Determination of O and OH Adsorption Sites and Coverage in Situ on

At low coverages OH adsorbs primarily in 1-fold coordinated atop sites. As the coverage ... Shraboni Ghoshal , Qingying Jia , Jingkun Li , Fernando Ca...
2 downloads 0 Views 481KB Size
Download PDF
8076

J. Phys. Chem. B 2005, 109, 8076-8084

Determination of O and OH Adsorption Sites and Coverage in Situ on Pt Electrodes from Pt L23 X-ray Absorption Spectroscopy M. Teliska,† W. E. O’Grady,‡ and D. E. Ramaker*,† Department of Chemistry, The George Washington UniVersity, Washington, D.C. 20052, and Chemistry DiVision, NaVal Research Laboratory, Washington, D.C. 20375 ReceiVed: January 11, 2005; In Final Form: March 1, 2005

The adsorption of atomic oxygen and hydroxide on a platinum electrode in 0.1 M HClO4 or H2SO4 electrolytes was studied in situ with Pt L2,3 X-ray absorption spectroscopy (EXAFS and XANES). The Pt L3 edge absorption data, µ, were collected at room temperature in transmission mode on beamline X-11A at the National Synchrotron Light Source using a custom built cell. The Pt electrode was formed of highly dispersed 1.5-3 nm particles supported on carbon. A novel difference procedure (∆µ ) µ(O[H]/Pt) - µ(Pt)) utilizing the L3 XANES spectra at different applied voltages was used to isolate the effects of O[H] (O or OH) adsorption in the XANES spectra. The ∆µ results are compared with results obtained from real-space full-multiple scattering calculations utilizing the FEFF8 code on model clusters. The experimental results, when compared with theoretical calculations, allow the adsorption site to be identified. At low coverages OH adsorbs primarily in 1-fold coordinated atop sites. As the coverage increases, O binds in the bridge/fcc sites, and at still higher coverages above 1.05 V RHE, O adsorbs into a higher coordinated n-fold or subsurface site, which is thought to be the result of Pt-O site exchange and oxide formation. These results are similar to those found in the gas phase. Direct specific adsorption of bisulfate anions in H2SO4 is spectroscopically seen in both the EXAFS and XANES data and is seen to impede oxygen adsorption consistent with previous reports.

Introduction The determination of the adsorption binding site, and its change with coverage, of atomic oxygen or hydroxyl groups, O[H] (O or OH indicated collectively by O[H]), on platinum is important to the fundamental understanding of the role of oxygen in many electrochemical and catalytic reactions. The adsorption of O[H] is, in some respects, more complex than that of atomic hydrogen, H, due to the existence of a subsurface oxygen species, and ultimately oxide formation; although even the adsorption of H on Pt goes through several binding site transformations.1 Although numerous studies in the gas phase have been reported for O[H] adsorption, there has been little information obtained in situ in an electrochemical cell. In this study we report detailed atomic level information on O[H] bindings sites as obtained from X-ray near edge absorption spectroscopy (XANES) data on a Pt electrode under working conditions in an electrochemical cell. Numerous studies of both O and OH adsorption on single crystal surfaces have been reported utilizing surface analytical techniques that are readily available in the gas phase. These techniques include scanning tunneling microscopy (STM),2 highresolution electron energy loss spectroscopy (HREELS),3 electron energy loss spectroscopy (EELS),3 and X-ray photoelectron spectroscopy (XPS), the latter both at the valence band and core level. These results show that O adsorption on platinum depends on a variety of factors including the crystal face, the coverage, the temperature and the presence of other adsorbed species. The nature of hydroxyl species resulting from the oxidation of water * Corresponding author. Telephone: 202-994-6934. Fax: 202-994-5873. E-mail: [email protected]. † George Washington University. M.T. e-mail: [email protected]. ‡ Naval Research Laboratory. E-mail: [email protected].

