Photoelectron spectromicroscopic study of the spreading behavior of MoO3 on titania and alumina model supports

Article abstract of DOI:10.1021/jp971600w

The spreading of micron-sized MoO₃ particles deposited on model supports, alumina and titania thin films, has been studied by scanning photoemission spectromicroscopy. It was shown that Mo species released from the MoO₃ crystals spre

Full text of DOI:10.1021/jp971600w

10004
J. Phys. Chem. B 1997, 101, 10004-10011
Photoelectron Spectromicroscopic Study of the Spreading Behavior of MoO3 on Titania and
Alumina Model Supports†
S. Gu1nther, M. Marsi, A. Kolmakov, and M. Kiskinova
Sincrotrone Trieste, Padriciano 99, 34012 Trieste, Italy
M. Noeske
Abt. Oberfla¨chenchemie und Katalyse, UniVersita¨t Ulm, Oberer Eselsberg, 89069 Ulm, Germany
E. Taglauer
Max-Planck-Institut fu¨r Plasmaphysik, Boltzmannstrasse 2, 85748 Garching, Germany
G. Mestl
Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14195 Berlin (Dahlem), Germany
U. A. Schubert and H. Kno1zinger*
Institut fu¨r Physikalische Chemie, Ludwig-Maximilians UniVersita¨t-Mu¨nchen, Sophienstrasse 11,
80333 Mu¨nchen, Germany
ReceiVed: May 14, 1997X
The spreading of micron-sized MoO3 particles deposited on model supports, alumina and titania thin films,
has been studied by scanning photoemission spectromicroscopy. It was shown that Mo species released
from the MoO3 crystals spread over the thin oxide film under conditions (heating to 720 K for 6 h in air)
similar to those used for catalysts on powdered supports. The changes in the surface morphology were followed
by elemental mapping and photoelectron spectroscopy from submicron spots where the intensities and kinetic
energies of the Mo 3d, Al 2p, Ti 3d, and O 1s photoelectrons were used as fingerprints for the chemical
composition of the surface. The high spatial resolution when imaging the spread phase on the laterally
inhomogeneous samples made it possible to throw light onto the active transport mechanism of the wetting
Mo oxide species during heat treatment of the model systems.
Introduction
evolution of surface mobility. This motivated the increasing
number of fundamental UHV-STM studies dealing with the
mobility of metal overlayers on metal surfaces, surface alloy
formation, or their segregation.4-6 Comparably well-defined
experiments, however, on the mobility of oxide overlayers on
oxide supports have not been conducted yet.
Many industrially important solid catalysts consist of at least
two different compounds: an oxidic support material and the
usually highly dispersed active phases, which are often metals,
oxides, or sulfides. At high temperatures and in the presence
of reactive gases, e.g., under catalytic conditions, these active
phases can develop surface mobility, which may result in
sintering of the active phase (catalyst deactivation) or in wetting
and spreading across the support surface (activation or reactiva-
tion). The following well-known phenomena are a consequence
of this surface mobility at high temperatures: (i) sintering and
redispersion of the active catalyst phase (often a function of
the reactive gas phase); (ii) strong metal-support interaction
(SMSI), where the active noble metal catalyst is partially
encapsulated by an overlayer of the support oxide, or (iii)
spreading of salts or oxides across support surfaces.1-3 Al-
though surface mobility strongly affects catalytic properties, e.g.,
catalyst aging and deactivation, it has not been considered
frequently in the literature. All these processes are strongly
related and are suggested to depend on the surface and interface
free energies of the phases present and the mean free paths of
the diffusing entities. The mechanisms on an atomic level that
lead to mobility are not fully understood yet, and a physico-
chemical theory is still missing that completely describes the
Xie and co-workers were the first to report that the XRD
reflections of MoO3 in physical mixtures with Al2O3 are
disappearing upon calcination at about 700 K.7-9 XPS inves-
tigations supported their interpretation of a Mo oxide monolayer
formation during this treatment. Stampfl et al. also concluded
from Raman and IR spectra that Mo oxide monolayers were
formed in physical mixtures of MoO3 with various support
oxides upon calcination in air at 720 K.10 ISS studies revealed
that MoO3 in physical mixtures with alumina or titania was
spontaneously spreading over the support surface at temperatures
close to its Tammann temperature and even in dry oxygen.11,12
Raman spectroscopy was additionally used to characterize the
surface Mo species formed under spreading conditions.11-13
After heat treatment at 720 K (and 800 K) in dry oxygen only
crystalline MoO3 could be detected, whereas under the presence
6-
of H2O vapor the formation of surface molybdates (Mo7O24
on γ-Al2O3) was observed. Thus, it was concluded that
spreading on one hand and the chemical conversion into
polymolybdates on the other were two independent processes.
