PCCP 2014 Rh

PCCP PAPER Cite this: Phys. Chem. Chem. Phys., 2014, 16, 13136 Spectral evidence for hydrogen-induced reversible segregation of CO adsorbed on titania-supported rhodium† D. Panayotov,*a M. Mihaylov,a D. Nihtianova,ab T. Spassovc and K. Hadjiivanov*a The reduction of a 1.3% Rh/TiO2 sample with carbon monoxide leads to the formation of uniform Rh nanoparticles with a mean diameter of dp E 2.2 nm. Adsorption of CO on the reduced Rh/TiO2 produces linear and bridged carbonyls bound to metallic Rh0 sites and only a few geminal dicarbonyls of RhI. The n(CO) of linear Rh0–CO complexes is strongly coverage dependent: it is observed at 2078 cm 1 at full coverage and at ca. 2025 cm 1 at approximated zero coverage. At low coverage, this shift is mainly caused by a dipole–dipole interaction between the adsorbed CO molecules while at high coverage, the chemical shift also becomes important. Hydrogen hardly affects the CO adlayer at high CO coverages. However, on a partially CO-covered surface (yCO E 0.5), the adsorption of H2 at increasing pressure leads to a gradual shift in the band of linear Rh0–CO from 2041 to 2062 cm 1. Subsequent evacuation almost restores the original spectrum, demonstrating the reversibility of the hydrogen effect. Through the use of 12 CO + 13 CO isotopic mixtures, it is established that the addition of hydrogen to the CO–Rh/TiO2 system leads to an Received 17th March 2014, Accepted 14th April 2014 increase in the dynamic interaction between the adsorbed CO molecules. This evidences an increase in the density of the adsorbed CO molecules and indicates segregation of the CO and hydrogen adlayers. When DOI: 10.1039/c4cp01136h CO is adsorbed on a hydrogen-precovered surface, the carbonyl band maximum is practically coverage independent and is observed at 2175–2173 cm 1. These results are explained by a model according to www.rsc.org/pccp which CO successively occupies different rhodium nanoparticles. 1. Introduction The vibrational spectroscopy of adsorbed CO is among the most used and powerful methods for identification of active sites and elementary mechanistic steps in heterogeneous catalytic reactions proceeding on the surfaces of both single crystals and supported metals.1 The spectral performance of adsorbed CO is determined by different factors: geometry of the bonding, nature of the bond and the coverage.1–3 For linearly bonded CO, the position of the carbonyl band depends mainly on the balance between the s- and p-bonds.3–6 The s-bonding leads to an increase in n(CO), while the p-bonding causes the opposite effect.3 With metal surfaces, the p-bonding usually predominates and, as a result, n(CO) is detected below the CO gas-phase stretching frequency (2143 cm 1). When CO bridges two metal atoms, rehybridization to sp2 hybrid orbitals occurs, a Institute of General and Inorganic Chemistry, Bulgarian Academy of Sciences, Sofia 1113, Bulgaria. E-mail: dpanayo@svr.igic.bas.bg, kih@svr.igic.bas.bg b Institute of Mineralogy and Crystallography, Bulgarian Academy of Sciences, Sofia 1113, Bulgaria c University of Sofia, Department of Chemistry and Pharmacy, Sofia 1126, Bulgaria † Electronic supplementary information (ESI) available: Fig. S1–S7 and details on the CO interaction with single rhodium surfaces. See DOI: 10.1039/c4cp01136h 13136 | Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 which leads to a lower C–O bond order and n(CO) is shifted to 2000–1880 cm 1. Multi-centered carbonyls formed on threefold and four-fold hollow sites are observed at even lower wavenumbers (below B1880 cm 1).7 The position of n(CO) for linear carbonyls is primarily affected by the coordination state of the adsorption site. The lower its coordination number, the greater the adsorption strength and thus, the frequency of the n(CO) mode normally appears at lower wavenumbers. Note that low coordinate atoms on edges, corners and defects are likely to represent a large proportion of all the available adsorption sites on small particles.5,6,8 Usually, the position of the carbonyl bands is coverage dependent. For most metal surfaces, n(CO) for linearly adsorbed CO is blue shifted with the increase of coverage. This shift is mainly caused by the cooperative effect of two phenomena: (i) dipole–dipole coupling between the adsorbed molecules (dynamic shift) and (ii) competition between adsorbate molecules for electrons back donated from the metal to form p-bonds (chemical shift).3 Rhodium is an important component in many catalysts for CO conversion, e.g. selective CO hydrogenation,9,10 reduction of NO by CO in automotive exhaust gas,11,12 hydroformylation reactions (aldehyde production from olefins, carbon monoxide, This journal is © the Owner Societies 2014 Paper PCCP and hydrogen),13,14 etc. The Rh–TiO2 system is well known as a Fischer–Tropsch catalyst,15 where the details on the CO and H2 adsorption and coadsorption have fundamental importance. There are many studies of CO and H2 adsorption and coadsorption on rhodium single crystals. On both Rh(111) and Rh(100) planes, CO forms coverage dependent ordered structures consisting of linear (on top) and bridge-bound species.16–21 Two-fold bridge-bound CO has been observed on a