A photoemission study of molybdenum hexacarbonyl adsorption and decomposition on TiO2(1 1 0) surface

Article abstract of DOI:10.1016/j.susc.2006.12.009

The adsorption and decomposition of molybdenum hexacarbonyl on (1 1 0) TiO₂ surfaces were studied using both core levels and valence band photoemission spectroscopies. It was found that after an adsorption at 140 K, when going back to room temperature, only a small part of molybdenum compounds, previously present at low temperature, remained on the TiO₂ surface. This indicates that the desorption temperature on such a surface is lower than the decomposition one. The use of photon irradiation to decompose the hexacarbonyl molecule was also studied. It was shown that during such a decomposition molecular fragments were chemisorbed on the surface allowing a higher amount of metal to remain on the surface. It was also shown that it was possible to get rid of adsorbed subcarbonyl groups and to organize the metal atoms by thermal treatments at temperatures as low as 400 K, i.e. much lower than the one needed to obtain the same structures using physical vapour deposition (PVD). Moreover, due to lower used temperatures, this chemical way of deposition allows a better control of the interface than during PVD growth.

Full text of DOI:10.1016/j.susc.2006.12.009

Surface Science 601 (2007) 1144–1152
www.elsevier.com/locate/susc
A photoemission study of molybdenum hexacarbonyl adsorption
and decomposition on TiO (110) surface
2
a
a,
b
c
a
*
J. Prunier , B. Domenichini , Z. Li , P.J. Møller , S. Bourgeois
a
LRRS, UMR 5613 CNRS-Universit e´ de Bourgogne, BP 47870, 21078 Dijon cedex, France
b
ISA, University of Aarhus, Ny Munkegade, DK-8000 Aarhus C, Denmark
Department of Chemistry, University of Copenhagen, DK-2100 Copenhagen, Denmark
c
Received 17 October 2006; accepted for publication 6 December 2006
Available online 3 January 2007
Abstract
The adsorption and decomposition of molybdenum hexacarbonyl on (110) TiO surfaces were studied using both core levels and
2
valence band photoemission spectroscopies. It was found that after an adsorption at 140 K, when going back to room temperature, only
a small part of molybdenum compounds, previously present at low temperature, remained on the TiO surface. This indicates that the
2
desorption temperature on such a surface is lower than the decomposition one. The use of photon irradiation to decompose the hexa-
carbonyl molecule was also studied. It was shown that during such a decomposition molecular fragments were chemisorbed on the sur-
face allowing a higher amount of metal to remain on the surface. It was also shown that it was possible to get rid of adsorbed
subcarbonyl groups and to organize the metal atoms by thermal treatments at temperatures as low as 400 K, i.e. much lower than
the one needed to obtain the same structures using physical vapour deposition (PVD). Moreover, due to lower used temperatures, this
chemical way of deposition allows a better control of the interface than during PVD growth.
ꢀ
2006 Elsevier B.V. All rights reserved.
Keywords: Growth; Supported nanostructures; Molybdenum hexacarbonyl; TiO
2
; Photoelectron spectroscopy
1
. Introduction
ing temperature [3] on the final morphology, crystallinity
and stoichiometry of deposits.
The growth mode of a metal deposited by physical va-
The main conclusions were that, for films thicker than a
monolayer, the interfacial reaction is less related to the ini-
tial state of the substrate (i.e. the initial surface stoichiom-
etry and morphology) than to kinetical parameters such as
deposition rate which strongly influences the amount of
oxygen atoms which can diffuse into the deposit and the
electronic exchanges between metallic molybdenum and
titanium ions. Especially, whatever the surface state is
and for low enough deposition rate, the interfacial reac-
tion, which occurs at early stage of the deposition, leads
to the growth of metastable molybdenum oxide film while
both DFT calculations [6] and simple demonstration from
macroscopic thermodynamical values [3] show that from a
coverage higher than one layer, molybdenum should be
metallic. Especially, DFT calculations show that for depos-
its lower than 1.5 equivalent monolayers (eqML), a special
pour deposition (PVD) technique on a metal oxide sub-
strate often appears hard to control because it is highly
related to the interfacial reactivity which is a function of
numerous parameters. As concerns to Mo/TiO (110) sys-
2
tem, previous works were carried out in order to determine
the most relevant parameters playing on both interfacial
reactivity and growth mode, these two processes being
highly linked [1]. Hence, we successively studied the effects
of TiO surface stoichiometry [2], TiO bulk stoichiometry
2
2
[
3], surface roughness [4], molybdenum deposition rate [5],
deposited molybdenum amount [6] and subsequent anneal-
*
Corresponding author.
E-mail address: bruno.domenichini@u-bourgogne.fr (B. Domenichini).
0
039-6028/$ - see front matter ꢀ 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.susc.2006.12.009

J. Prunier et al. / Surface Science 601 (2007) 1144–1152
1145
organization of molybdenum atoms as pyramidal stripes
should occur in the channels delimited by oxygen atom
rows which are exhibited along the [001] direction by
sharp (1 · 1) LEED pattern with a low background was
obtained and with valence band resonance spectra where
no contribution of the Ti3d defect band was observed
[12].
