J . Chem. SOC., Faraday Trans. I, 1988, 84(5), 1329-1340 A New Ti0,-attached Rhodium Metal Catalyst Catalyst Characterization and Non-SMSI Behaviour Kiyotaka Asakura, Yasuhiro Iwasawa" and Haruo Kuroda Department of Chemistry, Faculty of Science, The University of Tokyo, Hongo, Tokyo 113, Japan Ti0,-attached Rh metal cluster catalysts have been prepared by the reaction between Rh(v3-C,H,), and surface OH groups of TiO,, followed by H, treatments at different temperatures. This class of Rh/TiO, catalysts, which are very active for ethene hydrogenation and ethane hydrogenolysis, show no SMSI phenomena, unlike conventional Rh/TiO, catalysts prepared by a traditional impregnation method using aqueous RhCl,. The structural properties of these catalysts, such as a shortening of the Rh-Rh distance, their raft-like structure and their positively charged character, have been studied in relation to non-SMSI behaviour by means of EXAFS together with other physiochemical techniques.These properties of structure, electronic state, chemisorption and catalytic activity may be ascribed to the Rh-0-Ti bonds which are formed in during attachment of the Rh-ally1 complexes onto TiO, and are partially retained after high-temperature H, reduction. Transition-metal n-ally1 complexes readily react with surface OH groups of inorganic oxides such as SiO,, A1,0, and TiO, to form attached-metal catalysts with well defined structures and properties on a molecular level as well as to provide highly dispersed metal particles on the support surfaces by the subsequent reduction with H, under mild conditions.'* The attached noble-metal catalysts such as Pd/Si0,,3,4 Pt/Si0,,4 Rh/Si0,,5,6 Rh/Ti0,7 etc.have been prepared by this method. TiO, has attracted much attention as a support because the 'SMSI effect' appears when H, or CO are chemisorbed on noble metals when they are dispersed on TiO,.' We have prepared a Ti0,-attached Rh catalyst using Rh( r73-C3H5)3,7a which shows higher activities for ethane hydrogenolysis and ethene hydrogenation than a con- ventional (impregnated) Rh/TiO, catalyst.'" In addition to this difference in activity, they also exhibit different adsorption behaviour ; the impregnated Rh/TiO, catalyst almost loses the ability to chemisorb H, and CO after high-temperature reduction with H,(SMSI phenomenon),8 while the Ti0,-attached Rh catalyst does not.T.e.m. studies have shown that the average size of Rh particles in the attached catalyst is ca. 1.4 nm, i.e. smaller than the Rh particles prepared by the impregnation method (ca. 3.1 nm). The uniqueness of the present attached sample is not due to the small particle size because Haller et al. pointed out that the SMSI effect is more apparent for smaller Rh particle^.^ A different metal-support interaction should be considered in this attached Rh/TiO, system, which would reflect the bonding feature and electronic state of the Rh metal particles on TiO,. It has been demonstrated by many researchers that EXAFS is a powerful tool for investigating the local structure around a specific atom in supported fine metal particles." It is potentially possible to elucidate the metal-support Recently Rh/TiO, catalysts prepared by an ion-exchange method were studied by means of EXAFS and direct Rh-Ti bonding was observed in the SMSI state.14*15 XANES is 13291330 New Ti0,-attached Rh Catalyst also powerful in obtaining information about the electronic state of metal atoms.In the present work, the bonding features and electronic states of the Rh atoms in the Ti0,- attached Rh catalysts were studied by means of EXAFS and XANES together with