FTIR Spectroscopic Investigation of Zeolite-Supported Pd–Ag

Article pubs.acs.org/JPCC

FTIR Spectroscopic Investigation of Zeolite-Supported Pd−Ag Bimetallic Clusters Olga Terekhina and Emil Roduner* Institute of Physical Chemistry, University of Stuttgart, Pfaffenwaldring 55, D-70569 Stuttgart, Germany ABSTRACT: The catalytic properties of bimetallic supported metal catalysts depend not only on the type and proportion of the metallic elements but, in an essential way, also on their alloying behavior. Here, CO was used as a probe molecule to test the electronic properties of a PdAg bimetallic catalyst with different Pd/Ag ratios, supported on NaY zeolite. Although CO does not adsorb on Ag0, changes in IR spectra on addition of Ag to Pd samples provide clear evidence that the two metals form a nanoalloy. Furthermore, addition of Ag stabilizes the Pd clusters.

1. INTRODUCTION Bimetallic nanoparticles of various compositions have been a subject of great interest in the catalytic community over the past decade. These bimetallic systems have unique catalytic and electronic properties, distinct from those of monometallic ones. The selective behavior of such catalysts can be controlled by changing their composition.1 The distinct changes in activity and/or selectivity with composition are attributed to alloying effects.2 Because of the high activity of palladium, platinum, and silver in various important catalytic processes, bimetallic particles of Ag in combination with Pd and Pt have been studied in great detail.3 After earlier studies were mostly done on inorganic oxide-supported nanoparticles,4,5 novel techniques have been developed to study Pd−Ag and Pt−Ag alloys in colloidal suspensions,6,7 microemulsions,8 and nanoparticles immobilized on fibers.9 Transition-metal clusters supported on zeolites are currently a subject of significant scientific interest, primarily from the viewpoint of their catalytic characteristics.10 The regular channels and cages of zeolites provide perfect matrixes for the dispersion of metal particles. Although numerous studies have been conducted to prepare and characterize well-defined, small monometallic Pt, Pd, or Ag clusters in zeolites,11−17 only a limited number of experimental studies were devoted to the investigation of binary Pd−Ag and Pt−Ag alloys supported on zeolites.18,19 Specifically, the Pd−Ag alloys have generated increasing interest for use in hydrogen membranes and as selective © 2012 American Chemical Society

hydrogenation catalysts. In many cases, the Pd-based alloy catalysts exhibit higher selectivity than the pure Pd catalysts.4 Knowledge of structural or electronic effects of alloying is essential to catalyst optimization. Both, electronic or ligand effects (i.e., electronic factors due to the change in electron density) and ensemble effects (i.e., surface structure factors due to the location of different type atoms) can contribute to the synergistic effects often observed for bimetallic alloys.20 Theoretical work predicts that the electronic properties of bimetallic alloys can be obtained by interpolation between those of the individual components.21 This electron transfer between the components of a bimetallic system, which leads to an exchange in electron density dispersion, has also been called “initial state effect”.22 Furthermore, ionization potential and electron affinity, which determine the availability of electrons in redox reactions or for bond formation, can be tuned by several electronvolts also by the metal particle size.23 Regarding the structure, we have to distinguish between random mixing on an atomic level, clustering of an element, and, in particular, surface segregation, which leads to core−shell structures. Interestingly, metals that do not mix in the bulk (e.g., Au is completely insoluble in bulk Ni) may form a surface alloy.24 In the present work, we study NaY zeolite-supported Pd, Ag, and Pd−Ag systems by Fourier transform infrared spectroscopy Received: November 24, 2011 Revised: February 28, 2012 Published: March 5, 2012 6973

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

band near 3650 cm−1, confirming the formation of Si(OH)Al acidic bridging hydroxyl groups.14 Flowing CO was introduced at 298 K. Background spectra were recorded with a reduced wafer. The spectra for adsorbed CO were then obtained by subtracting the background.

(FTIR) of adsorbed CO. We use CO as a probe molecule since a large database exists in the literature regarding its adsorption on transition-metal surfaces. In most of these cases, the frequency of the CO stretching vibration in the adsorbed state is lowered with respect to its gas-phase value of 2143 cm−1. Depending on the adsorption site, it typically takes values between 2000 and 2100 cm−1 in the on-top position and 1700−2000 cm−1 in the bridge-bonded configuration.25 This reduction in CO stretching frequency is generally taken as evidence for M → CO backbonding: the donation of electron density from the metal nd orbitals into the antibonding 2π orbital of CO. This backbonding weakens the C−O bond and lowers the stretching frequency.26 In this way, the observed stretching frequencies of CO adsorbed on the metal can be used to identify its specific binding sites or oxidation states and obtain information about surface composition, which is the key to understanding the catalytic properties of bimetallic surface alloys. Comparison of IR spectra for binary samples with monometallic ones allows us to discuss the effect of alloying.

