Molecular Chemisorption of O2 on a PdO(101) Thin Film on Pd(111

Phone: 352-392-0869 . ..... For O-atom coverages between 0.25 and about 1 ML, only a small feature at about 120 K is evident in the TPD spectra. This ...
0 downloads 0 Views 1021KB Size
Download PDF
8324

J. Phys. Chem. C 2008, 112, 8324–8331

Molecular Chemisorption of O2 on a PdO(101) Thin Film on Pd(111) Jose A. Hinojosa, Jr., Heywood H. Kan, and Jason F. Weaver* Department of Chemical Engineering, UniVersity of Florida, GainesVille, Florida 32611 ReceiVed: January 9, 2008; ReVised Manuscript ReceiVed: February 29, 2008

We used temperature programmed desorption (TPD) to investigate the molecular chemisorption of O2 on a (101)-oriented PdO thin film grown on Pd(111) in ultrahigh vacuum (UHV) using an oxygen atom beam. Our results show that the molecular chemisorption of O2 is facile on the PdO film at 85 K, producing a saturation O2 coverage of 0.27 ML (monolayers). Experiments with co-adsorbed 16O2 and 18O2 further reveal that molecularly chemisorbed O2 dissociates negligibly on the PdO(101) surface under the conditions examined. The O2 TPD spectrum from the PdO(101) surface at O2 saturation exhibits two main features centered at 117 K and 227 K, as well as smaller features at 275 K and 315 K, associated with the desorption of molecularly chemisorbed O2. Comparison with O2 TPD obtained from clean Pd(111) demonstrates that a large fraction of the O2 molecules on the PdO(101) surface are more strongly bound than O2 chemisorbed on the metallic surface at 85 K and saturation of the respective O2 layers. We find that O2 molecules chemisorb only in small quantities (< 0.03 ML) on the p(2 × 2) and two-dimensional Pd5O4 phases of atomic oxygen on Pd(111), indicating that these phases have much weaker binding affinities toward O2 than the PdO(101) surface generated in our experiments. Finally, temperature programmed reaction spectra with co-adsorbed 18O2 and CO demonstrate that both PdO and molecularly adsorbed O2 actively participate in the oxidation of CO, with the atomic and molecular species exhibiting similar activities for the conditions studied. The results of this study may have implications for understanding Pd oxidation catalysis at high pressures given that we find relatively strong binding states of O2 on PdO and observe that these molecularly adsorbed species are active in CO oxidation. Introduction Palladium is an excellent catalyst for methane oxidation in lean gas turbines1–8 and exhibits high activity for the oxidation of CO and hydrocarbons in automotive exhausts.9–11 Significant effort has been devoted to characterizing the reactivity of oxygen states on Pd surfaces. Of particular interest has been to understand the catalytic behavior of PdO due to its role in lean catalytic combustion systems. Indeed, it is now well-established that PdO is highly active toward methane oxidation,1–3,7,12 and is responsible for the favorable performance of Pd catalysts under oxygen-rich conditions. PdO is also active for CO oxidation,13 with hydrated PdO exhibiting exceptional activity for this reaction below 373 K.14 Despite significant progress, however, controversy still exists in the literature regarding the relative CO oxidation activities of metallic versus oxidic Pd and Pt surfaces and the mechanisms for reaction. These discrepancies are particularly evident when comparing findings of high pressure studies with those obtained from ultrahigh vacuum (UHV) experiments. Difficulty in producing welldefined oxide surfaces of Pd and Pt has certainly contributed to the limited fundamental understanding of the chemisorptive and reactive properties of Pd and Pt oxide surfaces. In general, the results of UHV experiments indicate that oxide phases on Pd and Pt surfaces are less active toward CO oxidation than chemisorbed oxygen atoms, whereas recent high-pressure experiments suggest the opposite trend. For example, both temperature programmed reaction spectroscopy (TPRS) and isothermal rate measurements conducted in UHV show that bulk-like oxides on Pt(111)15 and Pt(100)16,17 are much less * To whom correspondence should be addressed. E-mail: weaver@ che.ufl.edu. Phone: 352-392-0869. Fax: 352-392-9513.

active toward CO oxidation than oxygen phases that exist at lower coverage, particularly chemisorbed oxygen atoms. Ultrahigh vacuum investigations of CO oxidation on Pd(111)9 and Pd(100)18 also report a decrease in CO oxidation activity for more fully oxidized surfaces. In contrast, results of in situ high pressure experiments suggest that oxidized Pt(110)19,20 and Pd(100)21 surfaces are actually more reactive for CO oxidation than their metallic forms. Indeed, multiple factors could be responsible for the contrasting findings between the low and the high pressure experiments. However, a key difference in experimental conditions is that the UHV studies all examined CO oxidation in the absence of gaseous O2, mainly because O2 is ineffective in oxidizing single crystal Pd and Pt surfaces under high vacuum conditions. Considering this difference, the interactions between O2 and oxidized Pd and Pt surfaces, as well as the specific role of O2 during CO oxidation by these catalysts, warrants further scrutiny. A possibility that has seldom been considered for these systems is that O2 chemisorbed molecularly on the oxide surface reacts directly with CO to produce CO2, most likely in parallel with CO reacting with surface oxygen atoms. For such a molecular pathway to contribute appreciably to the reactivity, it is necessary that O2 molecules bind sufficiently strongly to the relevant Pd and Pt oxide surfaces. If the binding is strong, then elevated O2 pressures could sustain relatively high concentrations of adsorbed O2 on the oxide surface and thereby enable the molecular reaction channel to make a significant, perhaps even dominant, contribution to the CO oxidation rate. Motivated by this idea, we investigated the molecular chemisorption of O2 on a PdO(101) thin film grown on Pd(111) in UHV using an oxygen atom beam. As far as we know, this is the first study to examine adsorption on a crystalline PdO surface

