STM and LEED observations of a c(2 × 2) Ge overlayer on Pt(1 0 0)

We have investigated surface structures formed by deposition of 0.2 and 0.5-ML Ge on Pt(1 0 0) by using scanning tunneling microscopy (STM) and low electron energy diffraction (LEED). In addition, their temperature dependence and reactivity to CO have been studied. We observed the formation of disordered domains for Ge adatom coverages below 0.25-ML and complete c(2 × 2) structures at 0.25 to 0.5-ML Ge after annealing at 600-1200 K. Deposition of 0.2-ML Ge on a clean, hexagonally reconstructed (5 × 20)-Pt(1 0 0) substrate at 400 K lifts the reconstruction and ejects excess Pt atoms from the first layer into the adlayer. After annealing this surface to 600 K, the deposited Ge formed Ge adatoms on flat terraces and on round Pt adislands with incomplete c(2 × 2) structures, in addition to the presence of clean (1 × 1)-Pt(1 0 0) domains that were several nanometers across. Some domains of the unreconstructed (5 × 20)-Pt(1 0 0) surface still remained. After the deposition of 0.5-ML Ge and annealing at 600 K, disordered Ge domains disappeared and a c(2 × 2) Ge overlayer was produced all over the surface. Square terraces with square domains of the clean (1 × 1)-Pt(1 0 0) surface extended for nanometers. Annealing this surface to 900 K produced disordered Ge domains, and this was associated with an increase in Ge vacancies. When surfaces with 0.2-ML Ge were heated to 900 or 1200 K, or when a surface with 0.5-ML Ge was heated to 1200 K, larger domains of (5 × 20)-Pt(1 0 0) were formed with the agglomeration of disordered Ge adatoms. Pt clusters were observed in the Ge domains, and we consider these to be composed of those excess Pt atoms formed by lifting the reconstruction of the (5 × 20)-Pt(1 0 0) surface upon Ge agglomeration during cooling. A paper published elsewhere [T. Matsumoto, C. Ho, M. Batzill, B.E. Koel, Physical Review B, submitted for publication.] describes Na⁺-ion scattering spectroscopy (Na⁺-ISS) and X-ray photoelectron diffraction (XPD) experiments that distinguish between Ge present in an overlayer from incorporation into the top Pt layer to form a surface alloy for the surface structures reported here. Furthermore, these investigations revealed that disordered Ge adatoms observed herein might be associated with incomplete c(2 × 2) structures. Therefore, our observations of the formation of complete and incomplete domains of c(2 × 2) Ge adatoms indicate that interactions between Ge adatoms are repulsive at nearest neighbor distances and attractive at second-nearest neighbor distances. Regarding the reactivity of these surfaces, CO does not chemisorb on a Pt(1 0 0) surface with a c(2 × 2)-Ge overlayer and no measurable CO uptake was observed under UHV conditions at 220 K.     

Structural evolution of sulfur overlayers on Pd(1 1 1)

Low energy ion scattering spectroscopy (LEISS) has been used to characterize the evolution of ordered structures of S on the Pd(1 1 1) surface during annealing. During exposure of the Pd(1 1 1) surface to 0.7 L H₂S at 300 K—conditions that produce the S(√3 × √3)R30 overlayer—the intensity of the Pd LEIS signal decreases and a feature assigned to adsorbed S appears as the adsorbed layer forms. When the surface is held at 300 K after exposure to H₂S is stopped, the LEIS Pd intensity partially recovers and the S signal weakens, presumably as surface S atoms assume their equilibrium positions in the S(√3 × √3)R30 overlayer. Subsequent annealing of the S(√3 × √3)R30 structure at 700 K causes it to convert into a S(√7 × √7)R19 overlayer, whose LEIS spectrum is identical to that of clean Pd(1 1 1). The absence of LEIS evidence for S atoms at the exposed surface of the S(√7 × √7)R19 overlayer is at odds with published models of a mixed Pd-S top layer. Despite the similarity of the LEIS spectra of Pd(1 1 1) and Pd(1 1 1)-S(√7 × √7)R19, their activities for dissociative hydrogen adsorption are very different—the former readily adsorbs hydrogen at 100 K, while the latter does not—suggesting that S exerts its influence on surface chemistry from subsurface locations.     