Rh(100) surface22 while CO bound at three-fold bridge (hollow sites) sites is characteristic for a Rh(111) surface.17 Hydrogen adsorbs dissociatively on both Rh(111)23 and Rh(100)24 surfaces (at B100 K), and forms no ordered structures observable by LEED. Co-adsorbed CO and hydrogen are found to segregate on the close packed Rh(111) surface.25 On the more open Rh(100) surface, CO and hydrogen adsorbates segregate when CO is preadsorbed, whereas a mixed adlayer is formed when H2 is adsorbed first.18,24 The chemistry of CO adsorbed on supported rhodium catalysts is different. Numerous studies have shown that CO provokes oxidative disruption of supported metal rhodium particles, forming isolated rhodium(I) geminal dicarbonyls.26,27 The disruption of Rh crystallites is assisted by surface hydroxyl groups28–30 or surface defects31,32 of the oxide support. The effect of hydrogen on CO preadsorbed on oxide-supported rhodium has also been studied.9,29,33–35 However, most of these studies were focused on the interaction of H2 with isolated geminal dicarbonyl species of RhI and the reverse aggregation of isolated RhI ions to metallic Rh crystallites.28,29,36 The production of water during the interaction of hydrogen with rhodium on reducible supports is an additional factor that complicates the spectra compared to the single crystal systems. Generally, a detailed spectroscopic investigation of supported rhodium systems towards the effect of hydrogen on the adsorbed CO layer has not attracted the attention that has been paid when the effect was studied with single crystals. It is of definite interest to establish the phenomena occurring with supported metal nanoparticles because they are characteristic of the real Rh-metal oxide catalyst systems. The aim of this work is to study the effect of hydrogen on the CO adsorbate layer formed on titania-supported rhodium nanoparticles. To clarify the nature of this effect, we used IR spectroscopy of adsorbed CO and a 13CO-enriched 12CO + 13CO isotopic mixture. 2. Experimental 2.1. Sample preparation The support used for the deposition of Rh nanoparticles was a commercially available Degussa P25 titanium dioxide material containing 80% anatase and 20% rutile.37 The Rh/TiO2 sample was prepared by incipient wetness impregnation of 1.0 g of TiO2 with 2.5 mL of a 1.4 wt% aqueous solution of RhCl3
xH2O (38.7% Rh, VEB Bergbau und Hüttenkombinat ‘‘Albert Funk’’) and dried at 373 K. The Rh/TiO2 sample thus obtained contained, nominally, 1.3 wt% of rhodium. Before the IR experiments, the sample was This journal is © the Owner Societies 2014 reduced in situ in the IR cell by CO, as described below. For the TEM analysis, the reduced Rh/TiO2 sample was passivated by successive additions of small doses of oxygen in situ in the IR cell before exposing to air and transferring into the TEM equipment. 2.2. Techniques FTIR spectra were recorded at ambient temperature with a Nicolet Avatar 360 FTIR spectrometer accumulating 64 scans at a spectral resolution of 2 cm 1. Self-supporting pellets (ca. 10 mg cm 2) were prepared from the sample powder and treated directly in a purposemade IR cell. The cell was connected to a vacuum-adsorption apparatus with a residual pressure below 10 3 Pa. Prior to the adsorption experiments, the sample was first oxidized (15.0 kPa O2, 673 K, 30 min), evacuated at 523 K for 2 min and then cooled to 298 K under a dynamic vacuum. After that, the sample was reduced by exposure to 4.0 kPa CO at 523 K for 1 min and then evacuated at 298 K. The CO coverage measured by the integrated IR absorbance of rhodium carbonyls under these conditions was adopted as the saturation CO coverage of exposed Rh sites. Partial CO coverages were obtained by the evacuation of CO-saturated Rh/TiO2 at chosen elevated temperatures. A liquid nitrogen trap served to remove traces of water vapour from the cell during the experiments. Carbon monoxide (>99.5% purity) was supplied by Merck and labelled carbon monoxide 13CO (99 atom% 13C, o5 atom% 18O), by ISOTEC. Oxygen (>99.999%) and hydrogen (>99.999%) were both purchased from Messer. All the gases (CO, CO isotopic mixture, O2 and H2) were additionally purified by passing through a liquid nitrogen trap. The histogram of the metal particle size has been obtained by counting about 300 particles. 3. Results 3.1. Sample characterization by TEM The TEM study was performed with an ex situ CO-reduced Rh/TiO2 sample. The low-magnification TEM images (Fig. 1A and B) revealed that the TiO2 particles (average size 25 nm) are decorated by Rh nanoparticles. The recorded one- and twodimensional HRTEM images confirmed the phase composition of the TiO2 support. The histogram in Fig. 1C shows a rather narrow size distribution of Rh particles with a mean diameter of dp E 2.2 nm. An estimate showed that the Rh nanoparticles are pseudohemispherical and possess an aspect ratio (height/diameter) of B0.3–0.4. The (100) and (111) lattice fringes of the rhodium particles are also registered. Moreover, the (100) lattice fringes of anatase coexist with the (100) lattice fringes of rhodium (Fig. 1D) while the (101) lattice fringes of anatase TiO2 coexist with the (111) lattice fringes of rhodium (Fig. 1E). These data suggest that Rh nanoparticles with different exposed faces preferentially grow on different TiO2 crystal planes. 