Modeling heterogeneous catalysts: metal clusters on planar oxide supports

Transcription

1 Topics in Catalysis Vol. 14, No. 1-4, Modeling heterogeneous catalysts: metal clusters on planar oxide supports C.C. Chusuei, X. Lai, K. Luo and D.W. Goodman Department of Chemistry, PO Box 30012, Texas A&M University, College Station, TX , USA Model catalysts consisting of Au and Ag clusters of varying size have been prepared on single crystal TiO 2 (110) and ultra-thin films of TiO 2,SiO 2 and Al 2 O 3. The morphology, electronic structure, and catalytic properties of these Au and Ag clusters have been investigated using low-energy ion scattering spectroscopy (LEIS), temperature-programmed desorption (TPD), X-ray photoelectron spectroscopy (XPS) and scanning tunneling microscopy (STM) and spectroscopy (STS) with emphasis on the unique properties of clusters <5.0 nm in size. Motivating this work is the recent literature report that gold supported on TiO 2 is active for various reactions including low-temperature CO oxidation and the selective oxidation of propylene. These studies illustrate the novel and unique physical and chemical properties of nanosized supported metal clusters. KEY WORDS: metal clusters; activity; quantum size effects; thin film oxides; Au; Ag; TiO 2 ;SiO 2 ;Al 2 O 3 ; TPD; XPS; LEIS; STM; STS 1. Introduction In its bulk form, Au has typically been regarded to be chemically inert as compared to other platinum group metals and received little attention as a catalyst. However, recent findings have shown that Au clusters when deposited as finely dispersed, small particles (<5 nm in diameter) on reducible metal oxides such as TiO 2,Fe 2 O 3 and Co 3 O 4 [1 3] dramatically enhance a number of industrially relevant reactions. Reactions that occur on TiO 2 supports include lowtemperature CO oxidation [1,2,4 21], hydrogenation and partial oxidation of hydrocarbons [2,3,21 23] and the selective oxidation of propylene to propylene oxide [3,24]. Similarly, small Ag clusters on metal oxides have been shown to promote selective oxidation of ethylene to ethylene oxide [25 27] and methanol to formaldehyde [28 30]. The catalytic properties of these small noble metal clusters vary widely depending on their particle size. Unsupported, nanosized metal particles are typically highly reactive and known to agglomerate (limiting their practical application to catalysis), but can be stabilized on solid metal oxide matrices. The oxides most commonly used to support Au particles (in addition to the aforementioned TiO 2 ) include SiO 2 [14,16,21,31,32] and Al 2 O 3 [21,33]. Ag particles supported on TiO 2 [34 38], SiO 2 [39 41] and Al 2 O 3 [27,33,42 49] have also been examined, showing cluster size-dependent catalytic activity. Although considerable effort has been directed to understanding activity observed in the high surface area catalyst systems, much remains to be understood regarding chemical interactions at the atomic level. Such investigations regarding structure size dependency on activity and selectivity are still in their infancy. Fundamental insight into the nature of the interaction of metal clusters with the oxide support is a necessary prerequisite to fine-tuning and improving catalytic performance. To whom correspondence should be addressed. Significant progress has been made in the past several decades using ultrahigh vacuum (UHV) surface sensitive probes, capable of providing detailed information about surface composition and structure [50 53]. However, since bulk metal oxides (ubiquitous in heterogeneous catalyst systems) are insulators, charging problems typically hamper their analysis. In our laboratory, these difficulties have been circumvented by synthesizing ultra-thin film metal oxides via hot filament deposition on planar supports under background O 2 pressure, using them as model planar supports for metal particle deposition to mimic the real-world systems. In addition to using single crystal TiO 2 (110) [36,37], thin films of TiO 2 [34,35,38,54], SiO 2 [55 59] and Al 2 O 3 [60 62] have been synthesized on single crystal refractory metal substrates, making them amenable for surface analysis. These films ( 50 Å thick) are thin enough to surmount the insulating problems associated with charged particles and yet substantive enough to retain the bulk chemical and electronic properties of the metal oxides. Since flat surfaces are used, these systems are also suitable for scanning probe microscopy to study morphology. In addition, difficulties related to inhomogeneous sample heating, which complicates the use of temperature-programmed desorption (TPD), are also circumvented by the thin-film oxide synthesis procedure. In this review, a summary of recent studies in this laboratory are described involving model catalysts of nanosized Au and Ag clusters supported on single crystal TiO 2 (110) and ultra-thin films of TiO 2,Al 2 O 3 and SiO 2. These systems have been investigated using low-energy ion scattering spectroscopy (LEIS), TPD, X-ray photoelectron spectroscopy (XPS) and scanning tunneling microscopy (STM) and spectroscopy (STS) specifically to establish correlations among physical, chemical and catalytic properties /00/ $18.00/ Plenum Publishing Corporation

2 72 C.C. Chusuei et al. / Modeling heterogeneous catalysts 2. Experimental ATiO 2 (110) single crystal (Commercial Crystal Laboratories) had been chosen as the oxide support chiefly due to its suitability for atomically-resolved STM and STS analysis [63]. The n-type semiconductor was found to be sufficiently conductive for STM and electron spectroscopy after cycles of Ar + bombardment and annealing to K. The TiO 2 (110) crystal was mounted onto a tantalum sample holder. The temperature was monitored using an optical pyrometer (OMEGA OS3700), the temperature scale of which was calibrated from voltages from a W-5% Re/W-26% Re thermocouple glued to the edge of the crystal using hightemperature ceramic adhesive (AREMCO). The TiO 2 (110) crystal could be resistively heated to 1500 K (or electron beam heated to 2400 K) and cooled via a liquid nitrogen (LN 2 ) reservoir thermally attached to the sample using Cu leads. TPD experiments were carried out using a line-ofsight quadrupole mass spectrometer (QMS). In order to reduce the TiO 2 (110) sample (making it conductive and hence suitable for STM/STS), the crystal was annealed to 1000 K for several hours in UHV. Stoichiometric TiO 2 (110)-(1 1) was prepared by Ar + bombardment (1.0 kv), followed by annealing to 1000 K in Torr O 2 pressure for 10 min [64]. The TiO 2 (110)-(1 2) reconstructed surface was prepared by annealing to 1300 K; low-energy electron diffraction (LEED) and STM verified both the (1 