3. RESULTS 3.1. Adsorption of CO on NaY. For reference, FTIR spectra of CO adsorbed on NaY at room temperature and 85 K are shown in Figures 1 and 2. At room temperature and a

2. EXPERIMENTAL SECTION 2.1. Preparation of Samples. NaY zeolite (Zeocat, Z6-0101 from CU Chemie Uetikon, with a Si/Al ratio of 2.31, unit cell composition Na58[(AlO2)58(SiO2)134]) was heated in air with a rate of 1 K min−1 and calcined further at 773 K for 14 h to burn off organic impurities. Pd/NaY samples were prepared in a flask containing 1 g of pretreated zeolite by ion exchange with 200 mL of 1 mM aqueous [Pd (NH3)4](NO3)2 solution, which was stirred at 298 K for 20 h. The loading of Pd in NaY was 1.81 wt %, as verified by elemental analysis, corresponding to roughly 2.70 Pd atoms per unit cell (abbreviated below as Pd2.70). Ag/NaY samples were prepared by stirring 1 g of NaY with 200 mL of 0.1 mM AgNO3 solution for 12 h at 298 K (the loading of Ag amounted to 0.21 wt % or 0.31 Ag atoms per unit cell). Both binary samples were prepared in two steps by first exchanging with 200 mL of 1 mM [Pd(NH3)4](NO3)2 solution. After 20 h at 298 K, 200 mL of AgNO3 (0.1 or 1 mM water solution) was added and stirred for another 12 h. The loading of Pd/Ag in binary samples was 1.73/0.34 and 1.85/2.42 wt %, respectively (corresponding to Pd2.66Ag0.52 and Pd2.80Ag3.61 loading per unit cell). From the elemental analysis of the transition-metal content, it is clear that, in all cases, less than 10 out of the originally 58 sodium ions per unit cell are replaced by transition-metal ions; that is, the vast majority of the charge balancing ions is still Na+. Exchanged samples were calcined in flowing O2 (20 mL min−1 g−1) using a ramping rate of 0.5 K min−1 from room temperature to 523 K, and then holding the temperature for 120 min. 2.2. FTIR Spectroscopy. FTIR spectra were obtained on a Magna-IR 560 spectrometer at a spectral resolution of 2 cm−1, accumulating 256 scans. Small amounts of the catalyst powder were pressed into 1 cm diameter wafers of 20−30 mg cm−2. The wafers were transferred to the IR cell and recalcined in O2 for 10 min. O2 was then replaced by N2. The absence of water was verified by the absence of bands around 1650 and 3650 cm−1. The wafer was heated up under hydrogen to 473 K at a rate of 8 K min−1, and then held for 20 min. The gas was then removed. After reduction, the FTIR spectra showed a strong

Figure 1. FTIR spectra recorded after CO adsorption (500 mbar) at 298 K on NaY, followed by evacuation at 298 K. p(CO) = 50 mbar (a), 2 mbar (b), and under vacuum (c).

Figure 2. FTIR spectra recorded after CO adsorption (100 mbar) at 85 K on NaY, followed by evacuation at 85 (a), 173 (b), 223 (c), and 273 K (d).

pressure of 50 mbar CO, we observe IR band progressions around 2166 and 2124 cm−1 assigned to the R and P rotational−vibrational branches of gaseous CO (Figure 1, spectrum a). The purely vibrational transition (Q band), which is expected at 2143 cm−1, is IR-inactive. The gas-phase bands can be easily removed by evacuation (spectra b and c). Weak 6974

