J. Phys. Chem. C 2008, 112, 18521–18530
18521
Carbon Monoxide Oxidation as a Probe for PtRu Particle Surface Structure Elena R. Savinova,*,† Francoise Hahn,‡ and Nicolas Alonso-Vante‡ l’Ecole Europe´enne Chimie Polyme`res Mate´riaux, UniVersite´ Louis Pasteur, UMR 7515 du CNRS-ULP-ECPM, 25, rue Becquerel, F-67087 Strasbourg, France, and Laboratory of Electrocatalysis, UMR-CNRS 6503, UniVersite´ de Poitiers, 40 AVenue du Recteur Pineau, F-86022 Poitiers, France ReceiVed: July 2, 2008; ReVised Manuscript ReceiVed: September 5, 2008
The carbon monoxide molecule was used to probe the surface properties of PtRu(1:1)/C catalysts with 5, 30, and 60 wt % metal. The spectroscopic IR study performed in the external reflection configuration clearly showed that linearly bonded CO dominates the spectra for all catalysts, although the presence of bridged adsorbates was also revealed. These nanostructures, consisting of metal grains interconnected via grain boundaries, and formed at high metal loadings on carbon support, strongly influence the catalytic oxidation of the adsorbed carbon monoxide. It is proposed that in the potential interval 0.3 < E < 0.45 V, O(H)ads formation occurs preferentially at the active sites which are presumably located at the emergence of the grain boundaries at the particle surfaces. Dissolved CO strongly inhibits CO oxidation on multigrained materials while for highly dispersed isolated PtRu nanoparticles it shifts the oxidation onset negative. The spatial distribution of the adsorbed CO molecules on the catalyst surface, and consequently the extent of their dipole coupling depend on the nanostructure. It is suggested that surface defects, created at the emergence of the grain boundaries to the surface, facilitate disintegration of compressed islands of adsorbed CO. 1. Introduction The catalytic activity of metals is known to depend substantially on the presence of structural defects, and well-defined surfaces have served as models to get an insight into their role in catalysis1 and electrocatalysis.2 In particular, oxygen reduction,3-6 carbon monoxide,5,7-9 and methanol oxidation5,10 are relevant examples giving evidence of the sensitivity of the electrocatalytic activity to the presence of structural defects, such as e.g. monatomic steps. On the other hand, electrocatalysts used for practical applications are usually composed of nanometersized (metal) particles supported onto carbon substrates. Numerous studies have been performed to unveil the dependence of their activities on the particle size,11,12 and on the presence of various structural defects. Relevant examples concern the oxygen reduction,13-19 methanol,20-23 and carbon monoxide oxidation reactions.24-28 Further phenomena influencing the electronic and adsorption properties of materials and affecting the (electro)catalytic behavior are the presence of bulk crystalline defects (twins, dislocations, or grain boundaries), microstrains, and foreign atoms. In this context, Tsybulya et al.29 have drawn the attention to the role of grain boundary regions between metal particles in heterogeneous catalysis. They have shown that the catalytic activity of Ag catalysts was greatly enhanced when particles were composed of nanometer-sized grains connected via grain boundaries. Recently, Waterhouse et al.30,31 proposed that oxygen adsorption on silver catalysts in the vicinity of grain boundaries results in the formation of strongly bound Oγ species which participate in selective oxidation of methanol to formaldehyde. Materials composed of nanometer-sized grains connected to each other via grain boundaries have acquired the name “nanostructured” or “nanocrystalline” and have long been the subject of attention of materials scientists due to their unique * Corresponding author. E-mail:
[email protected]. Phone: (0033) (0)390 24 27 39. Fax: (0033) (0)390 24 27 61. † Universite ´ Louis Pasteur. ‡ Universite ´ de Poitiers.