on platinum metal has received much attention; however, experimental and theoretical findings have yielded conflicting results.4 The lack of detailed geometric information regarding O[H] binding sites in an aqueous environment results in part from difficulties with the normal optical techniques because of absorption and interferences arising from the bulk electrolyte and complex electrochemical double layer. Over the past two decades, several techniques have been developed to avoid this unwanted absorption, or at least cancel it out. However, a good in situ spectroscopic technique that can provide site symmetry information is still needed. A promising technique that may be used to study adsorbate-metal surface interactions in situ is X-ray absorption spectroscopy (XAS). XAS is one of the few techniques that can be applied in situ because it utilizes X-rays, which although absorbed by the bulk solution at some energies, do not interfere with the Pt L23 edge. Previous X-ray absorption near edge spectroscopy (XANES) data on both the Pt-H and Pt-O systems have shown an increase in the white line area in the potentials where the H or O[H] is electrochemically adsorbed, indicative of a quantitative measure of the adsorbate coverage.5-8 Recent advances in XANES analysis has allowed for the qualitative determination of specific H binding sites in situ, i.e., atop, vs higher-fold bridged or fcc sites.1,9 These techniques are now utilized (in both Extended X-ray Absorption Fine Structure (EXAFS) and XANES) to provide adsorbate binding site information for O[H] on Pt. Here, an in situ examination of the cathodic oxidation of Pt is reported. Experimental Section Sample Preparation. Carbon supported (Vulcan XC 72) Pt/C electrocatalysts obtained from ETEK, Inc. (a division of De

10.1021/jp0502003 CCC: $30.25 © 2005 American Chemical Society Published on Web 04/02/2005

O Adsorption on Pt by XAS

J. Phys. Chem. B, Vol. 109, No. 16, 2005 8077

Nora, NA, Somerset, NJ) was utilized to prepare the electrodes in a procedure described previously.1 The nominal metal loading was 10 wt % Pt on the Vulcan carbon. The cell used in these measurements was similar to that described previously.1 The Pt/C electrode was flooded with 0.1 M HClO4 or 0.1 M H2SO4 and held in a PTFE gasket. The counter electrode was an uncatalyzed high-surface-area carbon electrode flooded with the same electrolyte as the Pt electrode and held in a PTFE gasket. Nafion or filter paper, saturated with the same electrolyte, was used as separators. The reference electrode was a Pd/H electrode, and all the data are reported relative to the reversible hydrogen electrode (RHE). The potential in the cell was controlled with a PARC 273 potentiostat. X-ray Absorption Experiments. The Pt XAS data were collected at room temperature in transmission mode on beamline X-11A of the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory. The NSLS storage ring operated at 2.52 GeV beam energy with ring currents between 40 and 200 mA. Three gas flow ionization detectors were used to monitor the intensities of the incident (I0) and transmitted X-rays (I) through the sample and the reference detector (Iref). The Si(111) double crystal monochromator was detuned by 30% at 200 eV above the Pt L3 edge (11564 eV) to minimize the presence of higher harmonics in the beam. The energy calibration was achieved by recording the XAS of a 7 µm thick Pt foil simultaneously with the sample. EXAFS and XANES Analysis. The EXAFS analysis was performed using the WINXAS code.10 The preedge background was removed using a linear polynomial and the many-body So2 factor fixed at 0.934, the value obtained using the FEFF8 code. The postedge background was removed using the normal spline techniques and smoothing criteria that leave the atomic XAFS in the χ function as described previously.11-14 Reference phase and amplitude functions for the Pt-Pt scattering were obtained from FEFF8 calculations on the Pt6 cluster discussed below. The XANES analysis procedure used here is similar to that described previously for isolation of the Pt-H scattering contributions.1,13,15-17 A brief summary is given here because significant modifications to that original method have been made for this work on O adsorption. The absorption coefficient, µ, is given as µ ) µo(1 + χ), where µo is the rather smooth background, and χ contains the EXAFS oscillations. The contributions to µ arising from the adsorbed O is isolated by taking the difference between the µ obtained from a reference, labeled Pt, and the µ in the presence of the adsorbed O, labeled O/Pt, hence ∆µ ) µ(O/Pt) - µ(Pt). In practice the reference is usually taken to be a relatively “clean” surface, namely the electrode at 0.54 V. In the double layer region around 0.54 V, there is no adsorbed H, O[H], or adsorbed ClO4- anions, making this potential the best for use as the reference. Although bisulfate ions are known to be directly absorbed on the Pt electrode in H2SO4, and water molecules are nearby in the double layer, these are not visible in the XANES below 0.6 V, apparently because these bisulfate ions and water are mobile or not specifically adsorbed (i.e., not having a specific bond length or site) on the Pt atoms at the surface. Below we will show that scattering from specifically adsorbed bisulfate is visible, but this occurs only at potentials well above 0.6 V when other ions such as OH are on the surface. The difference ∆µ can be expressed as9

