CO Adsorption on Pd–Au Alloy Surface: Reversible ... - ACS Publications

Dec 24, 2015 - Kazuhiko Mase,. ‡. Bongjin Simon Mun,. § and Hiroshi Kondoh*,†. †. Department of Chemistry, Keio University, 3-14-1 Hiyoshi, Koh...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/JPCC

CO Adsorption on Pd−Au Alloy Surface: Reversible Adsorption Site Switching Induced by High-Pressure CO Ryo Toyoshima,† Nana Hiramatsu,† Masaaki Yoshida,† Kenta Amemiya,‡ Kazuhiko Mase,‡ Bongjin Simon Mun,§ and Hiroshi Kondoh*,† †

Department of Chemistry, Keio University, 3-14-1 Hiyoshi, Kohoku-Ku, Yokohama 223-8522, Japan Institute of Materials Structure Science, High Energy Accelerator Research Organization, and The Graduate University for Advanced Studies, 1-1 Oho, Tsukuba 305-0801, Japan § Department of Physics and Photon Science and Ertl Center for Electrochemistry and Catalysis, Gwangju Institute of Science and Technology, Gwangju 500-712, Republic of Korea ‡

S Supporting Information *

ABSTRACT: The interaction between carbon monoxide (CO) and a Pd70Au30(111) alloy surface was investigated under CO pressures with a wide range from ultrahigh vacuum (UHV) to sub-Torr at room temperature by a combination of near-ambient pressure (NAP) X-ray photoelectron spectroscopy and density functional theory calculations. The adsorption site and surface coverage of CO are reversibly controlled by the CO pressure. Under UHV conditions, the CO molecules occupy bridge and hollow sites on contiguous Pd clusters in the Aurich surface layer. Exposure to sub-Torr CO gas induces site switching of the adsorbed CO on the contiguous Pd clusters from multiple-coordination (hollow and bridge) sites to single-coordination (top) sites, even though the latter sites are energetically less favorable. This behavior is explained by a pressure-induced entropic effect on gas-phase CO, which is in equilibrium with the adsorbed CO. This site switching highlights an important aspect of highpressure-induced adsorption behavior. atoms.9,15 However, in situ observation of chemical interactions between CO and the close-packed (111) surface under (near) realistic conditions has been quite limited so far, even though nanometer-sized industrial catalysts are mainly composed of (111) surfaces, due to its thermal stability.13 Adsorption of a dense monolayer of CO on PGM surfaces at lower temperatures inhibits dissociative adsorption of O2 and hence CO oxidation, which is so-called CO poisoning.20,21 Therefore, CO adsorption behavior is a key factor to determine lowtemperature CO oxidation properties. Synchrotron-based NAP X-ray photoelectron spectroscopy (XPS) is a surface-sensitive spectroscopy applicable to a wide range of gas pressure conditions. It allows chemical and structural characterization of adsorption systems on the basis of core-level shifts (CLS). This approach provides useful information on vapor/solid interfaces under near-realistic environments.22 Density functional theory (DFT) calculations were employed to estimate the CLS and thermodynamic stability, including the entropy term, for vapor/solid interfaces of interest.23,24 The combination of NAP-XPS and DFT calculations is a powerful tool to understand adsorption behavior under various pressure conditions.

