CO-Induced Diffusion of Ni Atoms to the Surface of Ni–Au Clusters on

ARTICLE pubs.acs.org/JPCC

CO-Induced Diffusion of Ni Atoms to the Surface of Ni
Au Clusters on TiO2(110) Samuel A. Tenney,† Wei He,† Christopher C. Roberts,† Jay S. Ratliff,† Syed Islamuddin Shah,‡ Ghazal S. Shafai,‡ Volodymyr Turkowski,‡ Talat S. Rahman,‡ and Donna A. Chen*,† † ‡

Department of Chemistry and Biochemistry, University of South Carolina, Columbia, South Carolina 29208, United States Department of Physics and NanoScience and Technology Center, University of Central Florida, Orlando, Florida 32816, United States

bS Supporting Information ABSTRACT: The growth, surface composition, and chemical activity of Ni
Au clusters on TiO2(110) have been studied by scanning tunneling microscopy (STM), low energy ion scattering (LEIS), and temperature-programmed desorption (TPD), as well as density functional theory (DFT) calculations and ab initio molecular dynamics simulations. STM images of similar coverages of pure Au and pure Ni on TiO2(110) illustrate that Au clusters are larger with lower cluster densities, indicating that Au is more mobile on the surface than Ni. Consequently, bimetallic Ni
Au clusters can be grown by nucleating Au at existing Ni clusters. A sequence of STM images acquired from the same region of the surface after various depositions of Au on Ni seed clusters demonstrates that new clusters of pure Au are not formed on the surface. Furthermore, the size of the existing clusters increases with each Au deposition due to the incorporation of incoming Au atoms. For bimetallic clusters of varying compositions with a total coverage of 0.25 ML, the addition of Ni has a minor effect in suppressing cluster sintering. LEIS studies indicate that the surface of the clusters are Au-rich (85
95% Au) for bulk Au fractions g50%. For annealed bimetallic clusters, the presence of Au at the cluster surface does not significantly inhibit the encapsulation of Ni by titania, while surface Au is not encapsulated. TPD investigations of CO desorption show that CO desorbs from pure Ni clusters in a molecular peak at ∼400 K and a recombinant peak at ∼790 K. Although CO does not adsorb onto titania or pure Au clusters at room temperature, significant CO desorption occurs from bimetallic clusters even for surfaces with only a small fraction of Ni at the surface; this result suggests that CO induces the diffusion of Ni to the surface of the clusters. DFT calculations for unsupported Ni1Au121 clusters confirm that in the presence of a CO molecule, the lowest energy structure involves CO bonding to a Ni atom at the surface. In contrast, in the absence of CO, the most stable cluster surface is pure Au with all of the Ni atoms in the interior of the cluster. Ab initio molecular dynamics simulations show that Ni will migrate to the cluster surface at 300 K in the presence of CO, but Ni migration to the surface does not occur even at higher temperatures in the absence of CO.

’ INTRODUCTION It has been well established that bimetallic systems can exhibit activity different than that of pure metals, and there are many examples in the catalysis literature illustrating the ability of a second metal to promote the desired catalytic activity and selectivity.1
6 Consequently, there is much interest in basic understanding of the chemical activity on bimetallic surfaces in order to develop catalysts with properties that can be tuned by changing compositions. In some cases, reactions are promoted via a bifunctional mechanism, in which the reaction requires the different activities provided by each metal.7
12 Furthermore, electronic effects associated with the formation of new metal
metal bonds may alter surface chemical properties, such as CO adsorption strength,6,13
16 hydrogenation activity,3,17
19 dehydrogenation activity,20,21 and reforming selectivity.3,22,23 Bimetallic surfaces may also provide mixed-metal sites with activity different from that of the pure metal sites, such as on the Sn
Pt alloy surfaces.24
26 In addition, interactions between the metal clusters and the oxide support may also be used to control surface chemistry on the clusters, with lattice oxygen participating in reactions on the r 2011 American Chemical Society