This observation was further confirmed by EXAFS.14 Raman
microscopy demonstrated that the transport of MoO3 across the
alumina surface occurred over several hundred micrometers.15
† Dedicated to Prof. Sir J. M. Thomas on the occasion of his 65th
birthday.
X Abstract published in AdVance ACS Abstracts, October 15, 1997.
S1089-5647(97)01600-3 CCC: $14.00 © 1997 American Chemical Society

MoO3 on Titania and Alumina
J. Phys. Chem. B, Vol. 101, No. 48, 1997 10005
In situ high-temperature Raman spectroscopy suggested that the
actual spreading phase was surface molten MoO3, which seemed
to consist of MoO3 monomers or small oligomers.16 This
surface molten phase was highly reactive toward H2O as proved
by in situ exposure to water vapor at room temperature which
led to the formation of surface polymolybdates. The reduction
of the surface free energy of the whole system, support/MoO3,
was discussed as the driving force for this solid wetting.1,2
One of the disadvantages of the previous wetting and
spreading studies was the disperse nature of the oxide powders
used as supports. In this case the amorphizity of the support
particles affects the spreading due to the existence of altered
surface free energies. This introduces uncertainty when studying
the spreading phenomena. That is why the experiments,
reported in this paper, were performed using model supports,
namely, thin Al2O3 and TiO2 films on Al or Ti foils. Using
these supports, we reduced the possible effects of the support
polydispersity and expect to get direct experimental evidence
for the molecular nature of the spreading of MoO3 across support
oxides. The high lateral resolution and analytical capabilities
of the used scanning photoelectron microscope allowed us to
image directly this spreading process and to examine the
variations in the local surface composition and morphology of
the spread layer.
Figure 1. Conventional XP spectra taken before (i) and after heat
treatment (ii) (660 K for 5 h) of model samples where a small amount
of MoO3 crystals is dispersed on an Al2O3-Al (A) or a TiO2-Ti (B)
support foil, respectively. For determining the binding energy in system
A charging of 2 eV had to be considered (see text). No charging
occurred for system B.
crystals in methanol was taken with a pipet and placed on a
moderately heated (320 K) support foil. The methanol evapo-
rated instantaneously, and the small MoO3 particles remained
fairly homogeneously dispersed on the surface. By this
procedure we obtained a model system that is comparable to
samples where supported MoO3 model catalysts were produced
by physical mixing of MoO3 crystals with Al2O3 support
powder.11-13 One of the advantages of our model samples is
the very defined geometry, because the support is a flat oxide
film. This makes the samples suitable for spectromicroscopic
studies, where the topography artifacts become important. The
MoO3 crystals stick well on the support so that the sample can
be transferred into or outside the SPEM without loss of material.
The spreading of molybdena on the support was investigated
by characterizing the samples before and after heat treatment
in air (660-800 K for 5-7 h), using either a standard XPS
system or the laterally resolving SPEM. For the latter measure-
ments, we placed small scratches as markers on the support
foil. With the help of an optical microscope (OLYMPUS BH2,
resolution <1 µm), the position of the MoO3 particles could be
determined relative to the markers. Thus, it was possible to
reproducibly position the X-ray beam of the SPEM onto specific
MoO3 particles and to image the Mo distribution on a selected
area of the model surface before and after the heat treatment.
The surface sensitivity of the laterally resolving XPS enabled
us to determine the evolution of very thin surface layers as
expected for spreading systems.