1. INTRODUCTION Oxidation of carbon monoxide (CO) proceeding on Pt-group metal (PGM) surfaces (especially Pt, Rh, and Pd) has been extensively studied as a prototypical catalytic surface reaction for understanding heterogeneous catalysis. Furthermore, this reaction has been widely used in industrial applications, for instance, three-way catalyst for exhaust gas. Recent improvement of surface science techniques enables us to overcome the “pressure-gap” problem and to investigate heterogeneous catalysts under near-realistic conditions. A number of studies on CO oxidation under near-ambient pressure (NAP) conditions shed light on the reaction mechanisms at the atomistic level under working conditions.1−7 Pd−Au bimetallic alloy is a promising material for lowtemperature CO oxidation reaction.7−10 Intermixing of Pd with Au alters the electronic and geometric structures of Pd surface, and CO adsorption behavior is also influenced. Preferential Au segregation is observed at clean Pd−Au alloy surfaces, irrespective of preparation methods such as electrochemical deposition, annealing of bulk alloy, or Au-decorated Pd single crystal and vice versa.11−15 The elemental composition of an alloy surface varies a great deal depending on the chemical environment. High-pressure gas exposure to alloy surfaces pulls up inner chemically active (inactive) species to the surface, sometimes leading to phase separation.6,7,16−19 In the case of Pd−Au alloys, high-pressure CO exposure to (100) and (110) open-surfaces results in surface segregation of internal Pd © XXXX American Chemical Society

Received: October 31, 2015 Revised: December 11, 2015

A

DOI: 10.1021/acs.jpcc.5b10661 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C In this work, pressure dependence of CO adsorption on a Pd70Au30(111) single-crystal surface up to 10 mTorr was studied by the NAP-XPS technique in parallel with DFT calculations. XPS data clearly indicate reversible switching of adsorption site between multiple-coordination (bridge and hollow) sites and single-coordination (top) sites, which originates from a pressure-dependent entropic effect on adsorption phase by the medium of equilibrium between gas phase and adsorption phase.

2. EXPERIMENTAL SECTION XPS experiments were performed at the soft X-ray beamline 13A/B at the Photon Factory of the High Energy Accelerator Research Organization (KEK-PF) in Tsukuba, Japan.25 Base pressures of the analysis and sample preparation chambers were on the order of 10−10 Torr. The single-crystal surfaces were prepared by repeated cycles of Ar+ sputtering, surface oxidation, and brief annealing (Pd, 1100 K; Au, 900 K; Pd70Au30, 1000 K). Surface orientation and cleanness were checked by low-energy electron diffraction (LEED) and XPS. Pd and Pd70Au30 surfaces exhibited the (1 × 1) pattern, whereas Au exhibited (1 × 1) with hexagonal satellites, indicating (22 × √3) herringbone reconstruction. Incident photon energies were tuned such that the photoelectron kinetic energies are approximately 95 eV and the relative cross sections of interest are maximized (Pd 3d, 430 eV; C 1s, 380 eV; Au 4f, 180 eV). Pure CO gas (99.95%) was exposed to the surfaces at 298 K. All XP spectra were normalized by average photoelectron counts of baseline and further normalized by the intensity of the bulk component of the Pd 3d5/2 level to cancel out the gas attenuation effect on the photoelectrons. XP spectra are curve-fitted by the convolution of Doniach−Šunjić and Gaussian functions and Shirley-type background.

Figure 1. Pd 3d and Au 4f XP spectra taken from cleaned Pd(111), Au(111), and Pd70Au30(111) single-crystal surfaces at 298 K. All XP spectra are curve-fitted, and each component is labeled by B, S, and S′, except for Au 4f from Pd70Au30(111). Details of the assignment are discussed in the text.

