Polarization-Dependent Total-Reflection Fluorescence XAFS Study of

Polarization-Dependent Total-Reflection Fluorescence XAFS Study of Mo Oxides on a. Rutile TiO2(110) Single Crystal Surface. Wang-Jae Chun,† Kiyotaka...
1 downloads 0 Views 204KB Size
9006

J. Phys. Chem. B 1998, 102, 9006-9014

Polarization-Dependent Total-Reflection Fluorescence XAFS Study of Mo Oxides on a Rutile TiO2(110) Single Crystal Surface Wang-Jae Chun,† Kiyotaka Asakura,‡ and Yasuhiro Iwasawa*,† Department of Chemistry and Research Center for Spectrochemistry, Graduate School of Science, The UniVersity of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan ReceiVed: April 28, 1998; In Final Form: July 17, 1998

The structure of well-dispersed Mo oxides on TiO2(110) was investigated by means of polarization-dependent total-reflection fluorescence X-ray absorption fine structure (PTRF-EXAFS) in three different directions of the single-crystal surface. A model sample for supported Mo oxide catalysts was prepared by impregnation of (NH4)6Mo7O24‚4H2O dissolved in ultrapure water, followed by calcination at 773 K. Direct comparison of the observed EXAFS and the calculated one was performed based on three-dimensional model structures. We found that anisotropic Mo dimers were preferentially formed on the TiO2(110) surface, with Mo-Mo distance (0.335 nm) parallel to the [11h0] direction of the surface. It was also found that the Mo dimers were attached to the surface, showing Mo-Ti (interfacial) distance at 0.296 nm.

Introduction

the incident X-ray as expressed by eq 1

Understanding and controlling oxide surfaces with active metals and metal oxides are the key issues for the development of industrial supported catalysts. New concepts regarding structure and its arrangements are conceived in the development of new catalysts. Extended X-ray Absorption Fine Structure (EXAFS) has been regarded as a useful method to obtain information about the local structure of metal sites1,2 in supported metal catalysts, oxide catalysts, and sulfide catalysts.1,3 In situ EXAFS studies have also been performed to reveal the dynamic structure change of active metal sites under reaction conditions and to elucidate reaction mechanisms.4-9 Structures of active sites in supported catalysts often depend on surface structure of the support and interaction with the support surface. Therefore, the interfacial structure and bonding feature between active metal site and well defined support surface are the important issues to understand supported metal chemistry and to develop new catalytic materials. In general, the metal-support interaction and the structure of supported metals and metal oxides are anisotropical and asymmetrical. It is not easy to obtain the metal-support interface structure directly by conventional EXAFS though there were EXAFS studies indicating the interfacial structures for metal particles on metal oxides10 and Nb-oxide species on SiO2.11 In Pt/Al2O3 and Nb/SiO2, the Pt-O and Nb-Si bondings at the interfaces were observed, but they appeared only after removal of the EXAFS oscillations of Pt-Pt for Pt/ Al2O3 and of Nb-O and Nb-Nb for Nb/SiO2. Thus the interfacial bondings are minor components and are obtained by subtraction of the stronger signals originating from the inner bondings such as metal-metal and metal-oxygen bonds from the whole EXAFS oscillations. The EXAFS oscillation χ(k) depends on the angle θi between the ith bond direction and the polarized electric-field vector of * To whom correspondence should be addressed. FAX: 81-3-5800-6892. E-mail: [email protected]. † Department of Chemistry. ‡ Research Center for Spectrochemistry.

χ(k) )

∑3χi(k) cos2 θi

(1)

where χi(k) is an EXAFS oscillation accompanying the ith bond. If the distances in the ith shell are equal, eq 1 can be simplified to

χ(k) ) )

∑i χi(k) ∑j 3 cos2 θij ∑i χi(k)N*i,

(2)

where N* is an effective coordination number defined as

∑j cos2 θj

N* i )3

(3)

If a single crystal oxide is used as a model support, structural information on the bonds parallel and normal to the surface can be obtained independently using eq 3. When the polarization vector of the incident X-ray is parallel to the surface (spolarization), the bondings which lie parallel to the surface mainly contribute to the EXAFS signal. The bondings normal to the surface, i.e., metal-support bonds, are preferentially observed when the polarization vector is perpendicular to the surface (p-polarization). However, there are two problems in applying the polarization dependent EXAFS to dispersed active species on a flat oxide substrate. First, concentration of the species at the surface is very low, and hence it is impossible to measure EXAFS spectra in a transmission mode. The concentration of monolayer metals on a flat surface ranges 1014-1015 atoms per cm2, which is 4-5 orders of magnitude lower than that necessary for transmission EXAFS measurements. Fluorescence yield detection is preferable for such dilute samples.12 Second, most of interesting elements as catalysts have X-ray

10.1021/jp9820368 CCC: $15.00 © 1998 American Chemical Society Published on Web 10/21/1998