other physicochemical techniques such as i.r., t.e.m. and temperature-programmed decomposition (t .p.d.). Experimental Preparations and Treatments of the Catalysts The attachment procedure of Rh(q3-C3H,), onto TiO, (partially dehydrated Degussa P-25) is illustrated in scheme 1 .7 a The TiO, surface was first decarboxylated by heating to 790 K, exposed to water vapour at 293 K, then treated at 473 K under vacuum in situ A B Scheme 1.The attachment of Rh(v3-C3H5), onto TiO, and surface transformations. before the attachment reaction. The attachment reaction was carried out at 273 K in a specially devised Pyrex rea~tor.~' The Rh loading was 2.0 wt YO as Rh/TiO,. The attached Rh species (species A) was treated with H, to form species B at 293 K. B was reduced with H, at 773 K, followed by oxidation with 0, at 673 K. Then the sample was reduced with H, at the desired temperatures. The first reduction-oxidation process was carried out to minimize the effect of aggregation of the Rh particles. For comparison, the Si0,-attached Rh (2.0 wt YO) catalysts were prepared in a similar manner using the reaction between Rh(q3-C,H,), and the pretreated SiO,.The impregnation Rh/Ti0,(2.0 wt%) catalysts were obtained by the usual impregnation method using an aqueous RhCl, solution, followed by calcination at 673 K. The impregnation catalysts were reduced with H, at 473 and 773 K and are denoted as I-Rh/Ti02(473) and I-Rh/Ti02(732), respectively. Surface Characterization of the Samples 1.r. spectra were recorded in an i.r. cell connected to a closed circulating system on a JASCO IR-8 10 spectrometer. T.p.d. measurements were carried out by heating the samples at a rate of 4 K min'l. The evolved gases were analysed by gas chromatography. EXAFS spectra were measured at the beam line 10B of the Photon Factory (KEK-PF).l' The samples treated in a closed circulating reactor were transferred to EXAFS measurement cells which were also connected to the same closed circulating reactor.Thus air contact was avoided completely. T.e.m. photographs were taken on a JEOL-200B electron microscope. Results and Discussion Surface Structures of Species A and B Fig. 1 shows the i.r. spectra of species A and B. The absorption bands of species A are observed at 3050 and 1462 cm-l, suggesting that the allyl ligand is still of x-ally1 character.,., After the exposure of A to H, at 9.3 kPa for 1 h at room temperature, the allyl ligand was completely removed and a new peak appeared at 2048 cm-l, as shownK. Asakura, Y. Iwasawa and H. Kuroda 1331 I 1 I 3000 2000 1500 wavenumber/crn-' Fig.1. 1.r. spectra of the species A(a) and B(b). The broken curve of (b) shows the species after deuterium exchange of species B. Table 1. Stoichiometries in attachment of Rh(q3-C3H,), and surface transformations. amount of Rh attached on TiO, 2.0 wt Yo no. of allyl ligands per Rh atom from t.p.d. 0.98 from H, reaction 1.10 in fig. 1. Gas-phase analysis showed that 1.10 molecules of H, per Rh atom were consumed in the reaction of the allyl ligands with H, as given in table 1. The i.r. peak observed with B shifted to 1495 cm-l after deuterium exchange, as shown in fig. 1. Thus, this peak is attributed to the Rh-H vibration of terminal hydride species B. No peaks corresponding to bridging hybrides or the allyl group were found for species B, unlike the previously reported SiO, 5y6'and Ti0,-attached catalyst^.