overlapping of core-level peaks of Pd 3d5/2 (typically 335 eV) and Au 4d5/2 (335 eV) and of Au 4f7/2 (84 eV) and Pd 4s (87 eV). However, the influences due to overlaps can be made negligibly small by tuning the photon energy such that the difference in cross section between the two core levels is maximized. As a result, we chose the photon energy as 430 eV for Pd 3d and 180 eV for Au 4f, which guarantees that the residual components of Au 4d and Pd 4s are sufficiently small, as shown in Figure 1 panels c and b, respectively. In fact, the branching ratios Pd 3d3/2/Pd 3d5/2 (0.67) and Au 4f5/2/Au 4f7/2 (0.71) obtained from the alloy are almost the same as those for pure Pd (0.67) and pure Au (0.69). The monometallic surfaces exhibit two components, assigned as surface (S) and bulk (B) components as shown in Figure 1a,d.26,27 The alloy surface exhibits three components in Pd 3d, which are assigned as Pd in the first layer (S), in the second layer (S′), and in bulk (B) as shown in Figure 1e. The atomic fractions of Pd atoms are estimated to be 32% ± 3% in the first layer and 33% ± 1% in the second layer. Note that the peak position of the bulk component (B) in the alloy is slightly shifted to the lower binding energy side from that for pure Pd, which is also seen in Au 4f level. The Au 4f XP spectrum shown in Figure 1f can be fitted with a single component, indicating the difference in chemical environment is less emphasized for Au atoms in the alloy, which results in small CLS for various chemical environments. These results confirm that the Pd−Au alloy surface consists of an Au-rich layer.7,11−15 This Au-rich surface is explained by the lower surface free energy of Au than that of Pd. Detailed discussion about assignments and atomic fraction analyses are given in Supporting Information. 4.2. CO Adsorption on Pd70Au30(111). Next, the clean Pd−Au alloy surface was exposed to gaseous CO at different pressures. Figure 2 shows Pd 3d5/2 and C 1s XP spectra for the CO-exposed alloy surfaces with curve-fitting results. The C 1s level taken from a CO-saturated surface under UHV (pCO < 10−8 Torr) (Figure 2b) appears as an asymmetric peak and is deconvoluted into two components. Based on XPS results on CO/Pd(111) (Figure S5) and calculated CLS (Table S4), these two components can be attributed to CO molecules at hollow

3. COMPUTATIONAL METHODS Computational simulations were carried out with the Vienna ab initio simulation package (VASP). The projector-augmented wave (PAW) method was applied to represent the electron− ion interaction, and the Perdew−Burke−Ernzerhof (PBE) approximation was used for the exchange−correlation functional. For bulk calculations, the equilibrium lattice constants were a0 = 3.95, 4.17, and 4.01 Å for Pd (A1), Au (A1), and Pd75Au25 (L12 crystal), respectively. Two-dimensional Brillouin-zone sampling was carried out on (6 × 6) Monkhorst− Pack k-point mesh for the (2 × 2) supercell. Coequal density meshes were used for the different sized supercells. Substrates were modeled by a four-layer slab. Adsorbed CO molecules and the upper two substrate layers were fully relaxed until the residual forces on atoms are reduced to less than 0.05 eV·Å−1, whereas the bottom two layers were fixed at the bulk lattice constant. Details of calculations of adsorption energy, core-level shift, and surface free energy are described in Supporting Information. 4. RESULTS AND DISCUSSION 4.1. Clean Surfaces of Pd(111), Au(111), and Pd70Au30(111). First, CO-free clean surfaces were characterized by Pd 3d and Au 4f XP spectra. Figure 1 shows the Pd 3d and Au 4f XP spectra taken from Pd(111) (panels a and b), Au(111) (panels c and d), and Pd70Au30(111) (panels e and f) surfaces under ultrahigh vacuum (UHV) conditions. A serious issue for measuring Pd−Au alloys with XPS is significant B

DOI: 10.1021/acs.jpcc.5b10661 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

favored in the order hollow > bridge > top, which is the same as those for Pd(111) and Pd−Au alloy surfaces.34−36 Top site occupation is not observed under UHV conditions (pCO < 10−8 Torr) but becomes observable under 10−6 Torr exposure, indicating that the gas-phase CO stabilizes the top CO. The multiple-coordination sites consisting of both Pd and Au atoms are energetically less favored compared to those consisting of pure Pd (see Tables S1 and S2). In the present system, CO molecules can be adsorbed at pure Pd sites from the energetic point of view. By increasing the CO pressure up to 10 mTorr, the adsorption state drastically changes, as shown in Figure 2e,f. Top CO significantly increases and becomes a dominant species on the surface, whereas bridge and hollow CO decrease instead. Since both the surface/bulk ratio of Pd and the Pd/Au ratio remain unchanged when the CO pressure rises from 10−6 Torr to 10 mTorr, high-pressure CO-induced surface restructuring such as segregation and clustering of Pd does not occur in the present system. Therefore, the significant increase in top CO population is not due to an increase of surface Pd population but due to site switching of adsorbed CO from multiple-coordination sites to single-coordination sites. This is somewhat surprising because the adsorption energy of the single-coordination site is much smaller than those of the multiple-coordination sites as shown in Table S1. In order to check CO population changes for each site, CO coverages are estimated from C 1s XPS at different CO pressures as shown in Figure 3. Total CO coverage continuously increases with

