A novel growth mode of Mo on Au(111) from a Mo(CO)6 precursor: An STM study

Article abstract of DOI:10.1021/jp0270405

The growth of a submonolayer of metallic Mo prepared by chemical vapor deposition (CVD) of Mo(CO)₆ on Au(111) was studied by scanning tunneling microscopy (STM). At low coverage, nanoscale Mo clusters grow at the elbow sites and in the fcc regions of the reconstructed Au surface. When the coverage increases, rather than decorating uniformly all elbows as found in physical vapor deposition (PVD), the Mo clusters aggregate but do not coalesce, forming ramified cluster islands. Within the islands, the clusters preferentially aggregate along the fcc troughs and the domain boundaries. At step edges, Mo clusters are found to grow at both upper and lower step edges. Differences between CVD and PVD are attributed to differences in the mobility of the nascent Mo species leading to growth. We hypothesize that the difference in mobility for the CVD process is due to the presence of CO from the precursor molecules. Oxidation of the Mo islands leads to their spreading, proving that the Mo clusters are either three-dimensional above or embedded in the surface of the gold substrate.

Full text of DOI:10.1021/jp0270405

1036
J. Phys. Chem. B 2003, 107, 1036-1043
A Novel Growth Mode of Mo on Au (111) from a Mo(CO)6 Precursor: An STM Study
Zhen Song,† Tanhong Cai,† Jose A. Rodriguez,† Jan Hrbek,*,† Ally S. Y. Chan,‡ and
Cynthia M. Friend‡
Chemistry Department, BrookhaVen National Laboratory, Upton, New York 11973, and Department of
Chemistry and DiVision of Engineering and Applied Sciences, HarVard UniVersity, 12 Oxford Street,
Cambridge, Massachusetts 02138
ReceiVed: September 20, 2002; In Final Form: December 4, 2002
The growth of a submonolayer of metallic Mo prepared by chemical vapor deposition (CVD) of Mo(CO)6 on
Au(111) was studied by scanning tunneling microscopy (STM). At low coverage, nanoscale Mo clusters
grow at the elbow sites and in the fcc regions of the reconstructed Au surface. When the coverage increases,
rather than decorating uniformly all elbows as found in physical vapor deposition (PVD), the Mo clusters
aggregate but do not coalesce, forming ramified cluster islands. Within the islands, the clusters preferentially
aggregate along the fcc troughs and the domain boundaries. At step edges, Mo clusters are found to grow at
both upper and lower step edges. Differences between CVD and PVD are attributed to differences in the
mobility of the nascent Mo species leading to growth. We hypothesize that the difference in mobility for the
CVD process is due to the presence of CO from the precursor molecules. Oxidation of the Mo islands leads
to their spreading, proving that the Mo clusters are either three-dimensional above or embedded in the surface
of the gold substrate.
Introduction
metal surfaces. The ratio of decomposition to desorption is
higher on metal surfaces that bind CO strongly, e.g., Ni,12
because this provides an additional thermodynamic driving force
for dissociation of the metal-CO bond. One limitation of metal
CVD using metal carbonyls is that CO released during the
deposition process may lead to incorporation of carbon and
oxygen impurities in the growing metal layer. Therefore,
optimization of the deposition conditions is critical for growth
of a clean metal layer.
Use of metal carbonyls in chemical vapor deposition (CVD)
has been attractive for both the synthesis of supported metal
catalysts and the fabrication of electronic devices. In catalyst
manufacturing, CVD using metal carbonyls has clear advantages
in preparation of zerovalent metal clusters of very small size
and high dispersion.1-3 Catalysts made this way show improved
performance such as higher selectivity or activity.4-9 Further-
more, metal carbonyls provide flexibility in terms of catalyst
composition since they are readily available for a broad range
of metals. Advantages of CVD of metals on semiconductors in
electronics applications are selectivity, conformal step coverage,
high throughput, and low cost.10 Many studies have focused on
finding the decomposition conditions of metal carbonyls for
obtaining naked metals and also on the structure and reactivity
of the deposited metal clusters or films.11,12 Recently, Argo et
al.13 showed that the metal framework structure of certain
multinuclear metal carbonyls, e.g., Ir4(CO)12, stays intact after
decarbonylation in inert atmosphere at elevated temperatures.
The use of this approach may allow the preparation of naked
metal clusters with the preserved original metal skeleton of the
carbonyls.