TiO (110) surfaces [1].
2
In these special conditions, annealing is thus necessary
to break the metastability, i.e. induce the reduction of
molybdenum oxide and then lead to equilibrium. How-
ever, during such a subsequent annealing, another key
parameter plays a major role: the substrate bulk stoichi-
ometry [3]. Indeed, although the film can be fully reduced
The Mo(CO) precursor (98%, Aldrich) was held in an
6
independently pumped glass tube at room temperature,
which was pumped out for purification. It was then intro-
duced in the vacuum chamber via a micrometric valve to
À5
control the pressure (up to 10 Pa). During exposures,
the sample was kept at a temperature around 140 K.
Experiments were carried out on SX700 beamline (bend-
ing magnet) at the ASTRID synchrotron source of the
in the case of an initial non-stoichiometric bulk TiO , the
2
deposit is fully oxidized if the substrate is initially stoichi-
ometric. This fact, which was explained by entropic effect
˚
Institute for Storage Ring Facilities (Arhus, Denmark).
[
3], avoids the possibility to make stripes growing on the
The incident angle of the photon beam and the detection
angle were respectively 45ꢁ and 0ꢁ from the surface normal.
The same setup was used for core level spectra (XPS) and
valence band spectra (UPS) thanks to a ZEISS SX700
monochromator. Such an apparatus has a wide usable en-
ergy range: from 20 to 700 eV. However, the photon beam
given by the beamline is not constant over the whole range.
surface of stoichiometric TiO . Hence, a way to produce
2
molybdenum stripes on stoichiometric surface could be
to avoid the initial reaction between the TiO2 surface
and the first deposited molybdenum atoms in order to
not have to break metastability through subsequent
annealing.
A solution may be deposition of molybdenum at cryo-
genic temperature in order to kinetically hinder any interfa-
cial reaction. According to this idea, the present study deals
with the possibility of depositing molybdenum atoms on
1
0
Typical flux values for the used settling are ca. 10 pho-
À1
9
À1
tons s
at 125 eV and ca. 10 photons s
at 600 eV.
Moreover, the resolution of the SX700 monochromator de-
creases as the photon energy increases. Because of these
reasons, the quality of the spectra decreases with increasing
of photon beam energy, which is used for spectrum
recording.
For experiments on core level spectra, Ti2p, O1s, C1s
and Mo3d, the photon energies were chosen in such a
way as to get photoelectrons having almost the same ki-
netic energies that is to say very close mean free paths
and to avoid phantom peaks coming from second order
diffraction line through monochromator. Moreover, in or-
der to be as sensitive to surface phenomena as possible,
these kinetic energies were chosen between 80 and 100 eV
which is the range for the mean free paths minima. Most
of valence band spectra were obtained with primary pho-
ton energy equal to 45 eV which corresponds to the 3p–
TiO (110) surface from a chemical vapour deposition
(
2
CVD) route using Mo(CO) as a precursor. Actually this
6
chemical way of deposition has been found to lead to inter-
esting results in catalysts preparation [7–9]. For example
Mo nanoparticles prepared using PVD methods are stable
on gold surfaces, whereas those prepared using the chemi-
cal vapour deposition of a molybdenum hexacarbonyl pre-
cursor are found to be mobile, inducing different catalytic
properties [10,11].
This paper is about the results concerning the first steps
of such a deposition, the precursor adsorption stage and its
decomposition.
2
. Experimental
3d titanium resonance: thanks to resonance phenomena
The TiO (110) sample (MaTeck) was cleaned using the
standard procedure which consisted of argon ion bom-
bardments to remove any residual contaminant followed
the sensitivity to Ti3d derived defects is higher for this en-
ergy value [12].
2
The ultrahigh vacuum system with a base pressure of
À11
by sample annealing up to 1000 K in ultra high vacuum
7.5 · 10
mbar was equipped with a quadrupolar mass
+
(
UHV). The role of the Ar bombardment was to create
spectrometer, low energy electron diffraction (LEED) and
an electron energy analyser in order to record photoemis-
sion spectra. The electron analyser was a VG CLAM II,
working with 30 eV pass energy. The energy calibration
(Fermi edge) was checked by reference to a metallic molyb-
denum thick film (>15 nm) which was obtained after
defects thanks to the preferential oxygen sputtering; the
one of the annealing was to re-organize the sample by dif-
fusion of the defects of the top layers into the whole crys-
tal. In all the experiments presented here, the crystal bulk
was under-stoichiometric with a homogenous distribution
of defects in the bulk, explaining its blue colour. This
deviation from stoichiometry was high enough to allow
sample bulk conductivity sufficient to avoid any charging
effects during the LEED and photoemission experiments.
After annealing, exposures to 100 L of oxygen were done
in order to get a fully stoichiometric (110) surface. The
quality of the surface was checked with LEED where a
molybdenum deposition on TiO substrate using Caburn-
2
MDC e-vap 100 evaporator. The deposition conditions
À1
(1.5 eqML min for 10 min) were specifically chosen to
lead to metallic molybdenum film [1]. The sample was
put on a Ta foil connected to a copper sample holder that
could be cooled down to 140 K with liquid nitrogen or
resistively heated up to 1000 K.