Experimental Section
The experiments were performed using a conventional XPS
system (VSW ESCA 100) and the scanning photoelectron
microscope (SPEM) on the ESCAMICROSCOPY beam line
of the ELETTRA synchrotron radiation source.17 In the SPEM
the photon beam, provided by an undulator and monochroma-
tized by a spherical grating monochromator, was demagnified
to a spot of 0.15 µm diameter by using a zone plate optical
system. The emitted photoelectrons (PEs) were collected by a
100 mm hemispherical analyzer with a large acceptance angle,
mounted at 70° with respect to the sample normal and the
incident photon beam. The microscope can work in imaging
mode, collecting photoelectrons with chosen kinetic energy
while scanning the sample, and in spectroscopy mode, measur-
ing energy distribution curves (EDC) with the beam focused to
a feature selected from the maps. The contrast in the maps
due to variation of the photoelectron intensity is a fingerprint
for the local chemical composition of the imaged surface. The
lateral resolution of the SPEM is limited by the spot size of the
demagnified photon beam. All photoemission measurements
in this study were performed with photon energy between 580
and 590 eV and energy resolution of 0.4 eV.
As support samples, Al and Ti foils (Goodfellow, purity
99.999% and 99.6%, respectively) were used which were
mechanically mirror polished and anodically oxidized following
a recipe described elsewhere.18,19 Subsequent heat treatment
of the foils (823 K for Al2O3 and 623 K for TiO2) for 4 h
provided the transformation of the amorphous oxide layer into
a thin, polycrystalline oxide film resulting in a ∼20 nm thick
Al2O3 film on the Al foil and a ∼50 nm thick TiO2 film (anatase
modification) on the Ti foil. These oxidized foils provide
surfaces comparable to those of real catalysts, having the
advantage not to charge up severely when being illuminated
with soft X-rays. While single sapphire crystals charge up to
∼200 V under the highly focused X-ray beam of the SPEM,
the charging on the thin Al2O3 oxide films was less than 10 V
whereas it was negligible on the TiO2 films.
Results
The first characterization of the model samples consisting of
a few MoO3 crystals placed on a flat Al2O3-Al or a TiO2-Ti
support substrate was performed using a conventional photo-
electron spectroscopy system. Figure 1 shows the Mo 3d core
level spectra taken from both model systems before (i) and after
annealing (ii) at 660 K for 5 h in air. A significant increase of
the Mo 3d core level peak intensity after the heat treatment is
clearly visible for both supports, the alumina (A) and the titania
(B) thin oxide film. The same effect is also visible, if the
experiment is repeated when the support material is replaced
by a sapphire single crystal (not shown here), which confirms
that the thin oxide films serve as a suitable support material.
Due to the surface sensitivity of integrative XPS, one can
attribute the enhanced intensity of the Mo 3d core level peak
In order to prepare a model surface for a monolayer type
catalyst, small molybdena particles were placed on the support
foil. For this purpose, a drop of a suspension of the MoO3

10006 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gu¨nther et al.
Figure 2. (a) Image of a big MoO3 crystal (length 40 µm) on an Al2O3-Al support foil; the analyzer was tuned to the Mo 3d core level peak. The
imaged area is identical before and after heat treatment (720 K for 6 h) of the sample. (b) Reversed image of the nonannealed sample when tuning
the analyzer to the Al 2p core level peak. (c) XP spectra taken from the big crystal (i) and from the support before (ii) and after annealing the
sample (iii).
to the existence of molybdena species released from the MoO3
crystals and spread over the support surface, because before
the spreading the MoO3 particles cover only a small fraction of
the surface. To clarify this picture, which is in agreement with
the previous investigations of powder catalysts,11-13 we followed
the spreading of the Mo species on a micron scale with the
help of photoelectron emission microscopy. Therefore, a sample
with small MoO3 particles dispersed on an Al2O3-Al foil was
imaged by SPEM before and after annealing at 720 K for 6 h
in air.
The SPEM experiments included two-dimensional elemental
mapping with the analyzer tuned to a selected photoelectron
kinetic energy and photoelectron spectroscopy from a submicron
spot positioned on the MoO3 particles or away from them. The
contrast of the maps was determined by the surface and near
surface concentration of the probed element and by topographic
artifacts due to the anisotropic angular distribution of the emitted
photoelectrons. For the present system, because of the large
height of the MoO3 particles the topographic effects are
significant. However, they affect negligibly the spectroscopic
measurements because the size of the MoO3 particles allowed
us to choose spots for taking EDC spectra away from the edges.
A fairly large MoO3 crystal (length 40 µm) can be seen in
the Mo 3d two-dimensional micrograph shown in Figure 2a.