Molybdenum hexacarbonyl (Mo(CO)6) interacts with metal
surfaces at low temperatures forming physisorbed layers of the
molecular carbonyl.11,12,14 Upon heating, most condensed car-
bonyl molecules sublime without decomposition with an activa-
tion energy about 14 kcal/mol and only a small fraction of the
first carbonyl layer dissociates on the Au substrate.15 Since the
gas-phase dissociation energy for cleavage of the first CO-
Mo(CO)5 bond is about 40 kcal/mol,16 thermal activation is
needed for an efficient decomposition of the carbonyl on noble
We have recently prepared Mo particles on Au(111) by
Mo(CO)6 CVD and studied their reactivity using synchrotron-
based photoemission spectroscopy (PES).15 The selection of Mo
on Au(111) is based on the idea of using a patterned substrate
for preparation of a model catalyst with well-defined structural
and size distributions that will allow the study of the relationship
between nanoscale structures and reactivity. Mo-based catalysts
are widely used in the petroleum and chemical industries for
olefin metathesis, alkene hydrogenation, isomerization, and
hydrotreatment of oil-derived feedstock (hydrodesulfurization,
hydrodenitridation, hydrodeoxidation).17-20
The reconstructed Au(111)-(22 × x3) surface was chosen
as a support due to its chemical inertness and its dislocation
network that can be used as a template for growing self-
organized cluster arrays.21-31 In principle, the preparation of
well-ordered arrays of monodispersed, nanometer-scale clusters
can be accomplished by using patterned substrates formed
spontaneously in heteroepitaxial systems or at reconstructed
surfaces.32,33 Strain relief in such surfaces causes formation of
periodic dislocation networks that are often ordered into regular
arrays. The herringbone reconstruction of the Au(111) surface
is one of the best known examples of the dislocation networks.21
The reconstruction is driven by the tensile stress in the top Au
layer and partial dislocation stripes form to separate alternating
fcc and hcp stacking regions. The surface, uniaxially contracted
along {11h0} directions, forms a superstructure with stress
* Corresponding author. E-mail: hrbek@bnl.gov.
† Brookhaven National Laboratory.
‡ Harvard University.
10.1021/jp0270405 CCC: $25.00 © 2003 American Chemical Society
Published on Web 01/07/2003

Novel Growth Mode of Mo on Au (111)
J. Phys. Chem. B, Vol. 107, No. 4, 2003 1037
domains alternating in a zigzag pattern. The dislocation stripes
meet at the domain boundaries to form edge dislocations, the
elbows. The binding energies and energy barriers for surface
diffusion of deposited adatoms may, therefore, be periodically
modulated, and, thus, may modify nucleation and growth
processes.
Previous studies of metal deposition on reconstructed Au(111)
have demonstrated that it is an excellent template for self-
organized growth of metal nanostructures. Several different
nucleation and growth modes have been observed for metal
growth using physical vapor deposition (PVD) on the Au(111)
herringbone. Preferential nucleation of metal clusters at elbows
of the herringbone reconstruction have been observed for
transition metals that have a higher surface free energy than
Au, such as Ni,22 Fe,23 Co,24-26 Rh,27 and Mo.28 High nucleation
probability at elbows was originally explained by the trapping
of Ni adatoms at attractive potential wells present at the elbow
sites.22 Recently, site-selective place exchange of Ni adatoms
and surface Au atoms, followed by preferential nucleation at
the exchange sites, has been proposed to explain the nucleation
at the elbows. On the other hand, noble metals, e.g., Au29 and
Ag,30,31 grow from the Au step edges into the fcc region to form
fingerlike rows. In contrast, Al clusters were found to nucleate
only between dislocation lines in the fcc areas of the Au
reconstruction implying a strong repulsion of adatoms at
dislocations.34,35 Similarly, electrochemically deposited Ru
clusters also nucleate selectively in the fcc regions of the
herringbone and grow there to form nanowires confined to the
fcc domains.36
We are interested in using CVD methods to grow organized
nanostructures on reconstructed Au(111). Herein, we present
STM results for the CVD of Mo on Au(111) using Mo(CO)6
as a precursor. We find that Mo nanoclusters grow at elbows
and fcc areas, and that these clusters aggregate to form ramified
islands along fcc troughs and domain boundary directions of
the Au(111)-(22×x3) reconstruction. The data reveal a novel
growth mode of Mo on the reconstructed gold in which Mo
clusters aggregate without coalescence. This finding is a sharp
contrast to the type of growth observed for PVD of Mo on
Au(111) recently.28 The large mobility of the Mo clusters with
remnant or re-adsorbed CO ligands is the major reason for the
observed morphology.
Figure 1. STM images of 0.1% Mo coverage (% of surface area
covered by clusters and islands) on Au(111) obtained by exposure to
16 L of Mo(CO)6 at 500 K and annealed to 600 K. The images were
taken at room temperature. Image sizes: (a) (200 × 200) nm2; (b) (75
× 75) nm2.
600 K for 5 min. The Mo(CO)6 was kept in a glass tube at
room temperature and briefly pumped before each dose. This
design of the doser could induce part of the Mo(CO)6 to
decompose while dosing. The exposures of Mo(CO)6 were
measured by an uncorrected ion gauge; part of the ion gauge
reading could be attributed to CO, which could lead to large
apparent exposures to get small coverages. STM images of the
dosed surfaces were acquired at room temperature using a W
tip. Molybdenum surface coverages on terraces were estimated
from STM images and expressed as a fraction (%) of the surface
area covered by Mo. For reasons discussed later, this estimate
does not account for the increase in cluster heights with
increasing coverage. At Harvard, the same dosing procedure
was used, with the exception that the dosed surface was briefly
heated to 600 K. A Pt(10%)-Ir tip was used for STM data
acquisition.