3.2. Adsorption of CO The IR spectrum of CO (4 kPa equilibrium pressure) adsorbed at 298 K on the CO-reduced Rh/TiO2 sample exhibits two sharp Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 | 13137 PCCP Paper Fig. 1 (A) Bright field TEM image of the 1.3 wt% Rh/TiO2 sample; (B) an enlarged view of the small Rh particles at the surface of TiO2; (C) a histogram of the apparent diameter of the Rh particles; (D) and (E) HRTEM images of B25 nm anatase particles bearing B2–3 nm Rh particles. bands at 2187 and 2078 cm 1 and a broad feature centered at 1870 cm 1 (Fig. 2A, spectrum a). The band at 2187 cm 1 is observed only in the presence of gas-phase CO and arises from CO bound to low-coordinated Ti4+ sites on the TiO2 support.38 The intense band at 2078 cm 1 is unambiguously assigned to linear Rh0–CO species while the broad band at B1870 cm 1 is attributed to bridged carbonyls formed on the Rh nanoparticles.8,36,38 The sample was evacuated at ambient and elevated temperatures (Fig. 2A spectra b–j). This resulted in the fast disappearance of the Ti4+–CO band at 2187 cm 1. The principal Rh0–CO carbonyl band decreased in intensity and was gradually shifted from 2078 cm 1 to 2026 cm 1. The dependence of the band position on the coverage is shown in Fig. 2B. The measured shift in n(CO) was Dn(CO) = 52 cm 1 for a CO coverage (yCO) change from B1 to B0.04.8 Computer treatment indicates a good approximation of the spectrum when the band of Rh0–CO carbonyls was deconvoluted into three components (see Fig. S1 of the ESI†). These components could be attributed to CO bound to Rh sites with different coordinations.5,8,39 For small Rh particles with mean sizes of 2–3 nm, the fraction of low-coordinated Rh sites at edge, step and corner positions may account for B50% of the total number of exposed rhodium.8 Unfortunately, it is difficult to estimate the fraction of various Rh sites from the IR spectra because of intensity transfer phenomena usually observed with CO adsorbed on metal surfaces.5,40,41 13138 | Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 Fig. 2 Panel (A): FTIR spectra of CO adsorbed at 298 K on CO-reduced Rh/TiO2. Equilibrium CO pressure of 4 kPa, spectrum a; evacuation at 298 K, b; and, sequential 10 min evacuations at 448 K (c–h) and at 523 K (i–j). The spectra are background corrected. Panel (B): dependence of n(CO) of the Rh0–CO species on the relative CO coverage. The dashed line is a guide for the eye. The changes in the region of the bridging carbonyls are more complex: the broad band consists of at least two main components. These features show different stabilities and the high-frequency component tends to red shift with decreasing coverage. Based on previous assignments,8 the bands in the 1910–1890 cm 1 region can be attributed to two-fold bridgebound CO while the bands in the 1870–1820 cm 1 region to CO bound to three-fold hollow sites. After evacuation at 298 K (Fig. 2A, spectrum b), a weak band at 2098 cm 1 becomes observable and is more discernible after a short evacuation at 448 K (Fig. 2A, spectrum c). This band is attributed to the symmetric modes of isolated RhI(CO)2 This journal is © the Owner Societies 2014 Paper PCCP geminal species,28,29,36,38 the respective antisymmetric modes being masked by the intense carbonyl band of the Rh0–CO species.29,36,38 The 2098 cm 1 band declined under a dynamic vacuum at elevated temperatures and vanished at 523 K. The position of the band was not affected by the coverage. Such behavior is consistent with the proposed assignment as long as the two CO ligands in the geminal species do not participate in dipole–dipole interactions.29,36 Separate experiments have revealed that the fraction of the geminal dicarbonyls is much more important with the hydrogen-reduced sample (Fig. S2, ESI,† spectrum b), in line with literature reports.36,38 RhI(CO)2 species were the principal carbonyls produced after CO adsorption on oxidized Rh particles after exposure to O2. In the latter case, however, some carbonates were produced after CO adsorption. The results evidence that the reduction with CO ensures that rhodium remains mainly in a metallic state in the presence CO at 298 K. After evacuation at 523 K, re-adsorption of CO (4 kPa) at 298 K practically reproduced the spectrum of the CO-saturated sample (shown in Fig. 2A, spectrum a), evidencing that no irreversible changes occurred with the Rh particles during the experiments. 3.3. Adsorption of H2 To obtain a clean surface, the reduced sample was evacuated at 548 K for 20 min. This procedure ensured the disappearance of all the carbonyl bands. Exposure of the sample to H2 gas at 298 K led to an immediate rise in the IR background absorbance, as shown in Fig. S3 (ESI†). The observed IR absorbance is featureless and increases exponentially from B4000 to 1000 cm 1. It is attributed to delocalized conduction band electrons that accumulate in TiO2 with the H2 exposure. Similar observations were recently reported for H2 adsorption on Au/TiO2.42 The results evidence that H2 dissociates on the metallic Rh particles and produces atomic hydrogen that spills over onto the TiO2 support. The H atoms protonate surface oxygen atoms to OH groups while injecting electrons into the conduction band of TiO2.15,42 These observations are consistent with the literature data on H2 adsorption on rhodium monocrystals where H2 was found to dissociate. More details on the sample interaction with H2 are presented in the ESI.