1 August 2005兲

We present a comprehensive multiwavelength Raman investigation of a variety of hydrogenated amorphous

carbons 共a-C : H兲, ranging from polymeric a-C : H to diamond-like a-C : H and ta-C:H, which allows us to
derive values for their bonding, density, band gap, hydrogen content, and mechanical properties. The Raman
spectra of a-C : Hs show two different trends. In one case, the G peak width increases with G peak dispersion.
In the second case, the opposite trend is found. In the first case, the Raman parameters vary with optical,
structural, and mechanical properties in the same way as in hydrogen-free carbon films. In the second case,
typical of polymeric a-C : H, the G peak width correlates with the density, while the G peak dispersion varies
with the optical gap and hydrogen content. This allows a unified picture of bonding and disorder of all carbon
films. UV Raman is particularly useful for a-C : Hs, as it gives clear measurements in the D and G peaks
spectral region even for highly hydrogenated samples, for which the visible Raman spectra are overshadowed
by photoluminescence. On the other hand, the slope of the photoluminescence background in visible Raman
spectra can be used to estimate the H content. UV Raman measurements also allow the detection of C u H
stretching vibrations.

DOI: 10.1103/PhysRevB.72.085401 PACS number共s兲: 78.30.Ly, 63.50.⫹x, 61.43.Dq

I. INTRODUCTION 共4兲 a-C : H with low H content 共less than 20 at. %兲. They
Diamond-like carbon 共DLC兲 is an amorphous carbon with have a high sp2 content. The gap is under 1 eV. We call these
a significant fraction of CC sp3 bonds.1,2 Tetrahedral amor- films graphite-like a-C : H 共GLCH兲. They are usually depos-
phous carbon 共ta-C兲 is the DLC with the maximum sp3 con- ited by PECVD at high bias,1,2,4 dc glow discharge 共GD兲
tent. We classify hydrogenated amorphous carbons into four systems,14 or magnetron sputtering 共MS兲.15
types, as shown in Fig. 1.3 Raman spectroscopy is a fast and nondestructive method
共1兲 a-C : H films with the highest H content for the characterization of carbon materials.16 All carbons
共40– 60 at. % 兲. These films can have up to 70% sp3.1 How- show common features in their Raman spectra in the
ever, most of the sp3 bonds are hydrogen terminated and this 800– 2000 cm−1 region, the so-called G and D peaks, which
material is soft and has low density.1 We call these films lie at around 1560 and 1360 cm−1, respectively, for visible
polymer-like a-C : H 共PLCH兲. Their band gap ranges from excitation, and the T peak at around 1060 cm−1, which is
2 eV to 4 eV.1 These films are usually deposited by plasma only seen for UV excitation.17–21 The G peak is due to the
enhanced chemical vapor deposition 共PECVD兲 at low bias bond stretching of all pairs of sp2 atoms in both rings and
voltage.1,2,4 chains. The D peak is due to the breathing modes of sp2
共2兲 a-C : H films with intermediate H content atoms in rings.17,22–25 The T peak is due to the C u C sp3
共20– 40 at. % 兲. Even if these films have lower overall sp3 vibrations.19–21
content, they have more C u C sp3 bonds than PLCH. Thus, In previous papers we have shown that multiwavelength
they have better mechanical properties.1 Their optical gap is Raman spectra can be used to distinguish the different types
between 1 and 2 eV.1 We call these films diamond-like of amorphous carbons and to derive their structural and me-
a-C : H 共DLCH兲. They are usually deposited by PECVD1,2,4 chanical properties. We discussed hydrogen-free amorphous
or electron cyclotron resonance 共ECR兲 or reactive sputtering
at moderate bias voltage.5–7
共3兲 Hydrogenated tetrahedral amorphous carbon films
共ta-C:H兲. ta-C:H films are a class of DLCH in which the CC
sp3 content can be increased while keeping a fixed H con-
tent, as in Fig. 1. Many films defined in literature as ta-C:H
are just DLCHs. However, due to the highest sp3 content
共⬃70% 兲 and 25– 30 at. % H, ta-C:Hs are really a different
category as indicated by their Raman spectra, their higher
density 共up to 2.4 g / cm3兲 and Young’s modulus 共up to
300 GPa兲.8 Their optical gap can reach 2.4 eV.9 These films
are deposited by high-density plasma sources such as elec-
tron cyclotron wave resonance 共ECWR兲9,10 and plasma beam
source 共PBS兲.11,12 An alternative way to produce ta-C:H was
also attempted by introducing H while depositing ta-C by FIG. 1. Ternary phase diagram for H free and hydrogenated
filtered cathodic vacuum arc 共FCVA兲.13 amorphous carbons.

1098-0121/2005/72共8兲/085401共14兲/$23.00 085401-1 ©2005 The American Physical Society

CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 2. 共a兲 Amorphization trajectory, showing the schematic variation of the G position for 244 and 514.5 nm excitation energies. 共b兲
Amorphization trajectory, showing the possibility of non-uniqueness in stages 2 and 3 for 244 and 514.5 nm excitation energies wavelengths
共Ref. 18兲.

carbons,17,18 amorphous carbon nitrides,26 and The three-stage model17,18 describes the evolution of the
nanodiamond.27 In this paper, we present a detailed study of Raman spectra of carbons in terms of an amorphization tra-
the Raman spectra of hydrogenated amorphous carbons. We jectory, starting from perfect graphite, Figs. 2共a兲 and 2共b兲.
apply and extend our previous three-stage model and we de- The Raman spectra are dominated by the sp2 sites. Thus, the
rive a unified picture of bonding and disorder in carbons, clustering and disorder of the sp2 phase is the main factor
from two Raman parameters: the G peak width and the G affecting peak positions, width, and intensity. In principle,
peak dispersion. the sp2 clustering can vary independently of the sp3 content.
This paper is organized as follows. In Sec. II we discuss This means one visible Raman spectrum could correspond to
the background concepts for this paper. In Sec. III we give films with different sp3 content. Combining a visible Raman
the experimental details on the samples, the characterization spectrum with a UV Raman spectrum helps in resolving this
problem. For UV Raman excitation the G peak position de-
techniques used to derive the relevant properties, and the
creases with increasing sp2 clustering. If two samples have
Raman measurements and fitting procedures. In Sec. IV we
similar G peak positions in visible Raman but different ones
present our main experimental observations and, in particu- in UV Raman, the sample with the lower G position in UV
lar, a set of useful correlations between single and multi- has higher sp2 clustering, Fig. 2共b兲. Thus, a multiwavelength
wavelength Raman parameters and structure of the films. In Raman analysis is able to fully characterize DLCs. The com-
Sec. V we extend the three-stage model to include hydrogen- bination of visible and UV Raman spectra can be used to
ated amorphous carbons. define the G peak dispersion, Disp共G兲, i.e., the rate of change
II. BACKGROUND
of G peak position with excitation wavelength.
The other two key Raman parameters to monitor carbon
We desire to use Raman spectroscopy to extract bonding bonding are the intensity ratio of the D and G peaks,
parameters of DLCs, such as density, sp3 fraction, or H con- I共D兲 / I共G兲, and the full width at half maximum of the G
tent. However, Raman spectroscopy is mainly sensitive to peak, FWHM共G兲. In amorphous carbons I共D兲 / I共G兲 is a mea-
the configuration of sp2 sites because of their higher cross sure of the size of the sp2 phase organized in rings.17 If
section. The way to derive these properties is a model that I共D兲 / I共G兲 is negligible, then the sp2 phase is mainly orga-
relates sp2 site disorder to the overall bonding, the so-called nized in chains, or, even if rings are present, the ␲ bonds are
three-stage model.17,18 not fully delocalized on the rings.17

