Surface characterization of amorphous hydrogenated carbon thin films containing nanoclusters of noble metals

In this work nanocomposite thin films of amorphous hydrogenated carbon
(a-C:H) doped with noble transition metals of 1B group (gold, silver, and copper) are
studied. The composite materials are obtained by combined magnetron sputtering
(MS) of a metal target by argon, and plasma-assisted chemical vapor deposition
(PACVD) of methane under vacuum conditions. Particular attention is devoted to the
low metal-content a-C:H samples, in which metallic inclusions have a form of isolated
nanoclusters. Our aim was to reveal surface cluster arrangement, i.e. to figure out
whether topmost metallic nanoclusters are covered with a layer of a-C:H or are bald
on the surface and hence exposed to the surrounding environment. We deposited
our samples onto substrates kept on the ground potential and on �150 V dc bias
voltage. The differences encountered in the surface structure and nanocluster arrangement
between samples, which differed in the deposition process in this parameter
only, provided an answer to the question of surface clusters coverage.
The experimental techniques used to reach this goal comprised in vacuo and
ex vacuo photoelectron spectroscopy (PES), direct imaging by atomic force microscopy
(AFM) and scanning electron microscopy (SEM), and grazing incidence smallangle
x-ray scattering (GISAXS). The majority of work is made in several photoemission
experiments, using both x-ray and UV-excited photoelectron spectroscopy (XPS
and UPS, respectively). In the course of the work, series of (grounded only) a-C:H/Au, a-C:H/Ag, and
a-C:H/Cu samples, with metallic content varying from zero to 100 at.%, have been
deposited and studied in vacuo by XPS and UPS (Section 3.1). The XPS results of
these series show that, with decreasing metal content below the percolation threshold
(at about 40-50 at.% of metal concentration), a shift in binding energy (BE) of
metal core levels towards higher BEs is observed. With non-carbidic metals like the
ones we used in our work, these shifts can be related to their isolated cluster structure
in the host matrix. Decrease of the total metal content in the sample is followed
by the decrease in the cluster size, which is reflected in increased binding energy of
electrons escaping from them (Paragraph 2.2.2 and Section 3.1). At the same time,
carbon C 1s core level is shifted in the opposite direction, towards lower binding energies,
and this shift rises as metal content increases in the sample. Negative shift in
C 1s binding energy reveals increased relative content of sp2-coordinated carbon in
the a-C:H matrix, which, we believe, should be attributed to the compressive stress
that metallic inclusions introduce in the host a-C:H, being higher with higher metal
content in the sample. Another possible reason is catalytic reduction of hydrogen in
the a-C:H matrix with increasing metallic content. UPS of the same series (Fig. 3.2)
showed increased relative contribution of metal valence band features with increased
metal content. The Fermi edge evolution from monocrystalline metal reference samples
to low metal-content a-C:H showed both decrease in the density of states near the Fermi level and its shift to the higher BEs, related to the cluster structure of metallic
inclusions and to the same direction-shifts in metal core levels.
Our further attention was focused on low metal-content a-C:H that is, typically
below 10 at.%, and to the differences between grounded and biased samples. The
reprint of the communication on these effects encountered in a-C:H/Au and published
in the Applied Physics Letters (Vol. 80, 2002, p.2863) is given in Section 3.2.
Direct imaging techniques (Sections 3.2 and 3.3) reveal that even small
amount of metal included in the a-C:H significantly changes the surface morphology
and increases roughness. For all three nanocomposite materials, biased samples
show similar surface morphology, characterized by relatively flat basis and isolated
bump structures, of about 30 nm in diameter and up to 15 nm in height. These
structures are attributed to the altered morphology of a-C:H component of a nanocomposite,
with metallic clusters concentrated on them. Grounded samples characterize
low roughness in a-C:H/Au and increased in a-C:H/Ag and a-C:H/Cu. The increase
of roughness in the latter two materials is explained by enhanced surface diffusion
of metal atoms and clusters coalescence into bigger islands.
GISAXS patterns (Section 3.4) showed isotropic cluster size and intercluster
distance distribution in grounded a-C:H/Au and a-C:H/Cu samples. With grounded a-
C:H/Ag sample, no spatial correlation could be revealed, probably again due to the
surface agglomeration of Ag clusters. These analysis of GISAXS patterns showed
that biased samples contain bigger clusters than grounded ones, and slightly flattened
in the grow direction. The differences between grounded and biased samples were first detected by
in vacuo XPS of a-C:H/Au in systematically lower, in average about 50%, Au content
in the biased than in the grounded case. This decrease is followed by higher positive
shift of the Au 4f7/2 core level binding energy in the biased case (Section 3.2 and
Paragraph 3.5.1). The similar situation was encountered in a-C:H/Ag samples, while
a-C:H/Cu showed higher Cu content in the biased than in the grounded case and no
difference in Cu 2p3/2 shifts from the reference BE. The decrease of the total metal
content in biased a-C:H/Au and a-C:H/Ag samples most probably results from increased
distance among surface clusters and their concentration on isolated bump
structures. The reason for the increased BE shift of Au 4f7/2 and Ag 3d5/2 core levels
in the biased samples should be searched in cluster baldness at the sample surface.
