Top Catal (2011) 54:219228 DOI 10.

1007/s11244-011-9639-4

ORIGINAL PAPER

Gold Nanoparticles on Yttrium Modied Titania: Support Properties and Catalytic Activity
J. Plata Antonio M. Ma rquez Javier Fdez. Sanz Jose nchez Avellaneda Francisca Romero-Sarria Rafael Sa a Isabel Dom nguez Miguel Angel Centeno Mar Antonio Odriozola Jose

Published online: 22 January 2011 Springer Science+Business Media, LLC 2011

Abstract A series of titanium oxide catalysts modied with yttrium has been prepared by solgel method and their structural properties have been studied. The incorporation of yttrium in the titania lattice favors the formation of oxygen vacancies while at low Y loadings the anatase structure is preserved. The catalytic activity of these solids for CO oxidation is found to be signicantly dependent on their physical properties. In particular the amount of dopant controls the number of surface oxygen vacancies created as well as the gold particle size, which directly affects the catalytic activity. Also, a linear relationship between the catalytic activity and the band gap values, which depend on the Y loading, is observed. Density functional theory based calculations show that Y atoms are incorporated at the TiO2 surface at substitutional positions only, while the preferred oxygen vacancies arise by removing the bridge surface oxygen atoms. These O-vacancies are the preferential adsorption sites for Au atoms and nanoparticles, acting as nucleation centers that favor the dispersion of the catalyst active phase over the support surface. In agreement with experiment, Y doping is found to decrease the band

gap of the support due to a destabilization of the valence band of the oxide. Keywords Au/TiO2-anatase Yttrium doping CO-oxidation Oxygen-vacancies DFT

1 Introduction The high activity shown by gold based catalysts in oxidation reactions has motivated an increasing in the number of works devoted to this subject in the last years [114]. The main parameters governing the processes catalyzed by gold nanoparticles (NPs) are: the size and shape, the oxidation state of the surface species, and the reducibility properties of the support. Although there is no agreement in the literature on which is the active gold species in the CO oxidation reaction [1517], it is well established that small gold NPs (23 nm) deposited on a reducible support like TiO2 or CeO2 are more active than for instance when supported on MgO or SiO2 [18]. In a previous paper, we have shown that the creation of oxygen vacancies in CeO2, results in a decreasing of the average size of the gold particles deposed on it [19]. Other authors [20] explain this fact arguing that the support defects (vacancies) act as nucleation centers for the incoming gold, binding it strongly and hindering the agglomeration. Moreover, these authors conclude that gold is weakly adsorbed on a stoichiometric (free of defects) surface of titania, presenting a high mobility and making it possible the gold to sinter. Similar results have been reported from theoretical calculations, showing that the presence of vacancies on a TiO2 surface inhibits the formation of large gold particles [21]. A controversial point in the understanding of the CO oxidation mechanism concerns the charge state of the

Electronic supplementary material The online version of this article (doi:10.1007/s11244-011-9639-4) contains supplementary material, which is available to authorized users.
rquez J. Fdez. Sanz (&) J. J. Plata A. M. Ma mica F sica, Facultad de Qu mica, Departamento de Qu Universidad de Sevilla, Sevilla, Spain e-mail: sanz@[Link] nguez R. S. Avellaneda F. Romero-Sarria M. I. Dom M. A. Centeno J. A. Odriozola nica e Instituto de Ciencia de Departamento de Quimica Inorga Materiales de Sevilla, Centro Mixto Universidad de Sevilla rico Vespucio, 49, 41092 Sevilla, Spain CSIC, Avda. Ame

123

220

Top Catal (2011) 54:219228

active gold NPs. For instance Minato et al. [22] proposed that the presence of Aud- species is necessary during the reaction because the oxygen is activated on them by charge transfer from gold NPs towards the 2p* levels of the O2 molecule. This provokes the weakening of the OO bond and the CO oxidation is favored. As a consequence of this electronic transfer and oxygen activation, the gold species are transformed into Aud? ones. A mechanism involving a similar charge transfer in a peroxo-like species has been theoretically substantiated on the basis of density functional calculations [23]. The coexistence of gold species in different oxidation states during the CO oxidation has been rz et al. [24] suggesting that CO is interpreted by Wo weakly adsorbed on the Aud- species (on a vacant site) and therefore, its negative charge is maintained. In contrast, CO adsorption on Au on a stoichiometric surface induces charge transfer from the metal particle to the support, generating a stable Aud?CO species. The stability, reducibility, mechanical and chemical properties of titania makes it a common choice for support in both heterogeneous catalysis and photocatalysis, and, as aforementioned, the activity relies on the presence of surface oxygen vacancies, i.e. the presence of Ti3? surface species. A straightforward possibility to foster the activity is by doping with a transition metal with a different oxidation number. This practice is commonly done as well in photocatalysis, where the introduction of metals in the titania lattice induces a modication in the band structure. An adequate choice of the metal may drive to a situation in which, the introduced dopant states are located in the band gap, thus changing their absorption properties. These localized states may come not only from the dopant d orbitals but also from the vacancies that the charge balance requires (3d states of Ti3?). This is for instance the case of Cr3?/titania doping, where a new strong absorption is observed in the visible region [25]. Doping titania with a trivalent cation like Y3?, naturally gives rise to structures with oxygen vacancies. Actually, for a given amount of yttrium the most stable phase of the mixed metal oxide is that of pyrochlore, Ti2Y2O7, that may be considered as a uorite-type structure with one eighth of the anions missing, that means, with oxygen vacancies [26]. The structure and properties of supports based on Y doped titania, as well as the catalytic activity of Au NPs deposited on such supports has been the subject of a systematic and extensive recent work [27]. In the present paper we report theoretical and experimental results on the structure of Y doped titania that correspond to different low Y loadings that preserve the anatase structure. Once the support has been analyzed, the structure of deposited gold NPs is considered. The experimental characterization is based on UVVIS and XRD data while the theoretical analysis is carried out using density functional theory,

