Faruday Discuss. Chem. SOC., 1990, 89, 91-105 In Situ Studies of Supported Rhodium Catalysts Peter Johnston, Richard W. Joyner,* Paul D. A. Pudney, Efim S. Shpirot and B. Peter Williamsf Leverhulme Centre for Innovative Catalysis, Department of Chemistry, and Surface Science Research Centre, Uniuersity of Liverpool, PO Box 147, Grove St, Liverpool L69 3BX Rhodium catalysts have been prepared, supported on y-alumina, vanadium( I 1 1 ) oxide, chromia and molybdena. The activity and selectivity of these materials in the conversion of synthesis gas, ( H2/C0 : 1/1, 523 K and 5 bar pressure), to methanol, ethanol and hydrocarbons has been studied. All catalysts showed similar activity except for chromia, which was almost inactive at the temperature chosen. Selectivity to higher oxygenates followed the trend: V203 > Moo3 > A1203 > Cr203.The alumina supported catalysts had the highest selectivity to methanol. Catalysts have been characterised in situ by extended X-ray absorption fine structure (EXAFS) and the alumina-supported material has been examined in the greatest detail. On reduction at 473 K, rhodium particles with diameter (1 nm are observed by electron microscopy, and EXAFS indicates that these contain cu. 10 rhodium atoms on average. There is good evidence that these particles may be bonded to oxygen of the support or residual chlorine from the impregnation pro- cedure. X-ray photoelectron spectroscopy shows that the surface stoichiometry is ca. C1/ Rh : 1.3/ 1. On exposure to carbon monoxide at room temperature the metal particles break up completely and form entities with the composition (CO),.Rh.C12 and bond lengths similar to those in crystalline Rh2CI,(C0)4.On exposure to synthesis gas at 323 K metallic particles are partly reformed. Under conditions of catalytic relevance only metallic particles are observed although some Rh-CI bonding persists on alumina. No rhodium- chlorine bonding is observed on vanadia-supported catalysts, although the metal particles have a similar average size to those found on alumina. There is some evidence that the particle size is much less uniform on vanadia than on alumina. Exposure to synthesis gas causes small changes in particle parameters, which may be related to the chemisyption of carbon monoxide. The average particle size is much larger, >30A diameter, on the chromia support.The relation between catalyst structure and performance is discussed. Rhodium probably has the most interesting catalytic chemistry of any metal, playing a major role both in homogeneous and heterogeneous catalysis. it is the catalyst of choice for the reaction CO+NO -+ N,+CO, which is the most difficult of the reactions carried out in auto-exhaust treatment, and it is also excellent in the homogeneous carbonylation of methanol. It is an important component of the metal gauzes used for the oxidation of ammonia, and its performance in catalytic CO hydrogenation is fascinating. It is the only metal which produces significant yields of oxygenates other than methanol without promotion. Rhodium is also one of the most studied metals in catalysis. +Permanent Address: Zelinsky Institute for Organic Chemistry, U.S.S.R. Academy of Sciences, 47 Lenin $: Now at: ICI Chemicals and Polymers, Billingham, Cleveland.Prospect, Moscow, U.S.S.R.92 In situ Studies of Supported Rhodium Catalysts An aspect of major interest concerns the relative importance of zerovalent metal compared with either Rh' or Rh"'. This is of particular importance in synthesis gas chemistry, as indicated by several studies.' Using extended X-ray absorption fine structure, (EXAFS) spectroscopy, it was demonstrated that small rhodium particles were converted to isolated Rh' (CO)? entities on exposure to carbon monoxide. The present study was undertaken with two aims. The first was to probe the equilibrium: "2 Rhf nRh'(CO)? ('0 in the presence of mixtures of carbon monoxide and hydrogen.This question is of special interest in oxygenate synthesis, since Ponec and co-workers have argued strongly that positively charged metal ions are the active species in alcohol synthesis.' The catalytic significance of Rh' has also been stressed in a recent infrared study by Knozinger and co-workers.3 A preliminary account of part of our work has been p ~ b l i s h e d . ~ We are also interested in the role of the support in determining selectivity in synthesis gas chemistry. We distinguish three classes of product which are formed by rhodium catalysts, methanol, higher oxygenates and hydrocarbons. This ability of rhodium to synthesize each of these is interesting from a mechanistic standpoint.The Fischer- Tropsch synthesis, where the first step is thought to be cleavage of the carbon-oxygen bond, provides a route to hydrocarbons. In contrast, the integrity of this bond is maintained in methanol synthesis, as demonstrated by Takeuchi and Katzer for a palladium catalyst.6 The mechanistic pathway to ethanol and other higher oxygenates is unclear. It is most easily pictured as insertion of CO into a surface methylene or methyl species,' but this remains a conjecture. It is widely recognised that the support has a major influence on both activity and selectivity, but the reasons are again unknown. The extent to