The violation of both above conclusions in a-C:H/Cu, i.e. higher Cu content in the
biased sample and equal shift of the Cu 2p3/2 core level in grounded and biased
sample is probably due to high relative increase of copper cluster size upon biasing,
which compensates the effect of their baldness at the surface. The topmost clusters
of grounded nanocomposite samples are, therefore, most probably covered with a
layer of a-C:H. The thickness of this layer must be below escape depth of metaloriginated
photoelectrons, i.e. less than about 2 nm. The negative shift of the C 1s
core level in the biased samples is induced by increased sp2/sp3 coordinated carbon
ratio due to the sample bombardment by Ar+ ions during the deposition process. To our belief, the same effect is responsible for surface metallic clusters baldness in the
biased samples.
These conclusions are supported by several following PES experiments, other
than XPS of as-deposited samples. First of them was in vacuo UPS of as-deposited
samples. In the grounded samples, the He I spectra mostly reproduce the characteristic
shape of a-C:H valence band, and only higher sensitivity He II UPS reveal the
presence of metallic inclusions. Upon biasing, even when total measured metal content
was lower than in the grounded case (a-C:H/Au and a-C:H/Ag), all spectra
clearly showed increased metallic features, evidencing on higher metal exposure at
the surface.
XPS at off-normal take-off angle of escaping electrons also confirmed our conclusions
on the surface clusters coverage. Increasing the tilting angle of a sample,
measured intensity ratio of a metal core level to appropriate C 1s showed in most
grounded samples monotonous decrease, and regular and steady increase in the
biased ones. These results support our conclusion on the surface clusters coverage
in grounded samples and their baldness in the biased ones. The higher metalcontent
grounded samples of a-C:H/Ag, however, did not show the expected decrease
in the intensity ratio, and that was the first indication that the effect of coverage
may be a particularity of small clusters only, i.e. low-metal content grounded
samples. This suspicious is confirmed in the next test experiment that we have undertaken,
by subsequent in situ low-energy Ar+ ion etching and PES analysis of a sample.
The same metal to carbon core level intensity ratio curves were measured
against the sputtering time. In the grounded samples, at the beginning of the sputtering,
an increase in the intensity ratio is observed, related to the thinning of the top
a-C:H layer. In most cases, after some time of sputtering, the maximum is reached
related to the total removal of the cover layer, and from that point onwards, Ar+ ion
etching erodes the metallic clusters as well. In biased samples, a monotonous decrease
of intensity ratio curves was observed throughout the experiment and is
clearly related to the bald surface clusters that are sputtered together with the a-C:H
matrix. The grounded samples intensity ratio curves showed one more important
regularity: in higher metal-content samples the maximum is reached after shorter
time, i.e. these samples need less time to be fully uncovered. As a special case, a-
C:H/Ag 32.3 at.% did not show any increase in the intensity ratio curve, but monotonous
decrease throughout the measurement. That encouraged the conclusion that
the coverage of the topmost metallic clusters of grounded samples with a-C:H is an
effect that is characteristics of small clusters in the host matrix, i.e. low metal-content
samples. With higher metal contents, there is no observable difference between
grounded and biased samples regarding surface clusters coverage.
Apart from the core level intensity ratio curves, the evolution of our samples
with in situ Ar+ ion etching is described in XPS and UPS spectra recorded at each
point of the sputtering time scale. Metal core levels in these figures remained either
unchanged or are slightly shifted towards higher binding energies. In the grounded samples, this is related to the thinning of the cover layer, and in the biased ones to
the decrease of the cluster size by Ar+ ion sputtering. Carbon C 1s core level in all
samples shows shift with sputtering time towards lower binding energies, which is
related to the further sp2-coordinated carbon favoring by the in situ Ar+ ion bombardment.
The UPS spectra evolution generally follows the trend described by core level
intensity ratio curves. That is, in grounded samples, the metal features in valence
band spectra rise to the point of total removal of the cover layer, and decrease further
to the end of the sputtering experiment. Biased samples, on the other hand, show
continuous decrease of the metal features. In both grounded and biased valence
band spectra, the development of the carbon π-states is observed throughout the
experiment, evidencing on the increase of sp2-coordinated carbon content with sputtering
time.