1146
J. Prunier et al. / Surface Science 601 (2007) 1144–1152
3
. Results and discussion
3
.1. Exposures
The Mo(CO) exposure experiments at 140 K were ana-
6
lysed first by core level photoemission spectroscopy. Espe-
cially, the intensity of Mo(CO) signal (through Mo3d line
6
signal) as well as substrate signal (through Ti2p line signal)
was followed as a function of exposure (Fig. 1). It is obvi-
ous that the evolution of these signal intensities is best fit-
ted with piecewise-linear variations. Hence, the change of
the slopes of the curves recorded for 14 Langmuir exposure
can be linked to the end of the filling of the first monolayer
and the beginning of the next one in a layer-by-layer
growth mode [13]. These experiments prove that molybde-
num carbonyl adsorption begins by completion of two
monolayers. After adsorption of these two monolayers,
the substrate signal intensity begins too low to exhibit a
clear behaviour. This point is certainly due to the very
low mean free path of detected electrons coming from the
substrate (see Section 2). In the same time and for the same
reason, the deposit signal intensity changes very slowly be-
cause the adsorbed film reached the maximum thickness
which can be analysed. After completion of two Mo(CO)6
monolayers, the rest of the growth can be no longer fol-
lowed by XPS signal ratio evolution.
Fig. 2. Evolution of the C1s line during an exposure to Mo(CO)
and 20 Langmuir at 140 K. Primary photon energy: 380 eV.
6
between
0
right part of lines can be attributed to Mo(CO) (x < 6)
x
formation through partial molecule decomposition with
CO departure. Such a hypothesis will be more discussed
in Section 3.3.
Concerning the O1s spectra (Fig. 3), before exposure, the
spectrum is composed of a single line at 530.0 eV which cor-
responds to the oxygen of the TiO surface. After the first
2
exposure to Mo(CO) another peak appears at around
6
533.5 eV. As in the case of the peak observed at 288.0 eV
in the C1s spectrum, the intensity of this peak increases with
In Figs. 2–4 are presented the evolutions of C1s, O1s
and Mo3d lines during exposures to molybdenum hexacar-
bonyl at 140 K up to 25 L in several steps. The correspond-
ing valence band evolution is shown in Fig. 5.
Concerning the C1s lines (Fig. 2), before exposure, no
C1s peak is observed, attesting of the absence of carbon
contaminants on the sample. After the first exposure to
increasing Mo(CO) coverage. This peak is related to the
6
oxygen of the Mo(CO) molecule. In the same time, a peak
6
appears at 5.4 eV higher binding energy (shown in Fig. 3 in-
set) which is characteristic of a shake-up satellite resulting
from the same process as the one of the C1s level [14,15].
The photoemission spectra of the Mo3d 5/2 and 3/2
lines are presented in Fig. 4. Although Mo3d lines exhibit
a rather wide FWHM which indicates that several molyb-
denum species are present in the same time at the substrate
surface, these lines also exhibit a progressive shift towards
higher binding energies during exposures up to ca. 15 L.
Mo(CO) , a component appears at a binding energy equal
6
to about 288.0 eV. This component increases with increas-
ing Mo(CO) coverage and should thus be attributed to the
6
carbon of the simple physisorbed molecule as shown in a
previous study [14]. Another peak, at 5.4 eV higher binding
energy, increases in the same way. It can be attributed to
shake-up satellite which has been reported for all the car-
bonyl spectra [14,15]. Besides, shoulder appears on the
low binding energy side of the main peak (between 286.6
and 287.6 eV). The appearance of such a shoulder on the
6
Fig. 3. Evolution of the O1s line during an exposure to Mo(CO) between
Fig. 1. Evolution of the intensities of the Ti2p and Mo3d lines as a
0 and 25 Langmuir at 140 K. The spectrum in the right hand corner has a
larger energy window in order to evidence the satellite peak. Primary
photon energy: 600 eV.
6
function of Mo(CO) exposure. The dotted line corresponds to the change
of the slopes of both curves.

J. Prunier et al. / Surface Science 601 (2007) 1144–1152
1147
Fig. 4. Evolution of the Mo3d line during an exposure to Mo(CO)
between 0 and 25 Langmuir at 140 K. Primary photon energy: 320 eV.
6
Fig. 5. Evolution of the valence band during an exposure to Mo(CO)
6
between 0 and 25 Langmuir at 140 K. The valence band spectrum
obtained after an exposure of 200 L at 140 K is also shown. Primary
photon energy: 45 eV.
For higher exposures, the lines begin to shift in the oppo-
site side that is to lower binding energies. This behaviour
should be compared to the fact that 15 L is close to the re-
corded exposure threshold (14 L) corresponding to comple-
tion of the first adsorbed monolayer. Results given by
Fig. 4 simply show that different processes occur during
completions of the first and the following layers.
The shift recorded during the completion of the first layer
is towards higher binding energies. Such a kind of shift is of-
ten due to oxidation process of analysed element, which in-
volves the decrease of screening effect around the atoms.
The shift recorded after the completion of first monolayer
is in the opposite direction and can thus be related to the
opposite phenomenon, i.e. molybdenum reduction process.