1) and (1 2) structures. A low binding energy shoulder in the Ti 2p 3/2 XP spectrum (indicative of surface reconstructed Ti 3+ )provided an additional diagnosis of the (1 2) structure [65]. Thus, the effects of both surface structures on metal cluster strong-metal support interaction were investigated using LEIS. Thin film TiO 2 had been deposited onto Mo(110), supported by 99.9+% pure Ta wire. Similarly, the SiO 2 and Al 2 O 3 supports were epitaxially grown as thin films deposited onto Mo(110) and Re(0001) substrates, respectively. W-5% Re/W-26% Re thermocouple wires were spot-welded onto the metal supports in order to measure temperature. The chemical compositions of these oxide supports were verified by XPS and their surface structures by LEED. Metal dosers used to deposit the noble metal clusters of interest (Au and Ag) were constructed by wrapping high purity wires (99.99%) of the metals around diameter tungstenfilaments (H&R Cross), which were then resistively heated in vacuum by passing current through the filament wires. These wires were melted and thoroughly outgassed to remove impurities prior to use. The Ag and Au fluxes, respectively, were calibrated via integrated TPD peak areas of the metals (depositing them onto a Re(0001) substrate) and also using the metal-to-substrate Auger intensity break to denote the first monolayer equivalent (ML) [66]. A 1 ML surface coverage was defined as atoms/cm 2. During use of the dosers for metal deposition, the overall UHV system pressure did not exceed Torr. The deposition rates used for Au and Ag dosing were and ML/min, respectively. Pressures in the HPC region, isolated from the STM-UHV system via gate valve, were measured with a Baratron gauge. Three UHV systems were used to carry out the deposition of Au and Ag onto the planar metal oxide supports and the subsequent surface spectroscopic analysis. The first chamber was equipped with XP and Auger electron (AES) spectroscopies, LEIS, TPD and LEED capabilities. XPS data were collected using a Mg Kα anode (hν = ev) operated at 300 W and 15 kv (Perkin Elmer PHI X- ray source) and a concentric hemispherical analyzer (PHI, SCA ) with an incident angle of ca. 30 from the surface normal. LEIS data were acquired using He + projectiles with a beam energy of 650 ev and a scattering incident angle of 45 from the surface normal. To minimize surface damage, only one sweep (across the kinetic energy range) per set of scans was acquired. A second UHV chamber (also equipped with XPS, AES and LEED) was used to acquire the STM and STS data (Omicron STM-1). STM data were typically collected with sample biases of 1 2 V and tunneling currents of na. All STM and STS data were obtained at 298 K. A constant current topographic (CCT) mode was employed for STM imaging. Highly ordered pyrolytic graphite (0001) was used to calibrate the scaling of the STM micrographs. The STM tip was made from tungsten wire (0.020 in diameter, H&R Cross) and prepared by electrolytic etching [67]. In order to remove the oxide overlayer formed from the etching, the tip was dipped in HF solution for 30 s, rinsed in deionized H 2 O, rapidly transferred into UHV, and then heated to 1000 K prior to use. The base pressure of these first two chambers were both Torr after bakeout. An elevated high pressure reactor cell (HPC) had been combined with the STM chamber to carry out reaction kinetics at elevated pressure. Using the HPC, described elsewhere [68,69], reactions of the prepared model catalyst surfaces (Ag/TiO 2 (110) and Au/TiO 2 (110)) with gas mixtures of CO and O 2 were carried out and subsequently studied with STM and STS. The model catalysts were transferred from the HPC to UHV using differentially-pumped sliding seals. Research grade CO was purified by storing it at liquid-n 2 temperature; O 2 was used as received. The CO : O 2 (2 : 1) mixtures were premixed prior to use. The third UHV system (having a base pressure of Torr after bakeout) was equipped with AES, TPD, an HPC reactor and a Varian 3400CX series gas chromatograph (GC) with a thermal conductivity detector (TCD) and Hyesept D, 40 foot column. 3. Results and discussion 3.1. Finite size effects and catalytic activity Figure 1(A) shows a CCT STM micrograph of 0.25 ML Au deposited onto single crystal TiO 2 (110)-(1 1) [22]. The deposition was performed at 300 K, followed by annealing the TiO 2 surface to 850 K for 2 min. Only the Ti

3 C.C. Chusuei et al. / Modeling heterogeneous catalysts 73 (A) (B) Figure 1. (A) A CCT STM image of a 0.25 ML Au deposited onto TiO 2 (110)-(1 1) prepared just prior to a CO : O 2 reaction. The sample had been annealed to 850 K for 2 min; (B) STS data acquired for Au clusters of varying sizes on the TiO 2 (110)-(1 1). An STS of the TiO 2 substrate, having a wider band gap than the Au clusters, is also shown as a point of reference. cations are imaged in the STM; the O atoms are not seen [63]. The inter-atomic distances between the [001] rows are separated by 0.65 nm, which can be observed along the terraces corresponding to the length of the unit cell along the [110] direction of unreconstructed TiO 2 (110)-(1 1). Three-dimensional (3D) Au clusters, imaged as bright protrusions, have average diameters of 2.6 and 0.7 nm height (corresponding to 2 3 atoms thick) and preferentially nucleate on the step edges. Quasi-two-dimensional (2D) clusters are characterized by heights of 1 2 atomic layers. Previous annealing studies revealed that the Au clusters form large microcrystals with well-defined hexagonal shapes [36]. Figure 1(B) shows STS taken at various points on the surface. The ST spectra had been acquired by positioning the STM tungsten tip at a desired point and interrupting the STM feedback loop. In the figure, the tunneling current (I) as a function bias voltage (V ) across the tip is measured. The I V curves are thus correlated with various Au clusters of varying sizes on the TiO 2 surface (figure 1(A)). The length of the observed plateaus at the zero tunneling current (figure 1(B)) denotes the band gap (along the bias voltage axis) of electrons tunneling between the valence and conduction bands between the surface (imaged Au particles) and the tungsten tip. The electronic character of these clusters vary between that of a metal and a nonmetal, depending on size. With increase in size, the clusters gradually adopt metallic character with an enhanced density of states at the Fermi level, characterized by the more abrupt slopes in the STS for larger clusters. Note that the cluster with the 2.5 nm 0.7 nm size has a larger band gap than that of the 5.0 nm 2.5 nm cluster (figure 1(B)); smaller Au clusters have a non-metallic character resulting in significant band gaps