085401-2
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

FWHM共G兲 and Disp共G兲 both measure disorder, however, seen for UV excitation.18 However, a typical signature of
FWHM(G) is mainly sensitive to structural disorder, while hydrogenated samples in visible Raman spectra is the in-
Disp(G) is mainly sensitive to topological disorder. Struc- creasing photoluminescence 共PL兲 background for higher H
tural disorder arises from bond angle and bond length distor- content. This is due to the hydrogen saturation of nonradia-
tions. Topological disorder arises from the size and shape tive recombination centers.32–34 The ratio between the slope
distribution of sp2 clusters. m of the fitted linear background and the intensity of the G
Samples with a sp2 phase entirely constituted of fully ␲ peak, m / I共G兲, can be empirically used as a measure of the
delocalized rings do not show any G peak dispersion.18 The bonded H content, as will be shown in Sec. IV.
G peak dispersion arises from the resonant selection of sp2 For H contents over 40– 45 at. %, the PL background ob-
chains of different sizes at different excitation energies.18,19 scures the Raman signal for visible excitation7,35–38 but not
By the size of the sp2 chain we mean the effective chain for UV excitation. This means that we will mainly refer to
conjugation length. In this framework, a long, strained sp2 the FWHM共G兲 for UV excitation in the following sections
chain, with defects interrupting the ␲ delocalization, is because it can be measured for every film.
equivalent to a set of smaller chains. Thus, the G peak dis-
persion is mainly a measure of topological disorder. In con-
III. EXPERIMENT
trast, the G peak width is a single wavelength parameter. It
probes those sp2 clusters resonant at a particular excitation We studied more than 60 a-C : H samples prepared with
energy. Clusters probed at a given excitation energy have different deposition systems: 共1兲 13.56 MHz rf PECVD sys-
roughly similar size. FWHM共G兲 would be small if the clus- tem, with pressure and rf power ranging from 0.02 to
ters were defect-free, unstrained or “molecular.” For a given ⬃ 20 mbar and 10– 250 W, respectively; 共2兲 ECWR
cluster size, a higher bond length and bond angle disorder source.9,39 The rf power was 600 W and the plasma pressure
lead to a higher FWHM共G兲. This implies that FWHM共G兲 is was ranging from 1 to 6 ⫻ 10−3 mbar.39 For these deposition
mainly a probe of structural disorder. Higher excitation en- systems we considered two gases, CH4 and C2H2; 共3兲 Dis-
ergies are resonant with smaller clusters. This allows smaller tributed electron cyclotron resonance 共DECR兲 plasma
differences between configurations and so FWHM共G兲 de- reactor.40–42 In this case only C2H2 was used as precursor;
creases with excitation energy.18,26,28 共4兲 Magnetron sputtering in an Ar/ H2 atmosphere from a
In hydrogen-free carbons, the structural and topological graphite target with flow ratios in the range 0.025–0.9, and
disorder vary in parallel along the amorphization trajectory.18 pressures from 2.5⫻ 10−5 to 1 ⫻ 10−3;15,43 and 共5兲 FCVA,
FWHM共G兲 and Disp共G兲 both increase as disorder increases, used with a H2 supply to produce hydrogenated samples, as
for all excitation energies.18,26 The increase of disorder is described in Ref. 13. Further data were derived from
linked to higher sp3 content, density, and mechanical prop- literature.4,6,7,43–50
erties, which causes the correlation between multi- For comparison, we also considered the Raman spectra of
wavelength Raman spectra and mechanical properties seen some H free carbons: nanocrystalline graphite 共nc-G兲, an
previously.18,28,29 amorphous carbon deposited by MS, and a ta-C deposited by
This still holds for a-C : H’s with H contents under FCVA, as described in Refs. 17, 18, and 28.
20– 30 at. %, such as ta-C:H, DLCH, and GLCH. By further All the samples produced in our laboratory were depos-
increasing the amount of hydrogen over 25% at., the overall ited simultaneously on silicon 共for Raman measurements兲
sp3 content can still increase, but not the C u C sp3 content. and on quartz 共for optical gap measurements兲. The substrates
Indeed, PLCHs have the smallest defect density of any car- were cleaned in ultrasonic bath.
bon films, together with the smallest mass density and stress. The density and hydrogen content 共±5 % 兲 of the samples
Structural disorder now decreases with H content, while to- were determined from nuclear reaction analysis 共NRA兲. The
pological disorder continues to increase,30 so that FWHM共G兲 optical gap 共Tauc gap, ET, and E04兲 was derived by UV-
and Disp共G兲 will have opposite trends. visible spectrophotometry. The Young’s modulus 共E兲 was
Hydrogen has a further important effect, which helps the measured by the laser acoustic wave technique or by surface
analysis of the Raman spectra. Contrary to the introduction Brillouin scattering.51,52
of nitrogen26 or to the effect of annealing,18,31 the introduc- Unpolarized Raman spectra were measured at 244 and
tion of hydrogen into an amorphous carbon links the amount 514 nm excitations using Renishaw micro-Raman 1000
and configuration of the sp2 phase with the overall sp3 con- spectrometers. The UV Raman spectra are collected with a
tent 共C u C + C u H sp3兲. In principle, this is quite fortunate, 40⫻ objective and a UV-enhanced charge-coupled device
as it implies that for a-C : H a single wavelength Raman camera. The spectral resolution is ⬃6 cm−1. All the UV Ra-
study could be enough to quantify the H and sp3 content, the man spectra are corrected by subtracting the background sig-
optical, and mechanical properties. However, we urge ex- nal due to the optics. This is done by measuring an Al mirror
treme caution when doing a single wavelength analysis, background signal and normalizing this signal so that the
since there are some deposition processes that can introduce atmospheric N2 peak at ⬃2332 cm−1, detected for the mirror
sp2 rings even in samples with high H content. In these measurement, has the same intensity of the corresponding
cases, multi-wavelength Raman spectroscopy is necessary to atmospheric N2 peak, detected on the sample. Contrary to
extract reliable information. ta-C, hydrogenated amorphous carbons are particularly sen-
Due to the overwhelming cross section of sp2 sites for sitive to UV excitation and can be easily damaged. In order
visible excitation, the C u H stretching modes can only be to avoid damage, the power on the sample is kept well below