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

bands at 2003, 2033, and 2073 cm−1 characterize CO adsorption on unspecified zeolite centers. At 85 K (Figure 2), we observe a very strong and narrow band at 2173 cm−1, which belongs to CO adsorbed on Na+ ions in SII positions of the zeolite.27,28 In addition, a weak band appears at 2123 cm−1. It can be assigned to the 13CO stretching mode; its isotopic shift coincides with the theoretical expectation.29 The intensity of the CO band at 2173 cm−1 is so high that instrumental limitations do not allow the exact determination of the maximum, even under vacuum conditions. Only by increasing the temperature does the intensity of the band decrease (traces a−d). The weak bands, assigned above to CO adsorption on zeolite centers, are not visible at the scale of Figure 2, but they remain unchanged (see inset in Figure 2). 3.2. Adsorption of CO on Ag, Pd, and Pd−Ag Samples. FTIR spectra of CO adsorbed at room temperature on metal centers in NaY are presented in Figure 3. For Ag0.31,

be separated into two peaks that are assigned to Pd+CO (2124 cm−1) and Pd0CO (2112 cm−1) complexes. There is also a possibility of the presence of Pdδ+CO species, with the Pdδ+ presumably occupying SII sites, while the CO ligand may be coordinated to it through the O6 ring (at 2120 cm−1).14 The bands observed with the binary samples Pd2.80Ag3.61 and Pd2.66Ag0.52 (spectra b and c in Figure 3) are quite similar to those of Pd2.70 (spectrum d). Adding silver mostly seems to shift the bands in the bridge-bonded and linear region relative to those of the Pd2.70 sample. We can still assign four bands at 1950, 1881, 1857, and 1789 cm−1 to multiply bonded CO on Pd. Clearly, the shifts are due to the effect of alloying, which will be discussed further below. We further distinguish two overlapping bands at 2123 cm−1, assigned above to Pd+CO, and 2080 cm−1, which must belong to terminally bonded CO in the alloy. FTIR spectra obtained at 85 K are presented in Figure 4, parts A (linear region) and B (bridge region, only for samples containing Pd). In comparison with the room-temperature

Figure 3. FTIR spectra recorded after CO adsorption (500 mbar) at room temperature on Ag0.31 (a), Pd2.80Ag3.61 (b), Pd2.66Ag0.52 (c), and Pd2.70 (d). The equilibrium pressure of CO is 100 mbar.

we observe a weak spectrum with a new broad band centered at 2183 cm−1 (Figure 3, spectrum a) in addition to the bands observed above with plain NaY. The new band contains two overlapping peaks, one at 2188 cm−1, which, according to the literature, belongs to Ag+CO,26,30,31 the only feature that relates to CO adsorption on silver. The component at 2175 cm−1 is again assigned to CO adsorbed on Na+ ions. At room temperature, it appears also in the spectra of the binary samples (Figure 3, spectra b and c), but not for the monometallic Pd2.70 sample (spectrum d). This peak is weak and superposes with the rotational−vibrational lines from gasphase CO. There are no peaks in the region of bridge-bonded CO in case of the Ag0.31 sample. For Pd2.70 (Figure 3, spectrum d), we observe a spectrum that is akin to the one obtained earlier by Sheu et al.14 for Pd13(CO)x clusters in NaY zeolite, suggesting that we have a very similar or identical structure. Four bands were assigned to multiply bonded CO ligands, namely, to CO adsorbed on Pd in 2-fold (bands at 1988, 1954, and 1900 cm−1) and 3-fold (band at 1824 cm−1) binding sites. In the region of terminal CO ligands, the broad strong absorption centered at 2118 cm−1 can

Figure 4. FTIR spectra recorded after CO adsorption (100 mbar) at 85 K, followed by evacuation at 85 K for 30 min on Ag0.31 (a), Pd2.80Ag3.61 (b), Pd2.66Ag0.52 (c), and Pd2.70 (d) in the linear (A) and the bridge-bonded regions (B). 6975

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

spectra in Figure 3, there are insignificant shifts of the major bands in the bridge region, and the assignment remains the same. The strong band at 2173 cm−1 in the linear region belongs to 12CO adsorbed on Na+ ions. The one at 2123 cm−1 corresponding to Na+(13CO) species is clear for Ag0.31 (Figure 4A, trace a), but for other samples, it overlaps with the band near 2127 cm−1 belonging to Pd+CO (Figure 4A, traces b−d). For the samples containing silver, we also observe the small peak at 2188 cm−1 characteristic for Ag+CO (Figure 4A, traces a−c), overlapping with the peak at 2173 cm−1. Since the 2188 cm−1 band is important for the characterization of adsorption on silver, the present low-temperature spectra were recorded after 30 min evacuation where the peak at 2173 cm−1 is less intense (Figure 4A). The amount of adsorbed and partly desorbed CO differs for each sample, and the spectra recorded at 85 K serve only for a qualitative analysis. From the traces a and c (Figure 4A), we can also notice a shoulder at 2197 cm−1, which comes from the small peak at approximately 2195 cm−1, possibly some Agδ+CO species. Lynn et al. have also observed a surprisingly high stretching frequency (2204 cm−1) for the Ag− CO complex.26 To obtain more quantitative information, the spectra recorded at room temperature were used to study the amount of actually adsorbed CO. The adsorption bands were fitted with Lorenzian lines. In Figures 5 and 6, the integrated relative