Mo Oxides on a TiO2(110) Single Crystal Surface

J. Phys. Chem. B, Vol. 102, No. 45, 1998 9007 Experimental Section

Figure 1. Schematic drawing of TiO2 (110) surface (a) and arrangement for PTRF-XAFS measurements (b). E: electric-field vector of the incident X-ray.

absorption edges at higher energies than 4 keV. The X-ray can penetrate deeply into the materials which generates scattering X-ray and hinder fluorescence X-ray signals from the active species on the surface. To reduce the scattering X-rays from the substrate and to improve the surface sensitivity, we adopted the measurements under a total-reflection condition. When the incident X-ray hits a flat substrate below a critical angle (δc), the X-ray is totally reflected and can penetrate only a few nanometers into the bulk, resulting in a tremendous decrease in scattering X-rays from the substrate.13 Combining the total reflection technique and the fluorescence detection technique, we can measure the polarization-dependent EXAFS spectra for active metal sites supported on a flat oxide substrate. The technique is called polarization-dependent total-reflection fluorescence EXAFS (PTRF-EXAFS) method hereinafter. We have determined the asymmetric structures of Co oxides on R-Al2O3 (0001), Cu oxides on R-quartz (0001), and Pt4 clusters on R-Al2O3 (0001) using the PTRF-EXAFS technique.14-19 TiO2-supported Mo oxide catalysts have drawn much interest in relation to catalytic selective oxidation processes.20,21 Structural characterizations of Mo oxides on TiO2 have been performed to reveal the plausible structure models by RAMAN and conventional EXAFS.22-25 In this study, we carried out PTRF-EXAFS analysis to investigate the three-dimensional surface structure of molybdenum oxides supported on a rutile TiO2(110) single crystal as a model system for Mo/TiO2 (powder). Particular interest was paid to the interaction between Mo and TiO2 surface which may determine the Mo oxide structure. A rutile TiO2(110) surface has an anisotropic structure with alternative alignment of the protruding one-dimensional oxygen rows and the grooves composed of one-dimensional rows of 5-fold coordinated Ti atoms along [001] axis as shown in Figure 1(a). TiO2(110) surfaces have been well characterized by various techniques such as LEED (low energy electron diffraction), XPS (X-ray photoelectron spectroscopy), MEED (medium energy electron diffraction), STM (scanning tunneling microscopy), SXRD (surface X-ray diffraction), and also ab initio calculations.26-31

A. Sample Preparation. A polished rutile TiO2(110) single crystal (20 × 40 × 1 mm3) (Earth Jewelry Co.) with optical grade was washed with ultrapure water (18 MΩ; Millipore Co.) and annealed at 823 K for 2 h in air. Molybdenum was supported by an impregnation method using an ultrapure aqueous solution of (NH4)6Mo7O24‚4H2O in a quartz cell, followed by calcination at 773 K for 3 h under O2. The Mo loading was estimated to be 0.2 ML (1 Mo atom/nm2) by XPS (Mg KR) measurements. Comparing the intensity ratio of Mo 3d to Ti 2p with those for reference powder samples with various Mo loadings which were prepared by impregnation of TiO2 (rutile) with aqueous solutions of given amounts of (NH4)6Mo7O24‚4H2O. The obtained sample is denoted as Mo/ TiO2(110). B. PTRF-XAFS Measurement. PTRF-XAFS spectra were measured at BL14A vertical wiggler line of the Photon Factory in the Institute of Material Structure Science High Energy Accelerator Research Organization (KEK-PF). The Synchrotron radiation was monochromatized by a Si (311) double-crystal monochromator. The storage ring was operated at 2.5 GeV with 200-350 mA. Mo K-edge EXAFS spectra were recorded in the range from 19 780 to 20 500 eV with an energy interval of 3 eV at room temperature. Critical angles(δc) for total reflection of hard X-rays region are in general several mrad. Particularly for Mo K-edge at 20 keV the δc is as small as 1.7 mrad. To adjust the incident angle θ around δc, a 4-axis high-precision goniometer with a minimum step of 0.17 mrad was used.32 Both fluorescent X-ray and reflected beam intensity were monitored to determine an appropriate θ value. The X-ray beam size was reduced to 0.1 mm φ in diameter in order to avoid unnecessary irradiation. The incident X-ray was monitored by a 50 mm-long ion chamber filled with Ar. When we detect the fluorescence X-ray stemming from the sample, a small-size NaI (Tl) scintillation counter (14 mm φ) was used in order to avoid the Bragg diffraction from the TiO2 bulk.33 Since the TiO2(110) surface has an anisotropic structure, PTRF-XAFS measurements were carried out in three different directions of the electric vector of the incident X-rays parallel to the [11h0], [001], and [110] axes of TiO2(110) as shown in Figure 1b. C. Data Analysis. PTRF-EXAFS spectra were calculated directly from the intensity ratio of the fluorescence X-ray detected by a scintillation counter (If) to the incident X-ray signal detected by an ionization chamber (Io) without any correction for self-absorption because all the absorber is on the surface. The EXAFS oscillation, χ(k), was extracted from the spectra by a spline smoothing method and normalized by the edge height using the EXAFS analysis program REX ver. 2.5. The energy dependence of the edge height was taken into account using the McMaster equation. The origin of kinetic energy of photoelectron was taken to be an inflection point of the edge jump. Analysis for the surface structure of Mo oxides on TiO2 (110) was carried out by comparing the observed data with those calculated by FEFF 6.0134-36 using realistic three-dimensional models. The FEFF calculations were carried out with the following parameters: S02(amplitude reduction factor) ) 0.90; Rmax(maximum calculation path length) ) 0.5 nm; NLEG ) 5 (up to five scattering paths with total distances less than 0.5 nm were evaluated). Square of Debye-Waller factors was first set to 0.000 036 nm2 and then optimized together with interatomic distances. Polarization dependence was also considered in the simulation.