Substitutional adsorption geometry for Pb(1 1 1)-(√3 × √3)R30°-K

Intermixed structures for alkalis (larger than Li) on close-packed substrates have previously been observed only on Al(1 1 1). This study shows that K forms an ordered intermixed structure on Pb(1 1 1). The structures of clean Pb(1 1 1) and Pb(1 1 1)-(√3 × √3)R30°-K were studied using dynamical low-energy electron diffraction (LEED). The clean Pb(1 1 1) surface at 47 K was found to be a relaxed version of the bulk structure, in agreement with an earlier study of the same surface [Y.S. Li, F. Jona, P.M. Marcus, Phys. Rev. B 43 (1991) 6337]. At room temperature, adsorption of K on this surface results in a (√3 × √3)R30° structure, which was shown using dynamical LEED to consist of K atoms substituted in surface vacancies. The K-Pb bond length was found to be 3.62 ± 0.3 Å, with no significant change to the Pb interlayer spacings.     

Growth of (√3 × √3)-Ag and (1 1 1) oriented Ag islands on Ge/Si(1 1 1) surfaces

We have studied the growth of Ag on Ge/Si(1 1 1) substrates. The Ge/Si(1 1 1) substrates were prepared by depositing one monolayer (ML) of Ge on Si(1 1 1)-(7 × 7) surfaces. Following Ge deposition the reflection high energy electron diffraction (RHEED) pattern changed to a (1 × 1) pattern. Ge as well as Ag deposition was carried out at 550 °C. Ag deposition on Ge/Si(1 1 1) substrates up to 10 ML has shown a prominent (√3 × √3)-R30° RHEED pattern along with a streak structure from Ag(1 1 1) surface. Scanning electron microscopy (SEM) shows the formation of Ag islands along with a large fraction of open area, which presumably has the Ag-induced (√3 × √3)-R30° structure on the Ge/Si(1 1 1) surface. X-ray diffraction (XRD) experiments show the presence of only (1 1 1) peak of Ag indicating epitaxial growth of Ag on Ge/Si(1 1 1) surfaces. The possibility of growing a strain-tuned (tensile to compressive) Ag(1 1 1) layer on Ge/Si(1 1 1) substrates is discussed.     

Structural investigation of the c(√2 × 5√2)R45° surface alloy formed by Ti deposition on Cu(0 0 1)

The alloying process of Ti deposited on Cu(0 0 1) was studied by means of XPS, LEIS, XPD and LEED intensity analysis. With the sample held at 570 K, a linear decrease of the Cu LEIS signal as a function of the amount of Ti deposited is observed in the early stages of deposition until a constant value is reached. At the onset of the plateau a c(√2 × 5√2)R45° LEED pattern starts to be visible. XPD and LEED intensity measurements were performed for the c(√2 × 5√2)R45° phase prepared depositing ca. 1.5 monolayer of Ti. The angle-scanned XPD curves measured for the phase c(√2 × 5√2)R45° reveal that Ti atoms substitute Cu atoms in the fcc lattice of the substrate. The polar XPD curves show that at least the first four layers of the substrate are involved in the alloying process. We found that the (3 1 0) plane of the Cu₄Ti alloy (D1_a type-structure) fits, without significant contraction or expansion of the lattice parameters, the c(√2 × 5√2)R45° structure. The intensity versus energy curves of the diffracted beams were calculated on the basis of this structural model using the tensor LEED method. The results of the LEED intensity analysis provide a further evidence of the formation of a slab of Cu₄Ti(3 1 0) layers.     

Initial growth of platinum on oxygen-covered Ni(1 1 0) surfaces

The initial growth of Pt on the Ni(1 1 0)-(3 × 1)-O and NiO(1 1 0) surfaces has been studied by coaxial impact collision ion scattering spectroscopy (CAICISS), low energy electron diffraction (LEED) and X-ray photoelectron spectroscopy (XPS). Prior to Pt deposition, the atomic structure of the near-surface regions of the Ni(1 1 0)-(3 × 1)-O and NiO(1 1 0) structures were studied using CAICISS, finding changes to the interlayer spacings due to the adsorption of oxygen. Deposition of Pt on the Ni(1 1 0)-(3 × 1)-O surface led to a random substitutional alloy in the near-surface region at Pt coverages both below and in excess of 1 ML. In contrast, when the surface was treated with 1800 L of atomic oxygen in order to form a NiO(1 1 0) surface, a thin Pt layer was formed upon room temperature Pt deposition. XPS and LEED data are presented throughout to support the CAICISS observations.     