∆µ ) ∆µo + ∆[µo χPt-Pt] + µo,O χPt-O

(1)

The first term includes the O induced changes in the AXAFS (∆µo ) µo,O - µo,ref). The second term reflects the changes in

the Pt-Pt scattering ∆[µo∆χPt-Pt] and exhibits the largest change with adsorption site symmetry and therefore allows a determination of the absorption sites. The third term is the additional Pt-O scattering (µo,OχPt-O) arising from the O adsorption. All XANES spectra were carefully energy calibrated before subtraction by aligning the Pt foil reference spectra taken simultaneously at each potential, and then shifting the sample spectra according to the foil data. This energy calibration eliminates any shifts due to drift of the photon beam, etc. Other than this careful energy calibration, the spectra were not aligned as was done in some of our previous work for H,15-17 although the procedure used here is comparable to that used for H adsorption on a Pt electrode.1 This modification to that earlier technique was necessary because, unlike the case for H adsorption, O adsorption does modify the L2 edge. Therefore alignment of the L2 edges, a technique used previously to remove the effects of core-hole screening, is not a satisfactory technique here. However, because the particles are relatively larger in this work (1.5-3.5 nm compared to 0.5-1.5 nm in the earlier work), the core-hole screening differences between µ(O/Pt) and µ(Pt) are negligible (i.e., the core-hole attraction is negligible in either case), and this allows a direct subtraction of the calibrated data without further alignment, as described above. Finally, one further aspect needs to be identified when the difference ∆µ is taken. The ∆µ differences taken in the potential region between 0.54 and 0.80 V (i.e., small oxygen coverage) are extremely small, reflecting magnitudes of 1-2% of the total XANES step height. Under these circumstances, the ∆µ difference across the step edge around 0 eV can reflect incomplete cancellation of the µo; i.e., the difference may include “step” contributions other than those in eq 1. These “step” contributions usually appear within 20 eV of the edge at 0 eV and are believed to arise from a change in the arctan cutoff rate at the Fermi level. This arctan cutoff normally goes as arctan[(E - Eo)/Γ], and apparently the electrode charging is able to alter the effective width or experimental broadening, Γ, from that at 0.54 V used as a reference. This then introduces a small background approximately equal to

δbackground (V) ) arctan[(E - Eo)/ΓV] arctan[(E - Eo)/Γref] (2) This background can be subtracted out by optimally varying ΓV and Γref usually around 8 or 9 eV. This value of 8 or 9 eV reflects a Pt lifetime core width of around 5 eV and experimental resolution of around 3 eV.18 Generally we find ΓV is slightly greater than Γref suggesting that charging or some other process above 0.54 V increases the width by a small amount, although the exact reason for this is unknown at this time. It may result from unequal charging of the Pt atoms due to the roughness of the surface (or Pt on edges or corners vs faces), which would broaden the spectrum. It is visible in some of the data when the ∆µ is extremely small (i.e., small O coverage) and has no consequence at higher O coverage. When its characteristic line shape is evident in the ∆µ, eq 2 is used to subtract it out. FEFF8 Calculations. FEFF8 calculations20 were performed to interpret the ∆µ spectra using a series of Pt6 and Pt6Ox clusters with geometries shown in Figure 1. This geometry is preferred19 because it contains both fcc and hcp sites, and we have used it previously for H adsorption with success.1,9 Thus throughout this work the indicated Pt6 cluster is utilized, with the Pt-Pt distance at 2.77 Å. The FEFF8 code performs ab initio self-consistent field, realspace, full mutiple- scattering calculations.21 It is known that

8078 J. Phys. Chem. B, Vol. 109, No. 16, 2005

Teliska et al.