oxide-supported clusters. For example, atomic carbon on Ni clusters recombine with lattice oxygen from the titania support to produce gaseous CO,27,28 and gaseous products containing lattice oxygen are observed in reactions on metal clusters supported on ceria.29
34 Also for noble metals on ceria supports, ceria plays an important role in oxygen storage in the three-way catalysts for the conversion of CO, NOx, and hydrocarbons into CO2, water, and N2.35
37 In other cases, it has been reported that chemical activity occurs at metal cluster
oxide interfacial sites.29,38
42 In order to probe the nature of metal
metal and metal
support interactions, we have chosen to study a model system consisting of vapor-deposited Ni
Au bimetallic clusters supported on rutile TiO2(110). In this model system, the relationships between morphology, composition, and chemical activity can be explored on a fundamental level. The Au
titania system is one of chemical interest due to the unusual catalytic properties of Received: February 13, 2011 Revised: April 14, 2011 Published: May 16, 2011 11112

dx.doi.org/10.1021/jp2014258 | J. Phys. Chem. C 2011, 115, 11112–11123

The Journal of Physical Chemistry C Au nanoclusters supported on titania for low temperature CO oxidation43
45 and propylene epoxidation46,47 reactions. The nature of the enhanced activity of the Au nanoclusters is not fully understood but has been discussed in detail in a number of review articles,48
53 and there is evidence that the Au
titania interface plays an important role.43,54,55 For example, the activity of Au nanoclusters is higher on TiO2 compared to on other oxides such as SiO2.56 Furthermore, DFT calculations propose that in CO oxidation reactions, O2 bond scission occurs at titania sites while CO adsorption occurs on Au.57,58 Recent studies of “inverse” catalysts consisting of titania clusters on single-crystal Au surfaces report that these surfaces also exhibit enhanced activity for the water gas shift reaction, suggesting that the Au
titania interface is the active site.38 Ni
Au clusters were chosen for study because of the unique chemical properties previously observed in Ni
Au bimetallic systems. Ni
Au catalysts have been found to be superior materials for the steam reforming reaction, in which methane or other hydrocarbons react with water to produce H2 and CO.59,60 While Ni is a common steam reforming catalyst, graphite is also formed on Ni as a reaction byproduct, and the presence of this carbon poisons Ni activity.59,61,62 Studies of Ni
Au clusters supported on MgAl2O4 and SiO2 report that for n-butane reaction, pure Ni is deactivated quickly whereas the loss in activity for Ni
Au is negligible.59,60 The presence of Au on the Ni reforming catalysts is believed to increase resistance to carbon formation because the Au atoms block edge and kink sites,60 which are the most active sites for carbon formation on the Ni particles.62 Ni
Au surfaces have exhibited activity different from that of the pure metals in other reactions, such as low temperature CO oxidation,63
65 as well as in the adsorption of CO, CH4, and D2.66 The presence of Au on the Ni surface inhibits the ability of the surface to dissociate methane, and the desorption energies of CO and H2 are reduced by 25
30 kJ/mol for Au coverages up to 0.7 ML.66 For Au deposited on Ni(111), the effect of Au is to promote low temperature oxidation of CO to CO2 by blocking the formation of surface carbonate.67 In the study reported here, we have shown that Ni
Au bimetallic clusters on titania can be prepared by the deposition of Ni followed by Au. Although the surfaces of the clusters are Au-rich at high Au fractions, there is a significant amount of CO that adsorbs to the surfaces at room temperature, suggesting that Ni diffuses to the cluster surface to form strong Ni
CO bonds. DFT calculations demonstrate that there is a large thermodynamic driving force for Ni to bind to CO, and ab initio molecular dynamics simulations confirm that the diffusion of Ni to the surface is not kinetically limited at room temperature.

’ EXPERIMENTAL METHODS Experiments were carried out in two ultrahigh vacuum (UHV) chambers, which have been described in detail elsewhere.11,27,28,68
70 In both chambers, the base pressures were below 7  10
11 Torr. STM and LEIS experiments were conducted in the first chamber11,28,68,69 while temperatureprogrammed desorption (TPD) experiments were carried out in the second chamber.27,68,70 Rutile TiO2 (110) crystals (1 cm  1 cm  0.1 cm) were purchased from Princeton Scientific Corporation and were cleaned by Arþ ion sputtering followed by heating to 950
1000 K for 3 min.68 This treatment results in the formation of an n-type semiconductor, as oxygen is preferentially removed