10.1021/jp800216x CCC: $40.75  2008 American Chemical Society Published on Web 05/02/2008

Molecular Chemisorption of O2 under UHV conditions. Our results show that O2 can molecularly chemisorb on PdO in high concentrations, and that a large fraction of the O2 molecules bind more strongly on the PdO(101) surface than on clean Pd(111). We also find that O2 molecules chemisorbed on the PdO surface are active in the oxidation of coadsorbed CO during TPRS. These findings provide compelling evidence that molecularly chemisorbed O2 plays an important role in the oxidation of CO on PdO at commercially relevant pressures. Experimental Details Previous studies22–25 provide details of the three-level UHV chamber utilized for the present experiments. The Pd(111) crystal employed in this study is a circular disk (8 mm × ∼1 mm) spot-welded to W wires and attached to a copper sample holder in thermal contact with a liquid nitrogen cooled reservoir. A type K thermocouple spot-welded to the backside of the crystal allows sample temperature measurements. Resistive heating, controlled using a PID controller that varies the output of a programmable DC power supply, supports maintaining or linearly ramping the sample temperature from 81 to 1250 K. Initially, sample cleaning consisted of sputtering with 600 eV Ar+ ions at a surface temperature of 900 K, followed by annealing at 1100 K for several minutes. Subsequent cleaning involved routinely exposing the sample held at 856 K to an atomic oxygen beam for several minutes, followed by flashing the sample to 923 K to desorb oxygen and carbon oxides. We considered the sample to be clean when we could no longer detect contaminants with X-ray photoelectron spectroscopy (XPS), could obtain sharp low energy electron diffraction (LEED) patterns consistent with the Pd(111) surface, and did not detect CO or CO2 production during flash desorption after oxygen adsorption. We generated various oxygen atom coverages on Pd(111) using a RF plasma source that generates atomic oxygen beams by partially dissociating pure O2 (BOC gases 99.999%). Specifically, the atomic-oxygen covered surfaces were prepared at 500 K with the sample located at approximately 50 mm from the terminus of the quartz tube that collimates the oxygen atom beam and rotated roughly 45° with respect to the tube axis. The incident atomic oxygen flux utilized was about 0.02 ML s-1 as determined using the procedure discussed previously.24 We define 1 ML as equal to the surface atom density of 1.53 × 1015 cm-2 of Pd(111). After each O atom beam exposure, we cooled the sample to 85 K and exposed it to various doses of molecular oxygen. Molecular 18O2 (Isotech 99.9%) was delivered without further purification to the Pd(111) surface using a leak valve and with the sample positioned about 50 mm from the end of a tube connected to the outlet of the leak valve. To minimize the uptake of background 16O2, it was necessary to shut off the RF plasma source during experiments with 18O2. In other experiments, 16O2 molecules were supplied to the sample surface at an incident flux of about 0.1 ML s-1 using a calibrated molecular beam doser. Typical 18O2 exposures ranged from 0.01 to 0.1 L (Langmuir), while larger 16O2 doses of ∼60 ML were used to saturate the surface. After adsorbing O2 on the sample surface, we performed TPD by first facing the sample toward the entrance of a quadrupole mass spectrometer ionizer at a distance of about 10 mm and then raising the temperature at a linear rate of 1 K s-1. To enhance the resolution of the TPD spectra, we usually monitored only the mass of the molecular O2 that was used for the low temperature dose. As discussed below, however, we did monitor the masses of all O2 isotopic combinations in selected experi-

J. Phys. Chem. C, Vol. 112, No. 22, 2008 8325 ments. We processed the TPD spectra by smoothing the data with a five-point adjacent averaging procedure and then performing a linear background subtraction. Absolute oxygen coverages were determined from integrated O2 TPD areas and assuming that saturation exposures of Pd(111) to O2 at 300 K produces an oxygen atom coverage of 0.25 ML. We followed the procedure described in our recent study26 to saturate the subsurface of Pd(111) with atomic oxygen prior to the TPD experiments and thereby minimized potential uncertainty in the coverage calibration caused by oxygen dissolution into the Pd bulk during heating. It is important to note that we state coverages of O2 in terms of ML of O atoms throughout this manuscript. Using a similar procedure as that for the O2 TPD measurements, we also performed preliminary TPRS experiments to determine if molecularly chemisorbed O2 on the PdO surface is reactive toward CO. In these experiments, CO (Praxair 99.99%) was supplied to the surface using the molecular beam doser. We estimated the CO and CO2 yields obtained from the TPRS measurements by scaling the desorption curves with approximate calibration factors determined from recent studies of CO oxidation on Pt(100) conducted in our laboratory.16,17 Specifically, we assumed that the relationships among the O2, CO, and CO2 scaling factors are the same in the present study as in our work with Pt(100) and then rescaled the CO and CO2 factors using the calibration factor that we determined for O2 during the present study. This approach may introduce error in the absolute CO and CO2 yields, but such error should largely cancel when considering relative values. Results and Discussion Growth of a PdO Film on Pd(111). We conducted most of the O2 adsorption experiments in this study using a PdO layer that was generated in UHV by exposing a Pd(111) sample held at 500 K to an oxygen atom beam (∼0.02 ML s-1) for 10 min. According to XPS and O2 TPD, this treatment produces a PdO layer containing 3 ML of oxygen atoms.26 LEED images further reveal that the 3 ML PdO film generated at 500 K is well-ordered and suggest that the PdO film completely covers the Pd(111) surface. We note that the quality of the LEED image obtained from the PdO film improves after annealing to 675 K, so the PdO surface investigated in the present study apparently had a relatively high concentration of domain boundaries. However, we have recently obtained nearly identical O2 TPD spectra from annealed and unannealed PdO films after molecular chemisorption of O2, which suggests that domain boundaries and other defects have only a small influence on the chemisorption of O2 on these PdO surfaces. As detailed elsewhere,27 the LEED pattern obtained from the PdO film exhibits three rotational domains of a rectangular unit cell and closely matches the simulated LEED pattern of bulk-terminated PdO(101). A recent computational study predicts that the bulk-terminated, nonpolar PdO(101) surface is a thermodynamically favorable facet of PdO and is only slightly less stable than the preferred polar PdO(100)-PdO surface.28 Considering the PdO(101) surface structure (Figure 1), we estimate that the 3 ML PdO film on Pd(111) is about 12 Å thick and consists of between four and five layers, each with ∼0.70 ML of oxygen atoms. The nonpolar PdO(101) surface is stoichiometric, but half of the Pd atoms, equivalent to 0.35 ML, are coordinately unsaturated (cus) and are expected to be more active toward binding molecules than the saturated Pd sites.