Figure 6. Integrated absorption of the CO band at 2112/2080 cm−1 (squares) and 2060/2047 cm−1 (triangles) for Pd2.70 (dashed line), Pd2.66Ag0.52 (black line), and Pd2.80Ag3.61 (gray line) during desorption at 298 K.

the corresponding peaks decreases slowly. Around 12.5 mbar, we observe a sudden decrease of the 2112/2080 cm−1 band area, while, simultaneously, a new peak grows in at 2060/2047 cm−1 for Pd2.70 and the binary samples (inset in Figure 6). A peak at 2127 cm−1 also disappears (not shown here), and at low CO coverage, we observe only a small new peak, which also decreases with further CO desorption. This phenomenon was already observed by Sheu et al.,14 who distinguished two states of high and low CO coverage, suggesting that the fully carbonylated Pd13 cluster locates near the supercage center, and the partially decarbonylated one interacts with the cage walls. Such behavior obviously does not change on adding Ag.

4. DISCUSSION: EVIDENCE OF ALLOYING 4.1. The State of Silver. CO is used as a probe molecule for the composition of the electronic properties and the surface composition of the binary metal clusters. To permit this, we have to make sure that the clusters do not disintegrate on CO adsorption, as it is the case for zeolite-supported Pt13 clusters.32 Furthermore, CO adsorption could pull strongly binding Pd atoms to the surface. In the nonalloyed case, one would expect the spectra of bimetallic samples to be superpositions of those of the monometallic ones, possibly accompanied with small shifts if electronic effects are transmitted through the zeolite lattice. Thus, the main criterion for alloying will be that there are major spectral differences compared with monometallic samples. The band observed at 2188 cm−1 (Figures 3, and 4A, trace a) indicates the presence of Ag+−CO species.30 A small fraction of isolated silver ions obviously survived H2 reduction at 473 K. The band is blue shifted by Δν̃ = +45 cm−1 from the free CO gas-phase value of 2143 cm−1, an effect that is commonly observed for so-called “nonclassical” carbonyl complexes, in particular, for d10 Ag(I) complexes.26 These systems exhibit less M → CO backbonding into the C−O 2π* orbital, but significant OC → M σ-donation from the CO antibonding 5σ orbital, resulting in the high stretching frequencies. After H2 reduction, and since CO is also a reducing agent, one should expect the presence of peaks that belong to Ag0CO species.

Figure 5. Integrated absorption of the CO band at 1824/1789 cm−1 (squares) and 1988/1950 cm−1 (triangles) for Pd2.70 (dashed line), Pd2.66Ag0.52 (black line), and Pd2.80Ag3.61 (gray line) during desorption at 298 K.

absorbance per gram (the band area) is plotted against the equilibrated CO pressure during the desorption process. Four bands are presented as examples, two of them in the bridge region at 1824(Pd2.70)/1789 cm−1 (for both binary samples) and 1988(Pd2.70)/1950 cm−1 (for both binary samples) (Figure 5). These bands have essentially constant areas down to a CO pressure of ≈10 mbar, indicating that the M−CO bonds remain stable. Below this pressure, the bonds break and the corresponding IR peaks decrease quickly and disappear. Such behavior was observed for all major bands in the bridge region. The bands in the linear region (Figure 6) at 2112(Pd2.70)/2080 cm−1 (for both binary samples) show a different behavior: with decreasing CO coverage, the integrated relative absorbance of 6976