Figure 1. Illustration of Pt6 clusters used for FEFF8 calculations and the 1-fold atop, 2-fold bridged, 3-fold fcc, and 4-fold hcp sites.

Figure 3. Fourier transform of the EXAFS data (1.5 < k < 15.5 Å-1, k2) and two shell (Pt-Pt and Pt-O) fit in R space (1.3 < R < 3.5 Å) at 0.84 V in 0.1 M HCLO4. The individual Pt-Pt and Pt-O components are also shown.

In all of the calculations, the Pt-Pt distance was 2.77 Å and the Pt-O distance was 2.0 Å, typical of that found in bulk Pt and Pt oxides.11-17 The ∆µ difference spectra are obtained by taking the difference µ(Pt6Ox) - µ(Pt6), giving theoretical results exactly comparable to that obtained experimentally. Comparison of these theoretical results, obtained by placement of O in the atop, bridged, 3-fold fcc, or 4-fold hcp sites as shown in Figure 1, with the experimental results, enables a determination of the adsorption sites. Results Figure 2. Current vs potential (volts vs the reversible hydrogen electrode, RHE) in the oxygen region for a Pt electrode in 0.1 M H2SO4. The potentials where the XANES data were taken are also indicated, and the voltage at 0.54 V RHE, which is used as the reference or the “clean” electrode, is highlighted. The horizontal lines show the regions where the experimental ∆µ spectra primarily reflect O[H] binding in the atop (solid), bridge/fcc (dashed) and subsurface (dotted) sites on the Pt surface, as indicated by comparison to the theoretically calculated ∆µ spectral signatures.

in the FEFF code, the results are quite dependent on the potentials used in the code. Previously it was found that the Dirac-Hara potential with an imaginary part of 5 eV was optimal for describing the Pt-Pt and Pt-O scattering;11,12 however, the emphasis in that optimization procedure was placed on the 25-150 eV region. Here we are interested in the region below 40 eV. Although a more complete analysis in this energy region is required, the Hedin-Lundquist potential is preferred and used in this work to describe the Pt-O scattering following the work of Ankudinov et al.20 and our previous work on Pt-H scattering.9

Figure 2 shows the cyclic voltammogram for a Pt/C electrode recorded in 0.1 M H2SO4. The potentials where the XAS data were recorded in H2SO4 and HClO4 are indicated in the figure. EXAFS Analysis. Results from EXAFS analyses of the data at each potential in the oxygen region are presented in Table 1. Figure 3 shows experimental data and a theoretical fit in R space at 0.80 V. Recently published results1 comparing the Fourier transform of the experimental EXAFS at 0.54 and 0.80 V show the near disappearance of the Pt-O peak around 1.5 Å at 0.54 V confirming that at 0.54 V the electrode is relatively clean of adsorbed oxygen species. Table 1 shows a decrease in Pt-Pt coordination as the cluster is oxidized above 0.54 V consistent with previous EXAFS results for the oxidation of Pt.7 This will be discussed further below. The magnitudes of NPt-Pt around 0.54 V indicate that the Pt particles are around 1.5-2.5 nm, moving up to even 3.5 nm in H2SO4, on the basis of model cluster calculations assuming spherical particles.21 The larger NPt-Pt in H2SO4 suggests a more spherical cluster compared to HClO4; this will be further discussed elsewhere.22

TABLE 1: Summary of EXAFS Resultsa potential (V, RHE)

NPt-Pt ∆N ) 0.3b

RPt-Pt (Å) ∆R ) 0.02b

σ2 (Å2)

0.54 0.80 1.00 1.05 1.14

8.30 7.23 5.56 4.95 3.61

2.73 2.74 2.74 2.73 2.72

0.007 0.007 0.007 0.007 0.007

0.54 0.74 0.94 1.14 1.24

10.97 11.20 9.54 8.23 7.81

2.75 2.75 2.75 2.75 2.76

0.007 0.007 0.007 0.007 0.007

NPt-O ∆N ) 0.3b

RPt-O ∆R ) 0.02b

σ2 (Å2)