physical and mechanical properties.32-34 Formation of nanostructured materials composed of interconnected nanometersized grains is especially characteristic of fuel cell electrocatalysis, where high metal loadings per unit support area are normally utilized to minimize the thickness of the catalytic layers. The importance of grain boundaries for electrocatalysis on Pt nanomaterials has recently been interrogated by Cherstiouk et al.26,27 and then by Maillard et al.35 Later Gavrilov et al.36 demonstrated that the presence of defect grain boundary regions is of importance also for bimetallic electrocatalysis. It was shown that the catalytic activity of PtRu(1:1)/C in carbon monoxide and methanol oxidation increases dramatically with the metal percentage (and thus with the metal loading per unit support surface area). This has been attributed to the formation of multigrained catalysts comprising ordered crystalline domains interconnected via grain boundaries. It was suggested that these defect grain boundary regions provide active sites for formation of adsorbed oxygen species which are of paramount importance in the oxidative catalysis. In the present work, bimetallic PtRu(1:1)/C catalysts, with 5, 30, and 60 wt % loadings, were investigated with FTIR spectroscopy, using carbon monoxide as a probe molecule. The electrochemistry of this molecule plays an important role in lowtemperature fuel cell reactions, since in the conversion of small organic molecules (e.g., CH3OH) it is formed as a stable intermediate. It also represents a valuable tool to monitor in situ the interfacial properties of extended and/or nanodivided materials. The mid-IR vibrational bands for adsorbed CO and for CO2 formed in the vicinity of the electrode were monitored as a function of potential and coverage of COads to evaluate the adsorptive and electrocatalytic properties of surface defects in PtRu nanomaterials. 2. Experimental Section Sample Preparation and Characterization. PtRu/C catalysts were prepared using homemade pyrolytic carbon Sibunit 19P
10.1021/jp805807u CCC: $40.75 2008 American Chemical Society Published on Web 10/31/2008
18522 J. Phys. Chem. C, Vol. 112, No. 47, 2008
Savinova et al.
TABLE 1: Structural Characteristics of PtRu/C Catalystsa wt % PtRub
metal loading,c mgPtRu m-2
D
djs, nm
dchem, nm
dCST, nm
Achem, m2 g-1
AEC, m2 g-1
ξ
5 30 60
0.70 5.9 21
0.59 0.16 0.10
2.1 -d -d
1.7 6.1 10.0
1.6 2.8 3.6
190 54 33
152 58 37
0.06 0.54 0.64
a See text for details. b Metal weight percentage. c Metal loading per unit surface area of support. structures consisting of ordered crystalline domains connected via grain boundaries.
(Omsk, Russia, BET surface area 72 m2 g-1). Carbons of Sibunit family are mesoporous materials produced by high-temperature pyrolysis of hydrocarbons.37 Low BET surface area Sibunit carbons have recently shown superior properties over conventional Vulcan carbon as supports for PtRu anodes for direct methanol fuel cell.38 In this work we utilize three Sibunitsupported bimetallic samples, containing 5, 30, and 60 wt % PtRu with the atomic ratio 1:1. The catalyst preparation procedure included cohydrolysis of chloride complexes of RuIII and PtIV and is described in ref 38. Electrochemical Measurements. Solutions were made from Milli-Q water (18 MΩ cm), H2SO4 (Suprapur, Merck). Working electrodes were prepared by pipetting 2-20 µL of aqueous suspension containing 1.5 mg mL-1 of PtRu/C (no Nafion) on a polycrystalline gold foil (cleaned with Aqua Regia and annealed in the hydrogen flame) or polished polycrystalline Au disk (cleaned with Aqua Regia) and dried under the Ar flow. The resulting films were stable, and showed reproducible cyclic voltammograms (CVs). After each experiment PtRu/C was removed from the gold substrate by wiping the surface under a water flow. To ensure that no sample remained on the gold surface from the previous experiment, CV of the Au electrode was recorded before each new deposition. Electrochemical measurements were carried out in a three-electrode cell at 298 K in Ar atmosphere. The counter electrode was Pt foil and the reference electrode a mercury sulfate electrode (MSE) Hg/ Hg2SO4/0.1 M H2SO4 or trapped hydrogen electrode connected to the working electrode compartment via Luggin capillary. Potentials were controlled with a Autolab PGSTAT30 potentiostat and are quoted vs reversible hydrogen electrode RHE (EMSE) 0.73 V vs. RHE). Infrared Spectroscopy. FTIR measurements were performed at room temperature under external reflection conditions in a thin layer configuration with a Bruker IFS66 vacuum FTIR spectrometer modified for beam reflectance at a 65° incident angle. The spectroelectrochemical cell was made from Duran glass and contained a CaF2 window at the bottom. The detector was liquid N2-cooled MCT. To obtain a single-beam spectrum, 400-800 interferograms acquired at spectral resolution 4 or 8 cm-1 were coadded and then Fourier transformed. The IR absorption spectra were calculated as changes in the reflectivity (Ri) using the formula:
∆R Ri - Rref × 100% ) R Rref
(1)
Here Rref corresponds to a reference single-beam spectrum. The electrode consisted of a polycrystalline Au disk (7 mm) connected to a glass shaft. The surface of the disk was polished to a mirror finish with Al2O3 paste (A5, 0.05 µm). Deposition of PtRu/C on the gold disk was performed as described above. CO was adsorbed at 0.1 V. For the experiments in CO-free electrolyte CO was bubbled through the electrolyte for 5 min, and then the electrolyte was purged with N2 for 55 min to ensure that no dissolved CO remained in the electrolyte. Then the electrode was pressed to the IR window, and the measurements