ARTICLE

from the crystal. Surface cleanliness was established using a combination of STM, XPS, LEIS, LEED, and Auger electron spectroscopy. Au was deposited from either a pure pellet heated by passing current through a tungsten wire cage or from pellets in a molybdenum crucible heated by electron bombardment, as described previously.68,71 In the first chamber, deposition of Ni was achieved using a pure Ni rod in a commercial electron beam evaporator (Oxford Instruments), and in the second chamber, pure Ni wire was wrapped around a heated tungsten wire support.27 Metal fluxes were measured with a calibrated69 quartz crystal microbalance (Inficon). One monolayer (ML) is defined with respect to the packing densities of the Ni(111) or Au(111) surfaces, which are 1.61  1015 atoms/cm2 and 1.40  1015 atoms/cm2, respectively. Ni fluxes were ∼0.1 ML/min in both chambers, and the Au flux was also ∼0.1 ML/min unless otherwise specified. For the STM experiments in which the same region of the surface was imaged during Au deposition, Au was evaporated from a pellet in a Mo crucible using an Omicron electron beam evaporator mounted on the STM chamber.72 Au was deposited with the sample in the STM stage although the STM tip was moved 200 nm away from the evaporator during deposition so that the surface would not be shadowed by the tip. STM images were acquired with a 0.1 nA tunneling current, and the sample was biased at þ2.3 eV with respect to the tip. Tungsten STM tips were prepared by electrochemical etching in sodium hydroxide solution69 and ion sputtering in a 1 keV Arþ beam. LEIS spectra were collected with a 600 eV Heþ beam at a scattering angle of 130°. For the Ni
Au cluster annealing experiments, acquisition parameters were chosen so that Ni and Au signals did not change by more than 10% over the collection of seven successive spectra. The relative sensitivities of Ni and Au were determined by depositing metal films that were thick enough (20 ML Au, 10 ML Ni) to completely attenuate the signal from the substrate. The LEIS signals from the films were corrected for differences in surface area, which was obtained from STM images of the films, as described previously.28 The sensitivity to Au was found to be 2.43 times higher than the sensitivity to Ni. TPD experiments were carried out using a shielded quadrupole mass spectrometer (Hiden) with the crystal biased at
100 V to prevent damage induced by electrons from the mass spectrometer filament.68 In the CO adsorption experiments, the chamber pressure was increased by 3.0  10
10 Torr for 3 min as CO was leaked into the chamber via a variable leak valve, and the crystal was positioned ∼2 mm from the end of a stainless steel directed dosing tube. Assuming 1 order of magnitude increase in the local pressure at the surface of the crystal during directed dosing, the CO exposure is estimated to be ∼0.5 L. Cluster heights and densities were measured with an in-house computer program designed to remove step edges from the images in order to measure cluster heights;73 this program was based on previous work by Jak et al.74 Cluster heights were used as an indicator of size instead of the diameters because the latter are known to be overestimated due to tip convolution effects.69

’ COMPUTATIONAL METHODS Scalar relativistic density functional theory (DFT) calculations were carried out with the VASP code75 using the generalizedgradient approximation with the Perdew
Burke
Ernzerhof (PBE) functional76 to describe the exchange-correlation effects. 11113

dx.doi.org/10.1021/jp2014258 |J. Phys. Chem. C 2011, 115, 11112–11123

The Journal of Physical Chemistry C

ARTICLE

Figure 1. Scanning tunneling microscopy images of the following metals deposited at room temperature on TiO2(110): (a) 0.10 ML of Ni; and (b) 0.10 ML of Au. All images are 1000 Å  1000 Å.

The plane-wave pseudopotential method implemented in the VASP package with the projected-augmented-wave (PAW) pseudopotentials77,78 was used to describe the electron
ion interaction. To avoid finite size effects, a single cluster containing 122 atoms was placed into a 22  22  22 Å3 supercell. The kinetic energy cutoff for the plane-wave expansion was set to 350 eV, and the conjugate-gradient algorithm was used to relax the structure by requiring all force components acting on each ion to be less than 0.01 eV/Å. The Brillouin zone was sampled with one Γ-point using a nonshifted 1  1  1 Monkhorst
Pack k-point mesh. Similarly, ab initio molecular dynamics (MD) simulations were also carried out for 13-atom clusters using the VASP code and PAW-PBE potentials with energy cut offs of 700 eV for carbon and oxygen, 270 eV for Ni, and 230 for Au; the cell size was 20  20  20 Å3, allowing 13.14
16.40 Å of vacuum between clusters in the imposed periodic boundary conditions. Two cuboctahedral structures for Au12Ni (one with Ni at the center and the other with Ni at the surface) were optimized so that the forces on each atom were less than 2  10
3 eV/Å. The interaction of CO and the lower energy isomer, which was the one with Ni at the center, was then studied to examine the effect of CO on migration of Ni from the core to the surface. For all simulations, the time step was 3 fs, and the system was allowed to evolve over 366 steps (1 ps).