whereas larger clusters have bulk-like metallic properties giving rise to essentially no band gaps. The 4.0 nm 1.5 nm and 3.0 nm 1.0 nm clusters, having intermediate-sized band gaps, also follow this cluster size band gap dependency. Similar changes in the band gap as a function of cluster size also have been observed in STM/STS analysis of Fe clusters epitaxially grown on GaAs(110) [70]. In these studies, Fe clusters with volumes of 1 nm 3 (85 Fe atoms) displayed fully metallic characteristics while volumes 0.15 nm 3 (13 atoms) were also found to be non-metallic. Also, the same trends were observed for ErP islands (20 50 nm in size) grown on InP(001) surfaces [71]. Semi-metallic behavior denoted by narrower band gaps measured by STS I V curves was observed for thick ErP islands (>3.4 nm) while (relatively wide) semi-conducting band gaps were observed for thinner ones (<3.4 nm). A more detailed discussion of quantum sizes effects and their relationship with structure-sensitive activity (due to differences in metal cluster island size) will be published elsewhere [72]. Thus, the electronic structure of the adsorbed Au particles can play an important role in the activity of Au on the surface. In this present study, a correlation is observed between Au cluster size and catalytic activity for the oxidation of CO on Au/TiO 2 (110)-(1 1). Figure 2(A) shows a plot of activity for CO oxidation (expressed as (product molecules)/((total Au atoms on surface sites) (second)) or turnover frequency (TOF)) at 350 K as a function of Au cluster size supported on the TiO 2 (110)-(1 1) substrate [22]. CO and O 2 (1:5 mixture of CO:O 2 ) had been reacted on the previously prepared Au(0.25 ML)/TiO 2 (110)-(1 1) surfaces at 40 Torr [5,22,23]. Thin-film TiO 2 epitaxially grown

4 74 C.C. Chusuei et al. / Modeling heterogeneous catalysts Figure 2(B) shows a plot of the STS band gaps measured over the same cluster size regime used for the CO : O 2 reactions (figure 2(A)). Maximum catalytic activity occurs in concert with the metal-to-nonmetal transition. The average Au cluster size at the divergence from metallic character is 3.5 nm in diameter and 1.0 nm in height, corresponding to approximately 300 atoms per cluster. Figure 2(C) shows a histogram indicating the relative distribution of the Au cluster sizes ranging from 2.0 to 4.0 nm in diameter at maximum catalytic activity, corresponding to STS measured band gaps from 0.2 to 0.6 V. Two-atom-thick clusters (with diameters between 2.5 and 3.0 nm) are characteristic of those optimally active for CO oxidation. These above studies thus demonstrate that cluster electronic properties play a crucial role in defining the catalytic reactivity of small clusters [22]. There has been considerable interest, theoretically and experimentally, in studying these size-dependent changes in electronic structure. In recent years, XP and ultraviolet photoelectron (UPS) spectroscopies have provided many examples of cluster size dependent electronic modifications from the discrete energy levels of free atoms to the continuous, k-dependent energy bands of bulk metals [74 77]. These methods (with typical spatial resolutions in the µm scale) are not useful for analyzing the metal particles within the nanometer range; however, STM and STS can be utilized Effects of O 2 exposure on admetal cluster size and distribution Figure 2. (A) The activity for CO oxidation at 350 K as a function of Au cluster size supported on TiO 2 (110)-(1 1) assuming total dispersion of the Au. A 1:5 CO:O 2 mixture had been used at a total pressure of 40 Torr. Activity is expressed as (product molecules)/((total Au atoms) (second)); (B) cluster band gaps measured by STS as a function of Au cluster size supported on TiO 2 (110)-(1 1). The band gaps were obtained while the corresponding topographic scan was acquired on various Au coverages ranging from 0.2 to 4.0 ML. (") Two-dimensional (2D) clusters, (E) 3D clusters, two-atom layers in height, (P) 3D clusters, three-atom layers or greater in height; (C) relative population of Au clusters (two-atom layers in height) that exhibit a band gap of ev, as measured by STS of Au/TiO 2 (110)-(1 1). on a Mo(100) substrate [73] followed by Au cluster deposition was used for reaction kinetics in the GC-UHV system. These kinetic studies were carried out in parallel with the UHV-STM Au/TiO 2 (110)-(1 1) single crystal experiments. The product (CO 2 ) was extracted from the reactor with a vacuum syringe, compressed and then analyzed with a GC. For each point in figure 2(A), a particular Au cluster size was prepared then subjected to the CO 2 :O 2 reaction. The cluster sizes of the Au particles and coverage of the surface sites (identical to that of the TiO 2 (110)-(1 1) single crystal) obtained from the parallel STM imaging experiments were used to calculate the TOF. The activity of the Au/TiO 2 catalyst exhibits a maximum 1.90 TOF at an average Au cluster diameter of 3.5 nm, decreasing with larger diameter. Ambient pressures of O 2, ubiquitous in real world catalyst preparation conditions, is an important variable to consider as it may affect the admetal s ability to wet the surface and thereby alter particle size and distribution. From abrønsted linear free energy interpretation of TPD data of Au adsorbed on TiO 2 (110), Bondzie et al. [78] postulated that small clusters of Au are able to dissociatively adsorb O 2 (at 10 5 Torr) more readily than large ones. Further investigations [22,23,37] show that smaller Au clusters become larger with an increase in O 2 pressure, effectively reducing catalytic activity. For example, after a 10 Torr O 2 exposure to 0.25 ML Au deposited onto TiO 2 (110)-(1 1) for 120 min [22,37], the Au cluster density decreased and the cluster size increased. These changes are attributed to sintering of the Au particles via a ripening process possibly involving AuO x ; however, thermodynamic data regarding Au oxide formation is lacking. Thus, this mechanism cannot be confirmed. Similar thermodynamic data, however, exists for Ag oxide formation. In an experiment analogous to those described for the Au/TiO 2 system, various cluster sizes of Ag on TiO 2 (110)- (1 1) have been deposited and studied by STM and STS. The metal-to-nonmetal transition for Ag (figure 3) occurs at a slightly larger cluster size ( nm diameter) than that of Au ( nm diameter). STM and STS had been carried out on the Ag/TiO 2 system before and after high pressure O 2 exposure. Figure 4(A) shows a CCT