d
Particles coalesce forming extended
were performed. This procedure ensured saturation COads coverage and allowed sufficient time for the adlayer to equilibrate. For the experiments in CO-saturated electrolyte, CO was admitted to the cell while the electrode was held at 0.1 V, and bubbled through the cell for 10 min, then the electrode was pressed to the window, and IR spectra were acquired. 3. Results and Discussion 3.1. Sample Characterization. Structural characterization of PtRu/Sibunit catalysts with metal percentage ranging from 5% to 60% using X-ray diffraction (XRD), high resolution transmission electron microscopy (HRTEM), energy dispersive X-ray analysis (EDX), and gas phase CO chemisorption was previously reported.36 Here we provide a brief description of the structural data which are essential for the further discussion. According to TEM data, 5 wt % PtRu/C contains small metal particles finely dispersed over carbon support. The average particle size was estimated from three independent methods: (i) TEM, as the surface average particle size djs ) ∑i nidi3/∑i nidi2, (ii) XRD, as the size of coherent scattering domain dCST, and (iii) from the metal dispersion D determined from the gas phase CO chemisorption. D is defined as the number of adsorbed CO molecules NCO normalized to the total number of metal atoms NPt and is inversely proportional to the average particle diameter dchem:
D)
NCO constant ) NPt dchem
(2)
The “constant” has dimensions of nanometers, depends on the stoichiometry of adsorption, atomic density, and the particle shape, and is usually determined experimentally. For finely dispersed Pt, Ru, and PtRu particles the empirical values of the constant were reported as 1.08,39 0.91,40 and 1.00 nm,38 correspondingly. Note that the empirical value given above for Pt is in excellent agreement with the value calculated from the atomic density, and assuming spherical particle shape, and stoichiometry of adsorption 1 CO per 1 surface atom (see e.g. ref 41). The values of djs, dCST, and dchem are compared in Table 1. One may see that for 5 wt % PtRu/C the average particle sizes determined with the three independent methods are in fairly good conformity. However, for the high loading 30 and 60 wt % samples dchem systematically exceeds dCST. This apparent inconsistency is rationalized by the results of TEM and HRTEM. The latter show that as the metal loading increases, isolated particles coalesce forming dendrite-like structures (see Figure 9 of ref 36). The coalescence degree increases with the metal loading. High-resolution images (Figure 10 of ref 36) show that these dendrite-like structures consist of small (a few nanometers) ordered crystalline domains interconnected via grain boundaries. The size of these crystalline domains corresponds fairly well to dCST calculated from the XRD line broadening (Table 1). Meanwhile, the physical dimensions of the particles are in accordance with the results of CO chemisorption. According
Probing the Surface Properties of PtRu/C Catalysts
J. Phys. Chem. C, Vol. 112, No. 47, 2008 18523
to our experience, multigrained structures are generic for fuel cell catalysts with high metal percentage. However, this can only be revealed if catalyst samples are characterized with both XRD and HRTEM. Specific surface area of the metal in this work was accessed by (i) gas phase CO chemisorption and (ii) electrochemical CO stripping. We are aware of the problems widely discussed in the literature (see e.g. ref 42) and associated with the superposition of CO oxidation and surface metal oxidation currents. However, as demonstrated below, our experimental data prove suitability of the CO stripping method for accessing specific metal surface areas, at least in the case of PtRu/Sibunit samples with constant Pt:Ru ) 1:1 ratios. Assuming particles with the aspect ratio 1 (spherical or cubic), specific surface area of the metal Achem (in m2 g-1) can be estimated with eq 3:
6000 Fdchem
Achem )
(3)
Here F is the density of the metal (in g cm-3), and dchem is the above-defined particle diameter (in nm) determined from CO chemisorption. The density of PtRu can be roughly estimated as an average of the corresponding values for Pt (21.45 g cm-3) and Ru (12.41 g cm-3).43 Thus calculated Achem values are given in Table 1. The “electrochemically active” surface area per unit mass of the metal AEC was calculated by using eq 4:
AEC )
Q mqML
(4)
Here m is the mass of the metal in the catalyst layer deposited on the electrode, and Q is the CO stripping charge, calculated as the area under CO stripping peak versus the background (the second scan). qML is the charge density corresponding to the oxidation of CO monolayer, which according to refs 44 and 45 was taken as 0.385 mC cm-2. The data of Table 1 prove that the values of Achem and AEC are in fairly good agreement. To minimize the experimental error, the currents were normalized to AEC determined for each electrode on the day of the measurement. The coalescence degree ξ may be defined as the ratio of the area of crystalline domains confined in the grain boundary regions to the total area they would have if they were not coalesced. ξ was estimated by using eq 5 and given in Table 1.