Experiment
The experiments were performed in two separate UHV
chambers, both of which were equipped with a scanning
tunneling microsope (STM), sputter gun, and LEED/Auger for
sample characterization. In the chamber at Brookhaven, an
Omicron STM was used to obtain images and a preparation
chamber was configured for sample cleaning, annealing, and
preparation using a chemical vapor deposition doser. The
Au(111) surface was cleaned by cycles of Ne+ sputtering (600
eV, 2 µA) at room temperature followed by 900 K annealing.
The Au samples were deemed clean when STM images
exhibited extended domains of periodic herringbone reconstruc-
tion. At Harvard, a beetle-type STM was used, which has been
described elsewhere.37 The Au sample was cleaned by cycles
of Ar+ bombardment (1 keV, 3 µA) and annealing to 900 K.
The surface was determined to be clean by Auger electron
spectroscopy and the observation of the herringbone reconstruc-
tion using both LEED and STM.
Results and Discussion
Metallic Mo clusters are deposited by decomposition of
Mo(CO)6 on the Au(111) herringbone surface at 500 K (Figures
1-3). The coverages of Mo deposited from integrated exposures
of 16, 32, and 60 L of Mo(CO)6 at ∼1×10-7 mbar are 0.1, 1.4,
Molybdenum deposition in the Brookhaven chamber was
achieved by exposing the Au(111) surface to ∼1 × 10-7 mbar
of Mo(CO)6 at 500 K, followed by annealing in a vacuum at

1038 J. Phys. Chem. B, Vol. 107, No. 4, 2003
Song et al.
comparison, temperatures around 600 K are needed to desorb
CO from Mo(110) at low coverages.40
In initial stages of Mo deposition, Mo clusters, imaged as
bright dots, are observed both on terraces and at steps (Figure
1). Most clusters on the terraces are located at elbows of the
herringbone reconstruction; however, some clusters juxtapose
(i.e., aggregate without coalescence) within the fcc stacking
regions to form cluster islands. In the rest of the paper we will
use the “island” term to mean an island composed of individual
Mo clusters. Clusters and islands that grow at step edges are
located at both the upper and the lower steps as is evident from
the brightness contrast. As the Mo coverage increases, the
number of Mo clusters and the size of the aggregated cluster
islands increase (Figure 2). At the Mo coverage of 1.4%, nearly
all clusters are incorporated into islands. The clusters within
these islands do not coalesce; rather, they are ramified with the
centers of islands located mostly at elbows. Increasing the
coverage further to 5.5% leads to larger aggregates (Figure 3).
Even at these higher coverages, the STM images show clearly
that the clusters do not coalesce. Instead, the clusters remain
ramified with preferred orientations along the fcc troughs and
the domain boundary directions. Decoration of both the upper
and lower step edges with cluster islands persists at these higher
coverages.
It is worth mentioning that the deposition of Mo on Au did
not alter the superstructure of the Au surface. Irregularities of
the herringbone pattern (i.e., uniaxial domain boundaries) seen
in Figure 3b are also typical of a clean surface and we did not
observe any Mo-induced changes for coverages investigated in
this work.
Figure 3c shows the cluster size distribution ranging from 1
to 3.5 nm with a maximum at 1.8 nm. As the cluster diameter
vs height ratio is constant for a given bias voltage, the upper
limit of cluster sizes was estimated by measuring the apparent
diameters of the clusters from their STM images. The absolute
height of the clusters in this system cannot be quantitatively
determined because it is bias voltage dependent (Figure 4).
While the topographic height difference between the herringbone
stripes and hcp area is 0.01 nm and quite independent of bias
voltage in the range of -0.4 V to -2 V, the apparent height of
the Mo cluster increases 5-fold (from 0.06 to 0.3 nm) over the
same voltage range. This effect may be caused by a band gap
near the Fermi level of the Mo cluster due to its small size.
Figure 2. STM images of 1.4% Mo coverage on Au(111) obtained
by exposure to 32 L of Mo(CO)6. Image sizes: (a) (280 × 280) nm2;
(b) (60 × 60) nm2.
and 5.5%, respectively. These results are consistent with our
previous photoemission study which showed that Mo(CO)6
decomposed on Au(111) above 400 K.15 No C or O was detected
by photoemission or Auger electron spectroscopy for low
coverages (∼0.15 ML) of Mo deposited from Mo(CO)6 on the
Au(111) at 500 K.