^ Fig.2 shows the t.p.d. spectrum of species A.17 The main products of t.p.d. were C,H6 and C,H,. The number of allyl ligands per Rh atom in species A was determined from the t.p.d. chromatogram to be 0.98, in good agreement with the result obtained from the reaction with H, as shown in table 1. From these results the amount of the allyl ligand in species A is unity per Rh atom. The production of C,H, in fig. 2 seems to be unusual, but we infer that the C,H, peak may be caused by the interaction of the allyl ligand with a Ti surface atom as previously reported." From i.r., t.p.d. and H, titration, the incipient surface Rh complex and subsequent Rh-hydride species may be illustrated as in scheme 1.Chemisorption Measurements Fig. 3 shows the results of H, chemisorption on the Rh/TiO, and Rh/SiO, catalysts. The A-Rh/SiO, catalysts showed 100% dispersion (H/Rh = I), all Rh atoms being available for H, adsorption. The dispersion of ca. 50 % for the A-Rh/Ti02(473) catalyst is still high as compared with 24 YO dispersion for the I-Rh/Ti02(473) catalyst. From the dispersion of the catalysts and the assumption that the shape of the Rh particle is spherical and that all surface Rh atoms are available for H, chemisorption to form Rh-H bonds, the Rh particle sizes (diameters) are estimated to be 2.9 and 6.0 nm for the attached and impregnated Rh/TiO, catalysts, respectively. The I-Rh/TiO, system largely lost the ability to adsorb H, after high-temperature reduction owing to the so-1332 New 730,-attached Rh Catalyst 273 473 t T/K 13 Fig.2. Temperature-programmed decomposition of the species A ; (---) C,H,, (- - -) C,H,, (-) c H , ( ...... ) C3H8 and (-. --) C,H8. 0 -0.5 \ W x w 4 - 1 . 0 - 1 . 5 ‘b 1 I I 1 473 573 673 773 873 T/K Fig. 3. H, chemisorption of the attached and impregnated Rh/TiO, and Rh/SiO, catalysts; 26.6 kPa H,; (a) A-Rh/SiO,, (6) I-Rh/SiO,, ( c ) A-Rh/TiO, and ( d ) I-Rh/TiO,.J . Chem. SOC., Faraday Trans. 1, Vol. 84, part 5 Plate 1. (a) and (b), for legend see following page. K. Asakura, Y. Iwasawa and H. Kuroda Plate 1 (Facing p . 1332)J . Chem. Soc., Faraday Trans. 1, Vol. 84, part 5 Plate 1. (c) and ( d ) , for legend see facing page. Plate 1 K. Asakura, Y. Iwasawa and H. KurodaJ .Chem. SOC., Faraday Trans. 1, Vol. 84, part 5 Plate 1 Plate 1. The TEM photographs of (a) TiO,(P-25), (b) A-Rh/Ti02(773), (c) A-Rh/Ti02(863), ( d ) I-Rh/Ti02(773) and (e) A-Rh/Si02(773). The magnification is 450000 x . Some Rh particles are indicated by circles and arrows. K. Asakura, Y. lwasawa and H. KurodaK . Asakura, Y. Iwasawa and H . Kuroda 1333 ~ 6 12 L 18 24 i L S i2 18 24 particle size/ lo-’ nm d 2 3 4 5 r (d 1 1 3 4 5 I tFj-li-ll particle size/nm Fig. 4. Particle-size distribution of Rh metal in the A-Rh/TiO, and I-Rh/TiO, catalysts determined from t.e.m. ; (a) A-Rh/Ti02(473), (b) A-Rh/Ti0,(773), (c) I-Rh/Ti0,(473) and (d) I-Rh/Ti02(773). Average particle diameters are 1.4 nm (a), 1.5 nm (b), 3.1 nm (c) and 3.4 nm (4. called ‘SMSI effect’, as shown in fig.3. In contrast to the traditional I-Rh/TiO, catalyst, the attached Rh catalyst showed no SMSI phenomenon, the dispersion remaining 40-50% in the reduction temperature range 473-773 K. The decrease in dispersion observed with A-Rh/Ti02(873) may be due to the sintering of metal particles as proved by t.e.m. measurements. T.E.M. Measurements Plate 1 shows micrographs of TiO,(P-25), A-Rh/Ti02(773), A-Rh/Ti02(873), 1-Rh/TiO2(773) and A-Rh/Si02(773). The Rh particles of A-Rh/Ti02(773) [shown by circles and arrows in plate 1 (h)] were very small compared with those of I-Rh/Ti02(773) in plate 1 ( d ) . The H, reduction of A-Rh/TiO, at 873 K led to an increase in the particle size as shown in plate 1 (c), which gives an Rh size of 1.9 nm on average.However, the Rh particles were resistant to agglomeration on reduction with H, up to 773 K. The