† 3.4. Adsorption of CO on the hydrogen pre-covered sample The next experiments were designed to check the effect of pre-adsorbed hydrogen on the adsorption of CO. Before the H2 exposure, the CO-reduced Rh/TiO2 sample was first evacuated at 548 K for 20 min and then cooled down to 298 K under a dynamic vacuum. Then, hydrogen (4 kPa equilibrium pressure) was introduced to the sample and small portions of CO were successively added to the system to increase the CO coverage, as shown in Fig. 3. The Rh0–CO band appeared at 2073 cm 1 for yCO = 0.27 and its position changed only slightly with the coverage, reaching a value of 2075 cm 1 for yCO = 0.67. Note that this value is very close to the value observed for full CO coverage in the absence of hydrogen (Fig. 2A, spectrum a). The band for bridging carbonyls which appeared at ca. 1900 cm 1 This journal is © the Owner Societies 2014 Fig. 3 FTIR spectra of CO adsorbed at 298 K on a H2-precovered Rh/TiO2 sample. The sample was then exposed to H2 (4 kPa equilibrium pressure) and small portions of CO were added at 298 K (a–d). The spectra are background corrected. The inset shows the same spectra normalized according to the intensity of the band at 2075–2073 cm 1. was expressed even at low coverages. As seen from the inset in Fig. 3, where normalized spectra are presented, the relative intensity of the bands due to linear and bridge species remains the same at different CO coverages. The observations that under H2 equilibrium pressure, the ratio of liner/bridging carbonyls as well as that the positions of the carbonyl bands remain constant with CO coverage can be rationalized by assuming that, in the presence of pre-adsorbed hydrogen, even small amounts of adsorbed CO form an adlayer with a high density of CO molecules. In this CO adlayer, the interaction between CO oscillators is similar to that in the CO-saturated layer on the hydrogen-free surface. This supposition is confirmed by the computer treatment of the spectra for linear Rh0–CO species, which revealed a good approximation when the bands were deconvoluted into three components with positions similar to the positions observed with full CO coverage in the absence of hydrogen (see Fig. S4, ESI†). 3.5. Effect of hydrogen on pre-adsorbed CO The first series of experiments were performed with a Rh/TiO2 sample that was saturated with CO and then evacuated at 298 K. Just before the admission of H2, the spectrum of the sample exhibited a band at 2066 cm 1 due to linearly bound Rh0–CO species and the estimated yCO was 0.87 (Fig. S5, spectrum a). At this high CO coverage, the effect of H2 admission on the carbonyl bands was practically negligible: after 60 min exposure to hydrogen, the band at 2066 cm 1 due to Rh0–CO species was slightly blue shifted to 2069 cm 1 (Fig. S5, ESI,† spectrum b). Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 | 13139 PCCP Upon evacuation, this band appeared at 2063 cm 1 (Fig. S5, ESI,† spectrum c), i.e. at a slightly lower frequency than that registered before H2 exposure. This small difference is attributed to the desorption of some CO during evacuation. Thus, the results evidence that hydrogen hardly affects preadsorbed CO when the CO coverage is relatively high. The next series of experiments with co-adsorbed H2 were performed at a partial CO coverage, i.e. at yCO = 0.48, which was achieved after 50 min evacuation of the CO-saturated sample at 448 K. The IR spectrum registered before H2 admission (Fig. 4, spectrum a) featured one main Rh0–CO band positioned at 2041 cm 1. Only a very weak feature around 1900 cm 1 was detected in the region of bridging carbonyls, evidencing that CO was essentially adsorbed in the linear form. Additions of hydrogen to the system at progressively increasing H2 pressure from 0 to 4 kPa, led to a gradual blue shift of the band due to linear Rh0–CO species up to 2062 cm 1 (Fig. 4, spectra a–c). Concomitantly, a doublet with resolved maxima at 1926 and 1904 cm 1 due to bridging CO species developed. Note that in the absence of hydrogen, bridging CO is rather typical of high CO coverage. A further increase in the hydrogen pressure to 6.4 kPa (Fig. 4, spectrum d) hardly affected the carbonyl bands. Computer treatment of the spectra for linear Rh0–CO species again revealed a good approximation when the Rh0–CO bands were deconvoluted into three components (Fig. S6A and B, ESI†). Subsequent evacuation of the sample (Fig. 4, spectrum e) almost restored the Fig. 4 FTIR spectra of adsorbed CO during a CO + H2 co-adsorption experiment carried out at 298 K with Rh/TiO2 which contained partial CO coverage, i.e. yCO = 0.48. Adsorption of CO (4 kPa) followed by 1 min evacuation at 473 K (a), introduction of H2 to the system at equilibrium pressures of 0.5 (b), 4.0 (c) and 6.4 (d) kPa, and after evacuation at 1  10 5 kPa (e). The spectra are background corrected. 