These observations can be interpreted through the electron-
donor character of the CO group. If molybdenum hexacar-
bonyl loses a CO group, it loses also some partial negative
charge and it is thus submitted to a kind of oxidation, Mo
electrons then being less screened, their binding energies
appearing higher. Such a hypothesis can be supported by
the appearance of the shoulder on the right part of C1s lines
the components of condensed Mo(CO) without interfacial
6
layer contribution, the partially decomposed molecules
seem to be located at the TiO surface only.
2
Before exposure, the valence band spectrum is com-
posed of the mainly O2p derived band between about 3
and 9 eV, characteristic of stoichiometric TiO , with almost
2
no contribution around 1 eV corresponding to the Ti3d
states [12]. Upon deposition, different features appear that
the intensities increase with increasing the deposited
amount. In the same time, the oxide valence band de-
creases. A clear assignment can be done for all these peaks
by comparison with the 200 L exposure: indeed the spec-
trum recorded in this case is in good agreement with the
ones previously reported in the literature [14]. The feature
at 2.4 eV binding energy can be assigned to mainly Mo4d
derived molecular orbitals. It should be noticed that this
peak exhibits a resonance (not shown here) for photon
energies around 49 eV which correspond to the 4p–4d
absorption edge [16]. In the binding energy range of 7–
which could also be attributed to Mo(CO) (x < 6) forma-
x
tion through the partial molecule decomposition.
10 eV the CO-derived degenerate 5r–1p states are present
The shift to lower binding energy is observed for expo-
sures beyond 15 L occurs actually after the filling of the
first monolayer, i.e. as new adsorbed molecules are no
with the associated satellite at 5.4 eV higher binding en-
ergy. At least, the CO 4r level can be observed at a binding
energy close to 12 eV. Looking more closely to the feature
at the lowest binding energy, i.e. to the one corresponding
to Mo4d levels, a shoulder can be observed on the lowest
binding energy side of the peak. The presence of this peak
is closely related to the shoulder appearing on the C1s and
O1s lines (Figs. 2 and 3). These peaks which are at lower
binding energies than the main lines come from Mo(CO)x
(x < 6) species [17], which should have their origin from
longer in direct contact with TiO surface. Such a second
2
shift seems simply to correspond to the disappearance of
the first recorded shift. It could be attributed as the result
of adsorption of new non-decomposed Mo(CO)6 mole-
cules, which hide the ones previously adsorbed and par-
tially decomposed. In these conditions, it seems that
partial Mo(CO) molecule decomposition can occur for
6
molecules initially in contact with TiO surface only.
2
the partial decomposition of Mo(CO) [18–22].
6
The valence band evolution upon exposure is presented
in Fig. 5. The spectrum corresponding to a 200 L exposure
is also shown in the same energy range for reference pur-
pose. In this high exposure case, no contribution from both
3.2. Decomposition by thermal treatments
underneath TiO layers and interfacial adsorbed molecules
is expected. Hence, the different features can be assigned to
The second step of the experiment was the decomposi-
tion of the precursor in order to leave molybdenum on
2

1148
J. Prunier et al. / Surface Science 601 (2007) 1144–1152
the surface. The sample was thus heated up successively to
room temperature (ca. 300 K), and at different tempera-
tures in the 340–820 K range. Such experiments were fol-
lowed from a sample which was previously exposed to
2
5 L of Mo(CO) at 140 K. Phenomena are then followed
6
through Mo3d line (Fig. 6), C1s line (Fig. 7), O1s (not
shown here) and valence band evolutions. The parts close
to the Fermi level only of the latter spectra are shown in
Fig. 8.
It can be clearly seen in Fig. 6 showing Mo3d line evo-
lution during annealing, that, at room temperature, only a
very small part of molybdenum, previously present at a low
temperature, remains on the surface. This indicates that the
desorption temperature is lower than the decomposition
one. Thus, the molecules of the precursor are slightly
bound to the substrate surface and desorb before being
decomposed as demonstrated in other cases [18,23]. One
should note that different attempts were done to adsorb
molecular species at temperatures between 140 and
Fig. 8. Evolution of the low binding energy part of the valence band
during thermal treatments. The sample was previously exposed to 25
Langmuir of Mo(CO) at 140 K. Primary photon energy: 45 eV.
6
binding energies during the annealing. Using the whole of
our previous results on Mo/TiO system [1] as spectro-
2
2
00 K, but Mo(CO) does not adsorb for temperatures
scopic fingerprint for the Mo3d core level, one can suppose
that after annealing the deposits correspond to oxidized
6
higher than 180 K. After annealing at temperatures higher
than 300 K, the Mo3d line exhibits a strong shift to higher
molybdenum species MoO with x < 2.
x
As concerns the evolution of C1s lines (Fig. 7), the only
components, which remains in a very small amount on the
surface after the return to room temperature, are those cor-
responding to Mo(CO)x species. This confirms that
Mo(CO) completely desorbs and/or is decomposed during
6
the heating up to room temperature. However, the inten-
sity of the C1s component related to Mo(CO) (x < 6) spe-
x
cies does not really increase during return to room
temperature. This point seems to indicate that there is no
additional Mo(CO) decomposition phenomenon during
6
heating up to room temperature. In other words, during
annealing, Mo(CO) simply desorbs and the presence of a
6
small amount of molybdenum on the TiO surface after re-
2
turn at room temperature is to be related to the presence,
before subsequent annealing, of Mo(CO) (x < 6).
x
Fig. 6. Evolution of the Mo3d during thermal treatments. The sample was
In order to get rid of the carbon-based species remaining
on the surface, the samples were submitted to subsequent
annealings at different temperatures in the 340–820 K
range. It should be observed that, after an annealing at
previously exposed to 25 Langmuir of Mo(CO)
energy: 320 eV.