The Mo(CO)6 decomposition on metal surfaces is a thermally
activated process: a Mo(CO)6 multilayer adsorbed at low
temperatures primarily desorbs with ∼2% of the first layer
decomposing via decarbonylation.15 It is unlikely that Mo(CO)6
decomposition is promoted by charge transfer between the
substrate and the adsorbed carbonyl molecules,16b,38 because
there is a low probability for electron transfer given the ∼2 eV
energy difference between the Fermi level of Au and the lowest
Mo(CO)6 antibonding state (4d*(1T1g)) (using 5.3 eV as the work
function of Au(111)).39
The rate of Mo(CO)6 dissociation is increased by heating the
gold surface to 500 K. At this temperature the energy imparted
to the molecule is sufficient for dissociation of the first CO-
Mo-(CO)5 bond which has a bond energy of ∼40 kcal/mol.16
Once one Mo-CO bond is broken, the carbonyl fragments can
bond to the surface and to each other to form clusters. The
remaining CO molecules can be stripped off thermally.12
However, at 500 K some CO could be adsorbed or re-adsorbed
on the Mo clusters and affect the growth of Mo.16b For
Over the range of Mo coverages studied, the sizes of clusters
within the islands do not increase noticeably with increasing
coverage. The narrow and constant size distribution suggests
that the growth of Mo clusters is self-limiting. A possible reason
for this behavior is the presence of CO, which could have size-
dependent adsorption behavior on Mo clusters and influence
their growth.
Although the size of the clusters is constant, the number of
clusters per island does increase with increasing coverage. Figure
3d shows the island size distribution in terms of the number of
clusters with an average size of 1.8 nm. For comparison, a
Poisson distribution of the island size is also shown in Figure
3d. It was calculated by assuming that the nucleation event has
an identical probability. The maximum of Poisson distribution
was chosen at 14 clusters per island, to be close to the
experimental island size distribution. The measured Mo island
size distribution is very broad with a long tail, ranging from 1
to approximately 90 clusters. Significant differences between
the measured and Poisson distributions of the Mo islands
therefore imply that larger islands have clearly higher probability
of capturing mobile Mo clusters.

Novel Growth Mode of Mo on Au (111)
J. Phys. Chem. B, Vol. 107, No. 4, 2003 1039
Figure 3. STM images of 5.5% Mo coverage on Au(111) obtained by exposure to 60 L of Mo(CO)6. Image sizes: (a) (485 × 485) nm2; (b) (54
× 54) nm2; (c) cluster size distribution; (d) islands size distribution in terms of cluster number in each island.
elbows and at step edges. Compact Mo clusters formed a well-
ordered array reflecting the pattern in the surface arising from
the reconstruction. Mo, the element with a higher surface free
energy than Au and higher heat of sublimation, follows the
prediction of preferential nucleation at elbows due to site
selective place exchange as proposed by Meyer et al.41
For the Mo/Au(111) system, the elbows are the nucleation
sites and their density on a reconstructed Au surface is about 6
× 1011cm-2. If the number of the islands is controlled by the
number of the available nucleation sites, then these two values
should be comparable. In the case of CVD growth of Mo on
Au(111), for the apparent coverages ranging from 0.1 to 5%,
the density of the cluster islands is approximately constant (∼5
× 1010 cm-2) and 1 order of magnitude smaller than the density
of elbows (see Figure 5). This result implies that the Mo clusters
formed in the CVD process do not interact strongly with the
elbow nucleation sites as in the PVD process.
The evolution of the Mo islands morphology with increasing
coverage suggests that the islands are formed by diffusion-
Figure 4. The apparent height of a cluster and the height of the
herringbone stripes vs the hcp trough as a function of the tip bias.
limited aggregation (DLA)42 of the clusters. In the original
computer simulation, randomly ramified islands are formed as
diffusing particles hit and stick irreversibly to a growing island.
Besenbacher et al.28 studied the growth of Mo on Au(111)
by PVD of Mo atoms. They found that Mo nucleated at all
Experimentally, a growth of dendritic islands was observed in
bimetallic systems,43 and anisotropy of either substrate or the

1040 J. Phys. Chem. B, Vol. 107, No. 4, 2003
Song et al.
Figure 5. Area density of Mo clusters and islands as a function of
coverage expressed as % surface area covered.
aggregating particles is crucial for dendritic growth.44 Metal
islands can be characterized by fractal dimension df determined
from the A1/2 vs P1/d plot, where A is the area and P is the
f
perimeter of the island. Both randomly ramified and dendritic
islands have a fractal dimension df close to 1.7 and its value
may provide information about the growth mechanism; e.g., if
there is an energy barrier for the aggregation, the clusters will
not stick immediately. This will lead to a reaction-limited cluster
aggregation and formation of more compact, polydispersed
islands with df ) 2.1 as shown in studies of phase transition
kinetics of colloids.45,46 The ramified Mo islands studied in this
work have df ) 1.7 and this value suggests that the underlying
growth mechanism can be described by DLA and the Mo
clusters are the mobile aggregating particles.