’ RESULTS Scanning Tunneling Microscopy Studies. The growth of bimetallic Ni
Au clusters was studied by STM, and Figure 1 shows a comparison of the growth of pure Ni and pure Au on TiO2 at room temperature for equivalent 0.1 ML coverages. The average height of the Ni clusters is smaller than for Au (4.0 ( 1.2 Å vs 7.7 ( 3.0 Å), and the cluster density for Ni clusters is 2.75 times that of Au (4.4  1012 clusters/cm2 vs 1.6  1012 clusters/cm2). Histograms of the cluster heights indicate that the Ni clusters are relatively small with a narrow distribution centered around 4 Å (Figure 2). In contrast, the distribution of heights for the Au clusters is much broader with peaks around 4 and 9 Å. Both the smaller cluster sizes and the greater cluster densities for Ni are consistent with a lower mobility of Ni atoms on the surface compared with Au. Furthermore, in both cases there is a tendency for the metal clusters to reside at the step edges, implying that the metal atoms are mobile enough on the surface to diffuse to the most favorable binding sites at the step edges; STM studies of a number of other metals on TiO2(110),

Figure 2. Histograms of cluster heights for (a) 0.10 ML of Ni and (b) 0.10 ML of Au deposited at room temperature on TiO2(110). The STM images corresponding to these histograms are shown in Figure 1a and 1b.

including Au,79,80 Ni,81 Cu,81,82 and Ag,79,83 have shown similar tendencies for the clusters to be located at the step edges when deposited at room temperature. This effect is more pronounced for Au compared to Ni clusters and again reflects the higher mobility of Au on the surface compared to Ni. Therefore, the growth of bimetallic clusters from sequential deposition of the metals should be carried out by depositing the less mobile Ni atoms first, followed by the more mobile Au. An STM experiment was carried out in which the same region of the surface was imaged, as Au was deposited on top of existing Ni clusters in order to demonstrate that bimetallic clusters are formed under these conditions. The surface initially has a coverage of 0.02 ML of Ni (Figure 3a), and five successive depositions of Au from 0.01 to 0.10 ML are shown in Figure 3b
f. Even at the highest Au coverage, very few new clusters are formed, and newly nucleated clusters account for only 7% of the total cluster density. Thus, deposition of Au results in the incorporation of Au atoms into the existing Ni clusters rather than the nucleation of new clusters of pure Au. The STM line profile in Figure 4 illustrates how both the height and diameter of a typical cluster on the surface (Figure 3a, marked with a triangle) grow with increasing Au deposition due to the incorporation of Au into the Ni seed cluster. Although the growth of existing clusters is the primary event observed during the deposition of Au, nucleation, coalescence, and disappearance of clusters are also observed, with the latter two involving only 8% and 2% of the total clusters on the surface, respectively. For example, the circles in Figure 3a and 3b indicate the nucleation of a new cluster, while the circles in Figure 3d and 3e show the coalescence of a cluster with increasing Au deposition. In Figure 3b and 3c, the disappearance of a cluster is marked with a square and is possibly due to interaction with the STM tip. Notably, for a higher initial coverage of Ni clusters of 0.10 ML followed by the deposition of 0.10 ML of Au, new clusters represent 50% that have very little Ni at the surface, and this behavior is explained by CO-induced diffusion of Ni to the cluster surfaces. One possible scenario is that the atoms within the clusters are mobile at room temperature, and at any instant in time, Ni atoms exist at the surface even for clusters with a high fraction of Au; in the presence of CO, a strong Ni
CO bond is formed that traps Ni at the surface. Another possible scenario is that the subsurface Ni atoms are electronically perturbed by the presence of the CO at the surface and diffuse to the top monolayer in order to create the strong Ni
CO bonds. Density Functional Theory Calculations. Our main interest is in examining the interactions between CO and the surface atoms of the Ni
Au clusters, and therefore the role of the

Figure 11. Optimized structures for the unsupported Ni1Au121 clusters with the Ni atom in the (a) second layer, (b) first layer, and (c) third layer. Relative energies calculated by density functional theory are given below each cluster. Au atoms are shown in yellow and Ni in blue.