5 C.C. Chusuei et al. / Modeling heterogeneous catalysts 75 Figure 3. Size-dependent metal-to-nonmetal transitions for Au and Ag clusters on TiO 2 (110)-(1 1). The transition size regimes are nm in diameter for Au clusters and nm in diameter for Ag clusters. STM image of 2.0 ML of Ag (the approximate coverage where the metal-to-nonmetal transition is observed) vapor deposited on a clean TiO 2 (110)-(1 1) single crystal surface. The features displayed in the micrograph correspond to Ti + cations. Direct tunneling into (or out of) the oxygen sites is unlikely since the O 2p state is 3 evbelow the Fermi level, beyond the operational range of the microscope. The 3D homogeneous hemispherical Ag clusters are observed on both the flat terraces and step edges. These particles have an average cross-sectional area of 4.8 nm 2.6 nm (diameter height), corresponding to 1900 atoms per cluster. After the Ag/TiO 2 surface had been prepared, the substrate was transferred into the HPC reactor and exposed to 10 Torr O 2 at 298 K for 120 min and then transferred back into the UHV for STM and STS. The resulting image is shown in figure 4(B). A bimodal distribution of Ag clusters is evident on the TiO 2 (110)-(1 1) surface after the O 2 exposure. Comparing with the original micrograph (figure 4(A)), some clusters enlarge while others diminish in size (figure 4(B)). In addition, there is a 5 15% increase in cluster density, indicative of Ag cluster redispersion on the surface. A histogram of the Ag (figure 5) shows a transformation from a unimodal dispersion with a mean cluster diameter of 5 nm to a bimodal distribution with mean cluster diameters of 3.5 and 6.8 nm. The single distribution has cluster sizes ranging from 2.0 to 6.5 nm; the bimodal distribution has one size domain ranging from 1.0 to 5.0 nm in diameter while the second domain ranged from 5.0to11nm.Thesmallerclustersinthe nm domain have a higher density and narrower distribution with an average size of 3.0 nm 1.1 nm ( 260 atoms/cluster). The larger Ag cluster domain with the lower density and broader size distribution had an average size of 6.7 nm 3.1 nm ( 4200 atoms/cluster). STS band gap measurements showed the nm domain electronic structure to be nonmetallic and the nm domain to be fully metallic. The calculated Ag cluster volumes (from STM data) before and after O 2 exposure were the same (within ±10% error). The redispersion of the Ag clusters is a ripening process attributed to two possible mechanisms: migration and coalescence of the metal atoms on the surface or intercluster and/or vapor phase transport. Regarding the second possibility, reduction of the total surface free energy by intercluster transport occurs such that the larger clusters grow at the expense of the smaller ones [79]; some clusters increase in size while others decrease, leading to the bimodal distribution. The data are more consistent with an Ostwald ripening mechanism than of simple coalescence. With O 2 exposure the following reaction is plausible: 2Ag(s) O 2(g) Ag 2 O(s) and is thermodynamically favorable at 298 K. The standard free energy of formation ( G)ofAg 2 Ois 11.2 kj/mol, allowing an estimate of the equilibrium partial pressure of O 2 required for the above reaction to be Torr, far lower that than the pressure used in the data of figures 4 and 5. For an average cluster diameter of 5.0 nm (r = 2.5 nm), the G 298 (r) value decreases to 22.7 kj/mol, taking into account the effect of the cluster curvature on the free energy. The driving force for Ag cluster oxidation is high at room temperature and likely leads to Ostwald ripening. Detailed calculations of these thermodynamic considerations are shown elsewhere [37,72]. The intrinsic electronic properties suggest that certain sizes of Ag clusters are more reactive to O 2 molecules than others, i.e., certain Ag cluster sizes would undergo Ostwald ripening more rapidly and eventually deplete. Other Ag cluster sizes would experience ripening more slowly due to reduced kinetics with O 2 to form Ag 2 O. Clearly, the relative size-dependent reactivity to O 2 also contributes to the bimodal distribution of Ag. To further study the effect of the substrate on Ostwald ripening of the Ag clusters, a different oxide support had been selected for comparison with TiO 2 (110). Ultra-fine particles of Ag supported on Al 2 O 3 has attracted recent interest for the oxidation of CO and reduction of NO showing high conversions ( 90%) of NO and CO on the high surface area catalyst systems [80]. Al 2 O 3 thin films were prepared on a Re(0001) single crystal. Ag was then deposited via Ag metal evaporator and then followed by a 10 Torr O 2 exposure. The chemical compositional and surface structures of Al 2 O 3 thin films grown heteroepitaxially on refrac-

6 76 C.C. Chusuei et al. / Modeling heterogeneous catalysts Figure 4. CCT STM images ( nm 2, 2.0 V, 1.0 na) of 2.0 ML Ag/TiO 2 (110). (A) Fresh 2.0 ML Ag/TiO 2 (110)-(1 1); (B) 2.0 ML Ag/TiO 2 (110)- (1 1) after exposure to 10 Torr O 2 for 120 min at 298 K. tory metal substrates have been well-characterized [60 62]. Compared to TiO 2,Al 2 O 3 is an irreducible and unreconstructed surface and relatively free of surface defects. Figure 6 shows a series of O 2 exposed Ag support CCT STM images. Image (A), shown for comparison, is an image of 2.0 ML of Ag deposited onto a cleaned TiO 2 crystal. Image (B) is a CCT STM of the same Ag coverage onto an Al 2 O 3 thin film. The Ag clusters on the Al 2 O 3,ascompared to the TiO 2, are slightly larger and have a smaller cluster density. Interestingly, the metal oxide support with higher defect density (TiO 2 ) results in smaller cluster sizes at a higher dispersion. (This effect of defect density on admetal dispersion and particle size will be further addressed in section 3.3.) After the Ag/Al 2 O 3 surface had been exposed to 10 Torr O 2 for 120 min at 298 K in the HPC, image (C) was obtained. Instead of a bimodal distribution, prevalent for the Ag/TiO 2 (110) system, a relatively homogeneous distribution of Ag on the Al 2 O 3 is evident. Ag cluster