particle size distributions were obtained by t.e.m. analyses in fig. 4. The average diameter of Rh particles was estimated to be 1.4 and 1.5 nm for A-Rh/Ti02(473) and A-Rh/ Ti0,(773), respectively. On the other hand, the average diameters were 3.1 and 3.4 nm for I-Rh/Ti02(473) and I-Rh/Ti02(773), respectively. The particle-size distributions were much narrower for the attached catalysts than for the impregnated catalysts, implying that the Rh atoms prepared from the attachment of Rh(q3-C3H,), via the1334 New Ti0,-attached Rh Catalyst h Y .r( 7 0 0.1 0.2 0.3 0.4 0.5 rlnm Fig. 5. Fourier transform of the Rh K-edge EXAFS spectrum of the attached Rh species (B) formed from the attached Rh species (A).chemically bonded surface species A (scheme 1) constitute relatively homogeneous and well dispersed small particles. The high dispersion of Rh metal in the A-Rh/SiO, catalyst is shown in plate 1 (e). The average size of Rh particles of A-Rh/Si02(773) was 1.3 nm. The impregnated Rh/Si0,(773) catalyst showed a wide distribution of Rh particle size (1 .O-5.9 nm) with an average diameter of 2.8 nm. Thus, well dispersed Rh metal with a narrow particle-size distribution appears to be formed from the surface atomic Rh species with the covalent bonds of Rh-0-Ti (-Si). EXAFS In order to investigate the local structures of Rh metal particles in A-Rh/TiO, and I-Rh/TiO,, we performed EXAFS analyses for these Rh catalysts during the preparation processes.We extracted EXAFS oscillation p ( E ) from the observed data by subtracting the smoothly varying part ,us(E) which was estimated from the cubic spline method." The oscillation thus obtained was normalized by the absorption coefficient of the free atom (1) Po(E) X(k) = ME) - , u s ( ~ ) I / P o ( E ) where k is the photoelectron wavenumber related to E by k = [2m/h2(E-EO)]t. (2) Eo is the threshold energy and is usually taken to be equal to the energy of the inflection point for convenience. We performed Fourier transformation of k3-weighted ~ ( k ) over the region 40-120 nm-l as shown in fig. 5 and 6. The peak in the Fourier transformation was filtered and then inversely Fourier- transformed back into k-space. The Fourier-K. Asakura, Y. Iwasawa and H.Kuroda 1335 4 6 8 10 12 k/ 10 nm-' 4 n Y x o .y - 4 I I I I 4 6 8 10 12 k/ 10 nm-' 8 n Y * o X - 8 I 1 I I I 4 6 8 10 12 k/10 nm-' r/ lo-' nm 0 1 2 3 1 5 r/ lo-' nm Fig. 6. EXAFS oscillations k ~ ( k ) and their Fourier transform (FT) of (a) A-Rh/Ti02(773), (b) A-Rh/Ti02(473) and (c) Rh metal. filtered data [X'(k)J were then analysed by means of a least-squares curve-fitting method using the theoretical EXAFS equation :la ~ ' ( k ) = SNF(k) exp ( - 2a2k2) sin [2kr + $(k)]/(kr2) (3) where N, Y and o represent the coordination number, bond length and Debye-Waller factor, respectively, these values being fitting parameters, together with Eo. $(k) and F(k) are the phase shift and amplitude functions, respectively, for which we used theoretically derived va1~es.l~ S is the amplitude reduction factor which arises from many-body effects and inelastic losses in the scattering process.This value can be regarded as a constant function of k because many-body effects and inelastic losses have opposite k- dependence.'O The results are given in tables 2 and 3. The S values of Rh-Rh and Rh-0 were determined from Rh metal and Rh,O,, chosen as reference materials. The Rh-Rh distance of bulk Rh metal thus determined was 0.266 nm, while 0.268 nm has been reported from X-ray crystallography. Fig. 5 shows the Fourier transform of an Rh1336 New Ti0,-attached Rh Catalyst Table 2. Curve-fitting results