085401-3
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

1 mW 共down to 0.005 mW兲 and the samples are placed on a

spinner rotating at high speed 共⬎3000 rpm兲. The acquisition
time is in the 5 – 60 s range, adjusted to minimize the noise
but still kept below the damage threshold. The change in
acquisition time is allowed here, since what matters is the
overall shape of the spectra, not their absolute intensity.
These accurate procedures ensure no visible damage is pro-
duced on the sample surface and no change of the spectral
shape during the measurements. The visible Raman spectra
are collected with a 100⫻ objective. The spectral resolution
is ⬃2 cm−1. The power on the samples is always kept below
4 mW 共down to 0.4 mW兲 and the acquisition time is always
less than 90 s in order to avoid damage.
There are two main options for spectral shape fitting: an
all Gaussians fit or a fit with a Breit-Wigner-Fano 共BWF兲
line shape for the G peak and a Lorentzian for the D peak. FIG. 3. 共Color online兲 E04 as a function of Tauc gap. The linear
For samples with a significant PL background, such as the fit is shown.
ones studied here, the BWF line is not the most appropriate
since the coupling coefficient is influenced by the PL slope rately, since they contain different information.
and thus the BWF tends to adjust its asymmetry to reproduce When our comments on the trends of the Raman param-
part of the PL slope. This does not allow reproducible fits of eters are meant to be general, we will not specify the exci-
the PL background. On the other hand, fully symmetric tation wavelength. When we refer to particular excitations,
Gaussian lines work better in the presence of a background. we will indicate this by adding @wavelength to the param-
They also allow a direct comparison with previous studies eter. For example, FWHM共G兲@244 indicates the FWHM of
using Gaussian fitting.7,45,49 In any case, the crucial point, the G peak measured at 244 nm.
once the fitting procedure is decided, is to consistently use it
for the whole set of samples one wishes to study. A different, IV. RESULTS
but properly converged and stable fitting procedure, will
yield the same trends, even if the absolute numbers can be The experimental relation between E04 and ET measured
slightly different. for our samples and from literature data7,30,44,47,48 is shown in
Room temperature micro-photoluminescence was mea- Fig. 3. By fitting the data in Fig. 3 we find
sured for 514 nm excitation using a Renishaw RM-1000
spectrometer fitted with a low dispersion grating E04 共eV兲 = 0.46 + 1.06ET 共eV兲. 共1兲
共600 groves/ mm兲. The spectra were recorded in the This was used to estimate ET from some literature data, when
520– 850 nm spectral range 共2.4– 1.46 eV兲. The main PL only E04 was reported or vice versa.
peak is fitted with one Gaussian. The PL intensity is normal- Figures 4共a兲 and 4共b兲 show the 514 and 244 nm Raman
ized to the Raman G peak intensity. On the other hand, the spectra measured on films representative of the four a-C : H
PL background is measured from the 514 nm Raman spectra. classes described in Sec. I. Note that for low H content, the
This is defined as the ratio between the slope of the Raman visible spectra do not exhibit a PL background, while the PL
spectra 共m兲 between 1050 and 1800 cm−1, and the intensity is so strong for the highest H content that the visible spec-
of the G peak, m / I共G兲. trum is featureless. In contrast, a Raman spectrum can al-
We also study the spectral region 2800– 3400 cm−1 of the ways be measured for UV excitation, irrespective of the H
UV Raman spectra. This contains information on the sp3 content. Indeed, UV Raman spectra with two well-separated
CHx 共x = 2 , 3兲 and sp2 aromatic and olefinic stretching features, such as the top spectra in Fig. 4共b兲, are an unam-
vibrations.53 Since the frequencies of these modes are very biguous signature of PLCH films.18
close and the spectra are broad, it is not possible to make a Figures 5共a兲 and 5共b兲 plot Pos共G兲@514 and
precise quantitative mode analysis as previously done in in- I共D兲 / I共G兲@514 as a function of the H content. If H
frared spectroscopy 共IR兲.54,55 However, by UV Raman spec- ⬎ 20 at. %, Pos共G兲@514 and I共D兲 / I共G兲@514 decrease with
troscopy one can assess films deposited on substrates unsuit- increasing H content. At low H content, there is no obvious
able for IR measurements. link between the Raman parameters and H content. This is
The following fit parameters are thus considered in this simple to understand if one considers that different structures
study: Pos共G兲, Disp共G兲, FWHM共G兲, m / I共G兲, I共D兲 / I共G兲, and of amorphous carbons are possible for H = 0, ranging from
the area ratio of the D and G peaks, A共D兲 / A共G兲. In this graphite to ta-C. In Fig. 5 and in the following figures a line
paper the area ratio is used only for convenience, since it indicates the extrapolation to low H content of the ideal trend
allows us to perform a direct comparison with some previ- for as-deposited hydrogenated samples. Similarly, the non-
ously published data, reported only as the area ratio.7,45,49 uniqueness regions are represented by shadowed areas.
However, it must be noted that the area ratio is not the best Both optical gap and visible Raman spectra change ac-
parameter to use, since it is equivalent to the product of the cording to the clustering of the sp2 phase.1,17,18 However,
intensity and width ratios, which are better studied sepa- PLCH and ta-C samples, although having a similar gap, have

085401-4
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 4. Raman spectra of template a-C : H films at 共a兲 514.5 and 共b兲 244 nm.

a very different Pos共G兲@514, as in Fig. 6. The same happens the average PL peak position as a function of the optical gap
for low gap a-C and nanocrystralline graphite 共nc-G兲. This for various films measured by us or taken from
results in the nonuniqueness region shown in Fig. 6. How- literature.33,34,36,58,59 The PL peak position is seen to saturate
ever, for H ⬎ 20 at. % there is a unique relation between vis- when the optical gap exceeds the excitation energy. On the
ible Raman spectra and optical gap. The same unique rela- other hand, Fig. 9共b兲 shows that, by normalizing to the Ra-
tion is observed between visible Raman spectra and H man intensity, the PL intensity continues to increase with H
content 共Fig. 5兲. Indeed, optical gap and H content are pro- content, even for subgap excitation. Normalizing to the G
portional for H ⬎ 20 at. %, as shown in Fig. 7. By fitting the peak intensity is reasonable since, if we neglect the possible
data in Fig. 7 共Refs. 7,45,48,50,56,57兲 we get a formula re- variation of 514 nm Raman cross section across the different
lating H content and Tauc gap for H ⬎ 20 at. % a-C : H samples, the Raman and PL intensities both have a
similar dependence on sample thickness, absorption coeffi-
cient, and laser penetration depth. Thus, for practical pur-
ET 关eV兴 = − 0.9 + 0.09 · H 关at. % 兴. 共2兲
poses, it is not necessary to measure PL over a wide spectral
The 514 nm Raman spectra also provide information on range, as in Fig. 8, but we can take the PL background under
the H content via the PL slope. Figure 8 plots the room the D and G peak spectral regions. Note that PL has some-
temperature PL data for films with H contents of times been seen in some H-free carbons 共Fig. 8 and Ref. 60兲.
20 to 35 at. % and some H-free amorphous carbons. The However, the PL intensity in H-free samples is at least four
spectra have been normalized to the G peak intensity. times lower than that of PLCH, see Fig. 9共b兲. Also note that,
Note that some samples can have Tauc gaps over 2.41 eV, for low density and porous carbons such as in Ref. 60, the
the laser excitation energy used for PL and Raman measure- PL may be associated with adsorbed water during deposition
ments. Thus, if the gap is over 2.41 eV, we only probe the in poor vacuum conditions or after exposure to air.61,62
PL of sub band-gap clusters. This nevertheless allows an Figure 10共a兲 shows that the normalized PL slope, m / I共G兲,
estimate of the H content, because the recombination effi- increases exponentially with the hydrogen content, for all
ciency of sub band-gap clusters also increases with H films examined. The data of Fig. 10共a兲 give a simple quanti-
content.32–34 This is demonstrated in Fig. 9共a兲 where we plot tative formula for the hydrogen content