DFT, calculations. The paper is arranged as follows: Sects. 2 and 3 deal with the experimental and theoretical techniques and methods used, respectively. The results are reported and discussed in Sect. 4, starting by some catalytic results, characterization data and, nally, the theoretical analysis.

2 Characterization 2.1 Supports The synthesis of the supports has been carried out following a solgel method. Titanium tert-butoxyde (Sigma, 97%) and yttrium nitrate (Y2(NO3)3.6H2O 99.9% Alfa Aesar) were used as metallic precursors. The adequate amounts of each reactive were dissolved in 125 mL of 1-butanol and stirred for 24 h. The resulting homogeneous solution was maintained under reux at 134 C for 4 h and then cooled to room temperature. After that, 7.45 mL of a H2O/NH3 (28.5/1 ratio) solution was added droply. After 24 h, a gel was obtained. It was centrifuged at 15000 rpm for 15 min and the remaining solvent eliminated by Soxhlet solidliquid extraction using ethanol (48 h, 96 C). The obtained product was dried at 100 C for 24 h and nally, calcined at 300 C during 2 h. The nal solid will be denoted as TiYx, being x the weight percentage of yttrium. 2.2 Gold Catalysts The gold containing catalysts were prepared by deposition precipitation. The adequate amount of HAuCl43H2O (Alfa, 99.99% pure) to obtain 1% wt of gold in the support, was dissolved in 100 mL of deionized water and the pH of the solution adjusted to 7.0 by addition of NaOH 0.1 M in an automatic system of measurement and titration (CRISON pH-Burette 24). The solution was heated to 70 C and then, the support was added and kept under continuous stirring for 1 h. The samples obtained were washed several times (until the disappearance of Cl- and Na? ions), then dried during 24 h at 100 C and nally calcined 2 h at 300 C. 2.3 Characterization Techniques BET specic surface area was measured by nitrogen adsorption measurements at liquid nitrogen temperature in a Micromeritics ASAP 2010 apparatus. Before analysis, the samples were degassed 2 h at 150 C in vacuum. The isoeletric point (IEP) of the support was determined by measuring (Malven Zetamaster) electrophoretic mobilities of aqueous dispersions as a function of pH, at constant

123

Top Catal (2011) 54:219228

221

ionic strength (10-2 mol dm-3 NaCl). The pH was varied by adding HCl or NaOH, as needed. X-Ray diffraction (XRD) patterns were recorded using a Siemens Kristalloex D-501 diffractometer working with CuKa radiation ) in continuous scan mode over a 2h-range (k = 1.5404 A of 2080 using a step size of 0.05 and a step time of 1.0 s. The mean gold crystallite sizes were estimated using the Scherrer equation. For that purpose, XRD diffractograms of gold samples were recorded from 2h 3545 using a step size of 0.02 and a step time of 4.0 s. A peak broadening due to the instrumental broadening of 2h = 0.11 was taken into account. Chemical composition of catalyst was determined by X-ray uorescence (XRF) with a Siemens SRS 3000 sequential spectrophotometer equipped with a rhodium tube. XRF measurements were performed onto pressed pellets (sample included 10% wt of wax). Diffuse Reectance UVVIS spectra of the solids diluted in BaSO4 were recorded at room temperature on a Varian Cary 100 conc spectrometer equipped with an integrating sphere and using BaSO4 as reference. 2.4 Activity Measurements The catalytic tests were carried out in a conventional continuous ow U-shape glass reactor working at atmospheric pressure. The sample (80 mg) was placed between two plugs of glass wool. A thermocouple in contact with the sample assures the right measure of temperature. The feed mixtures were prepared using mass ow controllers (Bronkhorst). The reaction was followed by mass spectrometry, using a Balzers Thermostar benchtop mass spectrometer controlled by the software Balzers Quadstar 422 with capabilities for quantitative analysis. The light-off curves of CO oxidation (300 C, 5 C/min) were obtained with a mixture of 3.4% CO (Air Liquide, 99.997% pure, \3 ppm H2O) and 21% O2 (Air Liquide, 99.999% pure, \3 ppm H2O) balanced by He (Air Liquide, 99.999% pure,\3 ppm H2O) at a total ow rate of 42 mL/min. Empty reactor shows no activity under such conditions. The catalysts were pre-activated in situ for 1 h at 300 C with a mixture of 21% O2 in He at a ow of 30 mL/min and then stabilized at room temperature before the light-off curve started.