ξ)1-
Achem AXRD
(5)
Here AXRD stands for the surface area determined by using the size of coherently scattering domains determined from XRD data. It should be noted that the coalescence degrees observed in this work are higher than those usually obtained for Vulcan XC-72-supported catalysts46 due to the lower specific surface area of the Sibunit 19P support. An important issue for bimetallic catalysts is their composition and the degree of component alloying. According to the EDX data, the composition of the samples with various metal loadings does not differ from the nominal 50:50 by more than (10%. XRD data suggest that a 5 wt % sample contains PtRu alloy particles, while the samples with 30% and 60% PtRu comprise also some amount of nonalloyed coarsely dispersed Pt, whose contribution to the overall surface area, however, does not exceed 3%.36 3.2. Electrochemical Properties. Figure 1a shows cyclic voltammograms (CVs) of PtRu/Sibunit samples in 0.1 M H2SO4
Figure 1. (a) Cyclic voltammograms and (b) CO stripping voltammograms acquired for PtRu/Sibunit samples with different metal contents at 298 K and 5 mV s-1 in 0.1 M H2SO4: 5 wt %, black; 30 wt %, blue; 60 wt %, red.
electrolyte. Currents are normalized to the metal surface area estimated from CO stripping charges. Considering the sensitivity of CVs to the surface composition of PtRu electrodes, the remarkable similarity of the base voltammograms strongly suggests that the surface composition of PtRu(1:1)/Sibunit samples is independent of the metal loading. Gasteiger et al.44,47,48 have systematically followed the influence of the surface composition of bulk PtRu alloys on the shapes of CVs in supporting H2SO4 electrolyte. Comparison of the CVs of this work with those reported by Gasteiger et al. suggests that the surface composition of our homemade PtRu/ Sibunit catalysts was close to their nominal bulk 50:50 composition. Using the data from ref 48, we constructed a plot of pseudocapacitance calculated from the double layer splitting of CV at E ) 0.4 V vs the surface fraction of Ru (not shown). This plot is linear in the interval from 0.07 e xRu e 1.0 and was used to estimate the surface composition of PtRu/Sibunit samples. It was found that the surface composition of 5 to 60 wt % PtRu/C samples is independent of the metal loading, and that the surface fraction of ruthenium xRu is equal to ca. 0.47. It should be noted, however, that this constant surface composition required that the potential limits were kept within 0.03 e E e 0.80 and that the electrode was not exposed to long-term cycling in the presence of CO. Note that Maillard et al.46 observed an influence of the metal loading on the distribution of Ru on the surface of PtRu particles supported on Vulcan XC-72. The differences between the results of this work and those of ref 46 may be ascribed to the catalyst preparation procedures. CO stripping voltammograms represented in Figure 1b evidence remarkable influence of the metal percentage on the catalytic activity in the oxidation of adsorbed CO. As the metal dispersion increases, both the oxidation onset and the peak potential shift positive. The difference between the peak potentials for 5 and 60 wt % PtRu/C is ca. 110 mV. One should also notice that as the metal dispersion increases, the current tailing becomes more pronounced at potentials positive of the current peak. Given the results of the structural characterization as well as similar shapes of the base voltammograms for the samples under study, we infer that the differences in the catalytic activity for CO monolayer oxidation are related to the structural
18524 J. Phys. Chem. C, Vol. 112, No. 47, 2008
Savinova et al.