for Rh-0 (and Rh-C) peaks in the EXAFS data of the attached Rh species sample coordination no. distance/nm DW factor/nm A 3.0 (0.3) 0.214 (0.002) 0.0076 (0.001) B 2.0 (0.3) 0.2 15 (0.002) 0.0065 (0.001) Table 3.The curve-fitting analyses of the EXAFS data for the reduced Rh metal catalysts sample coordination distance DW factor edge energy no. /nm /nm / e v A-Rh/Ti02(473) 7.7 (2.0) 0.262 (0.002) 0.0081 (0.0010) 4.0 (0.5) A-Rh/Ti02(773) 8.3 (2.0) 0.262 (0.002) 0.0083 (0.0010) 3.1 (0.5) I-Rh/Ti02(473) 12.0 (2.0) 0.266 (0.002) 0.0076 (0.0010) 0.4 (0.5) I-Rh/Ti02(773) 12.0 (2.0) 0.267 (0.002) 0.0078 (0.0010) 0.4 (0.5) A-Rh/Si02(473) 5.0 (2.0) 0.264 (0.003) 0.0089 (0.0010) 0.0 (0.5) A-Rh/Si02(773) 5.2 (2.0) 0.264 (0.003) 0.0087 (0.0010) 0.0 (0.5) Rh metal (12) 0.266 (0.002) 0.0040 (0.0010) 0.0 Errors are indicated in the parentheses. K-edge EXAFS spectrum for the attached Rh species B, which was obtained by the interaction of the v3-allyl-type surface species A with H, at 298 K for 1 h.Since the scattering power of hydrogen is very small, the Rh-H bond does not contribute to the EXAFS oscillation. Only one peak assignable to the Rh-0 bond was observed at a distance of 0.215 nm (table 1) and no peak for the Rh-Rh bond was observed in the range 0.1-0.5 nm, as shown in fig. 5. This suggests that the Rh atom of species B is monoatomically distributed on the TiO, surface. The coordination number was determined to be 2 for Rh-0 bonds, which supports the bonding of Rh atoms to the surface through the oxygen atoms in a bidentate form as shown in scheme 1. Aggregation of the metal particles does not occur in this stage. This behaviour is different from that of the Si0,-attached Pd, Pt and Rh ally1 complexes, in which the metal aggregated during room-temperature r e d u ~ t i o n .~ * ~ Table 2 also shows the bond lengths and coordination numbers of Rh-0 (and Rh-C) bonds of the Ti0,-supported Rh-ally1 species A determined by curve-fitting analysis of the EXAFS data. The three Rh-C bond lengths for an q3-allyl ligand in (DBM)Rh(C,,H,,) are known to be 0.21 1, 0.218 and 0.221 nm,2* i.e. close to the Rh-0 bond length. Thus EXAFS cannot distinguish between the bond lengths of Rh-C(q3-allyl) and Rh-O(Ti0,). Therefore, a one-shell fit analysis was performed. Since species A was determined to have one $-ally1 ligand per Rh atom by i.r., t.p.d. and chemical analysis, the total coordination number of the carbon and oxygen around an Rh atom would be expected to be five.However, the large deviation of the bond length of Rh-C(q3-allyl) might diminish the contribution of Rh-C bonding to the total coordination number owing to the large static disorder. Fig. 6 shows the k-weighted EXAFS oscillation [ k ~ ( k ) ] and the Fourier transforms of A-Rh/Ti02(473), A-Rh/Ti02(773) and Rh metal. The oscillation observed in the higher wavenumber region implies the presence of Rh-Rh bonds and a peak around 0.25 nm (phase-shift uncorrected) assignable to an Rh-Rh bond was observed in each Fourier transform. The peak was further analysed by a curve-fitting technique. The results are given in table 3. The Rh-Rh bond lengths of I-Rh/Ti02(473) and I-Rh/Ti02(773) were 0.266nm, the same as the value for Rh metal.However, the A-Rh/Ti02(473) and A-Rh/Ti02(773) catalysts were found to have a bond length of 0.262 nm, which isK. Asakura, Y. Iwasawa and H . Kuroda 1337 shorter than those of Rh metal and the I-Rh/TiO, catalysts. The bond lengths of A-Rh/Si02(473) and A-Rh/Si02(773) were intermediate. The coordination numbers of A-Rh/Ti02(473) and A-Rh/Ti02(773) were determined to be 7.7 & 2.0 and 8.3 