7 C.C. Chusuei et al. / Modeling heterogeneous catalysts 77 Figure 5. Size distribution of 2.0 ML Ag/TiO 2 (110)-(1 1) before (top) and after (bottom) a Torr exposure for 120 min at 298 K. A bimodal distribution results after the O 2 treatment. ripening, forming the bimodal distribution, is not observed until higher O 2 pressures are employed. There is an obvious substrate effect of the oxide support on the admetal cluster distribution and Ostwald ripening. Ag is apparently more stable and resistant to ripening when supported on Al 2 O 3 than on TiO 2 (110) due to its comparatively defectfree surface. It can be concluded that the greater number of surface defects/vacancies on the TiO 2 (110) is responsible for the relative ease of Ostwald ripening occurring at 10 Torr O 2. The relatively defect-free Al 2 O 3 support leads to reduced admetal reactivity with O 2, a requirement for ripening Cluster dispersion on TiO 2 (110)-(1 2) versus TiO 2 (110)-(1 1) As alluded to in section 3.2, the structure of the underlying metal oxide substrate support also has an effect on metal cluster dispersion on deposited Au and Ag clusters. Factors influencingnoblemetal dispersionare importantconsiderations for addressing practical catalyst preparation issues since higher dispersions on oxide supports generally leads to increased catalytic activity [81,82]. TiO 2 (110)-(1 1) and TiO 2 (110)-(1 2) were chosen as the substrates for the study since their surface structures are well characterized, in particular with respect to surface defects that can be produced on each surface via ion bombardment and heating [65,83 85]. Nanosized Au and Ag metal clusters had been deposited onto TiO 2 (110)-(1 1) and TiO 2 (110)-(1 2) at 300 K, respectively, using metal evaporators and their growths were monitored with LEIS. Surface coverages had been checked with AES and TPD. The (1 2) surface has relatively more defect sites than the (1 1) surface; the defects serve as nucleation sites for metal cluster growth. Hence, metal deposition on a rough (1 2) surface would likely differ from the (1 1) surface metal cluster nucleation and growth behavior. LEIS (in conjunction with STM) is ideally suited for probing the surface compositional structure of adsorbed metal clusters due to its high surface sensitivity (capable of probing the top nm of the substrate). Plotting the integrated LEIS peak areas (arbitrary units) as a function of adsorbed metal clusters (ML units) has proven to be an effective diagnostic tool for distinguishing various admetal growth modes: layer-by-layer (Frank van der Merwe, FM), 3D cluster growth (Volmer Weber, VW) or an initial monolayer, followed by 3D cluster growth (Stranski Krastanov, SK) mode. Au, Pd and Ag admetals have been found to nucleate and grow in a 3D fashion when adsorbed onto the single crystal TiO 2 (110) surface at coverages 1 ML. However, the metal grows as quasi-2d clusters at coverages <1 ML, as shown by both LEIS [64] and STM imaging [36]. An FM or SK growth is characterized by a linear increase in the integrated LEIS peak areas as a function of increasing adsorbate surface coverage, reaching a plateau upon completion of the first monolayer. The substrate intensity simultaneously decreases with increasing admetal intensity, which is then fully attenuated above 1 ML. This effect has been observed for Fe [86] and Hf [87] clusters deposited onto single crystal TiO 2 (110) at 160 K and room temperature, respectively. It should be noted that after the 1 ML coverage, the substrate signal cannot be seen and hence FM and SK modes cannot be readily distinguished from LEIS data alone. During VW growth, on the other hand, the LEIS intensity increases with a smooth non-linear function at increasing metal coverages since the substrate is still exposed. The non-linear increase continues even after a few monolayers of admetal are deposited. This same growth behavior had been observed for VW growth of Cu [88] and Pt [89] in TiO 2 (110). Based on LEIS and XPS characterization, Campbell and coworkers have proposed a kinetic model for Au cluster growth on TiO 2 (110) [90]: one-atom, 2D thick Au islands initially form until a critical coverage is reached [91], after which the growth switches to 3D islanding. From LEIS data, this critical coverage was found to be independent of the metal vapor flux, but increased with decreasing adsorption temperature and increasing defect density of the substrate. The underlying metal substrate surface structure was thus clearly shown to affect the growth behavior of the deposited admetals. To further investigate the phenomenon in this current study, differences between TiO 2 (110)-(1 1) and TiO 2 (110)- (1 2) with respect to Au particle dispersion are examined. Figure 7 shows stackplots of LEIS intensity at various Au cluster surface coverages (in ML units of the admetal) on both (A) TiO 2 (110)-(1 1) and (B) TiO 2 (110)-(1 2) surfaces. The Au clusters had been evaporated onto the TiO 2 (110) at 300 K. For coverages of 1 5 ML (for the (1 1) surface)), Au cluster diameters are in the nm range; the Au cluster sizes deposited are thus near the metal

8 78 C.C. Chusuei et al. / Modeling heterogeneous catalysts Figure 6. CCT STM images showing O 2 exposure effects on 2.0 ML Ag/Al 2 O 3 /Re(0001). (A) Freshly exposed 2.0 ML Ag on TiO 2 (110) shown as a reference (2.0 V, 1.0 na); (B) freshly exposed 2.0 ML Ag/Al 2 O 3 /Re(0001) (2.0 V, 0.26 na); (C) 2.0 ML Ag/Al 2 O 3 /Re(0001) exposed to 10 Torr O 2 for 120 min; (D) 2.0 ML Ag/Al 2 O 3 /Re(0001) exposed to 1000 Torr for 120 min. nonmetal transition region (where increased catalytic activity would be predicted). LEIS spectra of both TiO 2 single crystal substrates at 270, 460 and 580 ev denote He + scattering from O, Ti and Au sites, respectively. Note that in each case, the Au attenuates the TiO 2 substrate layer as it grows; the Ti and O intensities decrease as Au increases. In comparing the two LEIS stackplots, a relatively greater attenuation of the Ti and O by Au on the (1 2) surface than on the (1 1) is apparent for the same amount of Au surface coverage, indicating a greater density of islanding on the (1 2) surface. Figure 8 shows a plot of the integrated LEIS peak areas of the Au adsorbed divided by those of the underlying substrate Ti and O atoms of the (1 1) and (1 2) substrates. Since the chemical composition of the substrate remains unchanged as Au coverage increases, the substrate provides a way of normalizing the adsorbed Au LEIS intensity. Plots of the Au/Ti and Au/O ratios versus Au coverage further accentuates the various Au cluster dispersions found for the (1 1) and (1 2) surfaces. Note the non-linearity of the Au growth versus coverage; both the (1 1) and (1 2) surfaces exhibit VW crystal growth characteristics. The upward curvature for the (1 2) surface is steeper than that of the (1 1); this is particularly pronounced between 2 and 5 ML, indicating that the Au clusters are more completely dispersed on the (1 2). Clearly, the higher dispersion is surface structure related. The inter-atomic spacing as measured by STM between the Ti atoms in the [001] direction is nm on the (1 1) whereas it is 1.3nmonthe (1 2) [63]. Also, STM shows that the bare (1 1) surface as compared with the (1 2) has a larger numberof terraces and fewer isolated cluster structures [38]. Given the tendency of metal clusters to preferentially nucleate at step edge defects rather than flat terrace defect sites, a greater density of 3D islanding is expected for the (1 2) at a similar coverage.