In situ scanning tunneling microscopy studies of bimetallic cluster growth: Pt-Rh on TiO2(1 1 0)

The growth of Pt on clusters on TiO₂(1 1 0) in the presence and absence of Rh was investigated by scanning tunneling microscopy (STM) for Pt deposited on top of 0.3 ML Rh clusters (Rh + Pt). In situ STM studies of Pt growth at room temperature show that bimetallic clusters are produced when Pt is directly incorporated into existing Rh clusters or when newly nucleated clusters of pure Pt coalesce with existing Rh clusters. Low energy ion scattering experiments demonstrate that Rh is still present at the surface of the clusters even after deposition of 2 ML of Pt, indicating that Rh atoms can diffuse to the cluster surface at room temperature. Rh clusters were found to seed the growth of Pt clusters at room temperature as well as 100 K and 450 K. Furthermore, clusters as large as 100 atoms were observed to be mobile on the surface at room temperature and 450 K, but not at 100 K. Pt deposition at 100 K exhibited more two-dimensional cluster growth and higher cluster densities compared to room temperature experiments due to the lower diffusion rate. Increased diffusion rates at 450 K resulted in more three-dimensional cluster growth and lower densities for pure Pt growth, but cluster densities for Pt + Rh growth were the same as at room temperature.     

Growth of Ce on Rh(1 1 1)

We have studied the growth of cerium films on Rh(1 1 1) using STM (scanning tunneling microscopy), LEED (low energy electron diffraction), XPS (X-ray photoelectron spectroscopy) and AES (Auger electron spectroscopy). Measurements of the Ce films after room temperature deposition showed that Ce is initially forming nanoclusters in the low coverage regime. These clusters consist of 12 Ce atoms and have the shape of pinwheels. At a coverage of 0.25 ML (monolayer, ML) an adatom layer with a (2 × 2) superstructure is observed. Above 0.4 ML, Rh is diffusing through pinholes into the film, forming an unstructured mixed layer. Annealing at 250 °C leads to the formation of ordered Ce-Rh compounds based on the bulk compound CeRh₃. At a coverage of 0.1 ML, small ordered (2 × 2) surface alloy domains are observed. The exchanged Rh atoms form additional alloy islands situated on the pure Rh(1 1 1) surface, showing the same (2 × 2) superstructure as the surface alloy. At a coverage of 0.25 ML, the surface is completely covered by the surface alloy and alloy islands. The (2 × 2) structure is equivalent to a (1 1 1)-plane of CeRh₃, contracted by 6%. Annealing a 1 ML thick Ce layer leads to a flat surface consisting of different rotational domains of CeRh₃(1 0 0). The Rh needed for alloy formation comes from 50 Å deep pits in the substrate. Finally we show that LEIS (low energy ion scattering) is not suitable for the characterization of Ce and CeRh films due to strong effects of neutralization.     

Sulfur-induced mobilization of Au surface atoms on Au(1 1 1) studied by real-time STM

The interaction of sulfur with gold surfaces has attracted considerable interest due to numerous technological applications such as the formation of self-assembled monolayers and as a chemical sensor. Here, we report on the interaction of sulfur with Au(1 1 1) at two different temperatures (300 K and 420 K) studied by real-time scanning tunnelling microscopy, low energy electron diffraction and Auger electron spectroscopy. In the low coverage regime (<0.1 ML), S adsorption lifts the herringbone reconstruction of the clean Au(1 1 1) surface indicating a lateral expansion of the surface layer. An ordered (√3 × √3)R30° sulfur adlayer develops as the coverage reaches ∼0.3 ML. At higher S coverages (>0.3 ML) gold surface atoms are removed from regular terrace sites and incorporated into a growing gold sulfide phase. At 300 K this process leads to the formation of a rough pit and mound surface morphology. This gold sulfide exhibits short-range order and an incommensurate, long-range ordered AuS phase develops upon annealing at 450-525 K. In contrast, formation of an ordered AuS phase via rapid step-retraction rather than etch pit formation is observed during S-interaction with Au(1 1 1) surfaces at 420 K. Our results shed new light on the S-Au(1 1 1) interaction.     