085401-5
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 7. 共Color online兲 Tauc gap as a function of H content. The

ideal trajectory with H incorporation from nc-G is marked by a line.
The shaded region represents the non-uniqueness at low H contents.

trend as a function of E04 or H content in the range plotted in

Fig. 11. For lower gap or H content the PL intensity is neg-
ligible.
Figures 12共a兲 and 12共b兲 plot Disp共G兲 and
FWHM共G兲@244 as a function of the H content. For H con-
FIG. 5. 共Color online兲 共a兲 Pos共G兲@514.5 and 共b兲 I共D兲 / I共G兲@ tents over 20 at. %, Disp共G兲 increases linearly with H con-
514.5 nm as a function of H content. The ideal trajectories of the tent, while FWHM共G兲@244 shows a maximum at 25 at. %
Raman parameters with H incorporation from nc-G are marked by a H and then decreases. The decrease of FWHM共G兲 and the
line. The shaded regions represent the non-uniqueness at low H increase of Disp共G兲 is different from the case of H-free car-
contents. bon films, where FWHM共G兲 and Disp共G兲 increase together.

再 冎
This is an indication that structural and topological disorder
m evolve differently in highly hydrogenated samples. In fact,
H 关at. % 兴 = 21.7 + 16.6 log 关␮m兴 . 共3兲 PLCH 共⬃40 at. % H兲 and ta-C:H films both have a
I共G兲
relatively high Disp共G兲 共⬃0.3 cm−1 / nm兲, but the
Figure 10共b兲 shows that m / I共G兲 also increases exponen- FWHM共G兲@244 of ta-C:H is larger than that of PLCH with
tially with the Tauc gap for H ⬎ 20 at. %. H = 40 at. % 共165 and 120 cm−1, respectively兲. On the other
Figure 11 plots the spin density in DLCH and PLCH as hand, FWHM共G兲@244 is similar for GLCH and PLCH, but
a function of the gap 共E04兲 using data from Disp共G兲 is smaller for GLCH 共⬃0.10 cm−1 / nm兲. Thus, both
literature.11,44,48,63–68 It also plots the fitting line between Disp共G兲 and FWHM共G兲 values are essential to derive the
m / I共G兲 and E04 共bottom X axis兲 or H content 共top X axis兲.
carbon bonding in a-C : H.
Figure 11 clearly shows that there is a general relation be- Figure 13 plots Pos共G兲@244 as a function of H content.
tween the decrease of spin density and an increase of PL
Pos共G兲@244 increases with H content, moving up to
efficiency, since both spin density and m / I共G兲 have a similar

FIG. 6. 共Color online兲 Pos共G兲@514.5 as a function of Tauc gap. FIG. 8. Room temperature PL spectra for different amorphous
The ideal trajectory with H incorporation from nc-G is marked by a carbons: 共a兲 PLCH with 35 at. % H, 共b兲 DLCH with 30 at. % H, 共c兲
line. The shaded region represents the non-uniqueness at low H DLCH with 25 at. % H, 共d兲 ta-C, 共e兲 ta-C:H, and 共f兲 sputtered
contents. a-C.

085401-6
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 9. 共a兲 Variation of the PL energy with E04. 共b兲 I共PL兲 / I共G兲
as a function of H content. Note that a-C, DLCH with H
⬍ 25 at. % and ta-C:H show a very weak PL, compared with
PLCH. The lines are guides to the eye. FIG. 10. m / I共G兲 as a function of 共a兲 H content and 共b兲 Tauc gap.
The linear fits to the data are shown.
⬃1625 cm−1 for PLCH films. However, for ta-C:H,
Pos共G兲@244 is slightly higher 共⬃1615 cm−1兲 than the cor- lution of a-C : H films for increasing H content.
responding DLCH samples with the same H content. This is As in our previous study of carbon nitrides,26 the easiest
due to the higher fraction of C u C sp3 bonds in ta-C:H.18 way to visualize the modification due to heteroatoms is to
UV Raman also allows the direct detection of C u H compare the properties to the corresponding pure carbon
bonds. This is shown in Fig. 14, where a broad peak due to film. Since hydrogen can only bond at a terminal site, its
CHx stretching is seen at 2920– 2970 cm−1. The other band effects are in principle quite different from nitrogen.
at ⬃3220 cm−1 is the second order G peak, as demonstrated
by H-D isotopic substitution experiments.18 ta-C:H has a
very broad and small second-order G peak compared to
PLCH. The CHx band is quite broad and it is difficult to
properly deconvolute the separate contributions of sp3 CHx
共x = 2 , 3兲 and CH sp2 aromatic and olefinic vibrations.52 In-
frared spectroscopy is certainly preferable for a quantitative
analysis of CHx stretching modes.54,55 On the other hand,
Raman measurements can be performed on any substrate or
shape, such as metal cutting tools, which are not suitable for
IR measurements. It is thus useful to analyze the CH-related
UV Raman features. If the broad CHx peak is fitted with a
single Gaussian, its average position shifts from 2920 cm−1
for PLCH and GLCH to 2960 cm−1 in ta-C:H. This shift can
be interpreted as a lower CH sp3 content or a higher sp2
olefinic or aromatic contribution in ta-C:H, compared to
PLCH or GLCH.53–55
FIG. 11. Defect density and PL efficiency 关m / I共G兲兴 as a func-
V. DISCUSSION tion of gap 共bottom X axis兲 and H content 共top X axis兲. For clarity,
instead of plotting all the experimental data points as in Fig. 10, we
We now extend the three stage model for the Raman spec- only plotted the linear fit 共dotted line兲 representing m / I共G兲 as a
tra of amorphous carbons to fully describe the structural evo- function of gap or H content.