9008 J. Phys. Chem. B, Vol. 102, No. 45, 1998

Chun et al.

Figure 2. Mo K-edge PTRF-XANES spectra of Mo/TiO2(110) and references.

Results To evaluate the validity of the FEFF calculation and appropriate S02, we compared the theoretical EXAFS spectra generated by FEFF and the experimental spectra observed for K2MoO4. We found a good agreement between the theoretical EXAFS oscillation and experimental data. Comparing with the X-ray crystallographic data, we can determine the bond distance with (0.002 nm accuracy by the FEFF analysis. The S02 was also determined as 0.90. Figure 2 shows Mo K-edge PTRF-XANES spectra of Mo/ TiO2 and reference compounds. A preedge peak is assigned to 1s f 4d transition.37 The peak positions for Mo/TiO2(110) in the three directions were 20 006 eV which are similar to that for MoO3, indicating that the Mo atoms are situated in the hexavalent level. The edge peak intensity correlated with the symmetry of Mo species and the number and direction of Mod O double bond. The tetrahedral [MoO4]2- has the strong peak intensity, whereas the octahedral species gives a very weak peak. MoO3 has a distorted octahedral structure, showing a small 1s f 4d peak. The spectra of Mo/TiO2(110) show much smaller preedge peaks than that of [MoO4]2- which may exclude the possibility of tetrahedral structure for Mo/TiO2(110). Bianconi et al. reported that the 1s f 3d peak height in V2O5 sample depends on the direction of VdO.38 Shirai et al. determined the direction of VdO for V oxides on a ZrO2(100) surface by means of PTRF-XANES.39 In the XANES spectra of Mo/TiO2(110) the 1sf 4d peak intensity in the [110] direction was the smallest among those in the [11h0], [001], and [110] directions, suggesting that molybdenum-oxygen bonds in the [110] direction possess the smallest double bond character, and/or the coordination number of double bond oxygen is the smallest in the [110] direction. Figure 3 shows Mo-K edge PTRF-EXAFS spectra for Mo/ TiO2(110) with different orientations of TiO2(110). The spectra (a)-(c) are entirely different from each other indicating the formation of anisotropic structure of Mo-oxide species on TiO2(110). The short-period EXAFS oscillation in the [11h0] spectrum appears at the high k-region. Since a heavy element like Mo has a large backscattering amplitude of EXAFS oscillation at the high k-region, the oscillation at the high k-region in the [11h0] spectrum can be attributed mainly to a heavier neighboring atom. In polarization-dependent EXAFS analysis, the orientation and location of the surface species are usually determined by

Figure 3. k0-weighted, background-subtracted raw PTRF-EXAFS spectra of Mo/TiO2(110) at Mo K-edge: (a) E//[11h0]; (b) E//[001]; and (c) E//[110]. Dotted line: observed; solid line: smoothing-treated.

comparing effective coordination number N*. This analysis is available for the sample with the coordination shell composed of one kind of distance. However, the supported Mo oxides may have a complicated structure with a variety of Mo-O distances. Indeed the XANES spectra in Figure 2 suggest a distorted octahedral structure, and thereby it can hardly determine the Mo structure by comparing the N* in different directions. Thus we compared the observed EXAFS oscillations with the EXAFS oscillations calculated assuming threedimensional structures. The flowchart for the analysis procedure is summarized in Figure 4. The parameters used for the calculation are descried in the Experimental Section. The calculations are iterated until the residual factor 2 calculated according to eq 4 become less than 1 for the data in all the three directions simultaneously. N

2 )

∑1 ∑((χobs(k) - χcal(k))/error(k))2

We only took statistical error into account because the S/N ratio of the observed data was not so good. The analysis may thus strongly depend on the model structure which we choose. We used the local structures of known Mo oxides, such as MoO3, (NH4)6Mo7O24, and Na2Mo2O7. Because there is the short-period in the range of 60-80 nm-1 indicating the presence of Mo-Mo, we neglected the isolated Mo monomer structure for the simulation. At first MoO3 bulk structures with (100), (010), and (001) faces normal to TiO2(110) surface were considered, but the

Mo Oxides on a TiO2(110) Single Crystal Surface

J. Phys. Chem. B, Vol. 102, No. 45, 1998 9009

Figure 4. Flowchart for the analysis procedure.