8326 J. Phys. Chem. C, Vol. 112, No. 22, 2008

Figure 1. Three-dimensional representation of the stoichiometric PdO(101)-PdO surface. Pd and O atoms are shown as blue and orange spheres, respectively, and the atoms in the surface layer are labeled according to their coordination. The coordinately unsaturated (cus), 3-fold Pd atoms form close-packed rows and constitute half of the Pd atoms in the surface layer.

Figure 2. 16O2 TPD spectra (heating rate ) 1 K s-1) obtained from clean Pd(111) and from a 3 ML PdO(101) thin film after exposing the surfaces to 60 ML of 16O2 at 85 K. The resulting 16O2 coverages are 0.66 and 0.26 ML for Pd(111) and the PdO film, respectively.

Molecular Desorption of O2 from the PdO Thin Film. Shown in Figure 2 are O2 TPD spectra obtained from clean Pd(111) and from the PdO thin film after exposing these surfaces to saturation doses of O2 at 85 K. By considering differences in sample heating rates, the TPD spectrum from the O2-exposed metal surface agrees well with those reported in prior studies.29–31 For the metal surface, we observe two sharp O2 desorption features, labeled as R2 and R1, centered at 120 and 179 K, respectively, as well as a third broader feature (not shown) at about 750 K. Previous studies have established that the low temperature R1 and R2 features originate from molecularly chemisorbed O2 on Pd(111), while the desorption feature near 750 K results from the recombinative desorption of chemisorbed oxygen atoms.29,30 At 85 K, only molecularly chemisorbed O2 exists on Pd(111),32 but a fraction of these species dissociate as the sample is heated during the TPD experiment, giving rise to the recombinative desorption feature near 750 K. While we do not observe desorption peaks below 120 K, we note that other researchers have observed an O2 desorption feature, designated as R3, at temperatures below that of the R2 peak.29–31 It is possible that O2 molecules do not populate the R3 state

Hinojosa et al. under the conditions we examined or that such species convert to the more strongly bound states before they can desorb during TPD. From our data, we estimate a total oxygen coverage of 0.66 ML at O2 saturation of the metal surface at 85 K, with 0.42 ML evolving below 400 K. These values agree well with prior work by Guo et al.29 Notice also that we obtain a coverage of 0.24 ML of oxygen atoms produced by O2 dissociation, in excellent agreement with the known O-atom saturation coverage of 0.25 ML on Pd(111) obtained with O2 as the oxidant.33 After the O2 exposure at 85 K, the O2 TPD spectrum obtained from the PdO film exhibits two prominent features (β2 and β1) with maxima at 117 and 227 K, respectively, and shoulders, labeled as γ2 and γ1, centered at about 275 and 315 K, respectively. As discussed in more detail below, experiments using 18O2 show that these features originate from molecularly chemisorbed O2, and that a negligible amount of the adsorbed O2 molecules dissociates on the PdO surface during heating. Note that decomposition of the 3 ML PdO film begins to occur appreciably only above about 650 K and produces a sharp peak at 760 K in the TPD spectra. The O2 TPD spectrum below 400 K is qualitatively similar to that obtained from the metal surface in that two main features are observed in each case. The β2 feature is less intense than the analogous R2 feature obtained from the metal surface, but both features appear at about 120 K in the TPD spectra and each is rather narrow. In contrast, the β1 feature is much broader than the R1 peak and reaches a maximum at a temperature nearly 50 K higher than the R1 state. This latter observation is intriguing since it implies that a large fraction (∼75%) of O2 molecules chemisorbed on the PdO film is more strongly bound than O2 chemisorbed on metallic Pd(111). The data shown in Figure 2 yields an oxygen coverage of 0.26 ML for O2 that desorbs below 400 K from the PdO surface. This value is lower than that for the metal surface, but it is a relatively large coverage in an absolute sense. In fact, the saturation O2 coverage of 0.26 ML is nearly 75% of the concentration of cus-Pd atoms of the PdO(101) surface. It is interesting that the saturation O2 coverage, while high, is still lower than the concentration of cus-Pd sites since this characteristic could indicate preferential O2 binding at the cus-Pd sites. At the least, the large O2 coverage makes it difficult to explain the strong O2 binding states on PdO solely in terms of O2 adsorption on oxide defect sites such as oxygen vacancies. Prior work using high resolution electron energy loss spectroscopy (HREELS) reveals distinct differences in the O2 binding states on Pd(111).31,32,34 Specifically, molecularly chemisorbed O2 species on Pd(111) give rise to vibrational bands at about 650, 850, and 1035 cm-1 in the HREEL spectra, referred to as ω1, ω2, and ω3, respectively, with the relative intensities of the loss peaks depending on factors such as the O2 coverage and surface temperature. These vibrational peaks have been attributed to O2 molecules bound in a superoxo state (ω3) and two peroxo states, designated as “peroxo I” (ω2) and “peroxo-II” (ω1), where binding of the peroxo-I and peroxo-II species is thought to involve each O atom of an O2 molecule bonding to the same Pd atom versus a different Pd atom, respectively, that is, unidentate versus bidentate configurations.32 These HREELS studies also demonstrate that the peroxo-I species converts to the peroxo-II species through a thermally activated process and that the peroxo-II species is the precursor to O2 dissociation on Pd(111).31,32,34 The three vibrational bands observed with HREELS correlate well with the distinct peaks observed in O2 TPD spectra, which lead to the original suggestion that desorption of the ω1, ω2, and ω3 species