085401-7
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 14. UV Raman spectra showing the CH peak and second

order of the G peak for template PLCH, ta-C:H, and GLCH films.
FIG. 12. 共Color online兲 共a兲 Disp共G兲 and 共b兲 FWHM共G兲@244 as
a function of H content. The ideal trajectories of the Raman param- 共2兲 If H is introduced in ta-C, this increases the order and
eters with H incorporation from nc-G are marked by a line. The clustering in the structure.13 Thus, Disp共G兲 and FWHM共G兲
shaded regions represent the nonuniqueness at low H contents. decrease, while I共D兲 / I共G兲 increases 关Figs. 5共b兲, 12共a兲, and
12共b兲兴.
A. 0 ⬍ H ⬍ 20 at. % 共3兲 If H is introduced in a mainly sp2 amorphous carbon,
In this region the H content is too low to decouple the a-C, it can produce both disordering and ordering effects,
evolution of structural and topological disorder. The Raman and they tend to compensate.
parameters can be interpreted as in H-free films. Generally speaking, for low H contents, the non-
共1兲 If H is introduced in an otherwise nanocrystalline gra- hydrogenated sp2 phase can easily display all the non-
phitic carbon, it breaks bonds and increases the bond angle uniqueness effects typical of low sp3 carbons, as indicated by
and bond length disorder. This is reflected in an increase of the shadowed regions in Figs. 5–7 and 12.
both Disp共G兲 and FWHM共G兲@244, Figs. 12共a兲 and 12共b兲. B. H ⬎ 20 at. %
Thus, the size of the sp2 clusters decreases so that I共D兲 / I共G兲
decreases as in Fig. 5共b兲. GLCH samples can be seen as the For higher H contents, however, the situation is different.
result of H incorporation in nc-G. The ideal trajectories, de- Although the increase in H drastically reduces the size of
scribing the modification of the Raman parameters with a H ring-like sp2 structures, it steadily reduces the density and
incorporation in an otherwise nc-G, are indicated by a line in strain in the samples. The sp2 phase is still topologically
Figs. 5–7 and 12. disordered, but the structural disorder is lowered. Further-
more, as-deposited samples tend to evolve in a unique way
as a function of H content, thus minimizing hysteresis.
In this region, Pos共G兲@514 and I共D兲 / I共G兲@514 decrease
with H as in Figs. 5共a兲 and 5共b兲. Disp共G兲 continues to in-
crease with H, but FWHM共G兲 reaches a maximum for ta-
C:H and then decreases for PLCH 关Figs. 12共a兲 and 12共b兲兴.
Similarly, Pos共G兲@514 in PLCHs decreases to ⬃1520 cm−1,
before being overshadowed by the PL background, Figs. 4
and 5共a兲. In contrast, in ta-C Pos共G兲@514 reaches
⬃1570 cm−1, Fig. 5共a兲. PLCHs also show an higher Disp共G兲
than ta-C:H, Fig. 12共a兲. Within hydrogenated carbon films,
ta-C:H shows the highest FWHM共G兲, and ta-C has the high-
est FWHM共G兲 of all carbons, for any given excitation en-
ergy.
FIG. 13. 共Color online兲 Pos共G兲@244 as a function of H content. As expected, I共D兲 / I共G兲 goes to zero for both PLCH and
The line is a guide to the eye. ta-C and is small for ta-C:H, Fig. 5共b兲, since it is a measure

085401-8
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 16. Correlation between Disp共G兲 and total C u C + C u H

sp3 content in hydrogenated carbon films. Sp3 data from Refs. 4 and
FIG. 15. Correlation between Disp共G兲 and FWHM共G兲@244 for 17.
carbon films. The lines are guides to the eye.
From a linear fit of the data in Figs. 17共a兲 and 17共b兲, we
of sp2 rings and no rings are present in ta-C or PLCH. get the following simple relations for H ⬎ 20 at. %:
Thus, Disp共G兲 and FWHM共G兲 do not follow the same
trends as a function of H content. This differs from the be- ␳ 关g/cm3兴 = 0.257 + 0.011W 关cm−1兴, 共4兲
havior in H-free carbons, where an increase of Disp共G兲 is
always accompanied by a higher FWHM共G兲, but just reflects E 关GPa兴 = − 511 + 4.66W 关cm−1兴, 共5兲
the decoupling of structural disorder from topological disor-
der. These effects are summarized in Fig. 15, which is the where W = FWHM共G兲@244.
key result of this paper. Figure 18 shows the variation of mass density with sp3
For H-free carbons Disp共G兲 and FWHM共G兲 are always fraction for the different carbon films.8 For H-free carbons,
proportional and have the same trend as a function of the sp3 the density increases monotonically and linearly with sp3
content in Fig. 15 共Line 1兲. ta-C:H and GLCH also lie on fraction. A similar behavior is found for GLCH and DLCH
Line 1. In contrast, all the samples with H ⬎ 25– 30 at. %, up to a certain density. Then the ta-C:Hs continue to increase
from DLCH to PLCH, follow the opposite trend 共Line 2兲 in monotonically, whereas the PLCHs turn over and the density
Fig. 15. Thus, in general, the same FWHM共G兲@244, if decreases for higher sp3 values. The similar trends in Figs.
lower than ⬃170 cm−1, is associated to two different
Disp共G兲, depending on the H content. In contrast, the same
Disp共G兲, if over 0.2 cm−1 / nm, is associated with two differ-
ent FWHM共G兲@244 values. However, if one performs a
multi-wavelength investigation and combines Disp共G兲 and
FWHM共G兲@244, one can uniquely characterize the carbon
films.
We now relate the parameters FWMH共G兲 and Disp共G兲 to
the underlying bonding properties. Since FWHM共G兲 is influ-
enced by the structural disorder, we expect this to be the key
parameter to uniquely measure the structural and mechanical
properties, which ultimately depend on the amount of C u C
sp3 bonds for any a-C : H film. On the other hand, we expect
Disp共G兲 to be very sensitive to the evolution of the sp2 phase
in any hydrogenated carbon film, and to be uniquely related
to the optical properties and hydrogen content.
Since for low H contents Disp共G兲 and FWHM共G兲 have
the same trends, and for high H contents Disp共G兲 uniquely
evolves with H, and thus with the amount of C u H sp3
bonds, we expect a relation between Disp共G兲 and the total
sp3 content in a-C : H 共C u C sp3 + C u H sp3兲. This is dem-
onstrated in Fig. 16. Note that ta-C:H and ta-C lie outside the
a-C : H line in Fig. 16.
However, since density and Young’s modulus mainly de-
pend on the C u C sp3 content, in hydrogenated carbons they
cannot be measured by Disp共G兲, but by FWHM共G兲, Figs. FIG. 17. 共a兲 Density and 共b兲 Young’s modulus as a function of
17共a兲 and 17共b兲. FWHM共G兲@244. The linear fits are shown.

085401-9
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 18. Density as a function of the sp3 content for H free and
hydrogenated amorphous carbons 共Ref. 8兲.