Perdew et al. [33] and the electronic states were expanded using plane wave basis set with a cutoff of 400 eV. A spin polarized formalism has been used when we were dealing with gold clusters with odd number of atoms. Forces on the ions were calculated through the HellmannFeyman theorem, including the HarrisFoulkes correction to forces [34]. This calculation of the force allows a geometry optimization using the conjugated gradient scheme. Iterative relaxation of the atomic positions was stopped when the forces on the atoms were less than 0.05 eV/A. Experimental and theoretical studies have shown that the (101) plane is the most stable low index surface of anatase TiO2 [35, 36]. It is also known that surface properties converge faster with the number of TiO2 layers in the model in anatase than in rutile and slab models with four TiO2 layers adequately reproduce experimental data as surface energies [36]. For the initial analysis of the preferential substitution of Ti atoms by Y atoms a 2 9 2 four-layer thick supercell was selected. For Au clusters deposition, and to explore higher Y loadings, a larger 3 9 3 two-layer thick supercell was adopted (see Fig. 1). In all cases, the supercell includes . The calculations with the 2 9 2 9 4 a vacuum of 14 A supercells were performed using a 2 9 2 9 1 Monkhorst Pack set of k-points [37] while those corresponding to the larger, 3 9 3 9 2 supercell were sampled at the C point of the Brillouin zone. To analyze the charge transfer upon adsorption a Bader type analysis was performed [3840].

4 Results and Discussion 4.1 Catalytic Activity Figure 2 shows the CO conversion as a function of the temperature for the studied catalysts. It is observed that the non-doped solid presents around 90% conversion at 300 C. For the doped solids, total conversion of CO is achieved at much lower temperatures (125 C for TiY6 and 100 C for TiY11). For the sample with the highest Y loading, the activity measured at room temperature is

3 Computational Details To model the extended nature of these surfaces, periodic three-dimensional DFT calculations were carried out using the VASP 4.6 code [2830] with the projector augmented wave (PAW) method [31, 32]. In these calculations, the energy was obtained using the generalized gradient approximation (GGA) implementation of DFT proposed by

Fig. 1 3 9 3 supercell model of the anatase TiO2 (101) surface with 2 OTiO layers. The bottom layer is kept frozen and the top layer is fully optimized. Atoms colors: red O, gray Ti

123

222

Top Catal (2011) 54:219228 Table 1 Chemical compositions, IEP, BET surface area, and gold particle size of the studied catalysts Solid 100 Y/(Ti ? Y) Nominal TiY0 0 7 20 Measured 0 6 11 4.73 4.05 3.86 IEP SBET with Au (m2 g-1) 37 150 154 Dp (Au) (nm) 11 9

&20%. These results show that incorporating yttrium into the TiO2 structure has a positive effect on the catalytic properties either by modifying the surface physical properties, including the O-vacancy concentration, or by inuencing the gold particle size and hence the reactivity. 4.2 Experimental 4.2.1 Supports: Physical Properties (XRF, SBET, IEP) The physical properties of the considered solids are reported in Table 1. Starting with the amount of Y doping, one can see that in general, the higher the amount of yttrium used in the preparation, the higher the amount incorporated into the solid. At low yttrium contents, both values, the nominal and the real, agree. However, at high loadings, the yttrium quantity in the support reaches a maximal value (about 24% wt), suggesting the existence of some limitations, probably of structural type, to the incorporation of yttrium into the lattice (see also Fig. 3). ) is signiNotice that as the ionic radius of Y3? (0.93 A ), [41] the insertion cantly larger than that of Ti4? (0.68 A of high quantities of yttrium in the titania lattice is necessarily limited. This result is in agreement with the Y2O3 TiO2 phase diagram [42] showing a eutectic point for 10% Y2O3 at 1580 C. However, no phase associated to pure yttrium oxide has been detected by XRD in the solids calcined in the 3001200 C range. The textural properties of the solids were obtained from N2 adsorptiondesorption isotherms. The isotherms correspond to typical mesoporous materials with complex pore structures made up from interconnected networks of pores of different size and shape (not shown). The specic surface area of the solid (Table 1) reects the effect of yttrium in hindering the sintering of the titania particles upon calcination, gold deposition hardly modify the textural properties of the supports. The isoelectric points (IEP) of the prepared supports are shown in Table 1. This value indicates both the pH at which an immersed solid oxide surface has zero net charge and the pH resulting in electrically equivalent concentrations of positive and negative complexes [43]. The IEP

TiY6 TiY11

We were unable to size the TiY11 sample

Fig. 3 Relationship between the nominal and real (measured by XRF) yttrium content on the synthesized supports

value of the titanium oxide support is 4.73. This value agrees with than reported for a non-hydrated synthetic titania [43]. The modication of the acidbase properties of the solids from yttrium contents correlates fairly well with gold particle size. 4.2.2 UVVIS Spectra In general, titania doping with a transition metal induces the appearance of a new band gap (called secondary band gap) with lower values than the corresponding to the pure oxide (called main band gap) [44]. Moreover, it has been proposed that the doping in the titanium oxide structure favors oxygen vacancies [45]. These oxygen vacancies play an important role in oxidation reactions since oxidative species, such as superoxide and peroxide ones, are formed from them by reaction with molecular oxygen present in the reactive ow [46]. The UVVIS spectra of the supports are shown in Fig. 4. We observe a tail of absorption at low energies for the supports with yttrium contents between 0 and 11% wt, which has been attributed to the presence of dopants [16, 17]. The band gap value was calculated by plotting the square root of the KubelkaMunk function multiplied by the photon energy (KM 9 hm)1/2 versus the photon energy (eV) and extrapolating the linear part of the rising curve to zero as reported in Table 2 [13, 47]. In the solid without yttrium (TiY0) the observed tail of absorption is due to the presence of Ti3? [48], suggesting that, by itself, the synthesis method induces the