9 C.C. Chusuei et al. / Modeling heterogeneous catalysts 79 Figure 8. Plot of Au/O and Au/Ti ratios taken from integrated LEIS peak areas of Au cluster growth on TiO 2 (110)-(1 1) and TiO 2 (110)-(1 2). Au coverages ranged from 0 to 5.0 ML. Higher Au/O and Au/Ti ratios on the (1 2) surface denotes a wider cluster dispersion. Figure 7. Stackplots of LEIS spectra at Au cluster coverages ranging from 0 to 5 ML. Intensities of signals from O and Ti were multiplied by 2 for clarity. (A) Au cluster growth on TiO 2 (110)-(1 1); (B) Au cluster growth on TiO 2 (110)-(1 2). A combination of greater distances between the Ti rows in conjunction with the inherently greater surface roughness results in greater dispersion of the 3D islands (figure 8). Differences have also been seen for Ag cluster growth (unreacted) on TiO 2 (110)-(1 1) and TiO 2 (110)-(1 2). The 3D Ag crystallites are more widely dispersed on the (1 2) surface than on the (1 1) structure [38]. The LEIS uptake curves of the He + scattered Ag intensity relative to Ti and O on the (1 1) and (1 2) surfaces (not shown) also increased with smooth non-linear functions, but the increase for the (1 2) is greater (similar to that for Au cluster growth). The degree of the curvature for the (1 2) versus (1 1) is also greater. The same factors influencing dispersion in for Au cluster adsorption also apply for Ag cluster adsorption on the TiO 2 (110) surface; namely (1) more defects (roughness) and (2) increased spacing between the Ti rows on the (1 2). The cluster sizes (comparing Ag and Au) differ slightly, but the overall growth trends (as measured by LEIS) on the (1 1) versus (1 2) surfaces are the same. STM analysis show that at a 0.5 ML Ag exposure, metal cluster diameters (with heights in parentheses) on the (1 1) has a diameter of 3.5 4nm( nm); for the (1 2), it is nm(0.8 1 nm) [36]. The Ag clusters disperse to a greater degree on the (1 2) surface, further verifying the conclusions drawn from LEIS. The dissimilar admetal dispersion on the (1 1) versus (1 2) titania structures has implications for metal surface restructuring in defining catalytic activity. According to the chemisorption induced surface restructuring model (CISRM), proposed by Somorjai and coworkers [92 95], surface metallic atoms do not remain in the same surface locations as for the bare surface but move to new positions in response to chemical changes in the environment. During chemisorption, these displacements strengthen the chemical bond of the adsorbate to the surface, provoking a local surface strain and hence promote catalytic reactions. Substrate atom rearrangement is more pronounced at steps, open surfaces and low coordination sites. It is inferred from our STM and LEIS results that for TiO 2 (110), surface Ti atoms during chemisorption/catalytic reactions (involving nanosized noble metal clusters) prefer a rearrangement to a structure much more like TiO 2 (110)-(1 2) than TiO 2 (110)-(1 1)

10 80 C.C. Chusuei et al. / Modeling heterogeneous catalysts structure. The (1 2) configuration is more amenable for increased metal cluster dispersion and subsequent activity Metal oxide support effects on cluster electronic structure XPS is a useful tool for probing the electronic structure of interactions between the adsorbed metal particles and the underlying metal oxide support. Binding energy (BE) shifts due to final state effects are a result of screening of electrons from core level vacancies created by the photoemission process. As electrons relax to screen the hole, the emitted photoelectron is ejected with increased kinetic energy (hence lower BE). If a given adsorbed cluster is sufficiently small such that screening of its photoelectrons is less that that of the bulk substrate, a higher BE would result. Initial state effects may originate from a variety of factors such as (1) interfacial chemical reactions, (2) BE differences between surface and bulk atoms or (3) nucleation on various defect sites can also result in cluster size dependent core level shifts. Self-consistent field (SCF) calculations of the core ionization potentials (from the XPS 4f core levels) of small Pt clusters deposited on SiO 2 substrates show that the BE shifts and line-broadeningdependchiefly on cluster size and cluster substrate interactions [96]. Furthermore, SCF analysis shows that the core level shifts were not due to electrostatic effects of the unit positive charge remaining on the ionized cluster. Figure 9 (A) and (B) shows two sets of core level BE peak centers for the Au 4f 7/2 core level as a function of Au cluster coverage on (a) the TiO 2 (110) surface and (b) on thin film SiO 2 ( 2.5 nm thick) deposited on a Mo(110) refractory metal surface. It should be noted that in its bulk form, SiO 2 would be non-conductive and hamper electron spectroscopic analysis; it therefore, had been deposited as a thin film. All Au dosings were carried out at 300 K. On the TiO 2 surface, a 0.8 ev BE shift is evident from small clusters (0.02 ML, 2 nm diameter), shifting to the bulk value of Au 4f 7/2 = 84.0 ev with increasing cluster size (6 ML Au coverage, 5 nm diameter; figure 9(A)). Average cluster diameters (within parentheses above) had been measured by STM from Au fluxes on TiO 2 (110) [36]; the BE of the 4f 7/2 core level typically found for bulk Au is 84.0 ev [97]. For the Au/SiO 2 system, the corresponding BE shift is greater (figure 9(B), 1.6 ev). This lowering from high BE to the bulk value is to be expected for supported metal clusters as they increase in size and as the crossover from non-metal to metal character occurs. As reported in an earlier XPS study by Mason [98], relative BE shifts of Au clusters (of varying sizes) supported on SiO 2 and Al 2 O 3 have been shown to differ as a result of differences in the relative abilities of the substrate to screen (final state effects) the outgoing photoelectrons. In comparing the relative core level BE shifts, SiO 2 ( 1.3 ev shift from small cluster to bulk size) was found to have a slightly greater screening ability than Al 2 O 3 ( 1.1 ev shift). Differences in the magnitudes of these Au (A) (B) Figure 9. Plots of XPS BE peak centers of the Au 4f 7/2 core level as a function of Au cluster coverage (ranging from 0.02 ML to bulk) on TiO 2 (110) (A) and SiO 2 (B) surfaces. The BE shift for SiO 2 ( 1.6 ev) is more pronounced than for TiO 2. cluster core level BE shifts were interpreted to be a result of the relative abilities of the metal oxide supports to shield the final-state hole via extra-atomic relaxation. Accordingly, the