Figure 3. FTIR spectra of saturated CO adlayer on PtRu/C with different metal percentage: (1) 5 wt %; (2) 30 wt %; (3) 60 wt %. Spectra are taken in CO-free 0.1 M H2SO4 at 0.1 V vs RHE. Eref ) 0.8 V vs RHE for 5 wt % PtRu/C and 0.7 V for 30 and 60 wt % PtRu/C. Figure 2. FTIR absorption spectra for CO monolayer adsorbed on 60 wt % PtRu/C in CO and CO2 region in CO-free 0.1 M H2SO4. CO was adsorbed at 0.1 V. The reference spectrum was taken at 0.7 V. The catalyst layer on the Au substrate was established by pipetting from 2 to 40 µL of suspension containing 1.5 mg of sample per milliliter.
rather than compositional changes. This conclusion will be further corroborated by FTIR data. 3.3. Infrared Reflection-Absorption Spectroscopy. 3.3.1. CO ML Spectra Depending on the Catalyst Amount, Metal Dispersion, and CO CoWerage. FTIR spectra of CO adsorbed at saturation coverage on 60 wt % PtRu/C at 0.1 V vs RHE are shown in Figure 2. To obtain the absolute bands, the single-beam spectrum taken after the oxidation of adsorbed CO was used as the reference. The spectra are dominated by linearly bonded CO whose band is inverted. The occurrence of inverted IR bands for molecules adsorbed on metal particles dispersed on poorly to moderately reflecting substrates (e.g., carbon) has been documented for both external and internal reflectance configuration.49-51 This phenomenon has been explained and quantitatively described by Pecharroman et al.,52,53 who simulated the shapes of the spectra as a function of the volume fraction of metal particles in a thin film matrix of a moderately reflecting glassy carbon substrate. Park et al.54,55 suggested that the “anti-absorbance” features may be diminished or even suppressed by spreading a very thin film of carbon-supported metal nanoparticles on highly reflecting Au electrode. However, Figure 2 shows that the decrease of the amount of the catalyst suspension from 40 to 2 µL only influenced the signal-to-noise ratio but not the band shape. When the amount of suspension was decreased to 2 µL, the catalyst film over Au electrode was barely visible but still gave rise to the same band shape. Note that the electrodes were always rinsed with Milli-Q water after drying the catalyst film to remove particles loosely bonded to the electrode surface as suggested by Park et al.55 We assume that the inverted shape of the spectra in this work results from the IR light reflecting characteristics of the catalysts on low surface area Sibunit substrate, which according to the terminology of ref 52 have high “fill factors”. Figure 3 demonstrates that FTIR spectra of CO adsorbed on 5, 30, and 60 wt % PtRu samples in the CO spectral region are dominated by the vibration band at 2045-2064 cm-1, which red-shifts with the increase in PtRu dispersion. This band is
typical for a-top CO adsorbed on monometallic Pt and/or bimetallic PtRu surfaces and will be further addressed as COL(PtRu).56-60 According to Lin et al.,60 the wavenumbers for a-top CO are 2065-2075, 2055-2065, and 2000-2020 cm-1 for polycrystalline Pt, PtRu (1:1). and Ru, respectively. It has been documented in the literature (see e.g. refs 58 and 61) that the a-top CO band red-shifts with the amount of Ru. Depending on the experimental conditions, less intense bands at 1969-1997 and 1860-1882 cm-1 could also be observed. The former may be attributed to a-top CO on Ru atoms COL(Ru),62 while the latter to the bridge-bonded CO either on PtPt or PtRu sites (COB(PtRu)).56-60 According to the literature data, alloy surfaces usually show a single vibration band of a-top CO, while simultaneous observation of COL(Pt) and COL(Ru) is typical for Ru decorated Pt surfaces with relatively large domains of Pt and Ru.63-66 Thus, spectra observed in this work evidence rather “homogeneous” surface composition of the homemade catalysts whose surface could either be formed by well intermixed Pt and Ru atoms or by pure Pt. The latter hypothesis may, however, be ruled out on the basis of the CVs in supporting electrolyte which are typical for PtRu electrodes with 50:50 surface composition rather than for Pt-rich electrodes.44,48 Observation of a weak band of COL(Ru) may be attributed either to the formation of some amount of isolated Ru phase or to occasional Ru clusters on the surfaces of alloy particles. The second hypothesis seems more likely in view of the observed spectra evolution during stepwise CO oxidation and will be discussed in more detail further in the text. Component segregation in PtRu particles has been suggested in a number of publications67-69 and has been shown to depend on the experimental conditions and on the nature of the adsorbates. The band intensity in the infrared spectra is not directly related to the concentration of oscillators, being strongly affected by their dipole-dipole coupling,70,71 whose extent decreases with the decrease of the adsorbate coverage. To obtain more insight into the surface composition, CO was dosed onto different samples from dilute solutions. Figure 4 demonstrates the influence of CO fractional coverage XCO (the ratio of the actual CO coverage to the saturation coverage) on the spectroscopic characteristics of C-O vibration for 60 and 5 wt % PtRu/C.