It is interesting that the ramified islands, formed by clusters
aggregating in the CVD process, have island morphology similar
to those observed previously in mass-selected cluster deposition
experiments. In the latter case, the motion of clusters on surfaces
and the interaction among clusters are used to explain the island
morphology. Yoon et al.47 explained the dependence of island
morphology on Sb cluster size by invoking a critical cluster
size, R0, such that island shapes were compact for R < R0 and
ramified for R > R0. They assumed that there was a competition
between cluster aggregation and coalescence dictated by the time
interval between successive arrivals of clusters. Perez et al.48
used the Arrhenius law to fit the temperature-dependent diffu-
sion coefficient of Sb2300 and Au250 clusters by applying the
predictions of the deposition-diffusion-aggregation model to
the observed island morphologies, and found large preexpo-
nential factors. They pointed out the diffusion mechanisms of
these clusters must involve rotational movement of the cluster.
This could explain why two clusters become immobile when
they become attached, which is the mechanism for the growth
of a ramified island. Jensen49 and Binns50 in their recent review
articles analyzed the growth of nanostructures by direct cluster
deposition and several conclusions are applicable for CVD
growth discussed in this work.
Figure 6. Two consecutive STM images with highlighted changes in
surface morphology taken 7 min apart: full circle, changed; dotted
circle, disappeared; dashed circle, new feature. The sample was not
annealed after being exposed to Mo(CO)6 at 500 K. Image sizes: (400
× 400) nm2
immobile. A weaker interaction leads to higher mobility of
clusters that is reflected in a low density of islands. The island
morphology, on the other hand, reflects cluster-cluster interac-
tion since a cluster that arrives at an island can either coalesce
or remain separate to form a ramified island. The coalescence
of clusters is driven by thermodynamics as the system is trying
to minimize the surface free energy. The ramification is a kinetic
effect because the clusters’ arrival rate is faster than their
incorporation, thus preventing the free energy minimization.
Accordingly, the ramified cluster island structure in the CVD
growth of Mo on Au(111) implies that the Mo clusters form
and move around before they anchor on the surface.
The mobility of Mo clusters is demonstrated in two sequential
STM images (Figure 6). The sample was prepared without
annealing after dosing Mo(CO)6 on Au(111) at 500 K. Ap-
proximately 15% of the islands on terraces either change in size
or move to another location on a time scale of minutes at room
temperature. We did not observe analogous cluster mobility if
the sample was annealed to 600 K after deposition, indicating
that annealing reduces the mobility of the clusters. Therefore,
the observed mobility can be related to the presence of the
residual CO on the Mo clusters.16b As there is a large CO
pressure during the CVD growth, the Mo clusters are probably
covered by CO molecules. The adsorbed CO may reduce both
The ramified island morphology observed in our study clearly
indicates that the Mo clusters formed from Mo(CO)6 are mobile
on the surface during nucleation and growth. Previous work
clearly indicates that cluster mobility determines the types of
islands formed. In turn, cluster diffusion reflects the strength
of cluster-substrate interaction. If the cluster-substrate interac-
tion is strong when compared to kT, the clusters are relatively

Novel Growth Mode of Mo on Au (111)
J. Phys. Chem. B, Vol. 107, No. 4, 2003 1041
Figure 7. Holes and kinks (indicated by arrows) formed at the step
edges of Au(111) after depositing Mo at 500 K from Mo(CO)6 at lower
dosing pressure. Image size: (200 × 200) nm2.
Mo-Mo and Mo-Au interactions16b with the net result of Mo
clusters weakly bonded to the Au surface. Indeed, there is pre-
cedence for enhanced mobility of Pt induced by adsorbed CO.51
The limited decoration of elbows also indicates the weakening
of the Mo-Au interaction. In the PVD process, Mo clusters
are strongly bonded and energy dissipation into gold substrate
is facile. With clusters poorly coupled to the surface in the CVD
case, the attachment of additional Mo to the clusters may lead
to kinetic energy gain from the Mo-Mo bond formation. We
are currently investigating the underlying reasons for the
mobility of the clusters formed during deposition at 500 K.52,53
Another apparent difference in the Mo(CO)6 CVD of Mo
from the PVD is that Mo clusters grow at both upper and lower
step edges as seen in Figures 2a and 3a. In PVD of metal atoms
on metal substrates, clusters appear to grow only at the lower
step edges.54 There is an asymmetry in the surface potential
energy of a metal adatom or cluster on a substrate near step
edges: a trapping well at the lower terrace side and a repulsive
barrier at the upper one. Therefore, metal atoms or clusters are
usually trapped at the lower terrace side of step edges and grow
from there in the so-called step flow mode. The growth of Mo
clusters both at the top and bottom of step edges indicates that
there is a potential well associated with the step that leads to
the trapping of growing clusters. This finding may be the result
of the presence of the dipole field at step edges that causes
preferential adsorption of Mo-containing species at the upper
step edges.55
We have direct evidence (Figure 7) for removal of Au at steps
during growth of Mo clusters deposited from Mo(CO)6. The
removal of Au at the step edge is manifested by the development
of holes and kinks at steps of the Au surface during deposition
of Mo when using lower dosing pressure of Mo(CO)6. Under
these conditions Mo clusters grow predominantly at Au steps.