Reconstructions 3 × 3 and √3 × √3 on SiC(0 0 0 1) studied using RHEED

The 3 × 3 and √3 × √3 reconstructions on 6H-SiC(0 0 0 1) surface were obtained via depositing thin silicon layer and annealing it in ultrahigh vacuum (without Si flux). Rocking curves of reflection high energy electron diffraction (RHEED) were measured for integer and fractional order beams. They were fitted with results of many-beam calculation on the basis of dynamical theory of RHEED to determine structural parameters. For √3 × √3 superstructure, it was found that the occupancy of adatom states is 0.45 (incomplete coverage). In the sequence of Si-C double layers ABCACB, the lattice is terminated with the layer A. For 3 × 3 superstructure, the rocking curves support the model with twisted tetra-cluster. The best-fit twist is as half of that predicted in ab initio calculations; it is due to limited source of Si atoms to build up the superstructure. Larger twist correlates with higher occupancy of corner sites and with slower cooling rate of the sample after annealing.     

Structural analysis of Si(1 1 1)-√21 × √21-Ag surface by reflection high-energy positron diffraction

The atomic structure of Si(1 1 1)-√21 × √21-Ag surface, which is formed by the adsorption of small amount of Ag atoms on the Si(1 1 1)-√3 × √3-Ag surface, was determined by using reflection high-energy positron diffraction. The rocking curve measured from the Si(1 1 1)-√21 × √21-Ag surface was analyzed by means of the intensity calculations based on the dynamical diffraction theory. The adatom height of the extra Ag atoms from the underlying Ag layer was determined to be 0.53 Å with a coverage of 0.14 ML, which corresponds to three atoms in the √21 × √21 unit cell. From the pattern analyses, the most appropriate adsorption sites of the extra Ag atoms were proposed.     

Methylthiolate-induced reconstruction of Ag(1 1 1): A medium energy ion scattering study

Medium energy ion scattering (MEIS), using 100 keV H⁺ incident ions, has been used to investigate the structure of the Ag(1 1 1)(√7 × √7)R19° -CH₃S surface phase. The results provide the first direct evidence that this structure does involve substantial reconstruction of the Ag surface layer. The measured absolute scattered ion yields and blocking curves are in generally good agreement with a specific structural model of the surface based on a reconstructed layer containing 3/7 ML Ag atoms, previously suggested on the basis of scanning tunnelling microscopy (STM) and normal incidence X-ray standing wave (NIXSW) studies. However, the MEIS data indicate that any rumpling of the thiolate layer, is small, and probably ?0.2 Å. This value is smaller than the amplitude suggested in the STM and NIXSW studies, but could be entirely consistent with the earlier experimental data.     

Influence of the surface-termination of hexagonal SiC(0 0 0 1) on the temperature dependences of Ge growth modes and desorption

With the aim of comparing initial Ge adsorption and desorption modes on different surface terminations of 4H-SiC(0 0 0 1) faces, 3 × 3, √3×√3R30° (R3) and 6√3×6√3R30° (6R3) reconstructions, of decreasing Si surface richness, have been prepared by standard surface preparation procedures. They are controlled by reflection high energy electron diffraction (RHEED), low energy electron diffraction and photoemission. One monolayer of Ge has been deposited similarly at room temperature on each of these three surfaces, followed by the same set of isochronal heatings at increasing temperatures up to complete Ge desorption. At each step of heating, the structural and chemical status of the Ge ad-layer has been probed. Marked differences between the Si- (3 × 3 and R3) and C-rich (6R3) terminations have been obtained. Ge wetting layers are only obtained up to 400 °C on 3 × 3 and R3 surfaces in the form of a 4 × 4 reconstruction. The wetting is more complete on the R3 surface, whose atomic structure is the closest to an ideally Si-terminated 1 × 1 SiC surface. At higher temperatures, the wetting layer stage transiets in Ge polycrystallites followed by the unexpected appearance on the 3 × 3 surface of a more ordered Si island structure. It denotes a Si clustering of the initial Si 3 × 3 excess, induced by the presence of Ge. A phase separation mechanism between Si and Ge prevails therefore over alloying by Ge supply onto a such Si-terminated 3 × 3 surface. Conversely, no wetting is obtained on the 6R3 surface and island formation of exclusively pure Ge takes place already at low temperature. These islands exhibit a better epitaxial relationship characterized by Ge(1 1 1)//SiC(0 0 0 1) and Ge〈1 1 −2〉//SiC〈1 −1 0 0〉, ascertained by a clear RHEED spot pattern. The absence of any Ge-C bond signature in the X-ray photoelectron spectroscopy Ge core lines indicates a dominant island nucleation on heterogeneous regions of the surface denuded by the 6R3 graphite pavings. Owing to the used annealing cycles, the deposited Ge amount desorbs on the three surfaces at differentiated temperatures ranging from 950 to 1200 °C. These differences probably reflect the varying morphologies formed at lower temperature on the different surfaces. Considering all these results, the use of imperfect 6R3 surfaces appears to be suited to promote the formation of pure and coherent Ge islands on SiC.     