Molecular Chemisorption of O2

Figure 3. 18O2 TPD spectra (heating rate ) 1 K s-1) obtained from a 3 ML Pd16O(101) thin film after chemisorbing various amounts of 18O2 at a surface temperature of 85 K. The initial 18O2 coverages are listed in the figure.

produces the R1, R2, and R3 states in TPD, respectively. However, subsequent studies provide evidence that desorption of the ω2 species produces both the R2 and the R1 states, while the more strongly bound ω1 species predominantly dissociates rather than desorbs during TPD.10,31 An implication is that the ω1 species would desorb at a temperature above that of the R1 peak if it did not readily dissociate. It is therefore possible that the ω1 species on Pd(111) has a surface binding strength that is comparable to or even higher than the binding strengths of the O2 species on PdO(101) that desorb above 200 K. Thus, on the basis of the TPD spectra obtained from O2-covered Pd(111) and PdO(101), we can conclude only that the majority of the O2 states on PdO(101) are more strongly bound than the superoxo and peroxo-I states on Pd(111). The data obtained in the present study does not provide information for determining the O2 binding configurations on the PdO(101) surface. However, it is conceivable that O2 molecules bind on PdO(101) in configurations that are qualitatively similar to those reported for O2 on Pd(111), such as superoxo and peroxo states. An interesting observation is that the separation between cus-Pd sites on PdO(101) is only about 11% greater than the lattice constant of Pd(111);27 so, O2 binding in a bidentate configuration on PdO(101) seems geometrically feasible. As observed for Pd(111), O2 molecules bound in different configurations on PdO(101) are likely to have different surface binding strengths and would be expected to desorb at different temperatures during TPD. However, prior studies by Kolasinski et al.10,31 demonstrate that the binding strength of O2 in a given configuration on Pd(111) can depend sensitively on the local environment as well, such as the O2 coverage. Thus, the multiple peaks observed in the O2 TPD spectra from PdO(101) could reflect variations in both the O2 binding configurations and a dependence of the O2 binding strength on the local environment. We note also that desorption of a small concentration of O2 molecules that are strongly bound to defects on the oxide surface could be partly responsible for the small features (γ1 and γ2) in the O2 TPD spectra. Spectroscopic and computational work are clearly needed to further characterize the bonding of O2 on the PdO(101) surface. Figure 3 shows 18O2 TPD spectra obtained after exposing the 3 ML PdO film held at 85 K to various doses of 18O2. The resulting 18O2 coverages are stated in the figure. Initially, 18O2

J. Phys. Chem. C, Vol. 112, No. 22, 2008 8327 desorbs in a broad feature that appears to consist of about three components (β1, γ2, γ1) centered at 250, 275, and 315 K, respectively. The β1 feature, initially centered at about 250 K, intensifies considerably and shifts toward lower temperature as the 18O2 coverage increases, its maximum appearing at 233 K at a coverage of 0.14 ML. The downward temperature shift of the β1 peak suggests that O2 binding to the surface weakens with O2 coverage, possibly because of repulsive interactions between adsorbed O2 molecules in the β1 state. The γ1 and γ2 desorption features also intensify with coverage but to a lesser extent than the β1 feature. In the spectrum obtained from 0.14 ML of 18O2, the γ1 peak is evident as a distinct shoulder at 315 K while superposition between the γ2 and β1 features appears to produce a single broad feature that is skewed toward high temperatures. As the coverage increases beyond 0.14 ML, a new desorption peak, labeled as β2, appears at 117 K. Unlike the β1 feature, the β2 peak temperature does not shift appreciably with increasing coverage, suggesting that interactions among chemisorbed O2 molecules have a negligible influence on desorption from the β2 state. Both the β1 and the β2 peaks intensify concurrently with increasing coverage until the 18O2 layer saturates at 0.29 ML. The development of the high and low temperature desorption features is similar to that observed previously on clean Pd(111).29,30 On the metal surface, the R1 desorption peak initially evolves with increasing coverage, and then the R1 and R2 states grow simultaneously beyond a critical coverage. While this may suggest similarities in the O2 binding states on the metal and oxidized Pd surfaces, the high temperature TPD feature obtained from the PdO film is clearly composed of at least three distinct features. This suggests greater variability in the bonding environments or configurations for O2 adsorbed on the PdO(101) film compared with the metal. Mixed Isotope Experiments. To determine whether the low temperature O2 desorption features arise from molecularly or atomically adsorbed species on the PdO surface, we conducted TPD experiments using coadsorbed 16O2 and 18O2. In these experiments, we first dosed 18O2 onto the PdO thin film at 85 K to generate an 18O2 coverage below 0.14 ML and thereby suppressed population of the β2 state. We then exposed the surface to a relatively large 16O2 dose and performed TPD while monitoring masses 32, 34, and 36 amu. Representative TPD spectra are shown in Figure 4. First, 16O18O evolution is immeasurable below 400 K, confirming that the low temperature desorption features indeed arise from molecularly adsorbed O2. The figure also shows the desorption signals above 500 K. Decomposition of the oxide produces the sharp 16O2 feature centered at 760 K as shown in recent work.26 The small amount of 16O18O desorbing as the 3 ML PdO film decomposes corresponds to approximately 0.005 ML of 18O. This is consistent with the natural abundance of 18O and hence indicates that O2 molecules chemisorbed on the PdO surface dissociate to a negligible extent during heating. This is not surprising considering that Pd(111) cannot be oxidized with O2 at partial pressures typical of dosing in UHV. The TPD data also reveals that O2 molecules interchange among the various adsorbed states at temperatures as low as 120 K. Specifically, from integration of the TPD data, we estimate that the initial exposures produced 18O2 and 16O2 coverages of 0.05 and 0.14 ML, respectively. On the basis of the data shown in Figure 3, the initial 0.05 ML of 18O2 molecules populates only the β1, γ1, and γ2 states, yet a small fraction of the 18O2 from the mixed layer desorbs in the β2 peak (Figure 4). Thus, the addition of 16O2 to the initial 18O2 layer causes