Pointed out several times, the Ag surface clusters coalescence is confirmed in
experiment in which we compared XPS and UPS spectra of as-deposited samples,
after 20 hours residence in the ultra-high vacuum (UHV) conditions and after additional
20 hours in the air. Generally, all as-deposited spectra and after 20 hours in the
UHV were almost identical. After exposure to the air, in all samples carbon C 1s core
level is shifted towards lower binding energies. The most interesting differences after
residence in the air show metal core levels. The Au 4f7/2 core level remained practically
identical to the one measured in the UHV, revealing that air conditions do not
affect Au clusters, and that their size and arrangement remain fully determined by the
deposition process. That is not the case, however, with Ag clusters. The Ag 3d5/2
core levels of both grounded and biased a-C:H/Ag samples shift towards lower binding
energies. From the differences in UHV- and air-residence binding energy positions
of the Ag 3d5/2, it is estimated that the increase factor of cluster volume in
grounded samples is about 170, and the one of biased samples clusters � about 12.
In these rough figures one may find the cause of the specific behavior of the
a-C:H/Ag sample that we encountered in several occasions and assigned to the Ag
surface clusters coalescence: roughness revealed by AFM, lack of correlation in the
GISAXS patterns, pronounced Ag 4d features even in low metal content valence
band spectra, and negative shift of the Ag 3d5/2 in the Ar+ ion in situ in-depth profiling.
The last of our nanocomposites subjected to UHV- and air-dwell comparison was a-
C:H/Cu. Copper, however, oxidizes in the air, but nevertheless it provided in this experiment
one of most elegant evidences on the surface clusters coverage. The deconvolution
procedure applied to copper- and CuO-originated Cu 2p3/2 revealed that
relative content of oxidized copper is higher on the surface of biased sample (with
bald surface clusters), than on the grounded one (where surface clusters are covered
by a-C:H). The last in the series of experiments aimed to check our conclusions on the
surface clusters coverage was based on the prospective sulfur binding to noble metal
atoms. Our samples, together with appropriate monocrystalline reference samples,
were covered with a layer of liquid thiophene (C4H4S) and, after evaporation, subjected
again to the XPS analysis. The total amount of adsorbed sulfur was generally low, about 5 at.% or less. That results in noisy XPS spectra of the S 2p core levels
region, in spite of increased measurement statistics. The S 2p spectra adsorbed on
the reference samples were fitted with three S 2p1/2 � S 2p3/2 doublets assigned to S
bonds to a noble metal, S in C4H4S, and to S�O bonds (Figs. 3.32-3.34). Intercomparison
of our nanocomposite samples with reference ones showed that sulfur adsorbed
on surfaces originates predominantly from C4H4S itself. However, in biased
cases a higher relative contribution of the shoulder related to sulfur bonds to a noble
metal is observed in spectra, evidencing on higher metal exposure at the biased
samples surfaces. The exception of a-C:H/Cu is due to the oxidation of copper.
In conclusion, in several different PES experiments, by direct imaging of samples,
and using GISAXS technique, we have revealed that MS/PACVD-obtained low
noble metal-content amorphous hydrogenated carbon nanocomposites are characterized
with topmost metallic clusters covered with a tiny layer of a-C:H when deposited
on a grounded substrate, and bald surface clusters when substrate is biased with
�150 V dc. Beside this main result, we encountered few other effects, like e.g. increased
sp2/sp3 coordinated carbon ratio in the a-C:H matrix in the biased samples
and surface clusters coalescence in a-C:H/Ag (and to some extent in a-C:H/Cu)
nanocomposites. By changing one parameter only � the substrate bias voltage in deposition of
our grounded and biased �counterparts�, we have shown that surface clusters coverage
effect has an origin in the plasma deposition process itself. We believe that one
should look for its cause in the plasma afterglow, the state established in the ionized
gas immediately after switching off the plasma power supply.
From the applicative point of view, we have described, in principle, the mechanism
that may be employed to tailor the coverage of topmost metallic clusters embedded
in the a-C:H matrix. Metal inclusions in the a-C:H showed to improve the
wear resistance of the coatings, so one can also envisage the applications when the
coverage of surface metal clusters with a-C:H would be useful. In tribology, these
would be cases when incorporated metal reduces the lubricating properties, i.e. increases
the friction coefficient. In biocompatible materials the same would be necessary
when incorporated metals are toxic, like e.g. silver or copper. Vice versa, one
may also envisage applications when topmost cluster baldness would be desirable,
like e.g. with low-friction MoS2 and WS2 inclusions in a-C:H for tribological purposes.
In addition, surface clusters exposure to the surrounding environment probably influences
the optical and aging properties of solar selective coatings based on metal- or
metal carbide-containing amorphous hydrogenated carbon nanocomposites.