Hyperthermal scattering of Cs from Pt(1 1 1): the effect of image charge on the ion-surface energy transfer

We have studied the energy exchange between hyperthermal (5-100 eV) Cs⁺ projectiles and a Pt(1 1 1) surface by measuring the kinetic energy of the scattered ions. The scattering geometry was chosen to be in-plane with specular scattering angles, and the energy of the scattered ions was analyzed as functions of incidence energy and angle. For low incidence energy (<40 eV), the energy transfer to the Pt surface is substantially enhanced due to the attractive image charge force between Cs⁺ and the surface. The image charge effects are highlighted by the different energy transfer on Pt(1 1 1) and Si(1 1 1) surfaces. Analysis of the experimental results using two- and three-dimensional theoretical models revealed a well depth of 1 eV for the image charge potential. Hyperthermal Cs⁺ ions scatter from Pt(1 1 1) predominantly via double collisions with Pt atoms, though the scattering phenomena are insensitive to the impact site at the surface.     

A structural parallel between Ge- and Si-induced 4 × 4 and 3 × 3 reconstructions on SiC(0 0 0 1) drawn from comparative RHEED oscillations

The initial Ge growth stages on a (√3 × √3)R30°-reconstructed SiC(0 0 0 1) surface (√3) have been studied using a complete set of surface techniques such as reflection high energy electron diffraction (RHEED), low energy electron diffraction (LEED), atomic force microscopy (AFM) and photoemission and compared with similar Si surface enrichments in place of Ge. The investigations essentially focus on the wetting growth-regimes that are favoured by the use of the √3 surface as a starting substrate, this surface being the closest to a smooth and ideally truncated Si-terminated face of hexagonal SiC(0 0 0 1). Depending on temperature and Ge or Si coverages, varying surface organizations are obtained. They range from unorganized layer by layer growths to relaxed Ge(1 1 1) or Si(1 1 1) island growths, through intermediate attempts of coherent and strained Ge or Si surface layers, characterized by 4 × 4 and 3 × 3 surface reconstructions, respectively. RHEED intensity oscillation recordings, as a function of Ge or Si deposited amounts, have been particularly helpful to pinpoint the limited (by the high lattice mismatch) existence domains of these interesting coherent phases, either in terms of formation temperature or surface coverages. Prominently comparable data for these two Ge- and Si-related reconstructions allow us to propose an atomic model for the still unexplained Ge-4 × 4 one. It is based on a same local organization in trimer and ad-atom units for the Ge excess as admitted for the Si-excess of the 3 × 3 surface, the higher strain nevertheless favouring arrangements, for the Ge-units, in 4 × 4 arrays instead of 3 × 3 Si ones. Admitting such models, 1.25 and 1.44 monolayers of Ge and Si, should, respectively, be able to lie coherently on SiC, with respective lattice mismatches near 30% and 25%. The experimental RHEED-oscillations values are compatible with such theoretical ones. Moreover, these RHEED coverage determinations (for layer completion, for instance) inform us in turn about the initial Si richness of the starting √3 reconstruction and help us to discriminate between earlier contradictory atomic models proposed in the literature.     

Determination of deuterium adsorption site on palladium(1 0 0) using low energy ion recoil spectroscopy

Ion beam analysis has been recently applied to study the adsorption phenomena of some adsorbates on metal surfaces. In this paper, surface recoils created by low energy Ne⁺ ions are employed to study the adsorption site of deuterium (D) atoms on Pd(1 0 0). This technique is extremely surface sensitive with the capacity for atomic layer depth resolution. From azimuthal angle observations of Pd(1 0 0) specimen, it was found that at room temperature, D was adsorbed in the fourfold hollow site of Pd(1 0 0) at a height of 0.25 ± 0.05 Å above the surface. The adsorbate remains in the hollow site at all temperatures to 383 K though the vertical height above the surface is found to depend on coverage and for the first time evidence is found of a transition to a p(2 × 2) structure for the adsorbate. There is no evidence of D sitting in the Pd(1 0 0) subsurface at room and higher temperatures.     