Fig. 2 CO conversion versus temperature for the studied catalysts

123

Top Catal (2011) 54:219228

223

Fig. 4 UVVIS spectra of the synthesized supports

Fig. 6 XRD patterns of TiY11 and TiY13 supports calcined at 950 C (A = anatase; R = rutile; P = stoichiometric pyrochlore, Ti2Y2O7, Y/Ti = non-stoichiometric pyrochlore, TiYO2.085

Table 2 Band gap values for the catalytic supports Support Band gap (eV) Secondary TiY0 TiY6 TiY11 1.94 2.27 2.34 Main 3.13 3.05 3.08

Fig. 5 Relationship between the catalytic activity (T50) and the secondary band gap of the studied solids

creation of vacancies. The secondary band gap in this support is 1.94 eV, while its main band gap, 3.13 eV, is close to that reported for pure anatase (3.2 eV). For the solids with yttrium contents of 6 and 11% wt, the secondary band gap values are 2.27 and 2.34 eV respectively; while their main band gaps (3.05 and 3.08 eV) are slightly lower than that of TiY0. The decreasing of the main band gap has been attributed to the band structure modication of titania by the introduction of dopants [48]. The secondary band gap is associated to the defect structure. The activity measured through the T50 increases on increasing the yttrium content (see Fig. 5) and therefore on modifying the defect structure. A similar correlation between catalytic activity and band gap of the catalysts was previously observed for Au/CeO2Al2O3 catalysts [13]. 4.2.3 XRD All the supports calcined at 300 C are amorphous solids, which avoids extracting any information about their

structure. To obtain some clues about it, the samples were calcined at higher temperatures (up to 1200 C), collecting a diffractogram each 50 C. The structural differences observed between supports treated at the same temperature reect the differences in the local compositions and permit us to observe if some phase segregation is produced. As an example, Fig. 6 shows the diffractograms corresponding to TiY11 and TiY13 samples calcined at 950 C. These two supports have been chosen because in their range of composition (1113% wt yttrium), the secondary band gap strongly decreases while the main band gap raises up to that of pure anatase. The XRD patterns of both samples are markedly different. In both cases, although with different crystallinity and ratio, peaks due to anatase and rutile are detected. Besides these, diffraction lines characteristic of a stoichiometric pyrochlore (Ti2Y2O7) and/or a non- stoichiometric pyrochlore (TiYO2.085) structures are observed. The appearance of these last two structures (both containing defects) suggests the success of the insertion of yttrium in the titania network. On the other hand, a shift in the 2h position of the rutile and anatase peaks in both samples is observed, indicating a different modication of each support, related, probably, with the yttrium content. From the XRD study of the supports calcined at different temperatures (results not completely shown), the following observations have been extracted: (1) the anatase to rutile transformation is observed at about 900 C, higher than that of pure phase (500 C) [49], and, both, the anatase and/or rutile phases, are present in all the range of temperatures studied. This high temperature of anatase to rutile change of phase is provoked by the interaction of the titanium oxide with the dopant, as previously established [50]; (2) pyrochlore structures (stoichiometric and nonstoichiometric) are always detected and; (3) no phase associated to yttrium oxide or any other segregated yttrium structure, has been observed. These three observations conrm that yttrium has been incorporated into TiO2. However, in all cases, part of the titanium oxide is free of yttrium, as deduced from the joint detection of anatase and/ or rutile, phases together with pyrochlore ones. This fact

123

224

Top Catal (2011) 54:219228

may be understood considering that, as aforementioned, the ionic radius of yttrium is signicantly larger than that of titanium, making difcult the introduction of yttrium in the bulk of the titanium oxide. These observations suggest that yttrium interacts with the titanium oxide only on the surface, which could explain why no more than 24% wt yttrium may be introduced in the titania network, as observed by the XRF analysis. 4.3 Theoretical Calculations 4.3.1 Geometries and Electronic Structure To setup the model calculations, and in agreement with the XRD and XRF data, we assume that Y atoms are incorporated into the TiO2 network in substitutional positions, as the large atomic radius of Y will make an interstitial incorporation very unlikely. In a rst step we considered all the different possible substitutional sites in a 2 9 2 TiO2 supercell four-layer thick, exposing the (101) face. In such a model a single Y/Ti substitution represent a Y percentage of 6.25%, and allows us to distinguish four different sites: two surface sites (penta- and hexa-coordinated, hereafter labeled as Ti5c and Ti6c) and two subsurface sites. To keep the formal oxidation degree of the system a surface anion has been replaced by a OH- group. The most favorable structure corresponds to the substitution at the Ti5c surface site. Substitution of surface Ti6c atoms is only moderately unfavorable (0.3 eV), while substitutions at subsurface sites are very unlikely, as they are less stable by 0.80.9 eV. These results clearly demonstrate the preference of the Y atoms for surface substitutional sites and can be rationalized by the size differences between the two cations. Indeed, the larger size of the Y atoms requires a noticeable distortion of the TiO2 network that is much harder at subsurface or bulk positions. Having established the preference of Y atoms for surface substitutional sites, we have now to increase the Y loading in our models to progressively approach to the experimental range of Y content and to understand the observed experimental trends. To this end, we will progressively substitute a pair of surface Ti atoms by Y atoms into a 3 9 3 supercell model of the TiO2 (101) surface. To keep the formal oxidation states of the ions in this case, an O-vacancy is created each two Y/Ti atoms substitution. Taking into account all different ways of surface Y/Ti substitution and the relative position of the generated O vacant result in 39 different arrangements that have to be computed to rationalize the preferred distribution of Y atoms and O vacancies in the TiO2 (101) surface. The results show (see Supplementary Information) that at this level of Y-doping the most stable arrangements correspond, as expected, to having the two Y atoms occupying