Probing the Surface Properties of PtRu/C Catalysts
Figure 4. FTIR spectra of CO adlayers on 60 and 5 wt % PtRu/C in 0.1 M H2SO4 for different fractional CO coverages XCO as indicated in the figure. CO was dosed from dilute CO-containing electrolytes and then removed from electrolyte by N2 purging for 1 h. Spectra are taken at 0.1 V vs RHE. Eref ) 0.8 V vs RHE for 5 wt % PtRu/C and 0.7 V for 60 wt % PtRu/C.
Figure 5. Evolution of FTIR spectra upon CO monolayer oxidation as a function of the applied electrode potential taken with 50 mV interval for 60 wt % PtRu/C in CO-free 0.1 M H2SO4 in CO (a) and CO2 (b) spectral regions. Eref ) 0.75 and 0.05 V vs RHE, respectively.
Note that XCO was determined from the intensity of CO2 signal obtained after full oxidation of the adsorbed CO. The figure shows low COL(Ru) band intensity in the whole range of XCO. This allows “massive” segregation or phase separation to be ruled out, and suggests that Pt and Ru are indeed well intermixed on the surface. The surface composition of the samples being independent of the metal percentage, the red-shift of COL(PtRu) from 60 to 5 wt % may be attributed to the increase in the metal dispersion. Such red-shift has been documented for monometallic Pt particles by Park et al.54 3.3.2. Oxidation of Adsorbed CO: Influence of the Catalyst Structure. To investigate CO oxidation, the electrode potential was increased in a stepwise manner from 0.05 to 0.80 V. The evolution of FTIR spectra in the spectral regions of CO and CO2 stretching vibrations is shown in parts a and b of Figure 5 for 60 wt % PtRu/C. Figure 6 compares the normalized intensities for CO2 and COL(PtRu) bands for 5 to 60 wt % PtRu/C at saturation and fractional CO coverages as well as in the presence of dissolved CO. We first concentrate on the data obtained in the absence of CO in the electrolyte.
J. Phys. Chem. C, Vol. 112, No. 47, 2008 18525 In accordance with the electrochemical results, FTIR data confirm that an increase in the metal loading results in a remarkable increase of the electrocatalytic activity toward CO oxidation at the saturation coverage, which is evidenced by the negative shift of the CO2 onset. The decrease of XCO from 1 to 0.48 results in only a small displacement of the oxidation onset. This confirms that the differences between the oxidation onsets for 5%, 30%, and 60% PtRu/C cannot be explained by eventually different saturation coverages, and must be attributed to the structural changes occurring upon increasing metal percentage. Comparing CO stripping voltammograms of PtRu/ Sibunit catalysts with the literature data (cf. stripping voltammograms in, e.g., refs 36 and 72), we come to the conclusion that the particle size alone cannot account for the obserVed actiVity enhancement. For more detailed discussion the reader is referred to ref 36. We propose that the major activity enhancement is due to the formation of multigrained PtRu structures. It is likely that the emergence of the grain boundaries to the particle surfaces creates extended defect regions important in catalysis and electrocatalysis. The catalytic role of the surface defects on PtRu electrodes was highlighted by Hoster et al.,73 who demonstrated that ruthenium alloyed into smooth Pt(111) terraces had rather low electrocatalytic activity toward methanol oxidation. However, the activity experienced many-fold increase after the electrode surface was sputtered with Ar+. Thus, the role of surface defects in electrocatalysis on bimetallic electrodes is at least as important as that in electrocatalysis on monometallic surfaces.3,4,6-10,74,75 Note that the role of the grain boundaries may be much more complex than that of the steps and kinks on the extended surfaces. It has been shown that metal atoms located in the grain boundary regions have decreased coordination numbers and lower local densities. The latter may facilitate formation of subsurface oxygen species which may play an important role in the reaction dynamics.76-78 3.3.3. Oxidation of DissolWed CO: Influence of the Catalyst Structure. To obtain more insight into the catalytic properties of PtRu catalysts we oxidized CO in the presence of CO in the electrolyte. Figure 6 demonstrates that for the 60 wt % sample the onset of CO oxidation remains basically unchanged, while the main peak is strongly displaced to the positive potentials. Qualitatively similar changes are observed for the 30 wt % sample, although the overall magnitude of the positive displacement is smaller. On the contrary, for the 5 wt % PtRu/C sample no positive displacement is observed. Moreover, the onset of CO oxidation shifts