calculations based on these structures failed to reproduce the observed data. Preferential growth of MoO3 layers on TiO2 has been suggested in the literature.22 Therefore, next we calculated the layer structures with different orientations. Figure 5 shows the calculated EXAFS oscillations generated from MoO3 (010) one-atomic layer. The 2 were 4.5, 7.3, and 6.4, respectively, for E// [11h0], E//[001], and E//[110]. As a result, the observed data were not reproduced by this structure model. Similarly, we could not obtain 2 less than 1 with any other layer structures of MoO3. Model structures using polymolybdates such as Mo7O246- which was used as a precursor also failed to reproduce the experimental results as shown in Figure 6 (2 were 1.2, 1.3, and 1.2, respectively). Unsuccessful fitting results seemed to be ascribed mainly to the fact that the MoMo interaction is only found in the [11h0] direction, indicating the presence of the chain structures such as dimers, trimers, tetramers, etc. Na2Mo2O7 is known as an example of a simple dimer structure. But the Mo-Mo distance in Na2Mo2O7 is too long and did not fit well with the observed spectrum in the [11h0] direction. To analyze the EXAFS oscillations by dimer models at first, we have examined the possibility of dimer structures extracted from MoO3 with different arrangements. There are two types of Mo-Mo dimer units in MoO3. One is corner shared dimer and the other is edge shared dimer. We examined a corner shared Mo dimer on TiO2(110) as shown in Figure 7. This model did not reproduce the observed data at all, showing 2 ) 4.9, 4.3, and 28.5 for [11h0], [001], and [110] directions, respectively. On the other hand, an edge shared dimer model reproduced the observed EXAFS oscillations in two directions E//[11h0] and E//[001]as shown in Figure 8. However, the calculated oscillation in E//[110] direction did not fit the observed one. The EXAFS oscillation in E//[11h0] may strongly be affected by interfacial bonding between Mo dimer and TiO2(110) surface because the bonds normal to the surface can most contribute to the E//[110] EXAFS. Thus, we proceeded the analysis further by including the interfacial bondings between the Mo dimer and the TiO2 surface. Mo-Mo direction in the dimer structure should be parallel to the TiO2(110) surface because no Mo-Mo oscillation was observed in the [110] direction. One possible model is that Mo dimers weakly adsorb on TiO2(110) by van der Waals force. But it did not improve the fitting degree in the [110] direction. Therefore, we assumed

Figure 5. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from MoO3(010) layer structure (solid line): (a) E//[11h0]; (b) E//[001]; and (c) E//[110].

chemical bonding between Mo dimer and TiO2(110) in such a way that the two bridging oxygen atoms of the edge-shared Mo dimer are shared by TiO2(110) surface. The original Mo dimer structure extracted from MoO3 has a point symmetry as shown in Figure 9. In this case, if two bottom oxygen atoms (11 and 12) in the dimer structure in Figure 9 are located at the same height on TiO2(110). The Mo-Mo direction should be tilted to have the contribution to the E//[110] EXAFS. We inverted thus the six oxygen atoms around the Mo atom against the plane including the Mo-Mo axis and perpendicular to the [110] direction. Two possible locations of the Mo dimer on TiO2(110) surface are shown in Figure 10; one is the groove site (model A) and the other is the ridge site (model B). In both models the protruding oxygen atoms at the ridge site of TiO2(110) surface are shared by Mo dimer and the surface structure of TiO2(110) was assumed to be not much reconstructed except the position of two oxygen atoms shared by the Mo dimer. Figures 11 and 12 shows the EXAFS oscillations calculated from model A and model B after refinement by optimization

9010 J. Phys. Chem. B, Vol. 102, No. 45, 1998

Chun et al.

Figure 6. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from the Mo7O246- model (solid line): (a) E//[11h0]; (b) E// [001]; and (c) E//[110].

Figure 7. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from the corner-shared Mo dimer model (solid line): (a) E//[11h0]; (b) E//[001]; and (c) E//[110].

of the structure. We found that the calculation based on model B (Figure 12) (2 ) 0.8, 0.7 and 0.8 for [11h0], [001], and [110] directions, respectively) well reproduced the observed data for the three different directions. Model A (Figure 11) failed to obtain the value of 2 less than unity for the EXAFS oscillation in [110] direction (2 ) 3.2). We also carried out the theoretical calculations for trimer, tetramer, and linear structures of Mo oxides extracting from MoO3. Only a trimer model among them reproduced the EXAFS oscillations for E//[11h0] and E//[001] within the error bar. However, we were not able to find good fitting results in the E//[110] direction for any location of Mo trimer. As a result, model B structure was most plausible among the Mo oxides on TiO2(110) examined in this work. The determined Mo-dimer structure is illustrated in Figure 13. The Mo-O bond lengths are summarized in Table 1. Error bars were estimated by varying a given bond until 2 exceeds 1. The Mo-Mo distance was determined to be 0.335 nm along the [11h0] direction. Error bar of the Mo-Mo direction against the [11h0] direction of TiO2 (110) was estimated to be within (10°. The Mo-O and Mo-Ti distances at the interface were assessed