Holes and kinks at Au step edges are not present on clean
Au(111) or on CO-dosed Au(111) using the same dosing
parameters for Mo(CO)6 (data not shown). These data suggest
that while CO may facilitate the transport of atoms and/or
clusters on the Au surface, CO alone cannot be responsible for
the removal of Au atoms from the steps. One plausible
explanation is that the Au atoms removed are incorporated into
the Mo clusters. Another possibility is the step pinning by the
Mo clusters and consequential alternation of the step flow.
Alternatively, the step atom removal can be linked to the
vacancy exchange between the surface and the bulk. The
Figure 8. STM images of a MoO3-covered Au(111) sample acquired
after NO2 oxidation of the sample shown in Figure 3. Image sizes: (a)
(400 × 400) nm2; (b) (200 × 200) nm2.
surfaces are known to be good sink of bulk vacancies56 and the
step movement can be affected by the presence of the Mo
clusters at the step edges, thus leading to roughening of the
steps.
Considering the large differences of surface free energy of
Au and Mo57 one may assume that when all CO ligands detach
from a Mo center in Mo(CO)6, a place exchange of the substrate
gold atoms by Mo atoms may occur for decreasing the surface
free energy of the system. The molybdenum-gold place
exchange is predicted by first-principle density functional
calculations.51,53,58 A consequence of this place exchange
mechanism is that Mo clusters would be embedded in the Au
substrate.53 As Au and Mo are bulk immiscible, Au atoms
ejected from the substrate or detached from the steps are likely
to cap the Mo cluster. The absence of reactivity toward CO,
S2, thiophene, O2, and C2H415,59 for clusters formed by Mo(CO)6
decomposition on Au(111) is consistent with the capping of
these clusters by Au atoms. The most conclusive evidence in
support of this is found in oxidation reactions.59 Molecular
oxygen shows little reactivity, but reaction with NO2 leads to
formation of molybdenum oxides with a single Mo 3d doublet
and O 1s core levels characteristic of MoO3. As gold does not
react with O2, Mo clusters capped with Au should also be inert
to O2, in contrast to the behavior with atomic oxygen produced
by the decomposition of NO2.

1042 J. Phys. Chem. B, Vol. 107, No. 4, 2003
Song et al.
Oxidation of Mo clusters and islands with NO2 leads to the
formation of 2D MoO3 structures, details of which will be
described in a separate paper.60 Figure 8 shows STM images
of MoO3 structures obtained after NO2 oxidation at 500 K of
the 5.5% Mo-covered surface in Figure 3. A substantial increase
in the apparent adlayer coverage of the MoO3-covered surface
is observed, as compared to the Mo-covered surface. Further-
more, the PES intensity of the Mo 3d core levels of the MoO3-
covered surface was about 2.5 times higher than that of the
metallic Mo-covered surface, although both spectra were
measured from the same Mo/Au sample before and after
oxidation.59 The observed intensity increase was ascribed to a
morphology change induced by the oxidation. MoO3 is a layered
molecular crystal with layers held weakly by van der Waals
interactions. We propose that the MoO3 forms 2D islands with
the (010) facet parallel to the Au(111) surface. If the metallic
Mo clusters were 2D crystallites, the conversion from Mo(100)
or Mo(110) clusters to the MoO3(010) would result in an
increase of the adlayer coverage by a factor of 1.7 or 2.4,
respectively. From STM images, however, the oxidation brings
an increase of the coverage by a factor of 6. We therefore
conclude that the metallic Mo clusters deposited on Au(111)
from Mo(CO)6 are 3D clusters of about 2-3 layers of Mo thick
either above the surface or in part embedded in the gold
substrate.
References and Notes
(1) Psaro, R.; Recchia, S. Catal. Today 1998, 41, 139.
(2) Alexeev, O.; Gates, B. C. Top. Catal. 2000, 10, 273.
(3) Huh, S. H.; Oh, S. J.; Kim, Y. N.; Lee, G. H. ReV. Sci. Instrum.
1999, 70, 4366.
(4) Commereuc, D.; Chauvin, Y.; Hugues, F.; Basset, J. M.; Olivier,
D. Chem. Commun. 1980, 154.
(5) Hugues, F.; Besson, V.; Basset, J. M. Chem. Commun. 1980, 719.
(6) Hugues, F.; Dalmon, J. A.; Bussiere, P.; Smith, A. K.; Basset, J.
M.; Olivier, D. J. Phys. Chem. 1982, 86, 5136.