13140 | Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 Paper original spectrum of CO–Rh/TiO2: the band of linear Rh0–CO species was red shifted back to ca. 2030 cm 1 and was slightly broadened but its integral intensity remained the same; the 1926 cm 1 band of bridged carbonyls disappeared while that at 1904 cm 1 decreased in intensity. Another effect of hydrogen addition was the formation of a small amount of adsorbed water, as evidenced by a band that appeared at 1618 cm 1 (not shown). These traces of water are produced by the reduction of residual oxidized rhodium species and/or of the TiO2 support by H atoms produced by H2 dissociation on Rh. The band at 1618 cm 1 resisted evacuation at ambient temperature. Evidently, the presence of water led to a small red shift of the carbonyl bands (compare spectra a and e from Fig. 4). A similar effect has been reported for water–CO coadsorption on a Pt/TiO2 sample.43 The observation of a practically reversible blue shift in the frequency of the Rh0–CO band induced by H2 co-adsorption, as shown in Fig. 4, provides spectral evidence that hydrogen adsorption causes reversible changes in the CO adsorption layer. The results strongly suggest hydrogen-induced segregation of the CO adsorption layer. 3.6. Experiments involving a 12 CO + 13 CO isotopic mixture In order to confirm the hypothesis about the segregation of CO and hydrogen adlayers, we performed experiments involving a 12 CO + 13CO isotopic mixture enriched to 13CO. The experiments aimed at estimating the dynamic coupling of CO molecules when adsorbed either alone or co-adsorbed with hydrogen. In order for dynamic interaction to occur, the adsorbed molecules must: (i) be adsorbed on one plane and each close to the other, (ii) be parallel and (iii) vibrate with the same intrinsic frequency.3,44 Thus, the dynamic interaction will increase on increasing the density of molecules due to the enhanced vibrational coupling of adsorbed dipole molecules.3,44,45 The requirement that the molecules vibrate with the same frequency is the basis of the experimental determination of the dynamic frequency shift.44 If a 12CO molecule is surrounded by isotopically labelled 13CO molecules, it will not be involved in dynamic interactions. Therefore, the difference between the frequency ni(12CO) observed after the adsorption of 12CO, and the frequency nii(12CO) obtained with an isotopic 12CO + 13CO mixture containing a small amount of 12CO will be equal to the value of the dynamic shift, i.e. Dndyn = ni(12CO) nii(12CO). The Rh/TiO2 sample was reduced by exposure to a 12CO + 13 CO mixture (with an 1 : 9 molar ratio) at 523 K. The spectrum of the reduced Rh/TiO2 sample in contact with the 12CO + 13CO mixture at an equilibrium pressure of 80 Pa (Fig. 5, spectrum a) exhibits an intense IR band at 2022 cm 1 and a broad feature around 1850 cm 1. These two bands are assigned to Rh0–13CO and Rh0–(13CO)–Rh0 complexes, respectively. In addition, the band at 2022 cm 1 has a shoulder at 2042 cm 1 (the exact band position was obtained from the second derivative of the spectrum). When the CO coverage was reduced by evacuation at increasing temperatures (Fig. 5, spectra b–g), the 2022 cm 1 band and its satellite at 2042 cm 1 decreased in intensity and shifted to lower frequencies. Simultaneously, the shoulder was gradually converted This journal is © the Owner Societies 2014 Paper PCCP Fig. 6 FTIR spectra of the 12CO + 13CO isotopic mixture (1 : 9 molar ratio) adsorbed at 298 K on CO-reduced Rh/TiO2: after evacuation at 484 K for 20 min (a); after interaction with 4 kPa H2 at 298 K (b) and after evacuation at 298 K (c). Fig. 5 FTIR spectra of the 12CO + 13CO isotopic mixture (1 : 9 molar ratio) adsorbed at 298 K on CO-reduced Rh/TiO2. Equilibrium CO pressure of 80 Pa (a); evacuation at 298 K (b), and at 448 K with increasing time (c–g). into a separate band. Prolonged evacuation at 448 K resulted in two well resolved bands at 2018 and 1983 cm 1 (Fig. 5, spectrum g). The band at 2018 cm 1 is assigned to Rh0–12CO species. Because of the increase of the dynamic interaction between adsorbed 13 CO molecules at high coverage, the Rh0–13CO band is strongly shifted to high frequencies, thus approaching the value of the Rh0–12CO band. Let us first consider the spectra at low coverage (Fig. 5, spectra d–g). It is seen that, in this series of spectra, the Rh0–12CO band is practically coverage independent. This means that no measurable chemical shift of the band occurs at low coverage, which is consistent with the results reported by Linke et al.17 These authors concluded that on the Rh(111) surface at a coverage of up to 0.5 ML, all frequency shifts are due to dipole–dipole coupling while at larger coverages (yCO > 0.5), some chemical effects start to contribute. At a coverage of 0.33, the Rh0–13CO band was detected at 1983 cm 1 (Fig. 5, spectrum g). Considering the 12CO/13CO isotopic shift factor (0.97777),3 one can easily derive the expected position of the corresponding 12CO modes (i.e. 2028 cm 1) if the 12 CO molecules were involved in a dynamic interaction with the 13 CO molecules. However, the experimentally observed 12CO band (in 13CO surrounding) was