TiO2(110) support is neglected because the average size of the clusters is several hundreds of atoms. Although the interface of the bimetallic cluster with the TiO2(110) surface may have some effect on the electronic and geometric structure of the clusters, these effects are assumed to be confined to atoms that are within a few atomic layers of the support, and the surface atoms are assumed to be unaffected. In our analysis, the DFT calculations were carried out on unsupported 122-atom clusters with varying Ni
Au compositions. In the first case the structure of Ni1Au121 clusters is determined by optimizing the fcc structure for a pure Au122 cluster terminated by (111) facets, incorporating a single Ni atom, and then reoptimizing the structure of the resulting bimetallic cluster. The lowest energy structure is one in which the Ni atom resides in the second layer from the surface (Figure 11a), whereas the cluster with the Ni atom at the surface (Figure 11b) is 0.74 eV higher in energy. This result is consistent 11118

dx.doi.org/10.1021/jp2014258 |J. Phys. Chem. C 2011, 115, 11112–11123

The Journal of Physical Chemistry C with the greater surface free energy of Ni compared to Au, causing Ni to prefer the interior sites in the cluster. Furthermore, incorporation of the Ni atom into the third layer (Figure 11c) results in a structure with an energy 0.24 eV higher than that of the cluster with Ni in the second layer. It has been suggested by DFT calculations that Ni can donate charge to Au in small (3
8 atom) Ni
Au clusters.101 Thus, the Ni atom may prefer to occupy the second layer from the surface because charge transfer can be maximized when the nearest Au atoms are at the surface, where the atoms can accommodate greater charge due to their lower coordination number. Ni atoms also prefer to occupy the subsurface sites for higher Ni concentrations in Ni2Au120 and Ni38Au84 clusters. Calculations for the Ni2Au120 clusters suggest that there is no strong tendency for Ni atoms to aggregate within the interior of the cluster. The structure in which the two Ni atoms are in the nearest neighbor positions in the same layer is 0.003 eV lower in energy than when the atoms are in the second nearest neighbor positions in the same layer, and 0.072 eV lower in energy than when the Ni atoms are in adjacent layers (Figure S1, Supporting Information). For a relaxed Ni38Au84 cluster, all of the Ni atoms are found in the interior of the cluster; note that the Ni38Au84 clusters correspond to the maximum number of Ni atoms in which the surface layer for the clusters can still consist only of Au atoms. For both the Ni2Au120 and Ni38Au84 clusters, the energy required to bring one Ni atom to the surface of the cluster is approximately 0.7 eV. In contrast, when a single CO molecule is adsorbed on the surface of the Ni1Au121 cluster, the lowest energy configuration is the one in which the Ni atom is at the cluster surface and binds to the CO molecule (Figure 12). When the Ni atom is in the second or third layer with CO bound to surface Au atoms, the energy of the system is ∼1.26 eV higher. Therefore, it is thermodynamically favorable for Ni to reside at the surface in the presence of CO in order to form the strong Ni
CO bond. Molecular Dynamics Simulations. Given that it is energetically favorable for Ni to exist at the surface of the bimetallic clusters in the presence of CO, it is also important to understand the kinetics of Ni diffusion to the surface of the bimetallic clusters. To address this issue, molecular dynamics simulations were carried out on unsupported Ni1Au12 clusters. Although the lowest energy configuration for Ni1Au12 is a flat structure, simulations were carried out on the next lowest energy structure shown in Figure 13a in order to model the behavior of threedimensional clusters. In the relaxed Ni1Au12 structure, the Ni atom is initially situated in the interior of the cluster. MD simulations at 300 K show rearrangement of atoms in the cluster upon substitution of the Ni atom, but no diffusion events are recorded (Figure S2a, Supporting Information). Simulations at higher temperatures demonstrate that the cluster dissociates at 1000 K before revealing any Ni migration to the surface (Figure S2b, Supporting Information). Of course, it is difficult to draw definite conclusions from these simulations carried out at temperatures relevant to our experiments because the simulation time of even 10 ps is far too short to capture rare events like atomic diffusion. In the presence of a single CO molecule, the MD simulations illustrate that the Ni1Au12 cluster undergoes significant restructuring. Figure 13b shows a CO molecule bound to a Au surface atom in the relaxed Ni1Au12 structure. Over the course of the MD simulation at 300 K, restructuring of the cluster brings Ni to the cluster surface (Figure 13c), and CO eventually desorbs from