Composition and structure of chemically prepared GaAs(1 1 1)A and (1 1 1)B surfaces

The (1 1 1)A and (1 1 1)B surfaces of GaAs chemically treated in HCl-isopropanol solution (HCl-iPA) and annealed in vacuum were studied by means of X-ray photoelectron spectroscopy (XPS), low-energy electron diffraction (LEED) and electron energy loss spectroscopy (EELS). To avoid uncontrolled contamination, chemical treatment and sample transfer into UHV were performed under pure nitrogen atmosphere. The HCl-iPA treatment removes gallium and arsenic oxides, with about 0.5-3 ML of elemental arsenic being left on the surface, depending on the crystallographic orientation. With the increase of the annealing temperature, a sequence of reconstructions were identified by LEED: (1 × 1) and (2 × 2) on the (1 1 1)A surface and (1 × 1), (2 × 2), (1 × 1), (3 × 3), (√19 × √19) on the (1 1 1)B surface. These sequences of reconstructions correspond to the decrease of surface As concentration. The structural properties of chemically prepared GaAs(1 1 1) surfaces were found to be similar to those obtained by decapping of As-capped epitaxial layers.     

Structure and composition of the NiPd(1 1 0) surface

The NiPd(1 1 0) alloy surface was studied using low energy electron diffraction to measure the structure and composition of the first three atomic layers. The surface layer is highly enriched in Pd and has a significantly buckled structure. The second layer is also buckled, with displacements even larger than the surface layer. The second layer also exhibits intralayer segregation (chemical ordering), with alternate close-packed rows of atoms being Ni enriched and Pd enriched. The third layer has a structure and composition close to that of the bulk alloy. These results are compared with results for the other low index faces of NiPd, the extensive literature on NiPt alloy surfaces, and the growing body of theoretical literature for NiPd alloy surfaces.     

Surface phase transition and electronic structure of c(5√2 × √2)R45°-Pb/Cu(1 0 0)

The c(5√2 × √2)R45°-Pb/Cu(1 0 0) surface phase is investigated by means of angle resolved ultraviolet photoemission and low energy electron diffraction in the temperature range between 300 and 550 K. We identify and characterize a temperature-induced surface phase transition at 440 K from the room temperature c(5√2 × √2) R45° phase to a (√2 × √2)R45° structure with split superstructure spots. The phase transition is fully reversible and takes place before the two-dimensional melting of the structure at 520 K. The electronic structure of the split (√2 × √2)R45° phase is characterized by a metallic free-electron like surface band. This surface band is backfolded with c(5√2 × √2)R45° periodicity phase at room temperature, giving rise to a surface band gap at the Fermi energy. We propose that a gain in electronic energy explains in part the stability of the c(5√2 × √2)R45° phase.     

Probing the interactions of Pt, Rh and bimetallic Pt-Rh clusters with the TiO2(1 1 0) support

Pt, Rh, and Pt-Rh clusters on TiO₂(1 1 0) have been investigated by scanning tunneling microscopy (STM), soft X-ray photoelectron spectroscopy (sXPS), and low energy ion scattering (LEIS). The surface compositions of Pt-Rh clusters are Pt-rich (66-80% Pt) for room temperature deposition of both 2 ML of Pt on 2 ML of Rh (Rh + Pt) and 2 ML of Rh on 2 ML of Pt (Pt + Rh). Pt and Rh atoms readily diffuse within the clusters at room temperature, and although diffusion is slower at 240 K, intermixing of Pt and Rh still occurs. The binding energies of surface and bulk states for Rh(3d_5/2) and Pt(4f_7/2) can be distinguished in sXPS studies, and an analysis of these spectra indicates that the surface compositions of the Pt + Rh and Rh + Pt clusters are similar at room temperature but not identical. In addition to sintering, the pure Pt, pure Rh and Pt-Rh clusters become completely encapsulated by titania upon heating to 700 K. sXPS investigations show that annealing the clusters to 850 K induces reduction of titania support to Ti⁺² and Ti⁺³, with the extent of reduction being the greatest for Pt, the least for Rh and intermediate for Pt-Rh. We propose that TiO₂ is reduced at the metal-titania interface on top of the clusters, not at the base of the clusters. Furthermore, the extent of titania reduction is greater for metal clusters with weaker metal-oxygen bonds because oxygen atoms are less likely to migrate to the top of the clusters, and therefore the encapsulating titania is oxygen-deficient.