at 2018 cm 1, i.e. the dynamic shift amounts to 10 cm 1 at this CO coverage. Using the same approach, the dynamic shift was calculated to increase to 17 cm 1 at a coverage of E0.58. (Fig. 5, spectrum e). At higher coverages (Fig. 5, spectra a and b), the chemical shift substantially contributed to the position of the carbonyl bands, as evidenced by the substantial blue shift of the band at around 2020 cm 1. Irrespective of this phenomenon, the This journal is © the Owner Societies 2014 dynamic interaction could be calculated using the same approach. Thus, for the highest coverage (y E 1) it was found to be 26 cm 1 (Fig. 5, spectrum a), i.e. about half of the total shift. Indeed, it is expected that the dynamic shift will decrease with CO coverage, reaching zero when a single CO molecule is adsorbed, as shown in Fig. S7 (ESI†). Note that, due to superimposition of the bands, we were able to calculate the dynamic shift only for the main component of the carbonyl band. We consider now the interaction of hydrogen with a preadsorbed isotopic mixture, as shown in Fig. 6. An initial CO coverage of yCO = 0.52 was obtained by evacuation of the preadsorbed CO isotopic mixture at 484 K for 20 min (Fig. 6, spectrum a). At this coverage, the dynamic shift was calculated to be 13 cm 1. Subsequent introduction of hydrogen (Fig. 6, spectrum b) resulted in a blue shift of the Rh0–12CO and Rh0–13CO bands. However, the shift was much more pronounced with the 13CO band. Calculations showed that in the presence of hydrogen, the dynamic shift increased to 18 cm 1. As already noted, some H2O was produced after the hydrogen submission and was not removed by subsequent evacuation. Thus, the most correct way to estimate the effect of hydrogen is to compare the spectra before and after the evacuation of hydrogen. After evacuation of the sample, the dynamic shift dropped to 13 cm 1, the value obtained before hydrogen adsorption. Thus, it appears that although water traces slightly affect the carbonyl spectra, it practically does not influence the dynamic interaction between the adsorbed CO molecules. 4. Discussion 4.1. Adsorption of CO The results of this study clearly show that the reduction of Rh/TiO2 in a CO atmosphere at 523 K produces rhodium nanoparticles exposing mostly metallic Rh sites that at room temperature adsorb Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 | 13141 PCCP Paper CO molecules as linear ‘‘on top’’ species and bridged structures. Only a very limited number of isolated RhI(CO)2 species were formed. In contrast, the reduction of Rh/TiO2 in a H2 atmosphere at the same temperature produced Rh nanoparticles that are disrupted upon the adsorption of CO at 298 K to form isolated RhI(CO)2 species (Fig. S2, ESI†), consistent with previous reports.36,38 A possible explanation is that the reduction of Rh/TiO2 with H2 probably produces Rh particles with rough surfaces that are easily disrupted upon adsorption of CO. In contrast, the reduction of Rh/TiO2 with CO likely produces metallic Rh nanoparticles with a smooth surface which are stable towards oxidative disruption in a CO atmosphere. This point of view is supported by the TEM photographs showing exposed (100) and (111) rhodium crystal planes on the rhodium nanoparticles (see Fig. 1). The spectrum of CO adsorbed on our CO-reduced sample is dominated by the band of linearly bound carbonyl species. The maximum of this band is strongly coverage dependent and indicative of CO coverage. The experiments with a 12CO + 13CO isotopic mixture (Fig. 5) indicated that at low coverages, the shift is essentially caused by dipole–dipole coupling of adsorbed CO molecules. This coupling also largely contributes to the shift occurring at higher coverages. Bridging CO is rather typical of high CO coverage. Based on previous studies with single crystal17,18 and nanoparticulate Rh,8 the broad complex band in the region of bridging carbonyls with maxima appearing in the regions of B1910–1890 cm 1 and 1870–1820 cm 1 are assigned to CO bound at two-fold bridge sites exposed on Rh(100)18 facets and three-fold hollow sites on more compact Rh(111)17 facets, respectively. According to TEM analysis, the supported Rh nanoparticles have a rather narrow size distribution, as shown in the histogram of Fig. 1C. Thus, it might be suggested that the asymmetric broadening of the band for linear CO species to the low frequency side is not associated with a substantially broad distribution of Rh particles sizes on the Rh/TiO2 sample. Rather, the band broadening for linear species is related to the population of Rh sites with different coordinations, as previously suggested for supported metal nanoparticles exposing both well-coordinated Rh sites (at Rh crystal planes) and lowcoordinated Rh sites (at steps, edges and corners).8,39,46 Here, as already mentioned, one should take into account the possible intensity transfer. Based only on the deconvolution of the CO spectra of the linearly bound CO species obtained at different coverages (Fig. S1, ESI†), it is impossible to directly assign the deconvoluted spectral components to the vibrations of CO adsorbed at different types of Rh sites (corner, edge, plane), as long as these