ARTICLE

Figure 12. Optimized structures of the unsupported Ni1Au121 clusters with a CO molecule adsorbed on the surface and Ni in the (a) first layer, bound to CO, (b) second layer, (c) third layer. Relative energies calculated by density functional theory are given below each cluster. Au atoms are shown in yellow and Ni in blue.

Figure 13. Optimized structures for the unsupported Ni1Au12 clusters used in molecular dynamics (MD) simulations: (a) the initial structure without CO; (b) the initial structure with one CO molecule bound to a surface Au atom; (c) the final structure at 300 K after 366 steps (3 fs/ step) in the MD simulation; (d) the initial structure from (c) with CO bound to the Ni atom instead of a Au atom; (d) the final structure at 300 K after 366 steps (3 fs/step) in the MD simulation. Au atoms are shown in dark yellow, Ni atoms in blue, carbon atoms in light yellow, and oxygen atoms in red.

Au at longer times. Furthermore, if CO is then adsorbed on the Ni atom at the surface of the resulting structure (Figure 13d), the Ni atom remains at the surface after the MD simulation at 300 K (Figure 13e). The results of these simulations indicate that migration of Ni to the cluster surface is possible even at room temperature, with the strong Ni
CO bonding as the driving 11119

dx.doi.org/10.1021/jp2014258 |J. Phys. Chem. C 2011, 115, 11112–11123

The Journal of Physical Chemistry C

Figure 14. Density functional theory results for the change in the electron charge density in the Ni1Au12 system after CO adsorption onto a surface Au atom. The blue and red regions indicate the extra and the missing charge density, respectively. Au atoms are shown in yellow (larger balls), Ni atoms in blue, carbon atoms in light yellow (smaller balls), and oxygen atoms in pink.

force for the cluster restructuring. In the absence of CO, there is no such driving force for Ni migration to the cluster surface, and even at high temperatures (1000 K) Ni diffusion to the surface does not occur. Therefore, the presence of CO must induce diffusion of Ni to the cluster surface, and the Ni atoms are then trapped at the surface by the formation of strong Ni
CO bonds. An electron charge-transfer mechanism might contribute to CO-induced Ni atom diffusion. Because both Ni101 and CO102 donate charge when bound to Au, when CO approaches to the gold atom its donor charge will try “to push out” the charge donated by the Ni atom, which may lead to reduction of the Ni
Au bond strength and increase the probability for Ni atom migration. A Ni atom at the surface is more stable because of the strong Ni
CO bond, and the Ni atom donates charge to the CO molecule, contrary to the Au
CO case.103 DFT calculations have been carried out on the Au12Ni
CO clusters used in the MD simulations in order to investigate these possibilities. One significant change in the system after CO adsorption is substantial charge redistribution (Figure 14). Charge flow from the CO molecule causes the charge on the bimetallic cluster to increase by 0.17 electrons, and most of this charge (∼0.1 electrons) accumulates on the Au atom interacting with CO. A second more striking result is a dramatic increase in Ni
Au bond length from 2.66 Å to 3.34 Å after the CO molecule is adsorbed on the Au atom. This change is also accompanied by an increase in the C
O bond length from 1.135 Å to 1.146 Å. Thus, a possible mechanism for CO-induced diffusion of Ni may involve both charge transfer and an increase in Ni
Au bond length. When CO approaches a Au atom, the charge donated by the CO may weaken metal
metal bonding between the Au atom and a nearest neighbor Ni atom, increasing the probability for Ni diffusion. The slight preference observed in DFT calculations for Ni to occupy the first subsurface layer rather than deeper layers facilitates CO-induced diffusion of the Ni atoms to the surface. Furthermore, an increase of the Au
Ni bond length may lead to an instability of the Ni atoms that stimulates their migration to the surface.