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

However, according to the literature, adsorption of CO on Ag0 can form a very weak bond only at low temperatures, and this IR band was rarely observed.31 Other work reported that CO does not interact at all with Ag0 atoms under the present conditions.30 This is in full agreement with our observations, also in the alloyed samples. As opposed to the plain Ag sample (Figure 3, spectrum a), the Ag+−CO peak at 2188 cm−1 is weak and not detectable or masked by the gas-phase CO progression for the binary samples at room temperature (traces b and c), but it is clear in the low-temperature spectra (Figure 4A, traces b and c). There is a possibility that, in the presence of Pd, Ag+ could be more easily reduced to Ag0, which does not adsorb CO and is not visible in the IR spectra. A similar phenomenon was already observed by Michaelis et al.,7 who reported that Ag+ ions were reduced in the absence of palladium with an induction period of several minutes and, in the presence of Pd, practically instantaneously. This observation may be explained by Sanderson’s mean electronegativity Sm of the zeolitic framework, defined as the geometric mean of the atomic electronegativities Si. Sm has been shown to correlate with the chemical shift of Brønsted acid bridging OH groups, Si(OH)Al.33 It is a measure of the mean electronic properties that depend on the composition of a sample, reflecting the fact that these properties are not local but propagate through the zeolite lattice. Replacing sodium ions by silver ions increases Sm and will thus also affect the electron density on Pd. Although there is a significant change between the monometallic Pd2.70 and the bimetallic Pd2.66Ag0.52 sample, in particular, in the bridge-bonded region (Figure 4B, traces d and c), there are quite unexpectedly no significant changes when more silver is added (Pd2.80Ag3.61, trace b). Obviously, there is no further influence on the electronic properties, so the question is where the additional silver atoms are. The Ag+−CO signal at 2188 cm−1 seems to be even weaker at the higher silver content. A possible explanation is that any “excess” silver remains unalloyed or as neutral Ag atoms, remaining invisible in the IR spectra. 4.2. Evidence of Alloying from CO Adsorption on Pd. Characterization of Pd nanoparticles with FTIR studies of adsorbed CO was discussed in numerous works.34−38 Some of them are also dedicated to Pd clusters entrapped in NaY zeolite supercages.14,15 In the present work, we focus on the distinctions that appear when silver is added to the palladium sample and the resulting effects of alloying. On the basis of the IR spectra and by comparison with the earlier studies, we conclude that we have a Pd alloy in NaY. Sheu et al.14 reported a coordination number of 6 for their Pd13(CO)x cluster with a cuboctahedral structure and a diameter of 8.2 Å. The size of the NaY supercage is 11.8−12.3 Å (entrance window: 7.4 Å).15 The present spectrum for the plain Pd sample also shows CO ligands in terminal, doubly and triply bridging positions (Figures 3 and 4, trace d) and is very similar to theirs. In comparison with the monometallic Pd sample, the IR spectra of the bimetallic Pd2.66Ag0.52 and Pd2.80Ag3.61 samples at 85 K (Figure 4B, traces d, c, and b) show distinct changes in the bridge-bonded region. Although the overall appearance of this region is the same, we note a red shift of Δν̃ = −38 cm−1 relative to the peaks of Pd2.70 for three of the four bands; for example, the band in the triply bonded region at 1824 cm−1 shifts to 1786 cm−1 (the line width changes from 30 cm−1 halfwidth for Pd2.70 to 23 cm−1 for Pd2.66Ag0.52, which is perhaps not significant). At the same time, its relative intensity reduces.