Surprisingly, the image of this area was identical prior and after
annealing in air. Even the small, micron-sized particles covering
the big crystal were not affected by the heat treatment within
our resolution limit. Microspot core level spectroscopy proved
that these particles were molybdena crystals as well. The
shadow on the left side of the particles is a typical topographic
effect. Since the analyzer is mounted at a grazing angle on the
right side of the displayed image, the particles prevent the

MoO3 on Titania and Alumina
J. Phys. Chem. B, Vol. 101, No. 48, 1997 10007
photoelectrons emitted beneath their left edge to reach the
analyzer in direct line-of-sight.
mesoscopic support defects, such as holes or scratches, on the
island distribution can be detected within our resolution. The
spectromicroscopic data also ruled out a Mo concentration
gradient in the close vicinity of the deposited MoO3 particles:
within the resolution limit no dark circle around the small MoO3
particles can be seen in the Al 2p maps in Figure 3b.
Despite the strong topographic effects, the Al 2p image in
Figure 2b shows the expected reverse contrast, consistent with
the different chemical composition of the support and the
deposited active phase. This is confirmed by the microspot
photoelectron spectra shown in Figure 2c. The top two spectra
are measured on the large crystal (i) and on the support away
of it (ii) before annealing the sample. Before annealing, the
spectra from the MoO3 crystal (i) contain only the O 1s (not
shown in Figure 2c), the Mo 3d, and the O(KLL) Auger peaks.
The spectra taken from the uncovered support surface (ii) contain
only the O 1s (not shown), the Al 2s, Al 2p, and the O(KLL)
Auger peaks, which proves the absence of any Mo species
dispersed away from the deposited MoO3 crystals.
Experiments similar to those described above (for the Al2O3-
Al support) were performed with samples where MoO3 crystals
were deposited on a TiO2-Ti foil. Here a treatment was applied
where the crystals were intended to be modified on a lateral
scale detectable with the SPEM. A smaller initial amount of
MoO3 crystals and a more severe heat treatment of the sample
(annealing at 720 K for 4 h and increasing the temperature to
800 K for 3 h) were chosen. Since charging in this system is
very low, the energy position of the Mo 3d core level peak is
mainly determined by the chemical state of the surface species.
These effects will be discussed in more detail in the next section.
The Mo 3d PE maps of Figure 4a,b reveal the extinction of the
MoO3 microcrystals (size ∼2 µm) after the above-described heat
treatment of the sample. The long vertical line in the right side
of the larger area image is the end of a scratch that served as a
marker to find the same area, where the MoO3 particle appears
as a bright spot before heating. The zoomed images clearly
confirm that the MoO3 particle has completely vanished after
the heat treatment. Spot PE spectra (not shown here) proved
that no traces of the particle (enhanced local Mo oxide content)
were left at its former position. To rule out the observation of
an exceptional event, we confirmed this finding for six different
areas containing MoO3 particles. As in the case of a Al2O3-
Al support, local spectroscopy proved again that the TiO2-Ti
foil was completely covered by a thin spread molybdenum oxide
phase after the heat treatment. Again, mesoscopic defects of
the substrate, like the imaged scratch in Figure 4, do not seem
to promote inhomogeneities in the spread phase.
The different status of the sample before and after annealing
can be characterized by following the spectroscopic information
of Figure 2c. While spectrum i, taken from the big crystal of
Figure 2a, does not change after the heat treatment of the sample,
the spectra taken from the free support show distinct differences.
As was mentioned, we can rule out the presence of Mo species
on top of the support before the heat treatment (spectrum ii).
After annealing, the Mo 3d core level peak is clearly detected
on top of the support (spectrum iii). This indicates that a small
amount of Mo species is released from the MoO3 crystals and
spreads over the Al2O3-Al support foil as a thin film or as
highly dispersed MoOx species. Since the MoO3 particles
contain a lot of material, the loss of Mo species due to spreading
is not visible within the resolution limit of the microscope.