(7) Bielawa, H.; Hinrichsen, O.; Birkner, A.; Muhler, M. Angew. Chem.,
Int. Ed. 2001, 40, 1061.
(8) Close, M. R.; Petersen, J. L.; Kugler, E. L. Inorg. Chem. 1999, 38,
1535.
(9) de Bont, P. W.; Vissenberg, M. J.; de Beer, V. H. J.; van Veen, J.
A. R.; van Santen, R. A.; van der Kraan, A. M. Appl. Catal. A 2000, 202,
99.
(10) Green, M. L.; Levy, R. A. J. Metals 1985, 63.
(11) Cho, C. C.; Bernasek, S. L. J. Appl. Phys. 1989, 65, 3035.
(12) Xu, M.; Zaera, F. J. Vac. Sci. Technol. A 1996, 14, 415.
(13) Argo, A. M.; Odzak, J. F.; Lai, F. S.; Gates, B. C. Nature 2002,
415, 623.
(14) Malik, I. J.; Hrbek, J. J. Vac. Sci. Technol. A 1991, 9, 1737.
(15) Rodriguez, J. A.; Dvorak, J.; Jirsak, T.; Hrbek, J. Surf. Sci. 2001,
490, 315.
(16) (a) Lewis, K. E.; Golden, D. M.; Smith, G. P. J. Am. Chem. Soc.
1984, 106, 3905. (b) Liu, P.; Rodriguez, J. A.; Muckerman, J. T.; Hrbek,
J. Surf. Sci., in press.
(17) Tøpsoe, H.; Clausen, B. S.; Massoth, F. E. Hydrotreating Catalysis;
Springer-Verlag: New York, 1996.
(18) Startsev, A. N. Catal. ReV. Sci. Eng. 1995, 37, 353.
(19) Chianelli, R. R.; Daage, M.; Ledoux, M. AdV. Catal. 1994, 40,
177.
(20) Dellmon, B. Bull. Soc. Chim. Belg. 1995, 104, 173.
(21) Barth, J. V.; Brune, H.; Ertl, G.; Behm, R. J. Phys. ReV. B 1990,
42, 9307.
Conclusion
By monitoring the morphology change at different Mo
coverages with STM, the growth of Mo on Au(111) using
Mo(CO)6 as a precursor is summarized in the following scheme:
(22) Chambliss, D. D.; Wilson, R. J.; Chiang, S. Phys. ReV. Lett. 1991,
66, 1721.
(23) Voigtla¨nder, B.; Meyer, G.; Amer, N. M. Surf. Sci. 1991, 255, L529.
(24) Voigtla¨nder, B.; Meyer, G.; Amer, N. M. Phys. ReV. B 1991, 44,
10354.
(25) Padovani, S.; Chado, I.; Scheuer, F.; Bucher, J. P. Phys. ReV. B
1999, 59, 11887.
yMo(CO)6 f yMo(CO)x f Moy(CO)x f
z{Moy(CO)x} f z{Moy}
(26) Padovani, S.; Chado, I.; Scheuer, F.; Bucher, J. P. Phys. ReV. B
2000, 61, 72.
(27) Chado, I.; Scheurer, F.; Bucher, J. P. Phys. ReV. B 2001, 64, 094410.
(28) Helveg, S.; Lauritsen, J. V.; Lægsgaard, E.; Stensgaard, I.; Nørskov,
J. K.; Clausen, B. S.; Topsøe, H.; Besenbacher, F. Phys. ReV. Lett. 2000,
84, 951.
(29) Wang, Z.; Moskovits, M. J. Appl. Phys. 1992, 71, 5401.
(30) Dovek, M. M.; Lang, C. A.; Nogami, J.; Quate, C. F. Phys. ReV.
B 1989, 40, 11973.
(31) Chambliss, D. D.; Wilson, R. J. J. Vac. Sci. Technol. B 1991, 9,
928.
(32) Brune, H.; Giovannini, M.; Bromann, K.; Kern, K. Nature 1998,
394, 451.
(33) Hrbek, J.; Hwang, R. Q. Curr. Opin. Solid State Mater. Sci. 2001,
Adsorption of carbonyl on a gold surface at elevated
temperatures is followed by decarbonylation. Deposited carbonyl
fragments nucleate to form Mo nanoclusters {Moy}. The clusters
formed in the process are mobile on the surface, leading to
formation of ramified islands (z{Moy}).
There are important differences with respect to PVD of Mo
atoms on Au(111). In the CVD process, Mo(CO)6 molecules
decompose on a Au(111) surface at elevated temperature and
form Mo 3D nanoclusters. On terraces, the clusters start to
decorate elbow sites. However, with the increase of Mo
coverage, more clusters are formed and aggregate to develop
ramified islands rather than being trapped uniformly at other
elbows. The Mo clusters gain kinetic energy during their
formation, decreasing the probability of their trapping at the
elbow positions. Ramified islands are formed when clusters stop
their motions once they meet others. At step edges, the ramified
islands grow at both upper and lower terrace sides, in contrast
to the PVD growth at the lower step edge. The former can be
attributed to the finite residence of CO on the upper step edges,
which modifies the surface potential of Mo on Au step edges.