Ti5c sites but not in close contact each other. There are also arrangements where one Y atom occupies a surface Ti5c site and the second Y atoms a Ti6c site that is only modestly less stable (less than 0.4 eV). In all cases, the O-vacancy is more favorably created at the surface bridge sites, O2s. These ndings can be rationalized by the larger average in Y2O3) compared to that of YO bond distance (2.26 A TiO (1.95 A). Indeed, when the Y atoms enter into substitutional sites at the TiO2 surface, the local geometry must strongly relax to accommodate the Y atom. When the O-vacancy is created in a nearby O2s site the Y atom is able to slightly displace away from the surface and the resulting local geometries show YO distances similar to those of the Y2O3. To simulate yttrium loadings comparable to the experimentally prepared catalysts, up to six surface Ti atoms have been substituted by Y in the 3 9 3 model already mentioned. This represents a 17% of Y content in the model that is similar to those experimental catalysts having the maximum activity. Relying on the clear tendency of Y to be at the surface, only models where Y atoms are at surface Ti5c sites or with half of the Y atoms occupying Ti5c and the other half Ti6c sites have been considered. In all cases three O2s vacancies have been also created. The relative energies of the congurations computed are reported in Table 3. At this doping the most stable distribution of Y atoms on the TiO2 surface results in the formation of Y2O3 dimers at the surface, with one Y atom occupying a Ti5c site and the second one in a nearby Ti6c site (see Fig. 7). The rationale for this behavior is that when the Y content increases, the arrangement of Y atoms at neighboring Ti5c is energetically unfavorable (as was previously discussed). Instead, the two Y atoms prefer to occupy neighboring Ti5c and Ti6c sites, with a YY distance similar to that found in Y2O3. Moreover, these dimers are in turn also ordered on the TiO2 surface, generating an extended structure that seems to stabilize the whole system. These associations (see Fig. 7) can be related to coherentincoherent precipitations [51] of Y2O3 into the TiO2 surface matrix.
Table 3 Computed relative energies (eV) corresponding to 3 9 3 models of anatase (101) surface with six Y/Ti substitutions and three O vacancies Model 50%Ti (a) 50%Ti (b) 100%Ti (a) 100%Ti (b) 100%Ti (c) E/eV 0.00 0.92 1.45 1.45 1.82 Model 100%Ti (d) 50%Ti (c) 50%Ti (d) 100%Ti (e) 100%Ti (f) E/eV 1.84 2.08 2.34 2.70 2.93

Each conguration is denoted by the percent of surface Ti atoms substituted and a label to distinguish them

123

Top Catal (2011) 54:219228

225

Fig. 7 Top view of a 3 9 3 supercell model of the anatase TiO2 (101) surface with six Y/Ti substitutions and three O vacancies (denoted by crossed circles). Atoms colors: red O, gray Ti, and brown Y. The 3 9 3 supercell is duplicated in this gure

Fig. 8 Computed band structure of undoped (top) and Y-doped (6%middle and 16%bottom) models of the (101) anatase surface. The two most important indirect transitions of the material are also show

An analysis of the relative stability of the different sites and loadings allows us for a rational of the limitations in the incorporation of Y into the solid experimentally observed. At low Y content in the precursor, the Y atoms will preferentially substitute surface Ti5c atoms and O2s vacancies will be generated in nearby sites. At higher loadings, a limit situation is reached, with the formation of the extended structures of yttrium dimers shown in Fig. 7. From this point, further incorporation of Y in the solid requires a surface reconstruction that can be associated to the observation of diffraction lines characteristic of pyrochlore structures in the XRD diffractograms. An additional aspect that is worth commenting is related with the energetics of the oxygen vacancy formation. In general, the involved energies are large whatever the phase, rutile or anatase is considered [52, 53]. In particular, using the 3 9 3 model we estimate this energy to be 4.54 eV, in agreement with the value reported by Vittadini et al. [54]. However, removing oxygen atoms from this model after Y-doping turns to be an exothermic process, and actually even the third removal, which is the one that gives the correct stoichiometry, releases 0.08 eV. This relative stabilization of the titania O-vacancy that results from the interaction vacancy-dopant has already been reported in several cases involving both rutile [20, 55, 56] and anatase [57]. Moreover, it is worth mentioning the relationship between the easiness in the O-vacancy formation and the higher catalytic activity of the Y-doped titania, which is in agreement with recent work of Corma et al. [58] for the Fe-doped anatase. Several studies available in the literature have related the introduction of dopant elements into the titania structure with a reduction in the band gap of the resulting material with respect to that of the undoped titania [59, 60]. To understand the modications induced in the electronic structure by the presence of the dopant element and to be able to establish comparisons with the experimental UV VIS data, the band structure of the models has been computed. In Fig. 8 the band structure for the models with 0%