2.0, respectively, by EXAFS analysis. On the other hand, the Rh metal particles of both I-Rh/TiO, catalysts showed the same coordination number (12 k 2.0) as that of bulk Rh metal. Note that A-Rh/Si02(473) and A-Rh/Si02(773) showed very small coordination numbers, 5.0 & 2.0 and 5.2 k 2.0, respectively. The structure (morphology) of the Rh particles attached on TiO, can be discussed from the coordination number of Rh(EXAFS) and the particle size (t.e.m.) by the method reported by Greegor and Lytle.,, In this analysis only the first shell was used because the higher shells were too small to obtain the accurate coordination number, N i / N , values (Ni is the average coordination number of the first shell determined from EXAFS and N , is the bulk coordination number) were calculated to be 0.64k0.17 and 0.69 4 0.17 for A-Rh/Ti02(473) and A-Rh/Ti02(773), respectively, and the particle sizes were ca. 1.4 nm.According to the method of Greegor and Lytle, Ni/Nl should be ca. 0.89 in the case of a spherical particle with a diameter of 1.4 nm. The values for the A-Rh/TiO, catalysts fit well a raft-like particle model consisting of di- or tri-atomic Rh layers which should have Ni/Nl = 0.66 or 0.77, respectively. Similarly, A-Rh/Si02(473) is suggested to have a monolayer raft Rh structure from the particle size of 1.3 nm and N i / N , = 0.42 0.17.The same conclusion is valid also for A-Rh/Si02(773), indicating the stability of this structure attached on SiO,. These structures are compatible with the chemisorption data. Fig. 7 shows the near-edge structure of the Rh K-absorption spectra of Rh metal and A-Rh/Ti02(473). The absorption edge of the A-Rh/TiO, catalyst shifted towards higher energy compared to that of Rh metal. The reproducibility of this shift was confirmed by two independent X-ray absorption measurements. Table 3 shows the energy shifts of the Rh K-absorption edge of A-Rh/TiO,, I-Rh/TiO, and A-Rh/SiO, referred to the edge energy of Rh metal, where the edge energies are taken at the inflection point of the edge.The shifts of A-Rh/Ti02(473) and A-Rh/Ti02(773), 4.0 L- 0.3 and 3.1 f 0.3 eV, respectively, are quite large compared with 0.4 & 0.3 eV of I-Rh/Ti02(473) and (773), while A-Rh/SiO, showed no edge shift. There are two possibilities for the cause of the energy shift; one is the reduction of relaxation energy arising from a decrease of the particle size and the other is the large positive charge of the Rh particles. The edge energy of A-Rh/SiO, was the same as that of Rh metal (edge shift = 0.0 eV) although the Rh particle size of the catalyst (1.3 nm) was as small as that of A-Rh/TiO, (1.4 nm). Furthermore, the particle sizes of the I-Rh/TiO, catalysts (3.4 nm) were much larger than those of the A-Rh/TiO, catalysts as shown in fig.4. Therefore the change of the relaxation energy is unlikely to be important for the energy shift. Thus the positive charge of the Rh particles may be the primary cause of the positive edge shift. The Rh particles of A-Rh/TiO, are formed by H, reduction of the attached species B which have Rh-0-Ti bonds and are atomically dispersed at the TiO, surface, as shown by EXAFS. The controlled preparation procedure led to the formation of small Rh particles with a narrow size distribution. The small raft-like Rh particles would be stabilized partially due to the Rh-0-Ti bonds by which a positive charge is induced on the Rh assembly through electron transfer. It is quite difficult to estimate the correct amount of electron flow from the edge shift.However, we tried to guess roughly the charge of the Rh particles attached on the TiO, by comparing