higher Au coverage (20 ML of Au on SiO 2, figure 9(B)) required to reach the bulk BE value (as compared to 6 ML for the Au/TiO 2 system; figure 9(A)) would signify SiO 2 s greater screening ability. In addition, the XPS linewidth, which is also dependent upon relaxation, should increase with a decrease in screening. This is the case in this present XPS study. The full-width half-maxima (fwhm) of the Au 4f 7/2 peaks (not shown) increased with decreased Au coverage on bothoxidesurfacesand is thusindicativeoffinal state contributions. The fwhm at 0.02 ML and bulk Au coverages are 2.1 and 0.8 ev, respectively, for the Au/SiO 2 system; for the Au/TiO 2 (110) system these values are 1.7 and 1.0 ev, respectively. It should be noted, however, that the origin of metal cluster size dependent BE shifts is a controversial subject. The first observations of core-level shifts with metal cluster size were attributed to a size dependence from the initial-state electronic structure [99]. An increase of valence d electrons (i.e., in Pd) with increasing cluster size were thought to be the chief cause of the BE shift. Later work presented an

11 C.C. Chusuei et al. / Modeling heterogeneous catalysts 81 alternative interpretation that shifts in BE were not due to initial state properties, but rather to variations in final state relaxation processes [75,100,101]. In these current studies, there may be initial state contributions to the BE shift due to differences in Au adsorption to various defect sites. However, at present there is insufficient data (STM on the Au/SiO 2 to compare with the Au/TiO 2 system) to make specific comments in this regard. The BE shifts that is observed from Au deposited on the TiO 2 and SiO 2 substrates is likely a convolution of both initial and final state effects. Perhaps an interplay exists between the quantum size effect of the clusters and their interaction with the underlying metal oxide support to give rise to their unique activity. The selection of the metal oxide support is also an important factor for enhanced catalytic activity. Clearly, interactions of the TiO 2 and SiO 2 supports influence the Au cluster electronic structure Metal cluster sublimation energies TPD is a useful tool for obtaining detailed information on adsorbate surface bonding and on adsorbate adsorbate interactions, desorption kinetics and determining binding energies of metals adsorbed onto surfaces. TPD binding energy determinations also allow for comparative estimations of admetal cluster size on different oxide supports. In a series of TPD spectra taken of Au on SiO 2, marked decreases of (adsorbed) Au cluster binding energies, denoted by the peak temperature maximum (T m ), is observed and attributed to varying cluster size. A 3 K/s linear heating rate had been used to acquire these spectra. Figure 10 shows a family of TPD spectra taken of the Au clusters deposited onto the SiO 2 thin film previously examined by XPS (section 3.4). The leading edge of the TPD peak maxima shifts to higher temperatures as Au coverage increases. The inset shows a plot of the sublimation energy (E sub )asafunctionofau coverage, which have been determined using leading edge analysis [102]. At 0.2 ML, the E sub at 50 kcal/mol increases rapidly (with increasing Au coverage) to the bulk value at 90 kcal/mol at 5.0 ML. The decrease in E sub can be explained by the fact that an atom at the edge of a small cluster has fewer nearest neighbors than larger ones and hence desorbs more easily due to decreased surface tension. Rodriguez et al. [48] used differences in T m in the TPD leading edges to show that Ag clusters are larger (and reduced in density) when the same admetal surface coverages were deposited onto an O/Mo(110) surface as comparedtoal 2 O 3 /Mo(110). This effect of decreased adsorbate bond energy that accompanies decreasing cluster size has been demonstrated from Monte Carlo simulations of a model 400 atom metal cluster island. The T m of a family of the calculated TPD spectra shifted to lower temperatures with decreasing island size [103]. In another series of TPD experiments, obtained by Xu et al. [104] to measure bond energies of Au clusters deposited onto TiO 2 (001)/Mo(100), the same T m increase in the TPD leading edge (5 K/s linear heating rate) at larger Au coverages was observed from Figure 10. A set of TPD spectra of Au (m/e = 197) on a 2.5 nm thick SiO 2 thin film on Mo(110) at Au cluster coverages ranging from 0.2 to 5.0 ML. The inset shows a plot of E sub determined from leading edge analysis K at 0.2 ML to 1190 K at 2.0 ML. A slow decrease in LEIS TiO 2 intensity even at high Au coverages (5.0 ML) denoting 3D Au cluster growth on the surface was further verified by STM images obtained from this surface. The E sub reported for bulk Au in this Au/TiO 2 study was found to be at 90 kcal/mol, in agreement with current results observed for the Au/SiO 2 system (figure 10). The T m s from the Au TPD spectra obtained from both oxide supports are approximately equal. For instance, the T m s for Au on SiO 2 are at 1100 K at 0.2 ML and 1190 K at 2.0 ML. Differences in the T m determination of the 0.2 ML coverage (as compared to 1090 K on the Au/TiO 2 ) may be due to statistical variations that accompany low intensity TPD peaks. At coverages less than 2.0 nm thick, the morphology of the Au exists as a single layer. As the clusters grow to from 2.0 to 4.0 nm thick, a second quasi-2d layer develops on top of the first. It is precisely in this region where maximum catalytic activity is observed for CO oxidation (figure 2(B), quasi-2d clusters denoted by the diamonds). Above a 4.0 nm thick (1.0 ML) coverage, the clusters form 3D islands. This same 2-to-3D growth behavior had also been observed from LEIS measurements performed in the groups of Madey [64] and Campbell [78]. Since Au grows in a 3D fashion (at coverages 1.0 ML) on both TiO 2 and SiO 2 and from relative surface tension effect arguments (on outer perimeter cluster atoms), the TPD spectra suggest that the Au islands deposited are the same size for both surfaces at equivalent Au dosings. It should be noted that this size comparison from the TPD data is an approximate estimation. Detailed STM imaging for Au cluster on SiO 2 to verify this result are currently in progress. Nevertheless, this finding provides further support for the role of cluster substrate interactions (independent from cluster size factors) for influencing the admetal electronic structure and accounting for differences in the observed XPS core level shifts (figure 9).