at 0.220 and 0.296 nm, respectively. In model B the two oxygen atoms bridging two Mo atoms are originally the surface oxygen atoms in the protruding one-dimensional row on TiO2(110), which were distorted from their original position as shown in Figure 13. If those two oxygen atoms were placed at the original positions of TiO2(110), we could not reproduce the observed EXAFS oscillations, where the 2 values were 1.7, 1.4, and 0.8 for [11h0], [001], and [110] directions, respectively. Discussion A. Analysis Procedure. In this paper we propose the Mo dimer structure attached on TiO2(110) shown in Figure 13 by the direct comparison of the observed and calculated spectra based on three-dimensional structural models. The analysis had difficulties: one is that the analysis should have been carried out for a great number of model structures, which was timeconsuming processes. The other is that there was no guarantee that the determined structure is of unique solution. To overcome these difficulties we postulated the following guidelines.

Mo Oxides on a TiO2(110) Single Crystal Surface

J. Phys. Chem. B, Vol. 102, No. 45, 1998 9011

Figure 9. An edge-shared Mo dimer extracted from MoO3.

Figure 8. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from the edge-shared Mo dimer model (solid line): (a) E// [11h0]; (b) E//[001]; and (c) E//[110].

1. We started the calculation and the optimization initially from the known stable structure of Mo6+ oxides. 2. We checked whether interatomic distances of Mo-O and O-O are in the range of the corresponding bonds found in crystal. The structure in model B has 12 Mo-O bonds as shown in Table 1. These bonds lie in the range of 0.17-0.23 nm as found in MoO3. The O-O bond lengths should be longer than the atomic radius. The minimum O-O bond length in model B is 0.26 nm between O(3) and O(6) in Figure 13, which is a similar value to that in MoO3 (0.26 nm). 3. We checked the similarity of local structure between the determined structure and the known structure. The first nearestneighbor structure of the determined Mo dimer (model B) is very similar to the Mo dimer unit in MoO3. As mentioned above, since we carried out the reflection operation around on Mo oxide, the Mo dimer structure becomes point-symmetric against the center of two Mo atoms. Mirror-plane symmetric structure can be found in the Mo dimer extracted from[Mo7O24]6(Table 1). Thus, the model B has an intermediate structure between the Mo dimer units in MoO3 and [Mo7O24].6-

Figure 10. Two models for Mo dimer structures on TiO2(110). Model A: Mo dimer is located on the groove sites composed of fivecoordinated Ti atoms. model B: two Mo atoms are straddled on the protruding oxygen atoms.

B. Mo Dimer Structure. We first expected that several model structures can satisfy 2 value less than unity for fitting the observed data. However, we found that only model B could reproduce the observed EXAFS oscillations among many models we tested. It is probably because the rather strict condition that 2 [11h0], [001], and [110]) are less than 1 at once is required in the determination of a satisfied structure. It is most likely that the dimer structure is a major species on the TiO2(110) surface. Other minor structures may also be present on the surface, but they were not evidenced by the present EXAFS analysis in the three different directions. The features of model B are summarized as follows. (1) The structure is the edge-shared dimer which has an intermediate structure of the dimers in MoO3 and [Mo7O24].6The bond distances in the dimer are given in Table 2. (2) The Mo-Mo distance at 0.335 nm is perpendicular to the protruding one-dimensional oxygen rows and parallel to the TiO2(110) surface. The bond distance is a little shorter than that of the edge-shared Mo-Mo in MoO3(Mo-Mo ) 0.343 nm) and is similar to that in [Mo7O24].6-

9012 J. Phys. Chem. B, Vol. 102, No. 45, 1998

Chun et al.

Figure 11. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from the edge-shared Mo dimer (model A) (solid line): (a) E//[11h0]; (b) E//[001]; and (c) E//[110].

Figure 12. k0-weighted, background-subtracted raw Mo K-edge PTRFEXAFS spectra of Mo/TiO2(110) (dotted line) and FEFF simulations generated from the edge-shared Mo dimer (model B) (solid line): (a) E//[11h0]; (b) E//[001]; and (c) E//[110].