the edge shift with those of standard compounds. The energy shifts of Rh,0,(Rh3') and Rh,(CO),Cl,(Rh+) were observed to be 10.0+0.4 and 6.0f0.3 eV, respectively. The charge induced by the Rh-0-Ti bonding in the Rh particles may be well delocalized through the Rh-Rh metallic bonds. Therefore the oxidation number per Rh atom in the Ti0,-attached Rh particles is estimated to be < 1 + , i.e. the average number of Rh-0-(TiO,) bonds of an Rh atom is < 1. Because1338 New Ti0,-attached Rh Catalyst I ElkeV Fig. 7. Rh near-edge absorption spectra of Rh metal (--) and A-Rh/Ti02(773) (-).of this small coordination number, it was impossible to observe direct Rh-0 bonding from the analysis of EXAFS data. The edge energy of A-Rh/Ti02(773) was lower by 0.6eV than that of A-Rh/Ti02(473). This may be considered as a result of the contribution of the reverse electron transfer from Ti3+ produced by the high-temperature reduction, although the large positive shift of 3.1 eV is still observed as given in table 3. No edge shift was observed in A-Rh/SiO,, owing to the difference between A-Rh/SiO, and A-Rh/TiO, in the interaction of Rh and surface oxygen. The metal- oxygen interaction on SiO, is in general weak and possibly an ion (oxygen)-induced dipole interaction as Koningsberger and coworkers12 suggested. On the other hand, the Rh-0 bond in the Ti0,-attached Rh species must be a chemical bond which stabilizes the positive charge of the Rh particles.These Rh particles may be said to be ‘attached Rh metal clusters’ in the sense that the Rh assembly exists under a strong interaction with the surface oxygens as shown in fig. 8. Conventional Rh/TiO, catalysts nearly lose the ability to chemisorb H, and CO after high-temperature reduction (SMSI effect).’ Recent EXAFS studies on Rh/TiO, prepared by an ion-exchange method demonstrated that a direct Rh-Ti bond was produced in the SMSI state.14.15 The bond lengths reported by Haller et al.14 and Koningsberger et al.15 were 0.252 and 0.342 nm, respectively. The Rh-0 bond length given by Koningsberger et al. was 0.270 nm, similar to that in the Rh/A1,03 system.This means that the interaction between Rh and 0 is as weak as an ion-induced dipole interaction. We do not know why two reported Rh-Ti bond lengths were so different from each other, but it may be possible that these direct Rh-Ti bonds produce the SMSI state. Horlsey reported that the 0.7 electrons are transferred from Ti to Pt through the direct Pt-Ti bond on the basis of his Xa calculation^.^^ In contrast to the conventional catalysts, a ‘ Ti0,-attached Rh metal cluster ’ with a positively charged character shows no SMSI behaviour as has been mentioned above. The attached cationic Rh clusters had a dispersion of 50% (H/Rh = O.SO), which is much smaller than the H/Rh value expected from the particle size of 1.4 nm. Thus the adsorption of H, and CO does not show the extent of the metal dispersion in the charged Rh system. The ‘ Ti0,-attached Rh cluster’ catalyst which does not experience the SMSI phenomenon shows a high activity for the hydrogenolysis of ethane as compared with the conventional Rh/Ti02 catalyst, as shown in fig.9.7a The A-Rh/Ti02(773) catalyst showed a similar activity to the A-Rh/Ti02(473) catalyst, so the catalytic activity is capable of maintainingK. Asakura, Y. Iwasawa and H. Kuroda I 1.3nm Ir I I * I 1339 ( b ) t t o p Rh layer t s e c o n d Rh layer first Rh layer substrate 0 layer Fig. 8. A proposed structure of the attached Rh metal cluster on TiO,; (a) top view and (b) side view. reaction time/ min Fig. 9. Ethane hydrogenolysis on Rh/TiO, catalysts at 373 K. 