Fig. 9 Variation of the band gap of the material with the yttrium content in experimental synthesized supports (black) and theoretical computed band gaps (red) for the three surface models developed

(undoped anatase), 6% (two Y/Ti substitutions), and 16% (six Y/Ti substitutions) are presented with the two most important indirect transitions shown. The band gap of the material has been estimated from the D ? C transition, and plotted in Fig. 9 along with the experimental band gaps obtained from the UVVIS spectra. The agreement between the theoretical and the experimental trends is quite reasonable, taking into account the well-known trend of the GGA functionals to produce band gaps lower than the experimental values [6163]. Further analysis of the band gap structure shows that there are no new states in the anatase band gap introduced by the dopant element and that the band gap narrowing is a result of a destabilization of the valence band of the oxide (mainly composed of O 2p levels) that pushes this band towards higher energies. 4.3.2 Deposition of Gold Nanoparticles Let us now consider the interaction between gold nanoclusters and the support surface. To be able to establish comparisons, four different surface models have been developed. First clean undoped and Y-doped anatase and secondly, to account for the experimental working conditions, the presence of water has been modeled by adding hydroxyl groups and hydrogen atoms to these two surfaces.

123

226

Top Catal (2011) 54:219228

We began our study by examining the adsorption of a single gold atom in the above described model surfaces. On each surface model, different adsorption sites have been tested and the bonding energies of the most stable conguration are presented in Table 4. These results show that, at low gold coverage, hydroxylation of the undoped surface improves the interaction between the gold atom and the surface increasing the binding energy from -0.46 eV on the clean surface to -0.54 eV on the hydroxylated surface. In both cases the preferred adsorption site is the same, a bridge site between two O2- anions. Doping the surface, on the other hand, modies the preferred adsorption site that is now the O-vacancy site, and results in a much larger adsorption energy (-0.81 eV) for the gold atom. Considering now gold atom adsorption on a hydroxylated doped surface two possibilities arise: either the hydroxyl or the gold atom can occupy the O-vacancy site. In contrast with the increase in gold adsorption energy found on the undoped surface, hydroxylation of the doped surface reduces the strength of the gold-surface interaction. The computed binding energy is -0.47 eV when the gold atom sits on the O-vacancy and -0.42 eV when the hydroxyl lls the vacancy. Metal NPs dispersed at oxide support surfaces generally experiment sintering processes (form larger clusters) under reaction conditions. This results in a loss of catalyst activity due to the recognized correlation between catalyst activity and metal particle size [64]. On the undoped model surfaces it is found that the differences in gold atom adsorption energies are small between different sites (-0.46 to -0.29 eV on the dehydrated surface and -0.54 to -0.38 eV on the hydroxylated surface). We can, then, presume a high mobility for the gold atom in these surfaces and as result gold nanoclusters may easily diffuse and sinter in these surfaces. In contrast, on the doped-dehydrated surface, it is found that there is a larger difference in the binding energy of the gold atom at the vacancy or at nearby sites (-0.81 to -0.64 eV) with respect to other sites at the surface (-0.49 to -0.37 eV).

This means that the vacancies will play a double role. First they favor the dispersion of the metal NPs over the surface as the number of vacancies increases with yttrium loading and secondly they will act as nucleation sites that will strongly anchor those metal nanoclusters, reducing the sintering processes. However, in the presence of surface hydroxyls the energy differences between the different adsorption sites is reduced (-0.47 to -0.42 eV at the O-vacancy or at nearby sites against -0.31 to -0.27 eV when the hydroxyl lls the O-vacancy). This means that surface hydroxylation will increase the mobility of gold over the surface, as result of the competition between Au and OH for the O-vacancy site. Starting from the most favorable adsorption sites for isolated gold atoms, the adsorption of Au4 clusters has also been examined. The computed binding energies, also summarized in Table 4, show a large difference between undoped and doped surfaces. For the Y-doped case, the Au4-surface interaction is enhanced with respect to the undoped surfaces. This can be appreciated both in the computed interaction energies displayed in Table 4 as well as in the adsorbed nanoclusters geometries (see Supporting Information). While at the undoped surfaces the nanoclusters display a three-dimensional arrangement, with only one or two gold atoms interacting with the surface, at the doped surfaces, the gold nanoclusters seems to adopt a twodimensional shape, increasing the number of gold atoms that directly interact with the support. Besides acting as a structural promoter, facilitating the dispersion of gold nanoparticles over the surface and hindering their sintering under reaction conditions, the support also takes an active role by inuencing the electronic structure of the deposited metal NPs [65]. To elucidate the effect of the support on the electronic properties of the Au4 clusters, the total density of states (DOS) and gold projected density of states (PDOS) of the Au4-support catalysts models have been examined. As shown in Fig. 10 the overlap between the valence band of the support and the

Table 4 Computed adsorption energies (eV) for single gold atoms and Au4 nanoclusters on surface models of undoped and Y-doped anatase (101) surfaces Model E/eV Au TiO2 Hydrated TiO2 Y-doped TiO2 Hydrated Y-doped TiOa 2 Hydrated Y-doped TiOb 2
a b

Au4 -1.10 -1.36 -2.39 -1.96 -2.01 Fig. 10 Total density of states (black) and gold projected density of states (red) for Au4 cluster deposited over hydroxylated model of undoped (top) and Y-doped (bottom) anatase (101) surface