8328 J. Phys. Chem. C, Vol. 112, No. 22, 2008

Hinojosa et al.

Figure 4. TPD spectra of 16O2, 16O18O, and 18O2 obtained from a 3 ML Pd16O(101) thin film after sequential exposures to 18O2 followed by 16O2 at 85 K. The initial 18O2 and 16O2 coverages are estimated as 0.05 and 0.14 ML, respectively. Above 400 K, the 16O2 TPD spectrum is multiplied by a factor of 0.01 for clarity.

Figure 5. Saturation coverage of 16O2 obtained on Pd(111) at 85 K as a function of the initial 16O atom coverage. The initial 16O atom coverages were generated by exposing the Pd(111) sample to an 16O atom beam at a surface temperature of 500 K. The surface was then held at 85 K and exposed to 60 ML of 16O2.

some of the 18O2 molecules to populate the β2 state. Notice also that the intensity of the β1 feature relative to the γ features is higher in the 18O2 TPD spectrum of Figure 4 than that seen in the TPD spectrum obtained from a layer containing only 0.05 ML of O2 (Figure 3). This suggests that a fraction of the 16O2 molecules also displaces 18O2 molecules from the γ states. The interchange of O2 among adsorbed states appears to be relatively facile since some interchange must occur below the β2 desorption temperature of 120 K. However, differences in the β1 to β2 peak ratios in the 16O2 and 18O2 TPD spectra demonstrate that the interchange is not rapid enough to cause complete isotopic mixing. Interestingly, a similar isotopic mixing behavior has been reported for O2 on clean Pd(111).29,30 Molecular O2 Chemisorption As a Function of Initial Atomic Oxygen Coverage. To examine how the initial atomic oxygen phases on Pd(111) influence molecular O2 chemisorption, we conducted a series of O2 TPD experiments on surfaces with varying amounts of atomic oxygen. For these experiments, we first generated an atomic oxygen coverage on Pd(111) at 500 K using an 16O atom beam and then exposed the 16Ocovered surface held at 85 K to an 16O2 exposure of ∼60 ML, which is well beyond that needed to saturate the surface of the 3 ML PdO film with chemisorbed O2. Before discussing the results, it is useful to summarize the evolution of atomic oxygen states on Pd(111). Oxygen atoms initially arrange into a p(2 × 2) chemisorbed layer up to 0.25 ML,33 followed by formation of an ordered two-dimensional (2D) oxide, which saturates at approximately 0.7 ML.35,36 Prior studies show that the 2D oxide consists of a single layer of Pd and O atoms arranged on top of the Pd(111) surface.35,36 The 2D oxide has a Pd5O4 stoichiometry and is incommensurate with the underlying (111) lattice, and its structure does not match any lattice plane of crystalline PdO.36 Recently, Kan et al. have presented evidence that O atoms adsorb on top of the 2D oxide as the oxygen coverage increases beyond 0.7 ML and that these adsorbed O atoms react with the 2D oxide to produce threedimensional (3D) PdO clusters.26 The PdO clusters coexist with 2D oxide domains as the coverage increases from about 0.7 to 2 ML but then begin to coalesce and eventually produce a more continuous film once the coverage reaches about 3 ML.27 The oxidation rate at 500 K decreases significantly above 3 ML for