’ DISCUSSION Au Surface Segregation. For the Ni
Au clusters on TiO2, the surface is Au-rich for Au fractions above 50%, and DFT calculations confirm that Au prefers to reside at the surface with Ni in the bulk for unsupported Ni
Au clusters of 122 atoms. These results are consistent with bulk thermodynamics, which suggests that Au should segregate to the surface in Ni
Au systems, given the lower surface free energy for Au85,86 compared to Ni.85,87 Furthermore, there are many experimental and theoretical reports in the literature that support Au surface

ARTICLE

segregation. For example, studies of Ni films grown on Au(001) show that Au migrates to the surface of the film as Ni is deposited.104 Similarly, Ni films on Au{111} exhibit extensive segregation of Au to the surface at 510 K, and all of the Ni diffuses into the bulk after heating to 630 K.105 Investigations of Ni films on Au(111) also report diffusion of Ni into the bulk upon heating to 673 K,106 and LEIS studies of Ni50Au50 clusters on carbon establish that Au segregates to the surface of the clusters.107 Monte Carlo simulations with clusters containing 583 Ni and 100 Au atoms show that Au atoms reside exclusively at the surface, preferring edge and kink sites.60 Other Monte Carlo simulations with cluster sizes up to 2123 atoms at a 4% Au composition demonstrate that Au atoms are always located at the cluster surface and are evenly spread across the surface rather than aggregating in two-dimensional islands; no Au was observed in the bulk of the clusters.108 Semiempirical calculations using the embedded atom method report that for Au clusters initially embedded in a Ni surface, the Au atoms segregate to the surface after heating to 700
800 K, but Au clusters on top of the Ni surface do not diffuse into Ni.109 However, there is some mixing of Au and Ni at the surface for the Ni
Au clusters on TiO2(110) even for the highest Au compositions studied (75%), and therefore the bimetallic clusters are not true core
shell structures with Au exclusively at the surface and Ni in the core. This behavior is not surprising because surface alloys and alloys in nanoclusters are known to form despite the immiscibility of Ni and Au in the bulk. For Au deposited on Ni(110)88 and Ni(111)110,111 surfaces, STM experiments show that Au replaces Ni in the surface, forming Ni islands. Total energy calculations from effective medium theory support the incorporation of Au into the first Ni layer.88 The exchange of a Au adatom with a Ni surface atom to create the surface alloy is an endothermic process that is thermodynamically driven by entropic effects.112 Moreover, EXAFS studies on Ni
Au clusters on SiO2 and MgAl2O4 also suggest that Ni
Au alloy clusters can be formed.60,113 Ni films grown on Au exhibit a splitting of peaks in the Au(5d) valence band that is characteristic of Ni
Au alloys,114,115 and the same is true for Au deposited on stepped 5(001)  (111) Ni surfaces.116 CO-Induced Changes in Ni
Au Systems. Exposure of the Au-rich Ni
Au clusters on TiO2(110) to CO results in migration of the Ni atoms from the bulk of the cluster to the surface. DFT calculations on a Ni1Au121 cluster indicate that it is energetically favorable for Ni to diffuse to the surface in order to form strong Ni
CO bonds. Moreover, MD simulations on Ni1Au12 show that at room temperature, the Ni atom can migrate from the interior of the cluster to the surface in the presence of CO; this migration does not occur in the absence of CO even for temperatures as high as 1000 K. CO-induced changes in surface composition have also been observed for alloys formed from Au deposition on Ni. As previously discussed, the deposition of Au on Ni(111) results in the alloying of Au atoms into the Ni surface,88 and the surface alloy is confined to the topmost layer.117 Both effective medium theory and DFT calculations demonstrate that the formation of a surface alloy for Au deposited on Ni is energetically favorable.88,118 However, exposure of this surface to 750 Torr of CO at room temperature causes the decomposition of the alloy due to phase segregation as Au is expelled from the surface layer, forming Au islands.64 It was reported that Ni atoms are removed from the surface via the desorption of Ni carbonyl, and the presence of Au facilitates Ni removal by forcing the CO into more compressed regions on 11120

dx.doi.org/10.1021/jp2014258 |J. Phys. Chem. C 2011, 115, 11112–11123

The Journal of Physical Chemistry C the Ni surface. Monte Carlo simulations of Au on Ni(111) also show that phase separation of the Ni
Au surface alloy at room temperature and CO pressures of 5
750 Torr.119 CO-induced compositional changes in bimetallic surfaces have not been reported at the low exposures (