6
at 140 K. Primary photon
340 K, all the molybdenum subcarbonyl species observed
at room temperature disappeared but there is still some
remaining adsorbed carbon on the surface at a binding en-
ergy around 284.9 eV (Fig. 7). This carbon line is compara-
ble (energy, intensity) to the one obtained after the already
used deposition method (physical vapour deposition, PVD)
for the same deposited molybdenum amount. Thus, it can
be concluded that carbon observed after an annealing at
340 K corresponds to a kind of adventitious carbon ad-
sorbed on the surface as it is often the case in UHV exper-
iments even in the case of oxide systems. So, this proves
that, even though the Mo deposition method using molyb-
denum hexacarbonyl as precursor introduces some carbon,
a subsequent annealing at quite a low temperature induces
the desorption from the surface of all the carbon species
carried by the hexacarbonyl: after such an annealing, just
Fig. 7. Evolution of the C1s during thermal treatments. The sample was
previously exposed to 25 Langmuir of Mo(CO)
energy: 380 eV.
6
at 140 K. Primary photon

J. Prunier et al. / Surface Science 601 (2007) 1144–1152
1149
adventitious carbon is detected by photoemission. It should
also be noticed that no molybdenum carbide species are
present on the surface after annealings in opposition with
other studies carried out on the alumina substrate [24].
The evolution of the part closest to the Fermi level of the
valence band (Fig. 8) during annealing is remarkable: the
global intensity of the line recorded in the TiO2 gap
strongly decreases and just a component is detectable in
this line when the sample was at room temperature. This
single component can be attributed to Mo4d states, low
but detectable molybdenum amount remaining at the sur-
face. Moreover, such a component appears at the same en-
ergy like the component which was due to Mo in Mo(CO)x
(
x < 6). The one which was due to Mo in Mo(CO) actually
6
completely disappeared after going back to room tempera-
ture. After annealing at higher temperature, drastic
changes can be observed in the valence band region, the
peak recorded at room temperature disappeared while a
new component at about 0.8 eV appears. This line is very
similar to the one recorded for fractional Mo deposits [6]
carried out by the PVD method. In such a case, the peak
close to the Fermi level corresponds to Ti3d states due to
Fig. 9. Evolution of the low binding energy part of the valence band
during an irradiation to the zero-order light. An exposure of 25 Langmuir
Mo(CO) at 140 K was performed before irradiating. Primary photon
6
energy: 45 eV.
interaction between deposited molybdenum and TiO sur-
2
face, molybdenum being oxidized whereas titanium is re-
duced. A similar phenomenon seems to occur here. After
Mo(CO) (x < 6) decomposition, molybdenum is free to re-
x
act with substrate and this seems to begin from 340 K.
The data recorded on valence band are in agreement
with the ones recorded on Mo3d line: after annealing, i.e.
after total carbonyl species decomposition, the low amount
of molybdenum remaining at the surface reacts with TiO2
giving molybdenum oxide as well as reduction of titanium.
However, although it seems possible, using the decomposi-
tion of molybdenum hexacarbonyl by thermal treatments
to get rid of carbonyl species and to obtain molybdenum
oxide, the final deposited molybdenum amount is very
low. Besides, a noticeable point is the fact that whatever
the initial molybdenum hexacarbonyl exposure is, the
molybdenum amounts remaining at the surface is always
in the same fractional monolayer range. According to this,
it seemed to us that the key point was the balance between
Fig. 10. Evolution of the C1s line during an irradiation to the zero-order
light. An exposure of 25 Langmuir of Mo(CO) at 140 K was performed
6
before irradiating. Primary photon energy: 380 eV.
Attempts of irradiating the sample with photons were
simply done using photons provided by the zero-order
beam of the synchrotron radiation. This was followed
mainly from the evolutions of the valence band (Fig. 9)
as well as C1s line (Fig. 10) on a sample previously exposed
the Mo(CO) decomposition and desorption. Indeed, the
6
precursor amount which is not decomposed at cryogenic
temperature seems to simply desorb during annealing, the
remaining molybdenum amount being related to the for-
mation of subcarbonyl species at the early stage of the
deposition.
to 25 L Mo(CO) at 140 K.
6
The evolution of the low binding energy part of the va-
lence band upon irradiation of such a sample is shown in
Fig. 9. It can be seen in this figure that the shape of the line
previously attributed to Mo4d derived molecular is modi-
fied upon irradiation while its average binding energy is
strongly shifted to lower binding energies: a 0.3 eV shift
is evidenced after only 30 s of irradiation and it increases
up to 1.3 eV after the longest irradiation. Actually, the
appearance of such a broad and shifted line may be related
to the partial decomposition of the precursor, which is
more and more important upon irradiations.