This may be a geometry effect because the numbers of adjacent Pd surface sites, which are necessary for multiply coordinated CO, are diluted by Ag atoms. Such phenomena were also described in several works devoted to the investigation of Pdbased bimetallic alloys. For example, Abbott et al.21 reported that adding Au to Pd alloy suppresses bonding in bridge geometries and leads to a substantial increase in the intensity of the CO atop species; Khan et al.4 discussed an effect of Ag on Pd in the blocking of CO adsorption on multiply coordinated sites and the increase in linearly bonded CO on the particles; and Huang et al.18 reported the complete disappearance of bridge-bonded CO on PdAg alloy, due to isolation of Pd atoms by Ag atoms. A peak at 1896 cm−1 (Pd2CO) for Pd2.70 shifts to 1857 cm−1 (Δν̃ = −38 cm−1) in the case of the bimetallic samples, and the peak overlaps with another one, leading to a shoulder at 1876 cm−1 (Figure 4B). In contrast, two overlapping bands at 1987 and 1956 cm−1 for Pd2.70, also assigned to a Pd2CO stretching frequency, merge to one peak at 1949 cm−1 for the bimetallic sample (Δν̃(1987 − 1949) = −38 cm−1). In general, it can be stated that the region of bridge-bonded CO gets a relatively strong red shift by adding Ag to the Pd sample. These shifts to lower wavenumbers reveal a weakening of the CO bond in comparison with the plain Pd sample. The presence of Ag atoms around the Pd atoms could lead to an increasing electron density on Pd by electron transfer from Ag (electronic effect of alloying), resulting in a stronger M → 2π back-donation from Pd to the carbonyl C atom. It would possibly lead to a strengthening of the Pd−CO bond with simultaneous weakening of the CO bond. Relying on the IR spectra alone, it is hard to say precisely whether the CO molecules form bridge bonds between Ag and Pd atoms or whether these bonds are to two Pd atoms that are surrounded by Ag atoms. However, since the spectra for two samples that differ in Ag loading by a factor of 7 (Figure 4B, traces b and c) are quite identical, we conclude that CO is likely bridge-bonded on the same metal; that is, carbon monoxide molecules adsorb preferentially on Pd atoms. This assumption is in line with the work by Heinrichs et al.39 on SiO2-supported PdAg alloy and by Huang et al.18 on Na+-β-zeolite-supported PdAg catalysts, reporting that CO molecules form chemical bonds with Pd atoms only. The identical value for the red shifts of both Pd2CO and Pd3CO bands in bimetallic samples indicates the same effect of the presence of Ag and suggests a well-mixed alloy. The fact that this value is also equal for both bimetallic samples (i.e., independent of the Ag loading, though one should expect further red shifts or any other differences in the spectra with higher Ag loading), can be discussed from two points of view. The first possible explanation was already mentioned above: both binary samples may actually contain an equal amount of alloyed Pd and Ag atoms. The excess Ag atoms are isolated and can occupy the sodalite cages or the hexagonal prisms of the NaY zeolite. The second possibility consists of a “limited” effect of alloying (in our case, a weakening of the C−O bond with a strong red shift of the stretching frequencies): no matter how much Ag we add to the Pd-based alloy, the effect saturates already at low Ag loading. In favor of this interpretation is the good mixing ability of Pd and Ag, noted in the literature; because of the similarity of the two atoms, they generally form a well-mixed alloy.1,4,6,18 In the linear region, we have only two overlapping bands assigned to Pd+CO and Pd0CO. For Pd2.70, we observe them at 6977

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

2127 and 2118 cm−1 at 85 K (Figure 4A, trace d). For binary samples, the strong peak at 2127 cm−1 remains invariable (if we have isolated Pd cations, it is clear that the position of this band would not be changed on adding Ag), but one small and inhomogeneous peak appears at 2080 cm−1. This new peak could belong to the Pd0CO species, being red shifted from 2118 cm−1 (Δν̃ = −38 cm−1) due to the effect of the presence of Ag, as described for the bridge region. Alternatively, one could imagine that CO is still adsorbed mostly in terminal mode at Pd, but tilted toward Ag so that the CO π* system can accept electron density from the Ag, which would also lead to a weakening of the C−O bond and concomitant red shifts.40 From Figure 6, we can see that this terminal bond is not as stable as the Pd2CO or Pd3CO bonds. This fact is also in line with the other works,36,38 reporting that, at room temperature, CO adsorbs on Pd mainly in a bridge-bonded configuration. At low CO coverage, an adsorbed CO molecule binds to two, three, or more Pd atoms. At high and/or saturation coverage, the CO molecules become linearly bonded.38 Theoretical calculations also confirm that CO adsorption at 3-fold sites is energetically preferred over ontop sites, and the strongest CO bonding is at bridge positions.36 Clearly, by adding Ag, we dilute the number of Pd3CO positions, and in Figure 5, we can see a significant decrease of the band area for binary samples compared to the plain Pd sample (squares, 1824/1789 cm−1 band). For 2-fold positions, the value does not differ notably. As we observe from Figures 5 and 6, the characteristic behavior of the Pd cluster does not change significantly on adding Ag. In contrast to the bimetallic samples that showed high stability and reproducible IR spectra even after 2 years, the monometallic Pd samples were not stable and the cluster often appeared to disintegrate at higher CO pressure during CO adsorption. A similar reconstruction was described by Akdogan et al.32 for Pt13 clusters breaking up into Pt2(CO)m, which also led to changes in IR spectra where four separate and clear bands in the bridge region disappeared. In aged samples, we observed only one broad peak at wavenumbers lower than 2000 cm−1. We also did not succeed in reproducing the IR spectra for a one-month-old Pd2.70 sample. It is, therefore, concluded that, just as for Pt13, CO bonding to the plain Pd clusters competes with Pd−Pd bonds so that the clusters restructure or break up, and apparently alloying with Ag is able to stabilize them, perhaps because the stronger multiple bonds are disfavored.