(3) Two Mo atoms share the bridging oxygen atoms with TiO2(110) surface. The position of the bridging oxygen atoms are distorted and deviated from their original positions on TiO2(110). (4) The Mo dimer is located at the distances of 0.218 nm up from the surface Ti plane as shown in Figure 13. The Mo-O and Mo-Ti distances at the interface were observed at the distances of 0.220 ( 0.009 nm and 0.296 ( 0.009 nm, respectively. The Mo-Ti interaction was certainly contributed to the calculation. (5) The ModO bonds (Mo1-O9, Mo1-O10, Mo2-O4, and Mo2-O6 in Figure 13) are to the TiO2(110) surface. The XANES data suggested that more ModO bonds point to the direction parallel to the surface. Thus the EXAFS analysis coincides with the XANES expectation. The Mo dimer species have been reported to show high catalytic activities. Iwasawa et al. found Mo-oxide dimers on SiO2 to be active for ethanol oxidation and metathesis reactions.4,40-42 Niwa et al. reported that Mo dimers on SnO2 worked as active sites for methanol oxidation.43 Ichikawa et al. also reported that SiO2-supported Mo dimers derived from Mo2(OAc)4 and Mo2(NMe4)6 showed high activities in propene metathesis and ethene dimerization.44-46 Evans et al. observed high activities of Al2O3-supported Mo dimers derived from Li4Mo2Me8 and Mo2(OAc)4 for hydrogenation reactions.47 Thus Mo dimers may behave as active structures for many catalytic reactions. The Mo dimer found in this work was active for reaction with methanol.48

The formation of dimeric species on metal oxide surfaces has been reported by several groups. Gota et al. claimed that Cu dimer species grow on R-Al2O3(0001) from the consideration of Cu-Cu bond length.49 Burrows et al. suggested the chain of W-oxide dimers on TiO2 surface by TEM and EXAFS.50 Xu et al. reported STM image for Pd dimers on TiO2(110).51 Thus dimer structures seem to be stable on oxide surfaces. The Pd dimers on TiO2(110) are distributed with Pd-Pd bond parallel to the protruding oxygen ridge along the [001] direction.51 Onishi et al. suggested that Ni atoms on TiO2(110) grow along the protruding oxygen rows by interaction between Ni particles and the substrate.52 On the contrary, two Mo atoms were bonded along [11h0] direction to form Mo dimer structure attached on the ridge-oxygen atoms. It has been demonstrated that the protruding oxygen atoms on TiO2(110) are adsorption sites for positively charged metal atoms.53,54 It is most likely that Mo6+ ions preferably interact with the ridge-oxygen atoms. As shown in Figure 13, the bridging oxygen atoms shared with Mo atoms are deviated from the original positions by strong interaction with Mo6+ ions. The energy demerit by the distortion would be reversibly gained by chemical bonding of the bridging oxygen atoms with two Mo ions to increase the coordination number.55 The reduction of the distortion energy and the gain of the bond energy would be advantageous with Mo dimers rather than with Mo monomers. Thus the distortion energy of bridging oxygen may be a key issue to form the Mo dimers attached to the oxygen ridge of TiO2(110). Further, the anisotropic topography of the TiO2(110) surface (Figure 1) may prevent the Mo oxides from the aggregation to larger clusters.

Mo Oxides on a TiO2(110) Single Crystal Surface

J. Phys. Chem. B, Vol. 102, No. 45, 1998 9013 as the solvent for (NH4)6Mo7O24‚4H2O in the impregnation procedure. It was found that the sample was contaminated with Na and K probably from the distilled water. It is known that [MoO4] a tetrahedral structure is stabilized by the presence of alkali metals on oxide supports.57-61 The present results show a significant effect of alkali metal impurity in the impregnation solvent on the Mo-oxide structure produced on the oxide surface. Finally, we should also mention the sensitivity of PTRFXAFS technique. We can determine the structure of 0.2 ML Mo oxides with the accuracy in the bond length less than 0.01 nm for the first and second shell. Further improvements in the synchrotron source and detection system will make it possible to measure a change of PTRF-EXAFS in chemical processes such as catalysis with shorter time. Accumulated knowledge of anisotropic and asymmetric structures of active metal sites at oxide surfaces including the interfacial bondings may provide a new way to develop efficient catalytic systems and a new concept of structural chemistry at the interface. Conclusion

Figure 13. A proposed structure (model B) for Mo oxides on TiO2(110).

TABLE 1: Bond Lengths of Mo Dimer Structures: (a) Mo Dimer on TiO2(110); (b) Edge-Shared Mo Dimer Unit in MoO3; (c) Mo Dimer Unit in [Mo7O24]6-

We succeeded in measuring Mo K-edge PTRF-XANES and PTRF-EXAFS spectra for Mo oxides on a rutile TiO2(110) single-crystal surface in three different directions for the first time. It was concluded that Mo oxides were supported in a dimer structure, which stepped over the bridging oxygen row of TiO2(110), with the Mo-Mo distance at 0.335 nm along the [11h0]. An anisotropic Mo-oxide dimer structure on TiO2(110) was determined in such a way that the EXAFS oscillations calculated for several hundred structural models were directly compared with the observed EXAFS oscillations in the [11h0], [001], and [110] directions. Each molybdenum atom in the dimer is bound to three oxygen atoms and one Ti atoms at the TiO2(110) surface. Distortion around the bridging oxygen atoms on TiO2(110) is associated with the formation of the Mo dimer. Acknowledgment. This work has been supported by CREST (Core Research for Evolutional Science and Technology) of Japan Science and Technology Corporation (JST) and performed under the approval of the Photon Factory Advisory Committee (Proposal No. 94G029). References and Notes