the O atom beam flux (∼0.02 ML s-1) employed in our experiments. Figure 5 shows that the saturation coverage of molecularly chemisorbed O2 depends sensitively on the initial atomic oxygen coverage on Pd(111) and hence the oxygen phases that are present on the surface. First, the precipitous decrease in the saturation O2 coverage at low initial O-atom coverages indicates that chemisorbed oxygen atoms inhibit the molecular chemisorption of O2, which is consistent with prior observations.32 For example, the O2 coverage drops from 0.66 ML on clean Pd(111) to 0.03 ML for an initial O-atom coverage of about 0.25 ML. The O2 coverage remains at only 0.03 ML as the initial O-atom coverage increases from 0.25 to about 1 ML, indicating that O2 chemisorbs negligibly on the 2D oxide as well. The O2 coverage then increases nearly linearly with initial O-atom coverage as PdO particles grow on the surface from 1 to 2.7 ML and rises more slowly in the O-atom coverage range from 2.7 to 3.4 ML, reaching a maximum of about 0.28 ML on the 3.4 ML oxide. The O2 uptake behavior shown in Figure 5 clearly reveals that the PdO(101) surface has a significantly higher affinity toward binding O2 molecules than both the 2D Pd5O4 oxide and domains with chemisorbed oxygen atoms. The 2D Pd5O4 oxide and the PdO(101) surface must therefore offer chemically distinct binding sites and likely exhibit different catalytic properties as well. Representative TPD spectra obtained from saturated O2 layers generated at 85 K are shown in Figure 6 as a function of the initial O-atom coverage. For O-atom coverages between 0.25 and about 1 ML, only a small feature at about 120 K is evident in the TPD spectra. This feature is broader than the β2 peak that populates at high coverage on the PdO film, though they appear at a similar temperature. By considering that such a small amount of O2 desorbs in this regime, this O2 may originate from defects in the chemisorbed O-atom layer or the 2D oxide. It is also possible that a fraction of this O2 evolves from the edges or backside of the crystal. As the amount of PdO increases with increasing O-atom coverage from 1 to 3.4 ML, O2 TPD features centered at 120 and 230 K simultaneously intensify. In fact, above an atomic oxygen coverage of 1.8 ML, the β1 to β2 intensity ratio remains approximately constant with increasing O-atom coverage. Since PdO clusters grow in the presence of

Molecular Chemisorption of O2

Figure 6. Saturation 16O2 TPD spectra (heating rate ) 1 K s-1) obtained from Pd(111) initially covered with various amounts of 16O atoms. The initial 16O-atom coverages and the corresponding saturation 16 O2 coverages are listed in the figure. The surface preparation is the same as that described in the caption of Figure 5.

2D oxide domains, the available surface area of the PdO increases with the total O-atom coverage. Thus, the TPD data suggests that O2 molecules populate the same binding states on PdO clusters of varying size, and that the total O2 coverage scales with the surface area of the PdO present at a given O-atom coverage. CO Oxidation The present study was motivated largely by the idea that molecularly chemisorbed O2 could be a reactive species in catalytic processes that occur on oxidized Pd at commercially relevant pressures, including the catalytic oxidation of CO and CH4. Our finding that a large fraction of O2 molecules bind more strongly on PdO(101) than Pd(111) supports the idea that molecularly chemisorbed O2 can exist in appreciable concentrations on oxidized Pd at high pressure and temperature and that it is therefore a viable reactant. To initially explore the reactivity of chemisorbed O2 on PdO(101), we examined the oxidation of CO on the 3 ML PdO thin film using TPRS. For this study, we first examined the reaction of only CO with the 16O oxidized surface and then studied the reaction of CO coadsorbed with 18O on the 16O oxidized surface. For each experiment, we first 2 oxidized the Pd(111) sample at 500 K with an 16O beam to reach an 16O coverage of 3 ML. We then exposed the resulting Pd16O surface held at 85 K to 0.01 L of 18O2, followed by a 26 ML dose of C16O. The same CO exposure was used in each experiment but resulted in different initial CO coverages depending on whether 18O2 was initially present on the PdO surface. Using the scaling factors discussed in Experimental Details, we estimate that the 26 ML CO exposure produced initial CO coverages of 0.18 and 0.35 ML, respectively, on the PdO surface with and without coadsorbed 18O2. Thus, the adsorbed 18O2 appears to suppress CO adsorption on the PdO surface, probably by blocking surface sites. While the CO coverages are approximate, it is interesting that the saturation CO coverage obtained on the clean PdO surface is equal to the surface concentration of cus-Pd sites of PdO(101). Finally, after preparing the adsorbed layers, we heated the sample at a linear rate of 1 K s-1 while monitoring the partial pressures of C16O, 16O , 18O , C16O , C16O18O, and C18O . 2 2 2 2

J. Phys. Chem. C, Vol. 112, No. 22, 2008 8329

Figure 7. TPRS spectra obtained from a 3 ML Pd16O(101) thin film with (a) C16O adsorbed at 85 K and (b) C16O and 18O2 coadsorbed at 85 K. The masses of C16O, 16O2, 18O2, C16O2, C16O18O, and C18O2 were monitored in each experiment. Estimates of the initial coverages of C16O and 18O2 are stated in the figures.

Figure 7 shows representative TPRS traces obtained in these experiments. The top panel shows the C16O2 desorption trace during TPRS with only adsorbed C16O, while the bottom panel shows the TPRS traces of C16O2, C16O18O, and 18O2 obtained during experiments with coadsorbed C16O and 18O2. The C16O2 desorption trace obtained without coadsorbed 18O2 exhibits three main features centered at about 117, 355, and 520 K (Figure 7a). In prior studies, we have found that desorption from the sample mounting wires, as well as the initially high transient heating rate, is largely responsible for the sharp peak at 117 K.15,17 Nevertheless, a reaction between CO and the oxide likely contributes to the trailing edge of this signal. The feature at 355 K likely originates from reactions between CO and O atoms of PdO, while the feature at 520 K is characteristic of reaction between CO and O atoms adsorbed on metallic sites or near vacancies produced as the oxide is reduced. Both C16O2 and C16O18O evolve from the 18O2 covered surface, indicating that PdO and chemisorbed 18O2 are active in CO oxidation during TPRS. Recall that chemisorbed O2 dissociates to a negligible extent on the PdO surface (Figure 4) so the presence of coadsorbed CO must facilitate cleavage of the O-O bond, resulting in the addition of 18O to adsorbed C16O. Importantly, we find that the evolution of C18O2 is negligible in these experiments, which implies that CO oxidation on PdO does not involve C-O bond cleavage. In addition to the initial sharp peak, the C16O2 spectrum obtained from the surface with adsorbed 18O2 exhibits a broad feature centered at 320 K and only low intensity above 500 K (Figure 7b). Since the initial CO coverage was lower in this experiment, the majority of the CO molecules likely reacted below 500 K thereby resulting in the diminution of the CO2 signal observed at 520 K in the experiment without coadsorbed 18O2. It is interesting that the intermediate C16O2 feature appears at 320 K in the presence of 18O2 but is centered at 355 K and is more asymmetric when CO is adsorbed alone on the oxide. Neglecting possible CO coverage effects, this may indicate that the presence of chemisorbed 18O2 influences the reaction of CO with the PdO. The C16O18O spectrum also exhibits a broad feature at about 320 K as well as a less intense feature below 200 K. In this case, the intensity below 200 K, while small, is broad and overlaps the range of temperature over which chemisorbed 18O2 desorbs, suggesting that this CO2 originates mainly from reaction