vibrational bands reveal clearly that the CO molecules experience an electronic environment that is different from that of the monometallic samples and that the metals must be alloyed. Because of a surface ensemble effect, CO bonding in bridge geometries is slightly suppressed in alloyed samples, which is clearly noticeable for Pd3CO species. Furthermore, adding Ag increases the stability of the Pd cluster during the CO adsorption due to the electron density transfer from Ag to Pd (electronic effect of alloying). This also leads to the strong red shifts in IR spectra, which is interpreted as evidence for a weakening of the C−O bond in the ligand. Some uncertainty remains about the interpretation of the IR spectra of Pd2.80Ag3.61, where the atomic fraction of Ag has increased by a factor of 7 without affecting the spectra any further; in agreement with the literature, it was found that CO does not adsorb on neutral Ag to any significant extent. If only a fraction of Ag atoms were present in the alloy with Pd and the remaining Ag stayed unalloyed as isolated Ag atoms or even in small clusters, this could explain the similarity of the spectra.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS

■

REFERENCES

Thanks are due to Ms. H. Fingerle from the Institute of Technical Chemistry for elemental analysis.

(1) Lahiri, D.; Chattopadhyay, S.; Bunker, B. A.; Doudna, C. M.; Bertino, M. F.; Blum, F.; Tokuhiro, A.; Terry, J. Phys. Scr. 2005, T115, 776−780. (2) Rades, T.; Borovkov, V. Yu.; Kazansky, V. B.; Polisset-Thfoin, M.; Fraissard, J. J. Phys. Chem. 1996, 100, 16238−16241. (3) Srnová-Šloufová, I.; Lednický, F.; Gemperle, A.; Gemperlová, J. Langmuir 2000, 16, 9928−9935. (4) Khan, N. A.; Uhl, A.; Shaikhutdinov, S.; Freund, H.-J. Surf. Sci. 2006, 600, 1849−1853. (5) Ferrari, A. M.; Xiao, C.; Neyman, K. M.; Pacchioni, G.; Rösch, N. Phys. Chem. Chem. Phys. 1991, 1, 4655−4661. (6) Torigoe, K.; Esumi, K. Langmuir 1993, 9, 1664−1667. (7) Michaelis, M.; Henglein, A. J. Phys. Chem. 1994, 98, 6212−6215. (8) Wu, M.-L.; Lai, L.-B. Colloids Surf., A 2004, 244, 149−157. (9) Liz-Marzán, L. M.; Philipse, A. P. J. Phys. Chem. 1995, 99, 15120−15128. (10) Okumura, K.; Honma, T.; Hirayama, S.; Sanada, T.; Niwa, M. J. Phys. Chem. C 2008, 112, 16740−16747. (11) Liu, X.; Dilger, H.; Eichel, R. A.; Kunstmann, J.; Roduner, E. J. Phys. Chem. B 2006, 110, 2013−2023. (12) Akdogan, Y.; Vogt, C.; Bauer, M.; Bertagnolli, H.; Giurgiu, L.; Roduner, E. Phys. Chem. Chem. Phys. 2008, 10, 2952−2963. (13) Du, H.; Klemt, R.; Schell, F.; Weitkamp, J.; Roduner, E. Proceedings of the 12th International Zeolite Conference; Material Research Society: Warrendale, PA, 1999, 2655. (14) Sheu, L. L.; Knözinger, H.; Sachtler, W. M. H. Catal. Lett. 1989, 2, 129−138. (15) Jiang, Y.-X.; Weng, W.-Z.; Si, D.; Sun, S.-G. J. Phys. Chem. B 2005, 109, 7637−7642. (16) Ozin, G. A.; Baker, M. D.; Parnis, J. M. Angew. Chem., Int. Ed. Engl. 1983, 22, 791−792. (17) Baldansuren, A.; Roduner, E. Chem. Phys. Lett. 2009, 479, 135− 137.