3
.3. Decomposition by photon irradiation
In order to decompose the molybdenum hexacarbonyl
precursor in a more efficient way before annealing, i.e. be-
fore desorption of these molecules, attempts were done of
irradiating the sample with photons after the adsorption
step. Indeed this kind of procedure has been shown to be
efficient for Mo deposition on different metallic [25–29]
surfaces.
This hypothesis is evidenced in the C1s (Fig. 10) and O1s
lines (not shown here) where both the peaks attributed to

1150
J. Prunier et al. / Surface Science 601 (2007) 1144–1152
Mo(CO) and to its satellites decrease and shift to lower
binding energies upon irradiation because of decomposi-
tion to subcarbonyl species. Especially, after 25 min of irra-
denum on the substrate surface. It should be noticed that,
in the range of photons energies we were able to use (from
about 30 to 650 eV) no real influence of the excitation en-
ergy on the precursor decomposition was observed: only
the photon dose seems to play a role. This should certainly
not be the same using irradiations in the UV-range, closer
to the maximum in the adsorption spectrum of Mo(CO)6
[30], but unfortunately SX700 beamline does not allow to
reach such an energy range.
6
diation, no Mo(CO) species is not observable on the C1s
6
spectrum as shown in Fig. 10 and just Mo(CO) (x < 6)
x
species can be detected.
Moreover, after 25 min of irradiation, the main compo-
nent of the O1s line (not shown here) is clearly the one at
5
30.0 eV corresponding to the oxygen of the TiO surface.
2
This point is in agreement with the Mo(CO) decomposi-
It should also be noticed that each spectrum recording
implies irradiation of the sample and then induces low
but a noticeable decomposition of the precursor as evi-
denced in Figs. 9 and 10: after acquisition of some spectra,
shoulders which appeared on the right part of lines notice-
6
tion which leads to CO departure i.e. a decrease of the
CO component intensity of O1s line. In the same time, be-
cause CO molecules leave the surface, the substrate is less
hidden by adsorbed molecules and the intensity of TiO2
component of O1s line increases.
ably increased proving additional Mo(CO) (x < 6) forma-
x
Besides, after subsequent annealing at 420 K, no more
CO component of O1s line is observed while the C1s spec-
trum exhibits only one component related to adventitious
carbon species (Fig. 10). These observations prove that
after a mild annealing, the carbonyl decomposition and/
or desorption is completed. However, after such treat-
ments, one can note that a lot of molybdenum remains
tion. Such a decomposition can explain why a low amount
of Mo can remain on the TiO surface after going back to
2
room temperature even without specific zero-order irradia-
tion. Irradiation by the photon beam needed to record
spectra used to follow the reaction steps is enough to initi-
ate Mo(CO) decomposition. Actually, the decomposition
6
phenomenon revealed during the first adsorbed layer is
probably mainly due to spectrum recording. Indeed, in this
case, the adsorbed molecule amount is low (lower than one
monolayer) and the photon irradiation spent a long time
(several spectra were recorded for each adsorption step).
Moreover, for a same adsorption step, the precursor
decomposition is lower after the recording of the first spec-
trum than after the recording of the last one, i.e. after that
the sample was irradiated for a long time (see Section 3.1).
This fact can explain why the decomposition is more re-
vealed by a specific spectrum rather than by another one.
on the TiO surface as shown in Fig. 11e. This indicates
2
that in such a case the decomposition phenomenon domi-
nate the desorption one. Hence, it seems that after partial
Mo(CO) decomposition, subcarbonyl species cannot des-
6
orb during annealing, but should decompose. A simple
explanation of this phenomenon can be the change of
bounds between molecules and substrate as molybdenum
carbonyl is decomposed. In other words, although
Mo(CO)6 is slightly bound to the surface, Mo(CO)x
(
x < 6) is stuck to the TiO surface through a real chemical
2
bound, which avoids desorption during annealing and then
leads to a complete decomposition of the precursor.
A strong photon irradiation of the sample at cryogenic
temperature is thus an efficient way to make leaving molyb-
For instance, after an exposure of 15 L Mo(CO) , the
6
decomposition is less revealed by Mo4d additional compo-
nent in the valence band spectrum (Fig. 5) than by the shift
of the Mo3d lines towards higher binding energies (Fig. 4).
In such a case, valence band spectrum was recorded just
after adsorption, whereas Mo3d spectrum was recorded
afterwards.
Fig. 11 shows the Mo3d lines of some depositions car-
ried out from some Mo(CO) exposures with or without
6
zero-order irradiation. In this figure, it is clear that the
amount of molybdenum remaining at the substrate surface
is related to both exposition and irradiation. Moreover, the
binding energies of the Mo3d lines are functions of the
remaining molybdenum amount. For instance, the Mo3d
lines corresponding to the highest deposited Mo amount
is characterized by the binding energy of Mo3d5/2 line
equal to 227.8 eV (Fig. 11e). The valence band related to
the same sample is characterized by the states at the Fermi
level (see the topmost curve of Fig. 9). These two observa-
tions are in a good agreement with metallic molybdenum.