-0.46 -0.54 -0.81 -0.47 -0.42

One Au atom sits at the O-vacancy An OH is occupying the O-vacancy

123

Top Catal (2011) 54:219228 Table 5 Bader charges for Aun (n = 14) clusters interacting with surface models of undoped and Y-doped anatase (101) surfaces Model TiO2 Hydrated TiO2 Y-doped TiO2 Hydrated Y-doped TiOa 2 Hydrated Y-doped TiOb 2
a b

227

Au ?0.46 ?0.02 -0.14 -0.05 -0.01

Au2 ?0.02 -0.01 -0.09 -0.14 -0.08

Au3 ?0.56 ?0.55 -0.05 -0.10 -0.08

Au4 ?0.17 ?0.03 -0.14 -0.18 -0.08

One gold atoms sits at the O-vacancy An OH is occupying the O-vacancy

electronic levels of the gold nanocluster is improved by Y-doping, as result of a shift of the support energy levels towards higher energies. It is worth mentioning here that when the energy levels of deposited gold nanoclusters are close to the O 2p band a higher catalytic activity with respect to the CO oxidation is obtained [13]. The effect of the support on the electronic properties of the metal NPs can also be probed by analyzing the charge transfer at the interface. To this end, Bader charges for increasing size gold nanoclusters, Aun (n = 14), have been determined (Table 5). The calculated charges show a different behavior for Y-doped surfaces against undoped ones. While for undoped surfaces the support oxidizes the Aun nanocluster, in doped surfaces there is a charge transfer from the support towards the metal cluster.

experimentally found to the incorporation of the dopant into the titania support. The oxygen vacancies generated to balance the charge (bridge surface oxygen atoms) are found to be preferential adsorption sites for Au atoms and NPs, acting as nucleation centers that favor the dispersion of the catalyst active phase over the support surface. The interaction between the gold NP and the support is analyzed by examining the electronic structure of the system. It is found that, as result of Y incorporation into the titania surface, the valence band of the support, composed mainly of O 2p energy levels, is shifted towards higher energies. A straightforward consequence is that the band gap of the support is reduced, in agreement with experimental UV VIS spectra. A further consequence is that the overlap with the electronic levels of the gold NPs is improved, enhancing the interaction with the support.
Acknowledgements This work was funded by the Ministerio de n, Spain, (grants ENE2009-14522-C05-01, Ciencia e Innovacio MAT2008-04918 and CSD-00023, co-nanced by FEDER from a (project P08-FQMEuropean Union), and the Junta de Andaluc 03661). Romero-Sarria thanks the Spanish MEC for her contract n y Cajal Programme). The Red Espan ola de Super(Ramo n (RES-BSC) provided part of the computer time. computacio

References
1. Haruta M (1997) Catal Today 36:153 2. Meyer R, Lemire C, Shaikuthdinov SK, Freund HJ (2004) Gold Bull 37:72 3. Nijhuis TA, Weckhuysen BM (2006) Catal Today 117:84 4. Thompson DT (2003) Appl Catal A 243:201 `a H (2006) Tetrahedron 5. Abad A, Almela C, Corma A, Garc 62:6666 6. Jacobs G, Ricote S, Patterson PM, Graham UM, Dozier A, Khalid S, Rhodus E, Davis BH (2005) Appl Catal A 292:229 7. Kim CH, Thompson LT (2005) J Catal 230:66 `guez JA, Hanson JC (2005) J Chem Phys 8. Wang X, Rodr 123:221101 9. Silberova BAA, Mul G, Makkee M, Moulijn JA (2006) J Catal 243:171 10. Bond GC, Thompson DT (1999) Catal Rev Sci Eng 41:319 11. Bond GC, Louis C, Thompson DT (2006) Cataysis by gold. Imperial College, London 12. Centeno MA, Paulis M, Montes M, Odriozola JA (2002) Appl Catal A 234:65 13. Centeno M, Portales C, Carrizosa I, Odriozola JA (2005) Catal Lett 102:289 nez LM, Odriozola JA, Moreno S, Molina R, 14. Carriazo JG, Mart Centeno MA (2007) Appl Catal B 72:157 S, Scire S, Crisafulli C, Visco AM, Galvagno S (1997) 15. Minico Catal Lett 47:273 16. Pillai UR, Deevi S (2006) Appl Catal A 229:266 17. Arena F, Famulari P, Truno G, Bonura G, Frusteri F, Spadaro L (2006) Appl Catal B 66(12):81 18. Schubert M, Hackenberg S, van Veen AC, Muhler M, Plzak V, Behm RJ (2001) J Catal 197:113 nez LM, Centeno MA, Odriozola JA 19. Romero-Sarria F, Mart (2007) J Phys Chem C 111:14469

5 Conclusions The strong inuence of the support on the catalytic activity of gold-based catalysts in the CO oxidation reaction has been evidenced. The incorporation of yttrium into the titania lattice, that is accompanied by creation of oxygen vacancies, favors the conversion of CO at low temperatures. In order to clarify this point the catalysts have been characterized and interesting relationships between catalytic activity and physical properties have been established. The most important are: (a) the quantity of dopant introduced into the titania controls the number of surface oxygen vacancies created and the gold particle size, directly inuencing the catalytic activity; and (b) the band gap is also modied by Y-doping, existing a linear relationship between the catalytic activity and the band gap values found for the studied solids. A theoretical analysis based on density functional theory calculations allows us to interpret the main experimental ndings. The relative energies show that Y atoms are almost exclusively incorporated into the titania at substitutional surface sites. Only a limited amount of surface Ti atoms can be substituted, explaining the restrictions