8330 J. Phys. Chem. C, Vol. 112, No. 22, 2008 on the sample surface rather than the mounting wires. Finally, we note that the highest C16O18O desorption rate occurs at 320 K and mainly overlaps the trailing edge of the 18O2 desorption feature. Estimates of the product yields show that the oxidation of CO is relatively efficient on both surfaces studied. For example, on the surface without coadsorbed O2, we estimate that 57% of the adsorbed CO molecules react with PdO to produce CO2. In the presence of adsorbed 18O2, approximately 42% of the CO molecules reacts with the PdO, while 14% reacts with 18O to produce C16O18O. Interestingly, the total fraction of CO molecules that reacts is nearly equal on both surfaces for the conditions examined. The yields show that fewer CO molecules react with 18O than 16O of the oxide. However, to compare the reactivity of PdO with that of O2 chemisorbed on the oxide, it is necessary to quantify the amount of 16O and 18O atoms that are present at the surface. From the desorption yields of C16O18O and 18O2, we estimate that the initial 18O2 coverage was 0.16 ML of 18O atoms. This estimate implies that about 15% of the available 18O atoms evolved from the surface in the C16O18O reaction product and hence that 30% of the adsorbed 18O2 molecules participated in the reaction. We consider the PdO(101) structure to obtain an estimate of the atomic oxygen concentration in the PdO surface layer. The nonpolar PdO(101) surface layer contains 0.7 ML of O atoms,27,28 but the O atoms are inequivalent. In particular, half of the O atoms are 3-fold coordinated with Pd atoms and reside closest to the vacuum-solid interface, while the remaining O atoms are coordinately saturated and lie slightly below the surface layer (Figure 1). Thus, the concentration of “available” 16O atoms of the Pd16O surface was between 2.2 and 4.5 times greater than the 18O coverage during the TPRS experiments. Since the yield of C16O2 was approximately three times greater than the C16O18O yield, the probability for CO to react with an 18O atom originating from a molecularly adsorbed 18O2 species appears to be similar to the probability for reaction with an oxygen atom of the Pd16O surface for the conditions studied. These initial experiments clearly demonstrate that chemisorbed O2 molecules actively participate in the oxidation of CO on PdO(101). However, more thorough studies are needed to elicit the detailed aspects of this reactive interaction and to quantify the reactivity of various surface oxygen species. For example, it is conceivable that coadsorbed CO and O2 react directly on the PdO surface to produce CO2 and an O atom. Another possibility, however, is that the reaction of CO with the oxide creates oxygen vacancies that are effective in activating the O2 bond, even though O2 dissociation on the oxide is otherwise negligible under UHV conditions (Figure 4). Overall, the findings of these preliminary TPRS experiments provide motivation for conducting further investigations of the reactivity of O2 molecules on oxidized transition metal surfaces, particularly considering the possibility that O2 chemisorbed on oxide surfaces is an active catalytic species in commercial applications. Summary We used TPD and TPRS to investigate the desorption and reactivity of molecularly chemisorbed O2 on a PdO(101) thin film grown on Pd(111) in UHV. Our results show that O2 molecules chemisorb readily on the PdO(101) surface, reaching a saturation coverage of about 0.27 ML at 85 K. Experiments with coadsorbed 16O2 and 18O2 further reveal that O2 molecules dissociate negligibly on the PdO surface