Eo (eV)

(a) 0.1 M HCLO4 1.40 1.95 3.02 0.95 0.41

0.76 1.53 1.61 1.52

1.99 2.00 1.97 2.01

0.005 0.005 0.005 0.005

-3.45 0.84 -0.90 2.24

(b) 0.1 M H2SO4 3.01 3.12 2.83 1.61 3.67

0.82 0.76 1.04 1.56

2.28 1.96 1.94 1.96

0.005 0.005 0.005 0.005

8.01 -5.59 -4.48 1.01

Eo (eV)

a So2 fixed at 0.934 as calculated by FEFF8. b Although the absolute values of ∆N and ∆R may be larger than that indicated, the variation in the values of N and R with potential are believed to be meaningful down to the values indicated.

O Adsorption on Pt by XAS

Figure 4. XANES data (µ) taken in 0.1 M HClO4 at the potentials indicated, normalized to 1.0 at 50 eV and on an energy scale relative to the bulk Pt edge. These spectra were carefully energy calibrated but were otherwise not aligned or shifted.

Figure 5. ∆µ difference spectra µ(x.x) - µ(0.54 V), at the indicated potentials for XANES data taken in 0.1 M HClO4.

The EXAFS results in Table 1 were obtained by fixing the Debye-Waller (σ2) parameter at an optimum value for all potentials to make the resultant NPt-Pt more meaningful. The large linear dependence between N and σ2 gives a large variation in N if σ2 is allowed to vary. The Pt-X distances were allowed to vary at each potential but within experimental uncertainty appear to remain the same with applied voltage except for the large Pt-O distance of 2.28 Å at 0.74 V RHE in H2SO4. This large Pt-O distance may reflect directly adsorbed bisulfate anions at this potential where the Pt-O distance actually reflects the Pt-Obisulfate distance. This will be evident also in the XANES data to be discussed below. XANES Analysis. Figure 4 shows the L3 XANES data for a Pt/C electrode at several voltages in 0.1 M HClO4. The L3 edges as a function of potential change significantly as the Pt-O bonds are formed or removed from the surface. Clearly, the changes in these spectra are much larger than the noise level in the data. Figures 5 and 6 show the ∆µ difference spectra using the 0.54 V spectrum as the reference spectrum (i.e., using the data in Figure 4 in the case of HClO4 and similar data for H2SO4). A systematic trend is seen in these data; the amplitude generally increases with oxygen coverage (i.e., increase in potential) and the spectral line shape changes (i.e. shifts particularly around 5-15 eV) somewhat, suggesting different binding sites are being covered. FEFF8 Results. Figure 7 shows ∆µ results from theoretical FEFF8 calculations for O in the atop, bridged, fcc, and hcp sites on the Pt6 clusters shown, along with the individual components as defined in eq 1. It is most interesting to compare how these components change with O binding site. Further, because we

J. Phys. Chem. B, Vol. 109, No. 16, 2005 8079

Figure 6. ∆µ difference spectra µ(x.x) - µ(0.54 V), at the indicated potentials for XANES data taken in 0.1 M H2SO4.

Figure 7. Theoretical ∆µ ) µ(Pt6Ox) - µ(Pt6) difference “signatures” (i.e., spectral line shapes) obtained from the FEFF8 calculations on the clusters indicated in Figure 1, along with the indicated components as defined in eq 1.