Mo1-Mo2 Mo1-O5 Mo1-O7 Mo1-O8 Mo1-O9 Mo1-O10 Mo1-O12 Mo2-O3 Mo2-O4 Mo2-O5 Mo2-O6 Mo2-O7 Mo2-O11

(a) (nm)

(b) (nm)

(c) (nm)

0.335 ( 0.008 0.232 ( 0.008 0.192 ( 0.008 0.179 ( 0.005 0.178 ( 0.008 0.169 ( 0.008 0.220 ( 0.009 0.179 ( 0.005 0.169 ( 0.008 0.192 ( 0.008 0.178 ( 0.008 0.232 ( 0.008 0.220 ( 0.009

0.343 0.233 0.195 0.225 0.195 0.167 0.174 0.174 0.167 0.195 0.195 0.233 0.225

0.325 0.225 0.197 0.171 0.195 0.177 0.217 0.171 0.196 0.225 0.172 0.191 0.218

In the previous paper,56 we found that tetrahedral [MoO4]2was formed on TiO2(110) when usual distilled water was used

(1) Koningsberger, D. C.; Prins, R. X-ray absorption, Principles, applications, techniques of EXAFS, SEXAFS, and XANES; Wiley: New York, 1988. (2) Iwasawa, Y. X-ray absorption fine Structure for catalysts and surface; World Scientific: Singapore, 1996. (3) Bart, J. C.; Vlaic, G. AdV. Catal. 1987, 35, 1. (4) Iwasawa, Y.; Asakura, K.; Ishii, H.; Kuroda, H. Z. Phys. Chem. N. F. 1985, 184, 105. (5) Asakura, K.; Iwasawa, Y. J. Phys. Chem. 1989, 93, 4213-4218. (6) Asakura, K.; Bando, K. K.; Iwasawa, Y.; Arakawa, H.; Isobe, K. J. Am. Chem. Soc. 1990, 112, 9096. (7) Izumi, Y.; Chihara, T.; Yamazaki, H.; Iwasawa, Y. J. Phys. Chem. 1994, 98, 594. (8) Chun, W.-J.; Tomishige, K.; Hamakado, M.; Asakura, K.; Iwasawa, Y. J. Chem. Soc., Faraday Trans. 1995, 91, 4161. (9) Ressler, T.; Hagelstein, M.; Hatje, U.; Metz, W. J. Phys. Chem. B 1997, 101, 6680. (10) Koningsberger, D. C.; Gates, B. C. Catal. Lett. 1992, 14, 271. (11) Asakura, K.; Iwasawa, Y. J. Phys. Chem. 1991, 95, 1711. (12) Jaklevic, J.; Kirby, J. A.; Klein, M. P.; Robertson, A. S.; Brown, G. S.; Eisenberger, P. Solid State Comm. 1977, 23, 679. (13) Heald, S. M.; Keller, E.; Stern, E. A. Phys. Lett. 1984, A 103, 155. (14) Shirai, M.; Inoue, T.; Onishi, H.; Asakura, K.; Iwasawa, Y. J. Catal. 1994, 145, 159. (15) Shirai, M.; Asakura, K.; Iwasawa, Y. Chem. Lett. 1992, 15, 247. (16) Shirai, M.; Asakura, K.; Iwasawa, Y. Chem. Lett. 1992, 1037. (17) Asakura, K.; Shirai, M.; Iwasawa, Y. Catal. Lett. 1993, 20, 117.