To obtain information about the lateral distribution, we have
to tackle a problem. As can be seen for the case of MoO3 on
the Al2O3-Al support in Figure 2c, there was a notable
difference (up to ∼10 eV) between the kinetic energies of the
Mo 3d photoelectrons emitted from the large crystal and from
the spread Mo oxide phase. As will be discussed in the next
section, these energy shifts are due to a mixture of charging
effects and changes in the Mo oxidation state. The different
kinetic energy of the Mo 3d photoelectrons accounts for the
preserved contrast of the Mo 3d maps measured before and after
annealing: in both cases the analyzer was tuned to the Mo 3d
levels corresponding to the compact MoO3 particles. Since the
deposited large Mo oxide particles and the spread phase cannot
be imaged at the same time when tuning to the Mo 3d core
level, more precise quantitative information on the surface
morphology can be obtained from the contrast of the Al 2p
maps. The contrast of the Al 2p maps varies with the local
thickness of the spread molybdena phase. In the Al 2p images
of the support the darker areas correspond to stronger attenuation
of the Al 2p emission due to local enrichment of Mo oxide
species covering the support foil, as will be seen below.
Photoinduced Reduction of the Mo Oxide Species
As was mentioned already, the energy position of the PE core
level peaks have to be studied in detail. Due to the noncon-
ductive character of the active phase and the supports and the
morphology of our samples, effects such as differential charging
and Fermi level decoupling should be considered when evaluat-
ing the core level binding energies from the measured photo-
electron kinetic energies.20
First, we will focus on the peak positions of the spectra
obtained with a standard laboratory Al KR X-ray source where
(differential) charging is less severe. For this experimental setup
the work function of the analyzer was determined in a separate
experiment. Thus, an absolute energy scale is obtained and the
peak shifts due to charging can be evaluated. The charging of
the different substrates was measured very reproducibly from
experiment to experiment within a scatter of less than 0.4 eV.
The energy position of the Mo 3d core level peak was cross-
checked by measuring relative to the O 1s (530 eV for TiO2
and 531 eV for Al2O3)21 and relative to the C 1s peak (284.4
eV),21 in fair agreement with the determined absolute energy
scale. Therefore, an error of ∼0.4 eV for the peak position of
the Mo 3d core level (and thus the chemical state of the Mo
oxide species on the model surfaces) seems to be realistic.
For the system MoO3/TiO2 charging of the substrate and the
MoO3 crystals definitely remained below 0.5 eV. Referencing
to the binding energy of O 1s (530 eV) and C 1s (284.4 eV)
confirms a Mo 3d5/2 peak position at 232.6 eV for spectra taken
from samples with MoO3 crystals placed on the TiO2-Ti
support at room temperature and samples after heat treatment.
This proves an oxidation state of +6 for both Mo oxide species,
Figure 3a,b shows Al 2p PE maps of the nonannealed and
annealed sample in which the contrast shows a conglomerate
of one big and several small MoO3 particles. The larger area
image of the same region, displayed in Figure 3b, reveals the
appearance of darker areas on the support foil after the heat
treatment. In order to emphasize the contrast variation, inserts
i and ii show zoomed images of the areas marked in Figure 3b.
Figure 3c shows the spectra taken from the center of these two
areas. The inverse Mo 3d and Al 2p intensity from the darker
and brighter spots confirm the lateral inhomogeniety of the
spread phase. The dark spots in Figure 3b can be described as
islands, where the amount of Mo oxide species, covering the
support, is enhanced. The maps evidence that the islands are
homogeneously distributed. The exact nature of these islands
in the spread molybdena phase is unclear. No influence of

10008 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gu¨nther et al.
Figure 3. (a) Al 2p image (size 30 × 30 µm2) of a conglomerate of small MoO3 crystals on the Al2O3-Al support before annealing (720 K for
6 h). (b) The same area imaged after the heat treatment (size 50 × 50 µm2). Zoom i and ii emphasize the occurrence of dark spots on the support
foil. (c) XP spectra taken from the center of zoom i (bright area) and of zoom ii (dark spot).
the MoO3 crystals and the spread phase.21 Oxygen loss of the
Mo oxide species in vacuum or induced during the X-ray
irradiation were not detectable in the PE spectra.
In the MoO3/Al2O3 system the substrate and the MoO3
particles charged up to ∼2 eV. By referencing again to O 1s
(531 eV) and C 1s (284.4 eV), the peak position of the Mo
3d5/2 core level was determined as 232.6 eV for all samples.
Thus, again it is confirmed that the oxidation state of the MoO3
crystals and the spread phase on the Al2O3-Al substrate is +6.
As could be seen already in Figure 2c, the situation changes
if the samples are illuminated with the intense and highly
focused X-ray beam of the SPEM. For the system MoO3/TiO2,
the charging of the substrate was almost negligible (below 1
eV), and more importantly, no differential charging occurred.