have recently reported exactly similar calculations for H binding sites on the same Pt6 cluster,1 it is interesting to compare results for H and O binding. This comparison is organized in Table 2. As might be expected, the Pt-O scattering is much larger compared to Pt-H scattering, but in both instances the Pt-X (X ) O or H) spectral shape is relatively unchanging with binding site. In contrast, the change in Pt-Pt scattering varies significantly with binding site, negative for all binding sites, but increasing in magnitude and shifting to lower energy with increasing coordination n. This negative contribution has previously been attributed to a weakening of the Pt-Pt bonds below the adsorbate, and as the calculations show this weakening increases with n for both O and H bonding. This Pt-Pt bond weakening has been referred to as “d-orbital frustration”23 and produces a negative contribution in the ∆µ signature because less Pt-Pt scattering occurs under this bond weakened condition than in the “clean” case. The change in “atomic” (AXAFS) with binding site is a bit more complicated. It goes from small and negative for atop binding (suggesting small charge transfer from the adsorbate to the Pt) and then goes positive and generally increases in magnitude with n (suggesting increased charge transfer from the Pt to the adsorbate with increased Pt-X coordination). The AXAFS contribution is generally much larger for Pt-O bonding compared to Pt-H bonding as might be expected due to the more ionic bonding with O. Previously1 for Pt-H bonding, the change in AXAFS was ignored because we found better

8080 J. Phys. Chem. B, Vol. 109, No. 16, 2005

Teliska et al.

TABLE 2: Comparison of O and H Components in Theoretical ∆µ Signature with Binding Site component

O binding

H binding

Pt-X scattering, µo,XχPt-X change in Pt-Pt, ∆[µoχPt-Pt] change in AXAFS, ∆µo

relatively unchanging with n increasingly negative with n small & negative for atop; larger, positive and varying in magnitude with n

relatively unchanging with n; much smaller than for O increasingly negatively with n small & negative for atop; slightly larger and positive for fcc

Figure 8. Comparison of theoretical signatures. These signatures are generally moved by 2-5 eV for optimal alignment with experimental data.

agreement with experiment when the AXAFS contribution was deleted. This was justified because when the AXAFS arises from a delocalized inductive mechanism active for a covalent Pt-H bond, the theoretical AXAFS is a bit exaggerated by using such small Pt6 clusters. However, with the more ionic Pt-O bonding, the AXAFS contribution is larger and arises predominantly from the more local space-field mechanism. Thus it should be modeled appropriately with the small clusters used here, and we leave it in the total ∆µ signature for the Pt-O case. A much more complete discussion of the inductive vs space field affects on the AXAFS has been given elsewhere.11 Figure 8 compares the total ∆µ “signatures” for the different binding sites. The ∆µ signature for both atop and n-fold bonded oxygen are similar; however, the maximum in the ∆µ signature for the atop O occurs at around 2-4 eV, lower than that for the n-fold bonded O, making it easy to distinguish the atop adsorption site. Calculations for atop OH (not shown) are nearly identical to atop O because H is a weak scatterer. The difference between the n-fold (n > 1) sites is small; therefore we cannot distinguish between the higher n-fold binding sites. Summarizing, the theoretical ∆µ signatures clearly change (i.e., the maximum shifts from near zero to 3 eV) between the atop and n-fold adsorption site, enabling the assignment of adsorption sites in the experimental data by comparing the theoretical signatures with experiment. The trends with binding site are remarkably similar to that found previously for Pt-H bonding. It should be noted that these theoretical signatures do change some with O coverage (i.e., x in the Pt6Ox cluster), Pt-O bond length, and Pt6 cluster geometry; thus the signatures given in Figure 8 are representative only. In general, these theoretical signatures have to be shifted by around 2-4 eV for optimal agreement with experiment. This occurs because the FEFF8 edge energy is not always in perfect agreement with experiment, and the edge in the experimental data and theory are not defined in exactly the same manner. Nevertheless, the O binding site appears to be the dominant factor, enabling a determination of the O[H] binding sites. Discussion Oxygen Adsorption on Pt in the Gas Phase. Oxygen adsorption as mentioned in the Introduction has been extensively studied in the gas phase with a variety of surface analytical