Figure 2. Pd 3d5/2 and C 1s XP spectra taken from Pd70Au30(111) under different CO pressure conditions at 298 K. All the XP spectra are curve-fitted, and each component is labeled by B, S, S′, and Pdx in Pd 3d5/2 and h, br, and t in C 1s. Details of the assignment are discussed in the text.

(h) and bridge (br) sites. The corresponding Pd 3d5/2 level (Figure 2a) exhibits an appearance of two components at the higher binding energy side at the expense of part of the surface Pd component (S) due to CO adsorption. The newly appearing two components are assigned as surface Pd atoms occupied by 3-fold-coordinated (hollow) CO (Pd1/3) and 2-fold-coordinated (bridge) CO (Pd1/2), where the index numbers indicate the coordination number with respect to Pd atom. Occupation of the hollow and bridge sites is consistent with the C 1s result. It should be noted that the CO-free surface Pd component (S) remains observable even after CO saturation under UHV conditions. If we assume that 30% Pd atoms in the first layer are randomly distributed, it is supposed that the isolated (monomer) Pd and contiguous (dimer and trimer) Pd coexist with almost the same populations. CO molecules adsorb at the hollow and bridge sites, while the top sites are free from CO adsorption. Since the isolated (monomer) Pd provides the top sites, the CO-free surface Pd component (S) is associated with monomer Pd. Thus, CO adsorption on top of the monomer Pd is not favored, which is also supported by calculated adsorption energies shown in Table S1. Preferential occupation of the bridge and hollow sites is observed for CO/Pd(111) under UHV conditions26,28−30 and confirmed by the present experimental result (see Figure S5). Exposure to 10−6 Torr CO induces further adsorption of CO, as shown in Figure 2c,d. The C 1s level exhibits an additional shoulder structure at the higher binding energy side (Figure 2d). This shoulder structure is attributed to CO molecules adsorbed on top sites. There is a general trend for C 1s binding energy of CO adsorbed on PGM surfaces; the lowercoordination CO exhibits the higher binding energy,31,32 which is also confirmed by the present CLS calculations (Table S4). The top CO component appears broad due to a significant tail at the higher energy side, which is contributed from the vibration structure.33 The Pd 3d5/2 level clearly shows a new component at the expense of the CO-free surface monomer Pd (S). This new component is assigned to surface monomer Pd bound to single-coordinated (top) CO (Pd1). At this moment all the surface Pd atoms interact with adsorbed CO. The three types of adsorption sites are energetically

Figure 3. CO coverage changes deduced from C 1s XP spectra as a function of CO pressure. Total coverage is the sum of coverages of CO on bridge (br), hollow (h), and top (t) sites.

increasing CO pressure. Under UHV (pCO < 10−8 Torr) conditions, the total CO coverage of 0.077 monolayer (ML) is significantly small compared with that on Pd(111) (0.5 ML) because the fraction of surface Pd is approximately 30% of the first layer, and only half of the surface Pd atoms bind to CO, as mentioned before. At 10−6 Torr, the remaining half of surface Pd atoms, which are isolated (monomer) Pd atoms, are occupied by top CO. Above 10−6 Torr, the (positive) slope of increase in top CO (t) is almost twice as steep as the (negative) slope of decrease in bridge CO (br), which indicates that one bridge CO is replaced by two top COs. This means that the contiguous two Pd atoms accommodate two adjacent top COs at high pressures. When the CO gas was evacuated to