5. CONCLUSIONS Additional, element-specific information about the composition of alloys and, in particular, about coordination numbers and particle size would be helpful and could be obtained from extended X-ray adsorption fine structure (EXAFS) experiments. Unfortunately, since Pd and Ag are adjacent atoms in the periodic table, it is very difficult to distinguish between these two neighboring elements in the EXAFS fits and to determine whether they are alloying. For monometallic Ag and Pt clusters, EPR proved to be successful in the characterization of small paramagnetic clusters.11,12,17 The current bimetallic samples did not appear to contain a significant fraction of paramagnetic clusters or ions under any conditions studied. The proof that the two metals form a nanosized alloy comes from IR experiments of adsorbed CO, which is used as a probe molecule. On addition of Ag to Pd, large shifts of the 6978

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979

The Journal of Physical Chemistry C

Article

(18) Huang, W.; Lobo, R. F.; Chen, J. G. J. Mol. Catal. A: Chem. 2008, 283, 158−165. (19) Grass, M. E.; Yue, Y.; Habas, S. E.; Rioux, R. M.; Teall, C. I.; Yang, P.; Somorjai, G. A. J. Phys. Chem. C 2008, 112, 4797−4804. (20) Abbott, H. L.; Aumer, A.; Lei, Y.; Asokan, C.; Meyer, R. J.; Sterrer, M.; Shaikhutdinov, S.; Freund, H.-J. J. Phys. Chem. C 2010, 114, 17099−17104. (21) Hammer, B.; Nørskov, J. K. Adv. Catal. 2000, 45, 71−129. (22) Guczi, L. Catal. Today 2005, 101, 53−64. (23) Roduner, E. Chem. Soc. Rev. 2006, 35, 583−592. (24) Pleth Nielsen, L.; Besenbacher, F.; Stensgaard, I.; Laegsgaard, E.; Engdahl, C.; Stoltze, P.; Jacobsen, K. W.; Nørskov, J. K. Phys. Rev. Lett. 1993, 71, 754−757. (25) Marian, C. M. Chem. Phys. Lett. 1993, 215, 582−586. (26) Lynn, M. A.; Bursten, B. E. Inorg. Chim. Acta 1995, 229, 437− 443. (27) Cairon, O.; Loustaunau, A. J. Phys. Chem. C 2008, 112, 18493− 18501. (28) Bordiga, S.; Garrone, E.; Lamberti, C.; Zecchina, A. J. Chem. Soc., Faraday Trans. 1994, 90, 3367−3372. (29) Hadjiivanov, K.; Knözinger, H. Chem. Phys. Lett. 1999, 303, 513−520. (30) Agostini, G.; Usseglio, S.; Groppo, E.; Uddin, M. J.; Prestipino, C.; Bordiga, S.; Zecchina, A.; Solari, P. L.; Lamberti, C. Chem. Mater. 2009, 21, 1343−1353. (31) Wang, P.; Yang, S.; Kondo, J. N.; Domen, K.; Baba, T. Bull. Chem. Soc. Jpn. 2004, 77, 1627−1634. (32) Akdogan, Y.; Anantharaman, S.; Liu, X.; Lahiri, G. K.; Bertagnolli, H.; Roduner, E. J. Phys. Chem. C 2009, 113, 2352−2359. (33) Hunger, M. Solid State Nucl. Magn. Reson. 1996, 6, 1−29. (34) Bertarione, S.; Scarano, D.; Zecchina, A.; Johánek, V.; Hoffmann, J.; Schauermann, S.; Frank, M. M.; Libuda, J.; Rupprechter, G.; Freund, H.-J. J. Phys. Chem. B 2004, 108, 3603−3613. (35) Grill, C. M.; Gonzalez, R. D. J. Phys. Chem. 1980, 84, 878−882. (36) Yudanov, I. V.; Sahnoun, R.; Neyman, K. M.; Rösch, N.; Hoffmann, J.; Schauermann, S.; Johánek, V.; Unterhalt, H.; Rupprechter, G.; Libuda, J.; Freund, H.-J. J. Phys. Chem. B 2003, 107, 255−264. (37) Vannice, M. A.; Wang, S.-Y. J. Phys. Chem. 1981, 85, 2543− 2546. (38) Guerrero-Ruiz, A.; Yang, S.; Xin, Q.; Maroto-Valiente, A.; Benito-Gonzalez, M.; Rodriguez-Ramos, I. Langmuir 2000, 16, 8100− 8106. (39) Heinrichs, B.; Noville, F.; Schoebrechts, J.-P.; Pirard, J.-P. J. Catal. 2000, 192, 108−118. (40) Crabtree, R. The Organometallic Chemistry of the Transition Metals; Wiley-Interscience: Hoboken, NJ, 2005.

6979

dx.doi.org/10.1021/jp2113297 | J. Phys. Chem. C 2012, 116, 6973−6979