Hinojosa et al. under the conditions examined. The desorption of molecularly chemisorbed O2 from the PdO(101) film gives rise to two main features centered at 117 and 227 K as well as smaller features at 275 and 315 K, which is indicative of multiple O2 binding states on the PdO surface. A large fraction (∼75%) of the molecularly chemisorbed O2 on the PdO(101) surface is more strongly bound than O2 chemisorbed on Pd(111) at 85 K and saturation of the respective O2 layers. Interestingly, we find that O2 molecules chemisorb only in small quantities on the ordered 2D Pd5O4 oxide on Pd(111), demonstrating that the 2D oxide and PdO(101) are chemically distinct with respect to binding O2. Finally, TPRS spectra provide clear evidence that both PdO(101) and molecularly chemisorbed O2 on the oxide are active in the oxidation of adsorbed CO, with the atomic and molecular species exhibiting similar activities in this reaction. This final observation warrants further study as it suggests that molecularly chemisorbed O2 could play a role in catalytic oxidation processes occurring at high pressure. Acknowledgment. We gratefully acknowledge financial support provided by the Department of Energy, Office of Basic Energy Sciences, Catalysis and Chemical Transformations Division through Grant DE-FG02-03ER15478. References and Notes (1) Carstens, J. N.; Su, S. C.; Bell, A. T. J. Catal. 1998, 176, 136. (2) Datye, A. K.; Bravo, J.; Nelson, T. R.; Atanasova, P.; Lyubovsky, M.; Pfefferle, L. Appl. Catal., A 2000, 198, 179. (3) Farrauto, R. J.; Lampert, J. K.; Hobson, M. C.; Waterman, E. M. Appl. Catal., B 1995, 6, 263. (4) Gabasch, H.; Hayek, K.; Klotzer, B.; Unterberger, W.; Kleimenov, E.; Teschner, D.; Zafeiratos, S.; Havecker, M.; Knop-Gericke, A.; Schlogl, R.; Aszalos-Kiss, B.; Zemlyanov, D. J. Phys. Chem. C 2007, 111, 7957. (5) Hoflund, G. B.; Hagelin, H. A. E.; Weaver, J. F.; Salaita, G. N. Appl. Surf. Sci. 2003, 205, 102. (6) Lyubovsky, M.; Pfefferle, L. Catal. Today 1999, 47, 29. (7) McCarty, J. G. Catal. Today 1995, 26, 283. (8) Ribeiro, F. H.; Chow, M.; Dallabetta, R. A. J. Catal. 1994, 146, 537. (9) Gabasch, H.; Knop-Gericke, A.; Schlogl, R.; Borasio, M.; Weilach, C.; Rupprechter, G.; Penner, S.; Jenewein, B.; Hayek, K.; Klotzer, B. Phys. Chem. Chem. Phys. 2007, 9, 533. (10) Kolasinski, K. W.; Cemic, F.; Demeijere, A.; Hasselbrink, E. Surf. Sci. 1995, 334, 19. (11) Nakai, I.; Kondoh, H.; Shimada, T.; Resta, A.; Andersen, J. N.; Ohta, T. J. Chem. Phys. 2006, 124, 224712. (12) Gabasch, H.; Unterberger, W.; Hayek, K.; Klotzer, B.; Kleimenov, E.; Teschner, D.; Zafeiratos, S.; Havecker, M.; Knop-Gericke, A.; Schlogl, R.; Han, J. Y.; Ribeiro, F. H.; Aszalos-Kiss, B.; Curtin, T.; Zemlyanov, D. Surf. Sci. 2006, 600, 2980. (13) Oh, S. H.; Hoflund, G. B. J. Phys. Chem. A 2006, 110, 7609. (14) Oh, S. H.; Hoflund, G. B. J. Catal. 2007, 245, 35. (15) Gerrard, A. L.; Weaver, J. F. J. Chem. Phys. 2005, 123, 224703. (16) Shumbera, R. B.; Kan, H. H.; Weaver, J. F., submitted. (17) Shumbera, R. B.; Kan, H. H.; Weaver, J. F. J. Phys. Chem. C 2008, 112, 4232. (18) Zheng, G.; Altman, E. I. J. Phys. Chem. B. 2002, 106, 1048. (19) Hendriksen, B. L. M.; Frenken, J. W. M. Phys. ReV. Lett. 2002, 89, 046101. (20) Ackermann, M. D.; Pedersen, T. M.; Hendriksen, B. L. M.; Robach, O.; Bobaru, S. C.; Popa, I.; Quiros, C.; Kim, H.; Hammer, B.; Ferrer, S.; Frenken, J. W. M. Phys. ReV. Lett. 2005, 95, 255505. (21) Hendriksen, B. L. M.; Bobaru, S. C.; Frenken, J. W. M. Surf. Sci. 2004, 552, 229. (22) Gerrard, A. L.; Chen, J. J.; Weaver, J. F. J. Phys. Chem. B 2005, 109, 8017. (23) Shumbera, R. B.; Kan, H. H.; Weaver, J. F. Surf. Sci. 2007, 601, 235. (24) Shumbera, R. B.; Kan, H. H.; Weaver, J. F. Surf. Sci. 2007, 601, 4809. (25) Kan, H. H.; Shumbera, R. B.; Weaver, J. F. J. Chem. Phys. 2007, 126, 134704. (26) Kan, H. H.; Shumbera, R. B.; Weaver, J. F. Surf. Sci. 2008, 602, 1337.

Molecular Chemisorption of O2 (27) Kan, H. H.; Weaver, J. F. Surf. Sci. Lett. 2008, in press. (28) Rogal, J.; Reuter, K.; Scheffler, M. Phys. ReV. B 2004, 69, 075421. (29) Guo, X. C.; Hoffman, A.; Yates, J. T. J. Chem. Phys. 1989, 90, 5787. (30) Matsushima, T. Surf. Sci. 1985, 157, 297. (31) Kolasinski, K. W.; Cemic, F.; Hasselbrink, E. Chem. Phys. Lett. 1994, 219, 113. (32) Imbihl, R.; Demuth, J. E. Surf. Sci. 1986, 173, 395. (33) Zheng, G.; Altman, E. I. Surf. Sci. 2000, 462, 151.

J. Phys. Chem. C, Vol. 112, No. 22, 2008 8331 (34) Nolan, P. D.; Lutz, B. R.; Tanaka, P. L.; Mullins, C. B. Surf. Sci. 1998, 419, L107. (35) Gabasch, H.; Unterberger, W.; Hayek, K.; Klotzer, B.; Kresse, G.; Klein, C.; Schmid, M.; Varga, P. Surf. Sci. 2006, 600, 205. (36) Lundgren, E.; Kresse, G.; Klein, C.; Borg, M.; Andersen, J. N.; De Santis, M.; Gauthier, Y.; Konvicka, C.; Schmid, M.; Varga, P. Phys. ReV. Lett. 2002, 88, 246103.

JP800216X