In the contrary, lowest deposits are characterized by a
Mo3d5/2 line binding energy in the 228.8–228.0 eV range.
Such a phenomenon can be explained by metallic clusters
having a very low size which then induces a less efficient
screening during photoemission process and thus a shift
Fig. 11. Comparison of the Mo3d spectra as a function of the deposited
Mo amount. The spectra were all obtained after an annealing at 420 K.
Spectrum (a) was obtained after an exposure of 10 Langmuir Mo(CO)
6
at
140 K without subsequent zero-order irradiation; Spectrum (b) was
6
obtained after an exposure of 10 Langmuir Mo(CO) at 140 K with
subsequent zero-order irradiation for 30 s; other spectra were obtained
after an exposure of 10 (c), 15 (d) and 25 (e) Langmuir of Mo(CO) at
40 K followed by zero-order irradiation for 25 min. Primary photon
energy: 320 eV.
6
1

J. Prunier et al. / Surface Science 601 (2007) 1144–1152
1151
Mo3d5/2 line equal to 227.8 eV), whereas such a film can
be obtained after annealing at least at 870 K by the PVD
deposition technique [3].
4. Conclusion
It was shown that it is possible, using the decomposition
of molybdenum hexacarbonyl followed by thermal treat-
ments to obtain molybdenum based phases and to get rid
of carbon species. However, although simple annealing al-
lows to obtain molybdenum amounts in the submonolayer
range, one should note that the remaining Mo amount
should be related to photon irradiation occurring during
the recording of spectra which initiates the decomposition
of a low part of the adsorbed precursor molecules. To ob-
tain higher deposits, sample irradiations by a photon beam
can be used.
Fig. 12. Ti2p spectra for the TiO
solid line) molybdenum deposition. The deposition was carried out from
an exposure of 10 Langmuir of Mo(CO) at 140 K with subsequent zero-
order irradiation for 30 s followed by an annealing at 420 K. Primary
photon energy: 510 eV.
2
surface before (doted line) and after
(
6
of lines towards higher binding energies [31]. However, if
such a phenomenon can induce significant shifts [32], these
ones are always lower than 1 eV whereas the one record
here can reach this value. Moreover, the Mo3d shift occurs
at the same time as Ti2p spectrum change (Fig. 12). Indeed,
in Ti2p spectrum, after the deposition of a low amount of
molybdenum followed by an annealing at 420 K, addi-
tional component clearly appears on the right part of
Ti2p3/2 line. Such an additional component is well known
Moreover, using this kind of precursor decomposition
and annealing at 420 K, the same kind of Mo nanostruc-
tures can be formed on the surface as after PVD followed
by thermal treatments at much higher temperatures. One
of the interests of the method using Mo(CO) comes thus
6
from the annealing temperature needed to obtain the final
molybdenum phases. Actually, like for many metals depos-
ited on oxides, some diffusion phenomena between the sub-
strate and the deposit may occur upon annealing, changing
the final deposit state. In the case of PVD deposited molyb-
denum, an annealing at temperatures in the 700 K range,
which is the one needed to obtain some specific phases,
should induce the diffusion of oxygen from the bulk to
the deposit leading to the formation of molybdenum oxide
instead of the desired structures [1,3]. On the contrary, in
the method using molybdenum hexacarbonyl decomposi-
tion, the needed temperatures are rather low and thus pre-
vent any bulk diffusion phenomena.
[
34] and corresponds to the titanium reduction. The Mo3d
line shift should thus be due to concomitant molybdenum
oxidation. In the case of low molybdenum deposits, molyb-
denum oxide phases seem thus to be formed.
Similar conclusions were obtained from the deposits
performed by the PVD method [33] after annealing at
6
73 K. Hence, it seems that molybdenum phases formed
by decomposition of Mo(CO) followed by a subsequent
6
annealing at 420 K and those obtained by classical PVD
deposition followed thermal treatments are chemically
pretty close. However, in our previous studies upon the
Acknowledgments
growth mode of Mo deposited on TiO by PVD technique,
2
one of the main conclusions was that, for films thicker than
one monolayer, the interfacial reaction is less related to the
initial surface stoichiometry and morphology than to kinet-
ical parameters such as, for instance, deposition rate. A
consequence of this fact is that metastable state occurs dur-
ing PVD deposition at room temperature and that anneal-
ing step (at 673 K) is necessary to reach a kind of
equilibrium [33]. In the case of Mo deposition using
This research was done with the support of the Euro-
pean contract STRP NanoChemSens. The European Com-
mission is also gratefully acknowledged for the access to
ASTRID facilities (Contract No. HPRI-CT-2001-00122).
References
[
[
[
[
[
[
[
1] B. Domenichini, in: Charles P. Norris (Ed.), Surface Science
Research, Nova Science Publisher Inc., 2005.
2] B. Domenichini, S. Petigny, V. Blondeau-Patissier, A. Steinbrunn, S.
Bourgeois, Surf. Sci. 468 (2000) 192.