15 and 18 again confirm that FWHM共G兲 and Disp共G兲 are FIG. 20. Correlation between Raman parameters, Urbach en-
related with density and total sp3 content, respectively. ergy, and Tauc gap. The lines are guides to the eye.
Disp共G兲, being a measure of topological disorder, is also a
good measure of the band gap in hydrogenated carbons, as relations between optical gap and density. Note that DLCHs
shown in Fig. 19. Given also that the Urbach energy EU represent a junction between the two lines, as it was in Fig.
varies roughly linearly with gap, we expect Disp共G兲 and EU 15.
to be linearly related. Thus, Fig. 15 is also the equivalent of If we restrict ourselves to a-C : Hs with high hydrogen
the EU versus optical gap relation found for hydrogenated content 共i.e., on line 2兲, we can get other useful relations,
amorphous carbons.1,30 EU is defined as the inverse of the valid only in this regime.
slope of ln关␣共E兲兴 at the E03 energy, where ␣共E兲 is the optical Figure 22 plots the square of the refractive index mea-
absorption and E03 is the photon energy at which the absorp- sured as a function of the density. By fitting Fig. 22 we find
tion coefficient reaches 103 cm−1.1,30,69 In fully sp3 semicon-
ductors, EU increases with the amount of disorder and the n2 = − 1 + 3.38 · ␳ 关g/cm3兴. 共6兲
spin density,69 but for PLCH EU increases for decreasing By combining Eqs. 共4兲–共6兲 we also get
spin density.30 From Fig. 20, we see that EU and Disp共G兲
have the same trends, thus the same physical meaning. Fig- n2 = − 0.13 + 3.7 ⫻ 10−2W 关cm−1兴. 共7兲
ure 20 demonstrates that EU is a measure of structural disor- From Fig. 21 we get the relation between optical gap and
der, as in a-Si: H, only in low or no hydrogen content density
samples. On the other hand, for high H content samples, EU
is a measure of topological disorder, as suggested in Ref. 30. ET 共eV兲 = 4.6 − 1.6␳ 关g/cm3兴. 共8兲
It is then easy to understand why by plotting density ver- By combining Eqs. 共5兲–共8兲 we get
sus gap we get Fig. 21. As expected, we find two different
ET 共eV兲 = 4.2 − 1.6 ⫻ 10−2 W 关cm−1兴. 共9兲
It is interesting to note that the linear relation between n2 and
the density of Fig. 22 confirms that in hydrogenated amor-

FIG. 19. Disp共G兲 as a function of Tauc gap. The linear fit is FIG. 21. Density plotted as a function of the Tauc gap. The lines
shown. are guides to the eye.

085401-10
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

GLCHH. This seems to be a very peculiar case, but has been

found in literature.71,72 GLCHH is typically produced by
PECVD from a mixture of CH4 and H2 and very high bias
voltage.71,72 Typical Raman spectra of GLCHH measured at
244 and 514 nm are plotted in Fig. 23 and compared with
GLCH and PLCH samples. Figure 24 upgrades Fig. 12 to
include GLCHHs. The Raman parameters are very close to
GLCH, as expected from the more ring-like sp2 configura-
tion, which reduces Disp共G兲 with respect to PLCHs. How-
ever, the visible Raman spectrum shows a higher PL back-
ground than standard GLCH. Typically m / I共G兲 for GLCHH
is ⬃5 ␮m. Indeed, the H content is 30– 40 at. %,73 Figs.
24共a兲 and 24共b兲. The density of these samples can be as low
as 1.3 g / cm3, below the density of typical GLCH
FIG. 22. Square of the refractive measured at 633 nm 共with 共⬃1.6 g / cm3兲. Thus, in order to identify GLCHH films we
some data at 600 nm兲 as a function of density. need to combine the analysis of Disp共G兲 and FWHM共G兲
with the PL background.
GLCHH films have interesting technological applications.
phous carbons the product EGm* ⬃ constant, where m* is the Even though the mechanical properties of these samples can
interband effective electron mass and EG is the Penn gap, be very poor, they can show excellent friction behavior in
i.e., the average bonding antibonding splitting.8 A similar vacuum or dry nitrogen.71,72,74 Thus, H content per se is not
linear trend is reported in Ref. 70. the only parameter controlling the friction in a-C : Hs, in
agreement with Refs. 75–78. Indeed, PLCH films do not
show a friction coefficient as low as GLCHH with the same
C. H ⬎ 20 at. % and sp2 phase clustered in rings
H content.75–78 It is the optimal combination of high H con-
This case corresponds to PLCH containing sp2 rings or to tent and sp2 configuration, which gives the best frictional
a highly hydrogenated GLCH. We call this class of materials properties. This will be discussed in detail elsewhere.79

FIG. 23. Raman spectra of PLCH, GLCHH, and GLCH films at 共a兲 514.5 and 共b兲 244 nm.

085401-11
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

FIG. 24. 共Color online兲 共a兲 Disp共G兲 and 共b兲 FWHM共G兲@244 as

FIG. 25. 共a兲 Density and 共b兲 Young’s modulus as a function of
a function of the H content for amorphous carbons, including
[email protected]. The lines are guides to the eye.
GLCHH.

In a single wavelength analysis, the crucial parameters are

Disp共G兲 decreases for increasing sp2 clustering18 and is
共1兲 FWHM共G兲, still related with structural disorder and the
zero if the sp2 phase is ordered in rings. Thus, for GLCHH
C u C sp3 content, 共2兲 I共D兲 / I共G兲, correlated to the amount
this parameter cannot be related to the H content, as shown
of ring-like sp2 clusters, and 共3兲 the PL background.
in Fig. 24共a兲. Disp共G兲 is in any case related the optical gap,
Figure 26 plots FWHM共G兲@514 as a function of
since the optical gap is uniquely defined by the sp2 phase
A共D兲 / A共G兲@514. Note that in this particular case we use
clustering. Thus, for GLCHH, Eq. 共2兲 fails. However, the PL
A共D兲 / A共G兲 instead of I共D兲 / I共G兲 in order to compare with
background can still be used to estimate the H content.
literature data, which usually report only A共D兲 / A共G兲. This
figure is again analog to Fig. 15, if one considers the oppo-
VI. MULTI-WAVELENGTH OR SINGLE WAVELENGTH site trends of Disp共G兲 and A共D兲 / A共G兲. Thus, the density and
RAMAN MEASUREMENTS? Young’s modulus can be derived from FWHM共G兲@514 and
the optical gap from A共D兲 / A共G兲@514. Further information
From Secs. IV and V it is clear that a multi-wavelength
analysis is always recommended with a-C : H, as for any
other carbon. This can be restricted to only two wavelengths,
such as 244 and 514 nm.
However, single wavelength measurements of a-C : H can
give reliable qualitative information, especially when the H
content is over 20 at. %. Even if, so far, we mostly discussed
FWHM共G兲@244, we get similar trends by using
FWHM共G兲@514, or indeed any other excitation wave-
length, see Figs. 25共a兲 and 25共b兲. However, for 514 nm ex-
citation, due to the PL background, the fit is less accurate and
the data spread is larger than at 244 nm. Furthermore,
244 nm excitation is necessary in order to get a FWHM共G兲
for high H content PLCHs, which do not show almost any
measurable Raman peaks for 514 nm. On the other hand, a
featureless visible Raman spectrum with no clear Raman
peaks can be taken as an empirical signature of PLCH with
low density 共艋1.2 g / cm3兲 and low Young’s modulus FIG. 26. FWHM共G兲@514.5 as a function of A共D兲 / A共G兲@514.
共艋20 GPa兲. The lines are guides to the eye.