MoO3 on Titania and Alumina
J. Phys. Chem. B, Vol. 101, No. 48, 1997 10009
Figure 4. Mo 3d images of a MoO3 crystal on top of a TiO2-Ti support foil before (a) and after heat treatment (b) (720 K for 4 h + 800 K for
3 h).
Using as a reference the kinetic energy measured for the C 1s
peak (if present) and the O 1s core level (assuming a binding
energy of 531 eV for MoO3 and 530 eV for TiO2),21 we
evaluated a binding energy of the Mo 3d5/2 core level of ∼230
eV for the MoO3 particles before and after spreading. The
binding energy for the spread Mo oxide phase after annealing
was ∼228 eV. Using the results of the already-described
conventional XPS analysis, we can rule out oxygen loss due to
the vacuum environment. Therefore, we can state that the
reduction is initiated by the photon beam. The MoO3 particles
are reduced to an oxidation state of +4 whereas the spread phase
is almost completely reduced to its metallic state.21
The illuminating photon beam affects equally the Mo oxide
species on the Al2O3-Al support although the analysis is more
complex since differential charging occurs. While the charging
of the substrate is low (since the oxide film is only ∼20 nm
thick) the micrometer-thick crystals are charged up to ∼10 eV
relative to the substrate. Again, referencing against C 1s (if
present) or O 1s (binding energy 531 eV for both Al2O3 and
MoO3)21 proves that the MoO3 crystals (before and after
annealing) are reduced to an oxidation state of +4 (Mo 3d5/2 at
230 eV) while the spread Mo oxide phase reaches nearly its
metallic state (Mo 3d5/2 at 228 eV) when being illuminated.
The energy shift of the differential charging (up to +8 eV for

10010 J. Phys. Chem. B, Vol. 101, No. 48, 1997
Gu¨nther et al.
the MoO3 particles) and the different photoinduced reduction
of the islands and the spread phase (+2 eV for the MoO3
particles) account for the observed shift of the Mo 3d core level
in Figure 2c) which prevented reliable imaging of the spread
phase on the Al2O3-Al support when tuning the analyzer onto
Mo 3d core level photoelectrons.
the vicinity of the MoO3 crystals, in agreement with the
described experiments. In this case, it would still be a problem
to explain the fairly homogeneously distributed islands of
enhanced Mo density on the spread molybdena film, which
cannot be correlated to the position of the MoO3 crystals.
Therefore, the transport mechanism a as the only spreading
process seems to be unlikely for the systems investigated.
The important new finding in the present study is the lateral
inhomogeneity of the spread Mo oxide phase: formation of
islands of higher Mo oxide concentration of a size up to a few
microns. These islands are almost homogeneously distributed
with an island-island distance of 1-10 µm. In order to explain
these lateral inhomogeneities by the diffusion mechanism b, the
surface mobility or the stability of the initial island nucleus
should be taken into account. Since the island-island distances
are large on an atomic scale, this suggests fairly mobile surface
species or/and very unstable nuclei, i.e., the Mo oxide species,
travel rather long distances on the surface before being trapped
in an island. A common feature in diffusion-controlled surface
processes is that the mobility and the adsorption energy are
influenced by surface defects, resulting in an uneven island
density distribution, e.g., as step decoration. The absence of
variations in the island density due to visible surface defects of
our samples can be used as an argument that questions the
determining role of surface diffusion in the spreading process.
However, the nucleation centers are not necessarily the visible
micron-sized topographic features, but might be structural and/
or compositional defects on a nanoscale or atomic scale, not
detectable with our spatial resolution.
The fact that the MoO3 particles were reduced equally before
and after the heat treatment of the samples confirms that the
spreading process does not change the stability of this species
against photoreduction. On the other hand, the spread phase
loses oxygen more easily, resulting in a complete reduction when
illuminated by the focused X-ray beam. As a process inducing
the oxygen loss, we propose a Knotek-Feibelman mechanism
which has been found for WO3 which is structurally and
chemically comparable to MoO3.22 The reason why the spread
phase loses oxygen easier might be related to the amorphous
nature of this phase. The fact that before illumination the
oxidation state of the spread Mo oxide is +6 (as for MoO3)
still provides the possibility of a Knotek-Feibelman mechanism
although the oxygen atoms interacting with the spread Mo might
be bonded to the metal atoms of the substrate as well.