0, species B (30 kJ mol-'); 0, A-Rh/Ti02(473) (51 kJ mol-l); x , A-Rh/Ti02(773) (53 kJ mol-l); -0, I-Rh/Ti02(473) (69 kJ mol-'); 0, I-Rh/Ti02(773) (82 kJ mol-'); ., A-Rh/Ti02(863) (132 kJ mol-').1340 New Ti0,-attached Rh Catalyst its high level under any reaction conditions as well as any pre-reduction temperatures.Our present work suggests that the non-SMSI behaviour of the ‘Ti0,-attached Rh clusters’ is due to the Rh-0-Ti bonds which were produced in the well defined attachment steps in scheme 1 and partially maintained after the high-temperature reduction. In summary, the Rh particle of the A-Rh/TiO, catalyst exists as small, thin metal clusters 1.4 nm large and with 2 or 3 Rh layers. The particular features of the attached Rh metal clusters such as the positively charged character, the shortening of the Rh-Rh distance and the thin raft-like structure are derived from the Rh-0-Ti bonds which were formed in the attachment procedure of the Rh-ally1 complex and partially retained after high-temperature H, reduction.Such Rh-0-Ti bonds cause the characteristic behaviour in the chemisorption of H, and CO and the high activity for ethane hydrogenolysis. We are grateful to Dr M. Numura and the PF staff for their technical help in the EXAFS measurements. References 1 Y. I. Yermakov, B. N. Kuzunetsov and V. A. Zakharov, Catalysis by Supported Complexes (Elsevier, 2 Tailored Metal Catalysts, ed. Y. Iwasawa (D. Reidel, Dordrecht, 1986). 3 Y. I. Yermakov, B. N. Kuzunetsov, L. G. Karakchiev and S. S. Derbeneva, Kinet. Katal., 1973, 14, 4 G. Carturan, G.Facchin, G. Cocco, S. Enzo and G. Navazio, J. Catal., 1982, 76, 405. 5 M. D. Ward and J. Schwarz, J. Am. Chem. SOC., 1981, 103, 5253. 6 H. C. Foley, S. J. DeCanio, K. J. Chao, H. H. Onuferko, C. Dybnowski and B. C. Gates, J. Am. 7 (a) Y . Iwasawa and Y. Sato, Chem. Lett., 1985, 507; (b) M. D. Ward and J. Schwartz, J. Mol. Catal., 8 S. J. Tauster, J. C. Fung and R. L. Garten, J. Am. Chem. SOC., 1978, 100, 170. 9 D. E. Resasco and G. L. Haller, Metal-Support and Metal-Additive Effects in Catalysis, ed. B. Imelik, C. Naccache, G. Coudurier, H. Praliaud, P. MCriaudeau, P. Gallezot, G. A. Martin and J. C. Vedrine (Elsevier, Amsterdam, 1982). Amsterdam, 1980). 709. Chem. SOC., 1983, 105, 3074. 1981, 11, 397. 10 F. W. Lytle, G. F. Via and J. H. Sinfelt, J. Chem. Phys., 1977, 67, 3831. 11 K. Asakura, M. Yamada, Y. Iwasawa and H. Kuroda, Chem. Lett., 1985, 511. 12 J. B. A. D. Van Zon, D. C. Koningsberger, H. F. J. Van’t Blik, R. Prins and D. E. Sayers, J. Chem. 13 P. Legarde, T. Murata, G. Vlaic, E. Freund, H. Dexpert and J. P. Bournonville, J. Catal., 1984, 84, 14 S. Sakellson, M. McMillan and G. L. Haller, J. Phys. Chem., 1986, 90, 1733. 15 D. C. Koningsberger, J. H. A. Martens, R. Prins, D. R. Short and D. E. Sayers, J. Chem. Phys., 1986, 16 H. Oyanagi, T. Matsushita, M. Ito and H. Kuroda, KEK Report, 1984, 83-30. 17 Y. Iwasawa and K. Asakura, Homogeneous and Heterogeneous Catalysis, ed. Y. I. Yermakov and V. 18 B. K. Teo, Basic Principles and Data Analysis, Inorganic Chemistry Concepts 9 (Springer-Verlag, 19 B. K. Teo and P. A. Lee, J. Am. Chem. SOC., 1979, 101, 2815. 20 B. K. Teo, M. R. Antonio and B. A. Averill, J. Am. Chem. SOC., 1986, 105, 3751. 21 G. Pantini, P. Pacanelli, A. Immirzi and L. Porri, J. Organomet. Chem., 1971, C17, 33; DBM = 1,3- 22 R. B. Greegor and F. W. Lytle, J. Catal., 1980, 63, 476. 23 J. A. Horsley, J. Am. Chem. SOC., 1979, 101, 2870. Phys., 1984, 80, 3914. 333. 90,3047. Likholobov (VNU Science Press, Tokyo, 1986). Berlin, 1986). diphenylpropane- 1,3-dionato. Paper 7/084; Received 16th January, 1987