techniques. Due to the complex electrochemical double layer, we expect that some differences may exist between the gas phase and the electrochemical phase. A brief review of some of the gas phase data is given in the Appendix to enable a comparison with the electrochemical results to be determined below. These data indicate that when O is adsorbed onto the surface it initially populates the bridged sites at surface edges, and as the coverage increases, oxygen adsorbs in the 3-fold sites both near the edges and on the flat faces of the platinum cluster. At higher coverage, and after a place exchange mechanism forming an oxide-like layer, O can also desorb from bridged-bonded sites near edges and from 3-fold sites, but coming off at a lower temperature due to Pt-O surface bond weakening caused by the subsurface O. Recently we24 reported ADF calculations25 on a Pt6 cluster like that in Figure 1 and examined the OH and O binding preferences on this cluster. OH clearly prefers to bond in an atop position, consistent with its preference to be singly coordinated, and O prefers to bond in a bridged position consistent with its preference to be 2-fold coordinated. These results are consistent with very recent DFT calculations on similar clusters,26,27 with STM results and DFT calculations reported by Janin19 on Pt(110) single crystals, and also with calculated vibrational frequencies on Pt(111) by Feibelman.28 Therefore, we conclude that at low coverages, oxygen prefers to be bridge-bonded near cluster edges and therefore prefers at least 2-fold coordination, and OH prefers atop adsorption, or 1-fold coordination. Although the ∆µ technique is not able to distinguish OH from O adsorption directly because the H scattering is negligible, we attribute adsorption in the atop site, when visible with the ∆µ technique, to OH adsorption and that in the n-fold (bridged or fcc) sites to O adsorption in this work. Comparison of ∆µ: Theory and Experiment. Figure 9 shows the experimental line shapes obtained at 0.80, 1.00 and 1.05 V RHE in HClO4 and Figure 10 at 1.14 and 1.24 V in H2SO4. These line shapes are similar and the good agreement with the (nearly indistinguishable) theoretical signatures for n-fold O clearly shows that the O is bound in these sites at these potentials and coverages. On the basis of the conclusions above from the gas-phase data, this is not a surprise, and it suggests that the water of hydration is not affecting the adsorption site. An examination of the HClO4 data at 1.14 V in Figure 5 suggests a changing line shape (increasingly negative contribution around 0 eV), which we attribute to the formation of subsurface O expected at potentials certainly above 1.05 V and perhaps much lower in HClO4. This subsurface O has an increased coordination and is therefore consistent with Figure 8, showing a systematic upward shift in ∆µ with n; ∆µ should shift to still higher energy with a resultant negative contribution around 0 eV. Figure 11 compares the decrease in the coordination number NPt-Pt and increase of NPt-O from the EXAFS results in Table 1, along with the maximum amplitudes of the ∆µ from Figures 5 and 6. Note that the ∆µ amplitudes and NPt-O track quite nicely together with potential for both H2SO4 and HClO4 as one would expect, and that the sharp rise occurs at much lower potential in HClO4 (0.95 V RHE) compared with a more gradual increase in H2SO4 at 1.10 V RHE. This is well-known29-31 and

O Adsorption on Pt by XAS

Figure 9. Comparison of XANES data taken at 0.80, 1.00, and 1.05 V vs RHE in 0.1 M HClO4 with theoretical calculations for the bridged n-fold site. The theoretical results were shifted by 4 eV and scaled by 0.4 for optimal comparison with experiment.

Figure 10. Comparison of XANES data taken at 1.14 and 1.24 V vs RHE in 0.1 M H2SO4 with theoretical calculations for the fcc n-fold species. The theoretical results were shifted up by 4 eV and scaled by 0.5 for optimal comparison with experiment.

Figure 11. Comparison of NPt-Pt and NPt-O results from Table 1 with the experimental ∆µ amplitudes from Figures 5 and 6 for both 0.1 M H2SO4 and HClO4 electrolyte. The notation “sub” indicates those potentials where the ∆µ spectral line shape reflects a subsurface site signature.

simply reflects the “passivation” of the Pt surface from O[H] chemisorption by directly adsorbed (bi)sulfate anions in H2SO4. The O[H] coverage for HClO4 appears to reach a maximum at 1.00 V, but this occurs primarily because the ∆µ spectral line shape begins to switch over from the fcc site signature to the subsurface site signature, which causes the amplitude to broaden significantly without an increase in amplitude. Thus this represents the visible onset of the place-exchange mechanism in ∆µ for oxide formation around 1.05 V. The decrease of NPt-Pt in both HClO4 and H2SO4 is quite interesting. One would not expect NPt-Pt to decrease significantly until place exchange or oxide formation begins. Using X-ray