negative by ca. 150 mV and a “pre-wave” appears. Comparison of CO2 and COL(PtRu) vs. potential profiles for the three catalysts proves that in the presence of CO the differences between their activities are to a certain extent leveled off and the reaction becomes structure insensitive. Figure 7 shows that COL(PtRu) wavenumbers in the presence of CO are considerably higher than those in the CO-free electrolyte suggesting higher COads coverage in the former case. According to Lin et al.,60 for polycrystalline PtRu in 0.1 M HClO4 ∂ν/∂θCO ) 50 cm-1. Assuming that νCO vs CO coverage is linear, we roughly estimate that the saturation CO coverage increases by ca. 0.15-0.2 upon CO admission into the electrolyte. The influence of CO in electrolyte on its oxidation kinetics has been discussed in a number of publications, see e.g. refs 79-81. As proposed by Batista et al.,81 the influence of CO pressure on its oxidation can be explained by two oppositely directed effects. On the one hand, the competition of CO molecules with H2O for the adsorption sites leads to suppression of CO oxidation and to the concomitant positive displacement
18526 J. Phys. Chem. C, Vol. 112, No. 47, 2008
Savinova et al.
Figure 6. Normalized intensity of COL(PtRu) (left-hand side) and CO2 (right-hand side) formed in the thin layer upon CO oxidation on 5 wt % PtRu/C, 30 wt % PtRu/C, and 60 wt % PtRu/C in 0.1 M H2SO4. Empty triangles correspond to CO fractional coverage XCO ) 0.48, filled triangles correspond to the saturated CO adlayer, and squares correspond to the oxidation of CO dissolved in solution. CO adsorption was performed at Eads ) 0.1 V, and the reference spectra were taken at Eref ) 0.80 V for the oxidation of monolayer and submonolayer CO and at 0.85 V for the oxidation of dissolved CO. The intensities are normalized to the corresponding maximum values. The decrease of CO2 intensity after the maximum is due to its escaping from the thin layer.
Figure 7. Evolution of the wavenumber of COL(PtRu) upon stepwise oxidation of CO ML (a) and dissolved CO (b) on 5 wt % PtRu/C (filled triangles), 30 wt % PtRu/C (open triangles), and 60 wt % PtRu/C (stars) in 0.1 M H2SO4. Arrows mark the onset of CO2 bands for corresponding samples under the same experimental conditions. For other details see Figure 6.
of the oxidation peak. The positive shift of CO oxidation upon CO admission in the electrolyte was documented in a number of publications. Lin et al.60 observed positive shifts for Pt, PtRu, and Ru electrodes and attributed these to the competition of dissolved CO and water molecules for the adsorption sites. On the other hand, under some conditions, the onset of CO oxidation has been found to shift negatively upon CO admission in the electrolyte. This can be corroborated by considering an increase in COads coverage in CO-saturated compared to CO-free electrolytes. For example, Cuesta et al.82 determined the equilibrium CO coverage of Pt(111) electrodes at room temperature in 0.1 M H2SO4 as a function of the CO partial pressure. The saturation CO coverage increased from 0.68 in CO-free to 0.75
in CO-saturated electrolytes.83 Meanwhile, surface science studies have demonstrated a significant decrease of CO binding energy to metal surfaces with the coverage due to repulsive CO-CO interactions.84 Considering this, Batista et al.81 rationalized the oxidation of dissolved CO on Pt(111) at potentials as low as 0.4 V in terms of the steep decrease of CO desorption energy at high degrees of coverage. Considering the experimental results obtained in this work and the literature data, the following tentative conclusions can be made. Since the adsorbate bonding is known to become stronger with the decrease of the metal coordination number,85 we assume that in the low potential interval of 0.3 < E < 0.45 V, formation of active oxygen species further participating in the CO oxidation reaction occurs preferentially at defect sites, which are likely to be located at the emergence of the grain boundaries to the particle surfaces. Admission of CO in electrolyte blocks these active sites. Now, in order that CO oxidation starts, some CO molecules must be either desorbed from the surface or displaced by oxygen species. As a basis for discussion we would like to juxtapose the observed marginal structural sensitivity of “bulk” CO electrooxidation with the gasphase CO oxidation at high CO partial pressures. On platinum metals the latter is usually structure insensitive, which is attributed to CO desorption being the rate-determining step (rds).86,87 3.3.4. CO-CO Dipole Interactions: EWidence for COads Island Formation. We now