12 82 C.C. Chusuei et al. / Modeling heterogeneous catalysts 4. Perspective Model catalyst systems with surface sensitive methods are a valuable methodology for probing the electronic and morphological structure of supported metal clusters. From detailed STM/STS studies, a physical basis for understanding the enhanced catalytic activities of small, dispersed metal is developing. Ostwald ripening is apparent at elevated O 2 pressures. STM suggests that Al 2 O 3, has fewer defects than TiO 2 (110), resulting in increased resistance to cluster ripening upon exposure to similar O 2 pressures. The dispersions of Ag and Au depend upon the underlying surface structure (roughness, density of plateaus and defects). Relative differences in XPS core level BE shifts as a function of cluster coverage also reveal that the chemical composition of the underlying oxide support affects its electronic structure; cluster size estimations from TPD analysis provide further corroboration of the role of the support. The above results obtained thus far signify an interplay of admetal cluster size and the cluster substrate interactions responsible for catalytic activity. The use of metal clusters supported on thin oxide films provides new insights into the special electronic and chemical properties that govern their unique catalytic chemistry. Future studies toward a more in depth understanding of nanostructured supported clusters will undoubtedly lead to practical catalytic applications of these interesting materials. Acknowledgment We acknowledge with pleasure support of this work by the Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences and the Robert A. Welch Foundation. CCC gratefully acknowledges financial support from the Associated Western Universities, Inc. and the Pacific Northwest National Laboratories operated by Battelle. References [1] M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M.J. Genet and B. Delmon, J. Catal. 144 (1993) 175. [2] M. Haruta, Catal. Today 36 (1997) 153. [3] T. Hayashi, K. Tanaka and M. Haruta, J. Catal. 178 (1998) 566. [4] J.-D. Grunwaldt and A. Baiker, J. Phys. Chem. B 103 (1999) [5] M. Valden, S. Pak, X. Lai and D.W. Goodman, Catal. Lett. 56 (1998) 7. [6] J.-D. Grunwaldt, C. Kiener, C. Wögerbauer and A. Baiker, J. Catal. 181 (1999) 223. [7] Z.M. Liu and M.A. Vannice, Catal. Lett. 43 (1997) 51. [8] F. Boccuzzi, A. Chiorino, S. Tsubota and M. Haruta, J. Phys. Chem. 100 (1996) [9] E.D. Park and J.S. Lee, J. Catal. 186 (1999) 1. [10] H. Liu, A.I. Kozlov, A.P. Kozlova, T. Shido, K. Asakura and Y. Iwasawa, J. Catal. 185 (1999) 252. [11] Y. Iizuka, H. Fujiki, N. Yamauchi, T. Chijiiwa, S. Arai, S. Tsubota and M. Haruta, Catal. Today 36 (1997) 115. [12] M.A.P. Dekkers, M.J. Lippits and B.E. Nieuwenhuys, Catal. Lett. 56 (1998) 195. [13] S. Tsubota, T. Nakamura, K. Tanaka and M. Haruta, Catal. Lett. 56 (1998) 131. [14] Y. Yuan, K. Asakura, H. Wan, K. Tsai and Y. Iwasawa, Catal. Lett. 42 (1996) 15. [15] G.R. Bamwenda, S. Tsubota, T. Nakamura and M. Haruta, Catal. Lett. 44 (1997) 83. [16] S.D. Lin, M. Bollinger and M.A. Vannice, Catal. Lett. 17 (1993) 245. [17] S. Minicò, S. Sciré, C. Crisafulli, A.M. Visco and S. Galvagno, Catal. Lett. 47 (1997) 273. [18] M.A. Bollinger and M.A. Vannice, Appl. Catal. B 8 (1996) 417. [19] N.W. Cant and N.J. Ossipoff, Catal. Today 36 (1997) 125. [20] K. Fukushima, G.H. Takaoka, J. Matsuo and I. Yamada, Jpn. J. Appl. Phys., Part I 36 (1997) 813. [21] M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Azuma and M. Haruta, Catal. Lett. 51 (1998) 53. [22] M. Valden, X. Lai and D.W. Goodman, Science 281 (1998) [23] M. Valden and D.W. Goodman, Isr. J. Chem. 38 (1998) 285. [24] M. Haruta, in: 3rd World Congr. Oxidation Catal. (Elsevier, Amsterdam, 1997) p [25] S. Cheng and A. Clearfield, J. Catal. 94 (1985) 455. [26] V.I. Bukhtiyarov, A.I. Boronin, I.P. Prosvirin and V.I. Savchenko, J. Catal. 150 (1994) 268. [27] V.I. Bukhtiyarov, I.P. Prosvirin, R.I. Kvon, S.N. Goncharova and B.S. Bal zhinimaev, J. Chem. Soc., Faraday Trans. 93 (1997) [28] A.N. Pestryakov and A.A. Davydov, Appl. Catal. A 120 (1994) 7. [29] A.N. Pestryakov, Catal. Today 28 (1996) 239. [30] D. Herein, H. Werner, T. Schedel-Niedrig, T. Neisius, A. Nagy, S. Bernd and R. Schlögl, in: 3rd World Congr. Oxidation Catal. (Elsevier, Amsterdam, 1997) p [31] S. Lin, Catal. Lett. 10 (1991) 47. [32] Y.A. Kalvachev, T. Hayashi, S. Tsubota and M. Haruta, in: 3rd World Congr. Oxidation Catal. (Elsevier, Amsterdam, 1997) p [33] D.G. van Campen and J. Hrbek, J. Phys. Chem. 99 (1995) [34] D. Martin, F. Creuzet, J. Jupille, Y. Borensztein and P. Gadenne, Surf. Sci (1997) 958. [35] D. Abriou, D. Gagnot, J. Jupille and F. Creuzet, Surf. Rev. Lett. 5 (1998) 387. [36] X. Lai, T.P. St. Clair, M. Valden and D.W. Goodman, Prog. Surf. Sci. 59 (1998) 25. [37] X. Lai, T.P. St. Clair and D.W. Goodman, Faraday Discuss. 114 (1999) 279. [38] K. Luo, T.P. St. Clair, X. Lai and D.W. Goodman, J. Phys. Chem. B 104 (2000) [39] S.R. Seyedmonir, J.K. Plischke, M.A. Vannice and H.W. Young, J. Catal. 123 (1990) 534. [40] X. Li and A. Vannice, J. Catal. 151 (1995) 87. [41] H. Kudo and T. Ono, Appl. Surf. Sci. 121/122 (1997) 413. [42] G.R. Meima, M.G.J. V. Leur, A.J.V. Dillen and J.W. Geus, Appl. Catal. 44 (1988) 133. [43] G.R. Meima, L.M. Knijff, R.J. Vis, A.J. van Dillen, F.R. van Buren and J.W. Geus, J. Chem. Soc., Faraday Trans. I 85 (1989) 269. [44] G.R. Meima, M. Hasselaar, A.J. van Dillen, F.R. van Buren and J.W. Geus, J. Chem. Soc., Faraday Trans. I 85 (1989) [45] G.R. Meima, L.M. Knijff, A.J. van Dillen and J.W. Geus, J. Chem. Soc., Faraday Trans. I 85 (1989) 293. [46] G.R. Meima, R.J. Vis, M.G.J. van Leur, A.J. van Dillen, J.W. Geus and F. van Buren, J. Chem. Soc., Faraday Trans. I 85 (1989) 279. [47] C.-F. Mao and M.A. Vannice, Appl. Catal. A 122 (1995) 41. [48] J.A. Rodriguez, M. Kuhn and J. Hrbek, J. Phys. Chem. B 100 (1996) [49] J.-A. Wang, G. Aguilar-Ríos and R. Wang, Appl. Surf. Sci. 147 (1999) 44. [50] G.A. Somorjai, Chemistry in Two Dimensions: Surfaces (Cornell Univ. Press, Ithaca, NY, 1981). [51] G. Ertl and J. Küppers, Low Energy Electrons and Surface Chemistry, 2nd Ed. (VCH, Weinheim, 1985). [52] D.P. Woodruff and T.A. Delchar, Modern Techniques of Surface Science, 2nd Ed. (Cambridge Univ. Press, Cambridge, 1994). [53] G.A. Somorjai, Introduction to Surface Chemistry and Catalysis (Wiley, New York, 1994).