085401-12
RAMAN SPECTROSCOPY OF HYDROGENATED… PHYSICAL REVIEW B 72, 085401 共2005兲

can be derived by measuring the refractive index and using G dispersion is found to be a measure of topological disorder
Fig. 22. The gap can then be cross checked with Fig. 21. 共range of cluster/chains sizes兲 and is proportional to the total
共C u C + C u H兲 sp3 fraction, gap, Urbach energy and H
content.
VII. CONCLUSIONS The PL background measured for visible excitation can be
We have presented a comprehensive analysis of a variety used to estimate the H content, if H ⬎ 20 at. %. If UV Raman
of a-C : H films grown by different deposition techniques and is not available, visible Raman spectra can be used for a
conditions. We extended the three-stage model for the Ra- qualitative characterization, but high caution must be taken
man spectra of carbon films to the case of hydrogenated in doing so. Particular deposition processes or substrate an-
amorphous carbons. The most important parameters to inter- nealing can induce sp2 clustering in a-C : H, even with rela-
pret the a-C : H Raman spectra are the G peak dispersion and tively high H content. In this case, the relation between sp3
the FWHM共G兲 measured at 244 nm. The Raman spectra of content and sp2 configuration fails. In general, a multi-
a-C : Hs show two fundamentally different behaviors. In one wavelength analysis is always advisable in order to avoid
case the G peak width increases as the G peak dispersion confusion.
increases. In the other case, the G peak width decreases with ACKNOWLEDGMENTS
increasing G dispersion. In the first case the correlation of
the Raman parameters to the structural and mechanical prop- The authors would like to thank D. Batchelder and I. R.
erties of the films is the same as for hydrogen-free carbon R. Mendieta of the University of Leeds and D. Wolverson of
films. In the second case, typical of polymeric hydrogenated the University of Bath for UV Raman facilities; F. Piazza, S.
amorphous carbons, the G peak width is correlated to the E. Rodil, B. Kleinsorge, B. Popescu, A. Champi, A. Erder-
structural properties, while the G peak dispersion is the best mir, and J. Fontaine, for some of the samples and useful
parameter to probe the optical properties and the hydrogen discussions. The authors acknowledge D. Schneider for
content. The unified diagram of G width and G dispersion for Young’s modulus measurements and D. Grambole for hydro-
the various carbons maps onto the variation of density gen content and density measurements. A.C.F acknowledges
against sp3 fraction. This allows us to identify the G width as funding from The Royal Society. This work has been sup-
a measure of structural disorder 共bond angle and bond length ported by the European Community 共FAMOUS; Project No.
distortions兲 and, thus, of density and Young’s modulus. The IST-2000-28661兲.

14
*Electronic address: [email protected] B. Meyerson and F. Smith, J. Non-Cryst. Solids 35–36, 435
1 J. Robertson, Mater. Sci. Eng., R. R37, 129 共2002兲. 共1980兲.
2 P. Koidl, C. Wagner, B. Discheler, J. Wagner, and M. Ramsteiner, 15 B. Popescu, A. Tagliaferro, F. De Zan, and E. A. Davis, J. Non-

Mater. Sci. Forum 52, 41 共1990兲. Cryst. Solids 266–269, 803 共2000兲.
3 W. Jacob and W. Moller, Appl. Phys. Lett. 63, 1771 共1993兲. 16 A. C. Ferrari and J. Robertson, Philos. Trans. R. Soc. London,
4 M. A. Tamor, W. C. Vassell, and K. R. Carduner, Appl. Phys.
Ser. A 362, 2267 共2004兲.
Lett. 58, 592 共1991兲. 17
A. C. Ferrari and J. Robertson, Phys. Rev. B 61, 14095 共2000兲.
5 S. F. Yoon, K. H. Tan, Rusli, and J. Ahn, J. Appl. Phys. 91, 1634 18 A. C. Ferrari and J. Robertson, Phys. Rev. B 64, 075414 共2001兲.

共2002兲. 19 S. Piscanec, F. Mauri, A. C. Ferrari, M. Lazzeri, and J. Robertson,

6
T. Schwarz-Selinger, A. von Keudell, and W. Jacob, J. Appl. Diamond Relat. Mater. 14, 1078 共2005兲.
Phys. 86, 3988 共1999兲. 20 K. W. R. Gilkes, S. Prawer, K. W. Nugent, J. Robertson, H. S.
7 O. Durand-Drouhin, M. Lejeune, and M. Benlahsen, J. Appl.
Sands, Y. Lifshitz, and X. Shi, J. Appl. Phys. 87, 7283 共2000兲.
Phys. 91, 867 共2002兲. 21
V. I. Merkulov, J. S. Lannin, C. H. Munro, S. A. Asher, V. S.
8 A. C. Ferrari, A. Libassi, B. K. Tanner, V. Stolojan, J. Yuan, L. M.
Veerasamy, and W. I. Milne, Phys. Rev. Lett. 78, 4869 共1997兲.
Brown, S. E. Rodil, B. Kleinsorge, and J. Robertson, Phys. Rev. 22
F. Tuinstra and J. L. Koenig, J. Chem. Phys. 53, 1126 共1970兲.
B 62, 11089 共2000兲. 23 C. Castiglioni, E. Di Donato, M. Tommasini, F. Negri, and G.
9
N. A. Morrison, S. E. Rodil, A. C. Ferrari, J. Robertson, and W. I. Zerbi, Synth. Met. 139, 885 共2003兲.
Milne, Thin Solid Films 337, 71 共1999兲. 24 S. Piscanec, M. Lazzeri, F. Mauri, A. C. Ferrari, and J. Robertson,
10
M. Weiler, K. Lang, E. Li, and J. Robertson, Appl. Phys. Lett. 72, Phys. Rev. Lett. 93, 185503 共2004兲.
1314 共1998兲. 25 C. Mapelli, C. Castiglioni, G. Zerbi, and K. Mullen, Phys. Rev. B
11
M. Weiler, S. Sattel, T. Giessen, K. Jung, H. Ehrhardt, V. S. 60, 12710 共2000兲.
Veerasamy, and J. Robertson, Phys. Rev. B 53, 1594 共1996兲. 26 A. C. Ferrari, S. E. Rodil, and J. Robertson, Phys. Rev. B 67,
12 M. Weiler, S. Sattel, K. Jung, H. Ehrhardt, V. S. Veerasamy, and
155306 共2003兲.
J. Robertson, Appl. Phys. Lett. 64, 2797 共1994兲. 27 A. C. Ferrari and J. Robertson, Phys. Rev. B 63, 121405共R兲
13
B. Kleinsorge, S. E. Rodil, G. Adamopoulos, J. Robertson, D. 共2001兲.
28 C. Casiraghi, A. C. Ferrari, J. Robertson, R. Ohr, M. v. Grad-
Grambole, and W. Fukarek, Diamond Relat. Mater. 10, 965
共2001兲. owski, and D. Schneider, Diamond Relat. Mater. 13, 1480