Mo(CO) , any influence of deposition rate is suppressed
6
because the first step of the protocol is always the adsorp-
tion of the precursor and is well separated from the second
one, i.e. the decomposition. Thus, this method allows to
reach the equilibrium after annealing at temperature as
low as 420 K only, this temperature being efficient to to-
tally decomposed the precursor. The proposed deposition
technique allows thus to perform stable molybdenum
structures at lower temperature than by the PVD method.
For example, the highest deposit performed during this
study is metallic (characterized by the binding energy of
3] B. Domenichini, A.M. Flank, P. Lagarde, S. Bourgeois, Surf. Sci. 560
(
2004) 63.
4] S. P e´ tigny, H. Most e´ fa-Sba, B. Domenichini, E. Lesniewska, A.
Steinbrunn, S. Bourgeois, Surf. Sci. 410 (1998) 250.
5] B. Domenichini, V. Blondeau-Patissier, M.J. Casanove, G.D. Lian, S.
Bourgeois, J. Cryst. Growth 263 (2004) 256.
6] B. Domenichini, P. Kr u¨ ger, A. Brevet, S. Bourgeois, Z. Li, P.J.
Møller, Appl. Surf. Sci. 244 (2005) 403.
7] F. P e´ rez-Cadenas, C. Moreno-Castilla, F.J. Maldonado-H o´ dar,
J.L.G. Fierro, J. Catal. 217 (2003) 30.

1152
J. Prunier et al. / Surface Science 601 (2007) 1144–1152
[
8] A.M. Sica, I.M. Baibich, C.E. Gigola, J. Mol. Catal. A 195 (2003)
25.
9] L.C. Ord o´ n˜ ez, P. Roquero, P.J. Sebastian, J. Ram ´ı rez, Catal. Today
07–108 (2005) 46.
[21] S. Jaenicke, W.L. Loh, Catal. Today 49 (1999) 123.
[22] S.K. Clowes, E.A. Seddon, E.M. McCash, Surf. Sci. 464 (2000)
L667.
[23] Z. Jiang, W. Huang, J. Jiao, H. Zhao, D. Tan, R. Zhai, X. Bao, Appl.
Surf. Sci. 229 (2004) 43.
[24] M. Kaltchev, W.T. Tysoe, J. Catal. 196 (2000) 40.
[25] D.V. Potapenko, J.M. Horn, R.J. Beuhler, Z. Song, M.G. White,
Surf. Sci. 574 (2005) 244.
2
[
1
[
10] Z. Song, T. Cai, J.A. Rodriguez, J. Hrbek, A.S.Y. Chan, C.M.
Friend, J. Phys. Chem. B 107 (2003) 1036.
11] M.M. Biener, C.M. Friend, Surf. Sci. 559 (2004) L173.
12] Z. Zhang, S.P. Jeng, V.E. Henrich, Phys. Rev. B 43 (1991) 12004.
13] T.E. Gallon, Surf. Sci. 17 (1969) 486.
[
[
[
[
[26] H.H. Huang, C.S. Sreekanth, C.S. Seet, X. Jiang, G.Q. Xu, Surf. Sci.
365 (1996) 769.
14] E.W. Plummer, W.R. Salaneck, J.S. Miller, Phys. Rev. B 18 (1978)
1
673.
[27] Z.C. Ying, W. Ho, J. Chem. Phys. 94 (1991) 5701.
[28] S.K. So, W. Ho, J. Chem. Phys. 95 (1991) 656.
[29] T.A. Germer, W. Ho, J. Chem. Phys. 89 (1988) 562.
[30] N.A. Beach, H.B. Gray, J. Am. Chem. Soc. 90 (1968) 5713.
[31] W.F. Egelhoff Jr., Surf. Sci. Rep. 6 (1987) 253.
[32] H.-P. Steinr u¨ ck, F. Pesty, L. Zhang, T.E. Madey, Phys. Rev. B 51
(1995) 2427.
[
[
[
[
15] J. Bustad, H. Tillborg, C. Enkvist, A. Nilsson, S. Lunell, S. Osborne,
A. Sandell, N. Maetersson, S. Svensson, Chem. Phys. 179 (1994) 303.
16] H. Matsuoka, K. Ohtake, R. Yamamoto, M. Doyama, H. Sakamoto,
T. Mori, M. Fujisawa, K. Soda, S. Suga, Physica B 143 (1986) 189.
17] P. Liu, J.A. Rodriguez, J.T. Muckerman, J. Hrbeck, Surf. Sci. 530
(2003) L313.
18] Y. Wang, F. Gao, M. Kaltchev, D. Stacchiola, W.T. Tysoe, Catal.
Lett. 91 (2003) 83.
19] M. Kaltchev, W.T. Tysoe, J. Catal. 193 (2000) 29.
20] J. Evans, B.E. Hayden, Gang Lu, J. Chem. Soc., Faraday Trans. 92
[33] B. Domenichini, G.A. Rizzi, P. Kr u¨ ger, M. Della Negra, Z. Li, M.
Petukhov, G. Granozzi, P.J. Møller, S. Bourgeois, Phys. Rev. B 73
(2006) 245433.
[
[
[34] U. Diebold, Surf. Sci. Rep. 48 (2003) 53.
(1996) 4733.