spectral features are not well-resolved with coverage, as in the case of Pt/SiO2.39,46 4.2. Adsorption of H2 and co-adsorption of CO and H2 Hydrogen adsorbs dissociatively on both Rh(111) and Rh(100) single crystal surfaces, even at 100 K where H atoms occupy hollow sites.23,24,47 Dissociative adsorption of hydrogen on Rh nanoparticles supported on TiO2 has been observed at room temperature.48 In our IR study of H2 interaction with the Rh/TiO2 13142 | Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 sample at 298 K, we have observed an increase in the IR background absorbance, as shown in Fig. S3 (ESI†). Dissociation of H2 on metallic particles produces H atoms that can spillover onto the TiO2 support where they protonate the surface while injecting electrons into the conduction band of TiO2.42,48 When CO doses were adsorbed in the presence of hydrogen, the CO molecules formed a highly compressed CO adlayer with spectral characteristics typical of a CO-saturated sample in the absence of hydrogen. These observations indicate that CO successively saturates separate rhodium nanoparticles replacing adsorbed hydrogen. Small deviations of this model are responsible for slight variations in the CO frequency at different coverages (Fig. 4). These results differ from the findings with monocrystals, where the formation of mixed CO-hydrogen layers were observed for the (100) Rh plane even at 150 K.18,20 The results for hydrogen adsorption on a sample precovered by CO also indicate H2-induced changes in the density of the CO adlayer, resulting in a substantial shift in the CO stretching frequencies. The structure of the CO adlayer on the monocrystal surfaces strongly depends on the presence of co-adsorbed hydrogen. When CO is adsorbed first, the H atoms coming into the CO adlayer repel the nearby CO molecules which move away to reduce the repulsive interaction and thus the CO and H adlayers are segregated.20,24,25,49 When hydrogen was exposed to Rh nanoparticles highly saturated with CO (yCO E 0.87, Fig. S2, ESI†), no measurable changes in the vibrational spectra of both linear and bridgebound CO species were detected. These results were expectable, as far as no significant further compression of the CO adlayer was possible. However, when hydrogen was exposed to Rh nanoparticles partially covered with CO (yCO E 0.48, Fig. 2), a blue shift in the frequency of linear CO species was observed, signalling the compression of the CO adlayer. The observed H2-induced changes in the values of the dynamic shift for a given total CO coverage could be explained only by the change in the density of the adsorbed CO molecules. Therefore, the results of the 12CO + 13CO co-adsorption experiments strongly support the hypothesis of hydrogeninduced compression of the CO adlayer which enhances the dynamic interaction between the adsorbed CO molecules. Note, however, that the dynamic interaction remains weaker than that observed for the highest CO coverages in the absence of hydrogen. This indicates that the changes occur in the framework of the surface of a given rhodium particle and no transfer of CO between the metal nanoparticles occurs. As a result, the spectral characteristics of the compressed CO adlayer differ from those of a CO monolayer in the absence of hydrogen. Thus, the observations in this work clearly show that the repulsive interaction of adsorbed hydrogen with a CO adsorbate leads to the compression of the CO adlayer on each nanoparticle. The IR bands of linear CO species observed in the presence of a hydrogen adlayer were also deconvoluted into three spectral components (Fig. S4 and S6, ESI†), where each component was at a position that is blue shifted from that registered in the absence of hydrogen. This suggests that all the CO molecules that are adsorbed in the linear atop configuration at different This journal is © the Owner Societies 2014 Paper Rh sites (edges, corners, steps, and face sites) form dense islands separated from the islands formed by absorbed H, probably at (100), (110) and (111) facets. Under evacuation, the hydrogen adlayer is destroyed as hydrogen desorbs and CO coverage expands back over all the exposed Rh sites on the metal particle. Thus, a less dense CO adlayer is formed in which the dipole–dipole interaction between CO molecules is weaker and this reflects in a back red shift in n(CO). 5. Conclusions Rhodium particles obtained by CO reduction of Rh/TiO2 are highly resistant to CO-induced oxidative disruption: the adsorption of CO on the sample at ambient temperature results mainly in the formation of linear Rh0–CO and bridge-bound carbonyls whereas the amount of RhI(CO)2 geminal-dicarbonyls is negligible. The position of the carbonyl band due to linear CO species strongly shifts to higher frequencies with coverage increases. This shift is due to two effects: a dipole–dipole interaction between the adsorbed CO molecules (prevailing at low coverages) and competition of CO adsorbates for electrons donated from the solid (both effects contribute almost equally at high coverages). The adsorption of CO in the presence of hydrogen leads to successive occupation of rhodium nanoparticles. However, the adsorption of hydrogen on the surface of Rh particles partially pre-covered with CO leads to reversible segregation and compression of the CO adlayer on each rhodium nanoparticle. Acknowledgements The authors are indebted to the Union Centre of Excellence (Contract No. DCVP 02-2/2009 with the National Science Fund). The support of the European FP7 project ‘‘Beyond Everest’’ is gratefully acknowledged. Notes and references 1 G. Rupprechter, Annu. Rep. Prog. Chem., Sect. C, 2004, 100, 237–311. 