085401-13
CASIRAGHI, FERRARI, AND ROBERTSON PHYSICAL REVIEW B 72, 085401 共2005兲

共2004兲. 1974兲.
29 54
M. v. Gradowski, A. C. Ferrari, R. Ohr, B. Jacoby, H. Hilgers, B. Dischler, A. Bubenzer, and P. Koidl, Solid State Commun. 48,
H.-H. Schneider, and H. Adrian, Surf. Coat. Technol. 174–175, 105 共1983兲.
246 共2003兲. 55 J. Ristein, R. T. Stief, L. Ley, and W. Beyer, J. Appl. Phys. 84,
30 G. Fanchini and A. Tagliaferro, Appl. Phys. Lett. 85, 730 共2004兲. 3836 共1998兲.
31 A. C. Ferrari, B. Kleinsorge, N. A. Morrison, A. Hart, V. Stolojan, 56 G. Fanchini, P. Mandracci, A. Tagliaferro, S. E. Rodil, A. Vomi-

and J. Robertson, J. Appl. Phys. 85, 7191 共1999兲. ero, and G. Della Mea, Diamond Relat. Mater. 14, 928 共2005兲.
32
J. Robertson, Phys. Rev. B 53, 16302 共1996兲. 57
S. Sattel, J. Robertson, and H. Ehrhardt, J. Appl. Phys. 82, 4566
33
Rusli, J. Robertson, and G. A. J. Amaratunga, J. Appl. Phys. 80, 共1997兲.
2998 共1996兲. 58 M. Yoshimi, M. Shimizu, K. Hattori, H. Okamoto, and Y. Ha-
34 T. Heitz, C. Godet, J. E. Bouree, B. Drevillon, and J. P. Conde,
makawa, Optoelectron., Devices Technol. 7, 69 共1992兲.
Phys. Rev. B 60, 6045 共1999兲. 59 J. Wagner and P. Lautenschlager, J. Appl. Phys. 59, 2044 共1986兲.
35 B. Marchon, J. Gui, K. Grannen, G. C. Rauch, J. W. Ager, S. R. 60 S. J. Henley, J. D. Carey, and S. R. P. Silva, Appl. Phys. Lett. 85,

P. Silva, and J. Robertson, IEEE Trans. Magn. 33, 3148 共1997兲. 6236 共2004兲.
36 61
S. Xu, M. Humdhausen, J. Ristein, B. Yan, and L. Ley, J. Non- S. E. Rodil, S. Muhl, S. Maca, and A. C. Ferrari, Thin Solid Films
Cryst. Solids 164, 1127 共1993兲. 433, 119 共2003兲.
37 Q. Zhang, S. F. Yoon, Rusli, J. Ahn, H. Yang, and D. Bahr, J. 62 G. Fanchini, S. C. Ray, and A. Tagliaferro, Diamond Relat.

Appl. Phys. 84, 5538 共1998兲. Mater. 12, 1084 共2003兲.

38 63
S. F. Yoon, K. H. Tan, Rusli, J. Ahn, and Q. F. Huang, J. Appl. S. Schutte, S. Will, H. Mell, and W. Fuhs, Diamond Relat. Mater.
Phys. 89, 4830 共2001兲. 2, 1360 共1993兲.
39 64
F. Piazza, D. Grambole, D. Schneider, C. Casiraghi, A. C. Ferrari, G. Fanchini, A. Tagliaferro, B. Popescu, and E. A. Davis, J. Non-
and J. Robertson, Diamond Relat. Mater. 14, 994 共2005兲. Cryst. Solids 299–302, 846 共2002兲.
40 65
A. Golanski, F. Piazza, J. Werckmann, G. Relihan, and S. J. Ristein, J. Schafer, and L. Ley, Diamond Relat. Mater. 4, 508
Schulze, J. Appl. Phys. 92, 3662 共2002兲. 共1995兲.
41 66
F. Piazza, Y. Arnal, D. Grambole, F. Herrmann, M. Kildemo, A. R. Bouzerar, A. Zeinert, and H.-J. von Bardeleben, Diamond
Lacoste, G. Relihan, and A. Golanski, Thin Solid Films 383, Relat. Mater. 14, 1108 共2005兲.
196 共2001兲. 67 A. Zeinert, H.-J. von Bardeleben, and R. Bouzerar, Diamond
42 F. Piazza 共unpublished兲.
Relat. Mater. 9, 728 共2000兲.
43 B. Popescu, C. Verney, E. A. Davis, V. Paret, and A. Brunet- 68 D. Dasgupta, F. Demichelis, C. F. Pirri, and A. Tagliaferro, Phys.

Bruneau, J. Non-Cryst. Solids 266–269, 778 共2000兲. Rev. B 43, 2131 共1991兲.
44 S. R. P. Silva, J. Robertson, Rusli, and G. A. J. Amaratunga, 69 R. A. Street, Hydrogenated Amorphous Silicon 共Cambridge Uni-

Philos. Mag. B 74, 369 共1996兲. versity Press, New York, 1991兲.
45 M. A. Tamor and W. C. Vassell, J. Appl. Phys. 76, 3823 共1994兲. 70 P. Patsalas, S. Logothetidis, and P. C. Kelires, Diamond Relat.
46
E. Tomasella, C. Meunier, and S. Mikhailov, Surf. Coat. Technol. Mater. 14, 1241 共2005兲.
141, 286 共2001兲. 71 A. Erdemir, O. L. Eryilmaz, I. B. Nilufer, and G. R. Fenske,
47 R. Bouzerar, C. Amory, A. Zeinert, M. Benlahsen, B. Racine, O.
Diamond Relat. Mater. 9, 632 共2000兲.
72
Durand-Drouhin, and M. Clin, J. Non-Cryst. Solids 281, 171 A. Erdemir, O. L. Eryilmaz, and G. R. Fenske, J. Vac. Sci. Tech-
共2001兲. nol. A 18, 1987 共2000兲.
48 73
M. L. Theye, V. Paret, and A. Sadki, Diamond Relat. Mater. 10, J. A. Johnson, J. B. Woodford, X. Chen, J. Andersson, A. Er-
182 共2001兲. demir, and G. R. Fenske, J. Appl. Phys. 95, 7765 共2004兲.
49 74
Rusli, S. F. Yoon, H. Yang, J. Ahn, Q. Zhang, Y. S. Wu, and W. L. A. Erdemir, O. L. Eryilmaz, I. B. Nilufer, and G. R. Fenske, Surf.
New, J. Appl. Phys. 84, 5277 共1998兲. Coat. Technol. 133–134, 448 共2000兲.
50 75
C. Marechal, A. Zeinert, K. Zellama, E. Lacaze, M. Zarrabian, J. Fontaine, J. L. Loubet, T. Le Mogne, and A. Grill, Tribol. Lett.
and G. Turban, Solid State Commun. 109, 23 共1999兲. 17, 709 共2004兲.
51 76
D. Schneider, T. Schwarz, H.-J. Scheibe, and M. Panzner, Thin J. Fontaine, T. Le Mogne, J. L. Loubet, and M. Belin, Thin Solid
Solid Films 295, 107 共1997兲. Films 482, 99 共2005兲.
52
A. C. Ferrari, J. Robertson, M. G. Beghi, C. E. Bottani, R. Feru- 77
A. Erdemir, Surf. Coat. Technol. 146–147, 292 共2001兲.
lano, and R. Pastorelli, Appl. Phys. Lett. 75, 1893 共1999兲. 78 A. Erdemir, Tribol. Int. 37, 577 共2004兲.
53 79
F. R. Dollish, W. G. Fateley, and F. F. Bentley, Characteristic C. Casiraghi, A. C. Ferrari, A. Erdemir, J. Fontaine, and J. Rob-
Raman Frequencies of Organic Molecule 共Wiley, New York, ertson 共unpublished兲.

085401-14