In the last section we will discuss our data with regard to the
transport mechanisms that govern the spreading of the Mo oxide
species during the heat treatment of the model samples.
Discussion of the Transport Mechanism during
Spreading
As mentioned in the Introduction, the described experiments
aimed at providing deeper insight into spreading phenomena
by replacing the support powder by a flat support film. We
outlined in the previous section that spreading of Mo species
on the support foil could be observed by standard XPS and
additionally characterized by SPEM.
There are three possible transport mechanisms for Mo oxide
species that spread across the support surface after being released
from the MoO3 crystals:1 (a) Unrolling carpet mechanism: Mo
oxide species, diffusing on top of crystalline MoO3 particles,
are trapped at the edge of the crystal when reaching the support
surface. (b) Surface diffusion on the support: Mo oxide species,
detached from the MoO3 crystals, diffuse and nucleate on the
surface of the support foil and eventually recrystallize at certain
spots. (c) Transport via gas phase: Mo oxide species from the
MoO3 crystals evaporate in the gas phase and via subsequent
recondensation spread on the support foil and eventually
recrystallize at certain spots. (d) A combination of processes b
and c is also possible.
The unrolling carpet mechanism a was favored in ref 16,
because a surface melting of the MoO3 phase could explain the
recorded high-temperature in situ Raman spectra, whereas no
accordance was found with published spectra for the gas-phase
Mo oxide species. If mechanism a is the active transport
mechanism, we would expect the MoO3 crystals to wet the
support in the direct vicinity of a crystalline particle. Thus,
the radius of the wet surface around a crystal would increase at
the expense of the crystal height, which should result in a Mo
oxide concentration gradient moving away from the crystal. The
reported spectromicroscopic data do not provide any evidence
for a higher Mo oxide density around the MoO3 crystals within
our resolution. Thus, it seems unlikely that the unrolling carpet
mechanism a is the active one. Our data can be explained in
the frame of an unrolling carpet mechanism only assuming a
very efficient wetting mechanism of the moving Mo oxide
species that nearly instantaneously stops after the surface is
completely covered by a thin molybdena film. In this case one
would not expect a thickness gradient of Mo oxide species in
If the spreading process follows the gas phase transport
mechanism c, it is needed to find out whether the recondensation
of the Mo oxide species takes place at special locations on the
surface and how these can be related to the observed islands.
In order to distinguish unambiguously between the active
transport mechanisms, further experiments are necessary; e.g.,
the effect of the gas phase composition and the humidity during
heat treatment should be studied in detail.
Conclusion
The results on MoO3 spreading on Al2O3-Al and TiO2-Ti
support surfaces obtained by synchrotron radiation photoemis-
sion spectromicroscopy have manifested the power of laterally
resolved compositional surface analysis in understanding the
transport processes on supported catalyst surfaces. It has been
shown that, after heat treatment in air, the model samples,
containing MoO3 particles on an Al2O3-Al or a TiO2-Ti
support surface, respectively, were covered with a thin molyb-
dena phase that included islands of higher Mo concentration.
These islands are fairly homogeneously distributed on the
support, and their formation cannot be correlated with prefer-
ential nucleation at mesoscopic topographic defects of the
support or at the deposited MoO3 particles. The presented
results suggest that the unrolling carpet mechanism can be
considered as a transport mechanism controlling the spreading
only assuming a very efficient wetting mechanism of the moving
Mo oxide species that nearly instantaneously stops after the
surface is completely covered by a thin molybdena film. In
this case the island formation observed under the described
preparation conditions has to be governed by a second process.
Future experiments using the new SPEM technique might help
to separate the leading transport and formation mechanisms of
the Mo oxide species during spreading of supported MoO3
catalysts.
Acknowledgment. We thank the entire staff of the ELETTRA
synchrotron radiation facility, especially Diego Lonza and Gilio

MoO3 on Titania and Alumina
J. Phys. Chem. B, Vol. 101, No. 48, 1997 10011
(9) Xie, Y.; Gui, L.; Liu, Y.; Zhang, Y.; Zhao, B.; Yang, N.; Guo, Q.;
Duam, L.; Huang, H.; Cai, X.; Tang, Y. in Adsorption and Catalysis on
Oxide Surface; Che, M., Bond, G. C., Eds.; Elsevier: Amsterdam, 1985; p
139.