J. Phys. Chem. B, Vol. 109, No. 16, 2005 8081

Figure 12. Comparison of XANES data taken at 0.74 V vs RHE in 0.1 M H2SO4 with theoretical calculations for 1-fold atop (shifted by 1 eV) and n-fold bridged/fcc species (shifted by 7 eV). Both theoretical curves have been scaled by 0.04 for optimal comparison with experiment.

reflectivity techniques on Pt(111), You et al.32,33 showed that the Pt(111) surface is roughened right at 1.05 V in HClO4, which they suggested occurred from the movement of Pt atoms out of their original sites (i.e., Pt and O[H] place exchange). Obviously this occurs at even a higher potential in H2SO4. Yet Figure 11 shows that in both cases NPt-Pt begins to decrease well below this point. In HClO4, NPt-Pt begins its decrease very near where NPt-O begins its rise, and the sharpest change in slope of both NPt-Pt and ∆µ occurs together around 0.95 V. This suggests that some O may go subsurface almost immediately near the edges and corners in small particles, but that it is not apparent in the experimental ∆µ line shape until much higher potentials. Alternatively, it could suggest that the chemisorption of O causes the Pt particles to go flatter, which decreases NPt-Pt, but this is not likely. Further, Figure 11 shows that the directly adsorbed (bi)sulfate has a pronounced effect on the Pt making it more round (increasing NPt-Pt). Well before the significant chemisorption of O[H], the directly adsorbed (bi)sulfate begins to desorb, allowing the cluster to return to its original more oblate shape. This is believed to be the cause for the sharp decrease in NPt-Pt in H2SO4 between 0.90 and 1.00 V. The slower rate of decrease beyond 1.00 V is then attributed to the increasing presence of subsurface O, even though the ∆µ signature throughout reflects the dominant presence of fcc surface O. Radio tracer studies of (bi)sulfate adsorption indeed indicate that the magnitude of the (bi)sulfate contact adsorption reaches a maximum around 0.60 V and begins to drop from there all the way to 1.30 V when it goes to zero.34 The direct spectroscopic identification of n-fold O has been made from the ∆µ spectra. Is their similar spectroscopic evidence for atop O[H]? Figure 12 compares the experimental ∆µ at 0.74 V in H2SO4 with theoretical signatures for atop and 3-fold binding. The two-peak structure clearly suggest two types of O binding in this case, and the large separation in energy is consistent with atop OH and n-fold bonded O, as verified by the comparison with the corresponding theoretical signatures. The fact that O adsorption is partially impeded in H2SO4 apparently allows us in this case to capture the experimental signature for atop OH, in contrast to that in HClO4, where we already see only bridged/fcc bonded O at 0.80 V RHE. The presence of the 3-fold O is probably a direct reflection of the specifically adsorbed bisulfate, consistent with the long Pt-O bond length of 2.2 Å found from the EXAFS data at this potential in H2SO4. Apparently the bisulfate is adsorbed via the O atoms in 3-fold fcc sites on the Pt clusters consistent with previous theoretical calculations.35,36 The longer Pt-O distance may account for the larger positive shift (7 eV) required to

8082 J. Phys. Chem. B, Vol. 109, No. 16, 2005

Figure 13. Comparison of previously reported ∆µ data22 taken for “small” (0.7-1.5 nm) Pt clusters on carbon with that for the larger (2-3.5 nm) Pt clusters examined in this work.

Figure 14. Same data as in Figure 13, but over a smaller energy range near the ∆µ maxima. Horizontal lines indicate energy range where the ∆µ maxima should fall for the indicated adsorbate O[H] species.

achieve optimal alignment with the experimental data compared to the 4 eV required above. The presence of atop OH can be further confirmed on clusters with more corners and edges; i.e., smaller clusters. Similar ∆µ data taken on previously reported Pt L3 XANES data22 for Pt clusters with EXAFS NPt-Pt coordinations of 5.5-7.5 (clusters sizes of 0.7-1.5 nm) have considerably more corners and edges. Cluster model calculations assuming cubooctahedral clusters37 predict that such clusters have 30-60% of the Pt atoms at corners compared to