9014 J. Phys. Chem. B, Vol. 102, No. 45, 1998 (18) Asakura, K.; Tomishige, K.; Shirai, M.; Chun, W.-J.; Yokoyama, T.; Iwasawa, Y. Physica B 1995, 208&209, 637. (19) Asakura, K.; Chun, W. J.; Shirai, M.; Tomishige, K.; Iwasawa, Y. J. Phys. Chem B 1997, 101, 5549. (20) Vanhove, D.; Op, S. R.; Fernandez, A.; Blanchard, M. J. Catal. 1979, 57, 253. (21) Ono, T.; Kubogawa, Y.; Miyata, H.; Nakagawa, Y. Bull. Chem. Soc. Jpn. 1984, 57, 1205. (22) Machej, T.; Doumain, B.; Yasse, B.; Delmon, B. J. Chem. Soc., Faraday Trans. 1 1988, 84, 3905. (23) Volta, J. C.; Tatibouet, J. M. J. Catal. 1985, 93, 467. (24) Ng, K. Y. S.; Gulari, E. J. J. Catal. 1985, 92, 340. (25) Shimada, H.; Matsubayashi, N.; Sato, T.; Yoshimura, Y.; Nishijima, A.; Kosugi, N.; Kuroda, H. J. Catal. 1992, 138, 746. (26) Henrich, V. E.; Cox, P. A. The Surface Science of Metal Oxide Cambridge University Press: Cambridge, 1994. (27) Maschhoff, B. L.; Pan, J. M.; Madey, T. E. Surf. Sci. 1991, 259, 190. (28) Onishi, H.; Iwasawa, Y. Surf. Sci. 1994, 313, L783. (29) Onishi, H.; Fukui, K.; Iwasawa, Y. Bull. Chem. Soc. Jpn. 1995, 68, 2447. (30) Charlton, G.; Hows, P. B.; Nicklin, C. L.; Steadman, P.; Tayor, J. S. G.; Muryn, C. A.; Harte, S. P.; Mereer, J.; McGrath, R.; Norman, D.; Turner, T. S.; Thornton, G. Phys. ReV. Lett. 1997, 78, 495. (31) Ramamoorthy, M.; Vanderbilt, D.; King-Smith, R. D. Phys. ReV. B 1994, 49, 16721. (32) Satow, Y.; Iitaka, Y. ReV. Sci. Instrum. 1989, 60, 2390. (33) Chun, W.-J.; Shirai, M.; Tomishige, K.; Asakura, K.; Iwasawa, Y. J. Mol. Catal. A: Chemical 1996, 107, 55. (34) Rehr, J. J. Jpn. J. Appl. Phys. 1993, 32, 8. (35) Vaarkamp, M.; Dring, I.; Oldman, R. J.; Stern, E. A.; Koningsberger, D. C. Phys. ReV. B 1994, 50, 7872. (36) Zabinsky, S. I.; Rehr, J. J.; Ankudinov, A.; Albers, R. C.; Eller, M. J. Phys. ReV. B 1995, 52, 2995. (37) Cramer, S. P.; Hodgson, K. O.; Gillum, W. O.; Mortenson, L. E. J. Am. Chem. Soc. 1978, 100, 3398. (38) Stizza, S.; Mancini, G.; Benfatto, M.; Natori, C. R.; Garcia, J.; Bianconi, A. Phys. ReV. B 1989, 40, 12229. (39) Shirai, M.; Asakura, K.; Iwasawa, Y. Catal. Lett. 1994, 26, 229.

Chun et al. (40) Iwasawa, Y. AdV. Catal. 1987, 35, 187. (41) Iwasawa, Y.; Tanaka, H. Proceedings of the 8th International Congress on Catalysis Berlin IV; 1984; p 381. (42) Iwasawa, Y.; Ogasawara, S. J. Chem. Soc., Faraday Trans. 1 1979, 75, 1465. (43) Niwa, M.; Sano, M.; Yamada, H.; Murakami, Y. J. Catal. 1995, 151, 285. (44) Zhuang, Q.; Fukuoka, A.; Fujimoto, T.; Tanaka, K.; Ichikawa, M. J. Chem. Soc., Chem. Commun., 1991, 745. (45) Ichikawa, M.; Zhuang, Q.; Li, G. J.; Tanaka, K.; Fujimoto, T.; Fukuoka, A., Proceedings of the 10th International Congress on Catalysis; 1992; p 529. (46) Zhuang, Q.; Tanaka, K.; Ichikawa, M. J. Chem. Soc., Chem. Commun. 1990, 1477. (47) Evans, J.; Gauntlett, J. T.; Mosselmans, J. F. W. Faraday Discuss. Chem. Soc. 1990, 89, 107. (48) Yamaguchi, Y.; Chun, W.-J.; Suzuki, S.; Onishi, H.; Asakura, K.; Iwasawa, Y. Res. Chem. Intermed. 1998, 24, 151. (49) Gota, S.; Gautier-Soyer, M.; Douillard, L.; Duraud, J. P.; Lefevre, P. Surf. Sci. 1996, 352, 1016. (50) Burrows, A.; Kiely, C.; Joyner, R. W.; Kno¨zinger, H. K.; Lange, F. Catal. Lett. 1996, 39, 219. (51) Xu, C.; Lai, X.; Zajac, W.; Goodman, D. W. Phys. ReV. B 1997, 56, 13464. (52) Onishi, H.; Aruga, T.; Egawa, C.; Iwasawa, Y. Surf..Sci. 1990, 233, 261. (53) Onishi, H.; Aruga, T.; Egawa, C.; Iwasawa, Y. J. Chem. Soc., Faraday Trans 1 1989, 85, 2597. (54) Onishi, H.; Iwasawa, Y. Catal. Lett. 1996, 38, 89. (55) Chun, W.-J.; Asakura, K.; Iwasawa, Y. Chem. Phys. Lett. 1998, in press. (56) Chun, W.-J.; Asakura, K.; Iwasawa, Y. Catal. Today 1998, in press. (57) Kantschewa, M.; Delannay, F.; Delgado, E.; Eder, S.; Ertl, G.; Jeziorowski, H.; Knozinger, H. J. Catal. 1984, 87, 482. (58) O’Young, C. L. J. Phys. Chem. 1989, 93, 2016. (59) Martin, C.; Martin, I.; Rives, V. J. Chem. Soc., Faraday Trans. 1993, 89, 4131. (60) Martin, C.; Martin, I.; Rives, V. J. Catal. 1994, 147, 465. (61) Martin, C.; Martin, I.; Rives, V.; Malet, P. J. Catal. 1996, 161, 87.