2 A. Föhlisch, M. Nyberg, P. Bennich, L. Triguero, J. Hasselström, O. Karis, L. G. M. Pettersson and A. Nilsson, J. Chem. Phys., 2000, 112, 1946–1958. 3 K. I. Hadjiivanov and G. N. Vayssilov, Adv. Catal., 2002, 47, 307–511. 4 R. M. Hammaker, S. A. Francis and R. P. Eischens, Spectrochim. Acta, 1965, 21, 1295–1309. 5 P. Hollins, Surf. Sci. Rep., 1992, 16, 51–94. 6 C. Ruggiero and P. Hollins, J. Chem. Soc., Faraday Trans., 1996, 92, 4829–4834. 7 N. Sheppard and T. T. Ngyen, in Adv. Infrared and Raman Spectros., ed. R. J. H. Clark and R. E. Hester, Heyden, London, 1978, vol. 5, pp. 67–148. 8 M. M. Sean, M. Lundwall, F. Yang, Z. Zhou and D. W. Goodman, J. Phys.: Condens. Matter, 2009, 21, 474223. This journal is © the Owner Societies 2014 PCCP 9 S. D. Worley and C. H. Dai, Catalytic Activation of Carbon Dioxide, American Chemical Society, Washington, DC, 1988, vol. 363, pp. 133–146. 10 J. A. Anderson, M. W. McQuire, C. H. Rochester and T. Sweeney, Catal. Today, 1991, 9, 23–30. 11 B. E. Nieuwenhuys, Adv. Catal., 1999, 44, 259–328. 12 C. Dujardin, A. S. Mamede, E. Payen, B. Sombret, J. P. Huvenne and P. Granger, Top. Catal., 2004, 30–31, 347–352. 13 C. C. Miyagawa, J. Kupka and A. Schumpe, J. Mol. Catal. A: Chem., 2005, 234, 9–17. 14 S. M. McClure, M. J. Lundwall and D. W. Goodman, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 931–936. 15 T. M. Apple and C. Dybowski, J. Catal., 1981, 71, 316–319. 16 B. A. Gurney, L. J. Richter, J. S. Villarrubia and W. Ho, J. Chem. Phys., 1987, 87, 6710–6721. 17 R. Linke, D. Curulla, M. J. P. Hopstaken and J. W. Niemantsverdriet, J. Chem. Phys., 2001, 115, 8209–8216. 18 M. M. M. Jansen, J. Gracia, B. E. Nieuwenhuys and J. W. Niemantsverdriet, Phys. Chem. Chem. Phys., 2009, 11, 10009–10016. 19 I. Nakamura, Y. Kobayashi, H. Hamada and T. Fujitani, Surf. Sci., 2006, 600, 3235–3242. 20 L. J. Richter, B. A. Gurney and W. Ho, J. Chem. Phys., 1987, 86, 477–490. 21 M. M. M. Jansen, C. G. M. Hermse and A. P. J. Jansen, Phys. Chem. Chem. Phys., 2010, 12, 8054–8062. 22 M. M. M. Jansen, O. Caniaz, B. E. Nieuwenhuys and J. W. Niemantsverdriet, Langmuir, 2010, 26, 16239–16245. 23 J. T. Yates Jr, P. A. Thiel and W. H. Weinberg, Surf. Sci., 1979, 84, 427–439. 24 Y. Kim, H. C. Peebles and J. M. White, Surf. Sci., 1982, 114, 363–380. 25 E. D. Williams, P. A. Thiel, W. H. Weinberg and J. J. T. Yates, J. Chem. Phys., 1980, 72, 3496–3505. 26 S. Trautmann and M. Baerns, J. Catal., 1994, 150, 335–344. 27 H. F. J. Van’t Blik, J. B. A. D. Van Zon, T. Huizinga, J. C. Vis, D. C. Koningsberger and R. Prins, J. Am. Chem. Soc., 1985, 107, 3139–3147. 28 F. Solymosi and M. Pasztor, J. Phys. Chem., 1986, 90, 5312–5317. 29 P. Basu, D. Panayotov and J. T. Yates, J. Am. Chem. Soc., 1988, 110, 2074–2081. 30 M. I. Zaki, B. Tesche, L. Kraus and H. Knözinger, Surf. Interface Anal., 1988, 12, 239–246. 31 M. Frank, R. Kühnemuth, M. Bäumer and H. J. Freund, Surf. Sci., 1999, 427–428, 288–293. 32 J. Evans, B. Hayden, F. Mosselmans and A. Murray, Surf. Sci., 1992, 279, L159–L164. 33 F. Solymosi, A. Erdöhelyi and M. Kocsis, J. Catal., 1980, 65, 428–436. 34 F. Solymosi and M. Pásztor, J. Catal., 1987, 104, 312–322. 35 S. D. Worley, G. A. Mattson and R. Caudill, J. Phys. Chem., 1983, 87, 1671–1673. 36 Z. L. Zhang, A. Kladi and X. E. Verykios, J. Mol. Catal., 1994, 89, 229–246. Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 | 13143 PCCP 37 M. Addamo, V. Augugliaro, A. Di Paola, E. Garcia-Lopez, V. Loddo, G. Marci, R. Molinari, L. Palmisano and M. Schiavello, J. Phys. Chem. B, 2004, 108, 3303–3310. 38 D. A. Buchanan, M. E. Hernandez, F. Solymosi and J. M. White, J. Catal., 1990, 125, 456–466. 39 M. J. Lundwall, S. M. McClure and D. W. Goodman, J. Phys. Chem. C, 2010, 114, 7904–7912. 40 F. M. Hoffmann, Surf. Sci. Rep., 1983, 3, 107. 41 F. Maugé, C. Binet and J. C. Lavalley, in Catalysis by Metals, ed. A. Renouprez and H. Jobic, Springer, Berlin Heidelberg, 1997, vol. 6, pp. 1–17. 42 D. A. Panayotov and J. T. Yates Jr, J. Phys. Chem. C, 2007, 111, 2959–2964. 13144 | Phys. Chem. Chem. Phys., 2014, 16, 13136--13144 Paper 43 K. Hadjiivanov, J. Chem. Soc., Faraday Trans., 1998, 94, 1901–1904. 44 K. Hadjiivanov, B. M. Reddy and H. Knözinger, Appl. Catal., A, 1999, 188, 355–360. 45 R. M. Hammaker, S. A. Francis and R. P. Eischens, Spectrochim. Acta, 1965, 21, 1295. 46 R. K. Brandt, M. R. Hughes, L. P. Bourget, K. Truszkowska and R. G. Greenler, Surf. Sci., 1993, 286, 15–25. 47 M. Fukuoka, M. Okada, M. Matsumoto, S. Ogura, K. Fukutani and T. Kasai, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 75, 235434. 48 J. C. Conesa and J. Soria, J. Phys. Chem., 1982, 86, 1392–1395. 49 L. J. Richter, T. A. Germer and W. Ho, Surf. Sci., 1988, 195, L182–L192. This journal is © the Owner Societies 2014