5, 67.
(34) Fischer, B.; Brune, H.; Barth, J. V.; Fricke, A.; Kern, K. Phys.
ReV. Lett. 1999, 82, 1732.
(35) Brune, H. Surf. Sci. Rep. 1998, 31, 121.
(36) Strbac, S.; Magnussen, O. M.; Behm, R. J. Phys. ReV. Lett. 1999,
83, 3246.
(37) Clark, P. G., Jr.; Friend, C. M. J. Chem. Phys. 1999, 111, 6991.
(38) Ying, Z. C.; Ho, W. J. Chem. Phys. 1991, 94, 5701.
(39) Skriver, H. L.; Rosengaard, N. M. Phys. ReV. B 1992, 46, 7157.
(40) Chen, J. G.; Colaianni M. L.; Weinberg W. H.; Yates, J. T., Jr.
Chem. Phys. Lett. 1991, 177, 113.
(41) Meyer, J. A.; Baikie, I. D.; Kopatzki, E.; Behm, R. J. Surf. Sci.
1996, 365, L647.
Acknowledgment. A part of this research carried out at
Brookhaven National Laboratory was supported under Contract
DE -AC02-98CH10086 with the U.S. Department of Energy,
Division of Chemical Sciences. The research at Harvard was
supported by the National Science Foundation through the
Nanoscale Science and Engineering Center at Harvard, No.
PHY-011-7795. A.S.Y.C. also gratefully acknowledges a Ma-
terials Research Science and Engineering Center Postdoctoral
Fellowship funded by NSF Grant DMR-98-09363.
(42) Witten, T. A., Jr.; Sanders, L. M. Phys. ReV. Lett. 1981, 47, 1400.
(43) Hwang, R. Q.; Schro¨der, J.; Gu¨nther, C.; Behm, R. J. Phys. ReV.
Lett. 1991, 67, 3279.
(44) Brune, H.; Roder, H.; Bromann, K.; Kern, K.; Jacobsen, J.; Stolze,
P.; Jacobsen, K.; Norskov, J. Surf. Sci. 1996, 349, L115.
(45) Weitz, D. A.; Huang, J. S.; Lin, M. Y.; Sung. J. Phys. ReV. Lett.
1985, 54, 1416.
(46) Anderson, V. J.; Lekkerkerker, H. N. W. Nature 2002, 416, 811.
(47) Yoon, B.; Akulin, V. M.; Cahuzac, Ph.; Carlier, F.; de Frutos, M.;
Masson, M.; Mory, C.; Colliex, C.; Bre´chignac, C. Surf. Sci. 1999, 443,
76.

Novel Growth Mode of Mo on Au (111)
J. Phys. Chem. B, Vol. 107, No. 4, 2003 1043
(48) Perez, A.; Melinon, P.; Dupuis, V.; Jensen, P.; Prevel, B.; Tuaillon,
J.; Bardotti, L.; Martet, C.; Treilleux, M.; Broyer, M.; Pellarin, M.; Vaille,
J. L.; Palpant, B.; Lerme, J. J. Phys. D: Appl. Phys. 1997, 30, 709.
(49) Jensen, P. ReV. Mod. Phys. 1999, 71, 1695.
(55) Pascual, J. I.; Jackiw, J. J.; Kelly, K. F.; Conard, H.; Rust, H.-P.;
Weiss, P. S. Phys. ReV. B 2000, 62, 12632.
(56) McCarty, K. F.; Nobel, J. A.; Bartelt, N. C. Nature 2001, 412,
622.
(50) Binns, C. Surf. Sci. Rep. 2001, 44, 1.
(57) Mezey L. Z.; Giber, J. Jpn. J. Appl. Phys. 1982, 21, 1569.
(51) Kalff, M.; Comsa, G.; Michely, T. Phys. ReV. Lett. 1998, 81, 1255.
(52) Chan, Ally, S. Y.; Friend, C. M. To be published.
(53) Liu, P.; Rodriguez, J. A; Muckerman, J. T.; Hrbek, J. Phys. ReV.
B, submitted.
(54) Gu¨nther, C.; Gu¨nther, S.; Kopatzki, E.; Hwang, R. Q.; Schro¨der,
J.; Vrijmoeth, J.; Behm, R. J. Ber. Bunsen-Ges. Phys. Chem. 1993, 97,
522.
(58) Ruban, A. V.; Skriver, H. L.; Nørskov, J. K. Phys. ReV. B 1999,
59, 15990.
(59) Chang, Z.; Song, Z.; Liu, G.; Rodriguez, J. A.; Hrbek, J. Surf. Sci.
2002, 512, L353.
(60) Song, Z., et al. To be published.