discuss the influence of the electrode potential on the COL(PtRu) wavenumber. In the potential interval of COads stability, νCOL(PtRu) increases with potential. ∂ν/ ∂E shows marginal dependence on the PtRu dispersion, and amounts to 22 cm-1 V-1 for 60 and 30 wt % samples, and 27 cm-1 V-1 for 5 wt % PtRu. These values are in agreement with the literature data for bulk polycrystalline bimetallic PtRu (32 cm-1 V-1 60) and monometallic Pt (28 cm-1 V-1 60) electrodes. The influence of the electrode potential on the adsorbate vibration frequencies has been widely discussed in the literature. It is attributed to a combination of the chemical and the Stark effects. The first is related to the influence of the electrode potential on the CO bonding to the metal (the extent of donation
Probing the Surface Properties of PtRu/C Catalysts and back-donation), while the second to the interaction of the adsorbed dipole with the electric field.88-90 For platinum metals the strength of CO bonding to the surface decreases with increasing positive charge density, resulting in the blue-shift of C-O vibration frequency. It is of interest to analyze the changes of νCOL(PtRu) after the onset of CO oxidation (Figure 7). For 60 wt % PtRu νCOL(PtRu) drops down right after the onset of CO oxidation. This is obviously due to the decrease of CO coverage and the concomitant lifting of CO-CO dipole coupling. A similar behavior was observed for bulk PtRu alloys.60 Increase of νCOL after reaching a minimum has also been observed by Stamenkovic et al.91 and attributed to the anion adsorption on the surface free sites and concomitant compression of COads domains. Analysis of Figure 7 suggests that the plots of νCOL(PtRu) vs the electrode potential are very sensitive to the structure of PtRu/C catalysts. Thus, an increase in the metal dispersion results in the increase of the gap between the onset of CO oxidation (determined as the potential at which CO2 is detected in the thin layer) and the potential at which νCOL(PtRu) starts decreasing. Thus, for 5 wt % PtRu CO oxidation commences above 0.45 V, but νCOL continues increasing, and does not drop until the electrode potential reaches ca. 0.65 V. The observed growth of νCOL unambiguously suggests that despite an oxidation of an amount of CO, its local coverage remains high. In the literature such behavior has been interpreted within the model of CO oxidation at the perimeter of compressed COads domains (islands).92,93 The formation of compressed CO islands at overall low CO coverage has been suggested on the basis of infrared studies at solid/gas94,95 and solid/liquid electrified interfaces96,97 for extended surfaces94-96 as well as for nanoparticles.98 Imaging techniques such as STM,99 PEEM (Photoemission Electron Microscopy), EMSI (Ellipsomicroscopy for Surface Imaging), FEM (Field Electron Microscopy), and FIM (Field Ion Microscopy) have also confirmed the existence of segregated CO and oxygen islands on the surfaces of single crystals100,101 and platinum particles.102 CO island formation at the interface between carbon-supported Pt nanoparticles and liquid electrolytes was discussed by Maillard et al.,103 who concluded that dipole coupling of CO molecules adsorbed on nanoparticles was stronger than that on extended surfaces. Several possible interpretations can be suggested to account for the increased stability of CO islands with high local coverage on small PtRu particles. These are the (i) increase of the strength of COads bonding to the surface, (ii) increase of the strength of anion bonding to the surface of small particles and concomitant compression of CO islands, and (iii) decrease of the concentration of structural defects which facilitate disintegration of CO islands. Analysis of Figures 6 and 7a suggests that the CO adlayer blocks the surface for anion adsorption, and the latter recommences only when XCO drops down to ca. 0.2. Thus, anions are unlikely to influence the COads adlayer dynamics at high CO coverages. The influence of the surface crystallography on the kinetics of dissipation of ordered COads domains was studied by Chang et al.104 The authors have found that ordered domains of adsorbed CO are stable on Pt(111) over prolonged periods of time (∼1 h), while on Pt(100) fast dissipation of ordered CO domains was observed by FTIR spectroscopy. Markovic et al.105,106 have stressed the influence of surface defects and anions on the stability and on the domain size. For example, the stability/domain size of p(2 × 2)-COads structure on Pt(111) was found to increase from KOH (∼3 nm between 0.05 < E < 0.3 V), to HClO4 (∼14 nm between 0.05 < E