Sandrin for their excellent technical support. G. Mestl thanks
the Alexander-von-Humboldt Foundation for the Feodor-Lynen
Fellowship which made possible his research stay at Texas
A&M University. This work was financially supported by an
EC grant under Contract ERBCHGECT920013, the Deutsche
Forschungsgemeinschaft (SFB 338), and Sincrotrone Trieste
SCpA.
(10) Stampfl, S. R.; Chen, X.; Dumesic, J. A.; Niu, C.; Hill Jr., C. G.
J. Catal. 1987, 105, 445.
(11) Margraf, R.; Leyrer, J.; Kno¨zinger, H.; Taglauer, E. Surf. Sci. 1987,
189/190, 842.
(12) Leyrer, J.; Margraf, R.; Taglauer, E.; Kno¨zinger, H. Surf. Sci. 1988,
201, 603.
References and Notes
(13) Leyrer, J.; Zaki, M. I.; Kno¨zinger, H. J. Phys. Chem. 1986, 90,
4775.
(1) Kno¨zinger, H.; Taglauer, E. Catalysis; The Royal Society of
Chemistry: Cambridge, 1993; Vol. 10, p 1.
(2) Kno¨zinger, H.; Taglauer, E. In Handbook of Heterogeneous
Catalysis; Ertl, G., Kno¨zinger, H., Weitkamp, J., Eds.; VCH-Verlag:
Weinheim, 1997; Vol. 1, p 216.
(14) Kisfaludi, G.; Leyrer, J.; Kno¨zinger, H.; Prins, R. J. Catal. 1991,
130, 192.
(15) Leyrer, J.; Mey, D.; Kno¨zinger, H. J. Catal. 1990, 124, 349.
(3) Stevenson, S. A.; Dumesic, J. A.; Baker, R. T. K.; Ruckenstein, E.
Metal-Support Interactions in Catalysis, Sintering and Redispersion; Van
Nostrand Reinhold: New York, 1987.
(4) Chambliss, D. D.; Wilson, R. J.; Chiang, S. J. Vac. Sci. Technol.
A 1992, 10, 1993.
(5) Nielsen, L. P.; Besenbacher, F.; Stensgaard, I.; Laegsgaard, E.;
Engdahl, C.; Stolze, P.; Jacobsen, K. W.; Norskov, J. K. Phys. ReV. Lett.
1993, 71, 754.
(6) Chang, S. L.; Wen, J. M.; Thiel, P. A.; Gu¨nther, S.; Meyer, J. A.;
Behm, R. J. Phys. ReV. B 1996, 53, 13747.
(7) Liu, Y.; Xie, Y.; Li, C.; Zou, Z.; Tang, Y. J. Catal. (China) 1984,
5, 234.
(8) Xie, Y.; Gui, L.; Liu, Y.; Zhao, B.; Yang, N.; Zhang, Y.; Guo, Q.;
Duan, L.; Huang, H.; Cai, X.; Tang, Y. In Proceedings of the 8th
International Congress on Catalysis, Berlin, 1984; Dechema: Frankfurt,
1984; Vol. 5, p 147.
(16) Mestl, G.; Verbruggen, N. F. D.; Lange, F. C.; Tesche, B.;
Kno¨zinger, H. Langmuir 1996, 12, 1817.
(17) Casalis, L.; et al. ReV. Sci. Instrum. 1995, 66, 4870.
(18) Linsmeier, Ch.; Kno¨zinger, H.; Taglauer, E. Surf. Sci. 1992, 275,
101.
(19) Linsmeier, Ch. Ph.D. Thesis, Universita¨t Mu¨nchen, 1994; p 13.
(20) Barr, T. L. In Practical Surface Analysis; Briggs, D., Seah, M. P.,
Eds.; John Wiley & Sons: New York, 1983; Vol. 1, p 283.
(21) Moulder, J. F., Stickle, W. F., Sobol, P. E., Bomben, K. D.,
Chastain, J., Eds.; Handbook of X-ray Photoelectron Spectroscopy; Perkin-
Elmer Corporation Physical Electronics Division: Eden Prairie, MN, 1992;
p 213 ff.
(22) Ramsier, R. D.; Yates Jr., J. T. Surf. Sci. Rep. 1991, 12, 243-378
and references therein.