123

228 guez JA, Sanz JF (2008) J 20. Graciani J, Nambu A, Evans J, Rodr Am Chem Soc 130:12056 21. Parker SC, Grant AW, Bondzie VA, Campbell CT (1999) Surf Sci 441:110 22. Minato T, Susaki T, Shiraki S, Kato HS, Kawai M, Aika KI (2004) Proceedings of the 22nd European Conference on Surface Science, Prague, vol 566568 (Part 2), p 1012 guez JA (2006) J Am Chem Soc 23. Cruz N, Sanz JF, Rodr 128:15600 rz AS, Heiz U, Cinquini F, Pacchioni G (2005) J Phys Chem 24. Wo B 109:18418 25. Liu Z, Cui ZL, Zhang K (2005) Mater Charact 54:123 26. Panero WR, Stixrude L, Ewing RC (2004) Phys Rev B 70(5): 054110 27. Avellaneda RS (2010) Estudio e inuencia del Y y otros modicadores superciales en el comportamiento de catalizadores soportados en TiO2. Ph.D. thesis, Universidad de Sevilla, Sevilla, Spain 28. Kresse G, Hafner J (1996) Phys Rev B 47:558 ller J (1996) Comput Mater Sci 6:15 29. Kresse G, Furthmu ller J (1996) Phys Rev B 54(16):11169 30. Kresse G, Furthmu chl PE (1994) Phys Rev B 50(24):17953 31. Blo 32. Kresse G, Joubert D (1999) Phys Rev B 59(3):1758 33. Perdew JP, Chevary JA, Vosko SH, Jackson KA, Pederson MR, Singh DJ, Fiolhais C (1992) Phys Rev B 46(11):6671 34. Harris J (1985) Phys Rev B 31(4):1770 35. Diebold U (2003) Surf Sci Rep 48(58):53 36. Lazzeri M, Vittadini A, Selloni A (2001) Phys Rev B 63(15): 155409 37. Monkhorst HJ, Pack JD (1976) Phys Rev B 13(12):5188 nsson H (2006) Comput Mater 38. Henkelman G, Arnaldsson A, Jo Sci 36:354 39. Sanville E, Kenny SD, Smith R, Henkelman G (2007) J Comput Chem 28:899 40. Bader R (1990) Atoms in molecules: a quantum theory. Oxford University Press, Oxford 41. Malati MA, Wong WK (1984) Surf Technol 22:305 42. Shcherbakova LG, Mamsurova LG, Sukhanova GE (1979) Russ Chem Rev 48:423

Top Catal (2011) 54:219228 43. Parks G (1965) Chem Rev 65:177 44. Radecka M, Zakrzewska K, Wierzbicka M, Gorzkowska A, Komornicki S (2003) Solid State Ionics 157(14):379 45. Serpone N (2006) J Phys Chem B 110(48):24287 46. Carrettin S, Hao Y, Aguilar-Guerrero V, Gates BC, Trasobares S, Calvino JJ, Corma A (2007) Chem Eur J 13:277771 nguez MI, Navio JA, Odriozola 47. Centeno MA, Hidalgo MC, Dom JA (2008) Catal Lett 123:198 48. Thomas AG, Flavell WR, Kumarasinghe AR, Mallick AK, tzel M, HenTsoutsou D, Smith GC, Stockbauer R, Patel S, Gra gerer R (2003) Phys Rev B 67(3):035110 49. Gole JL, Prokes SM, Gelmbocki OJ (2008) J Phys Chem C 112:1782 50. Lin J, Yu JC (1998) J Photochem Photobiol 116:63 (112) 51. Fischer J, Hollomon J, Leschen J (1950) Ind Eng Chem 44:61324 52. Ganduglia-Pirovano MV, Hofmann A, Sauer J (2007) Surf Sci Rep 62:219 53. Oviedo J, San Miguel MA, Sanz JF (2004) J Chem Phys 121:7427 54. Vittadini A, Selloni A (2002) J Chem Phys 117:353 lvarez LJ, Rodr guez JA, Sanz JF (2008) J Phys 55. Graciani J, A Chem C 112:2624 56. Graciani J, Ortega Y, Sanz JF (2009) Chem Mater 21:1431 57. Roldan A, Boronat M, Corma A, Illas Francesc (2010) J Phys Chem C 114:6511 58. Carrettin S, Hao Y, Aguilar-Guerrero V, Gates BC, Trasobares S, Calvino JJ, Corma A (2007) Chem Eur J 13:7771 59. Karvinen S (2003) Sol State Sci 5:811 60. Cavalheiro AA, Bruno JC, Saeki MJ, Valente JPS, Florentino AO (2008) J Mater Sci 43:602 61. Perdew JP, Levy M (1983) Phys Rev Lett 51:201884 ter M (1983) Phys Rev Lett 51:201888 62. Sham LJ, Schlu 63. Lany S, Zunger A (2008) Phys Rev B 78(23):235104 64. Campbell C, Parker S, Starr D (2002) Sience 298:811 65. Molina LM, Hammer B (2003) Phys Rev Lett 90:20206102

123