which well characterised rhodium surfaces dissociate carbon monoxide is a subject of controversy.x High-dispersion rhodium catalysts have therefore been prepared on a range of supports, characterised in situ by EXAFS and their activity in synthesis gas conversion measured.Experimental Catalysts with 1% rhodium loading by weight have been prepared by impregnation of rhodium trichloride (Johnson Matthey PLC) onto commercially available supports (Strem Chemicals), using the incipient wetness method. The requisite amount of RhCI,.xH,O was dissolved in the minimum volume of distilled water. Supports were used as received, except for alumina which was calcined at 873 K before impregnation. Following impregnation, the catalysts were dried in air, heated in vucuo for 12 h and pressed into discs 13 mm in diameter and cu. 1 mm thick. Some discs were used directly for EXAFS analysis while the remainder were sieved (600-1000 p m mesh), for catalytic activity measurements.Activity Testing A charge of 0.20 g was tested in a laboratory scale microreactor (Labcon Ltd, Croft-on- Tees), having a diameter of 3 mm. The reactor was purged with hydrogen and the catalyst reduced in situ at 673 K for 20 h (1 bar,? GHSV 3000 h-'). The catalyst was cooled to 523 K under hydrogen and the atmosphere changed to an equimolar CO-H2 mixture. Reaction conditions were either 5.0 or 30.0 bar pressure, with GHSV of ca. 3000 or ca. 18 000 h - I , respectively. .) 1 bar= 10' Pa.0.8 - 0.6 v) c1 .- C e 2 0.4 v x C .- C C .- 0.2 0.0 n 2 4 6 0 10 R I A Fig. I . ( a ) The EXAFS spectrum of dirhodium tetracarbonyl dichloride: (-) experiment, weighted by multiplication by k ' ; ( - - - ) calculated using the parameters given in table 1 .( b ) Fourier transform of the spectrum shown in fig. l ( a ) : (-) experiment, weighted by multiplication by k'; ( - - - ) calculated using the parameters given in table 1 .94 In situ Studies of Supported Rhodium Catalysts Table 1. Parameters used to fit the EXAFS spectrum of Rh,CI,(CO), shown in fig. 1 neighbour coordination number interatomic distance/ A Debye- Waller factor/ A' carbon carbon chlorine oxygen rhodium rhodium 1 .o 1 .o 2.0 5.8" 1 .o 1 .o 1.77 1.85 2.35 2.84 3.12 3.3 1 0.015 0.015 0.009 0.005 0.0 14 0.020 " See text. On-line product analysis was performed with a dual-column, parallel-injection gas chromatograph (Pye Unicam 4500 series). Permanent gases were separated by a 1 m x 3 mm column packed with Carbonsieve S (100-120 p m mesh) and detected by thermal conductivity.The hydrocarbon and oxygenate products were separated on a 30 m megabore capillary column and detected by flame ionisation. The peak areas were corrected using literature detector response factors.y Catalyst Characterisation EXAFS measurements were performed using station 9.2 on the wiggler beamline at the Daresbury synchrotron radiation source and an in situ apparatus which has been described previously.'" Catalysts in the form of 13 mm diameter discs were treated in appropriate gas mixtures at pressures up to 1 bar. During these experiments the source operated at 2 GeVt with circulating electron currents in the range 150-250 mA. Data collection procedures have been described previously,"' and data were anal ysed by standard procedures, which are discussed in detail elsewhere.' I The significance of all shells reported has been confirmed using a previously published statistical test,'? and random errors are quoted.Phase shifts have been obtained from the Daresbury data base and optimised by study of suitable standard compounds (rhodium metal, RhCl,, Rh2(C0)"C12 and Rh203). Our fit for metallic rhodium has been publi~hed.~ Fig. 1 shows that obtained for the dicarbonyl dichloride using the crystallographic parameters from Dahl et a l l 3 Six shells of neighbours were included, as indicated in table 1; the values used are taken from the crystal structure with the exception of the oxygen coordination number and the Debye- Waller factors. The linear Rh-C-0 bonding cause a 'focussing effect' and results in an exaggerated Rh-0 coordination number.Thermal vibrational parameters are not the same in EXAFS, where correlation of atomic motions is included, and X-ray diffraction, where it is not. X-ray photoelectron spectroscopy measurements were performed on a commercial spectrometer (Vacuum Generators ESCA3), using A1 K,, radiation. The adventitious carbon peak was assumed to have a binding energy of 285.0 eV and used for calibration purposes. Catalysts could be pretreated at pressures up to 1 bar in suitable environments. Relative stoichiometries were obtained from standard equations using excitation cross- sections due to Scofield.'" Results Catalytic Performance The performance of four 1% rhodium catalysts on different supports is summarised in table 2.Catalysts showed similar activity, with the exception of the chromia-supported + I eV.= 1 . 6 0 2 ~ 10 ' " J .Faraday Discuss. Chem. Snc., 1990, vol. 89 Plate 1. Scanning transmission electron micrograph of the as-received Rh/AI,O, catalyst. The bar at the top of the picture represents 100 A. P. Johnston ef al. ( Facing p. 9 5 )P. Johnston et al. 95 Table 2. Performance of rhodium catalysts in synthesis gas conversion, H,/CO:l/l at 523 K Plbar G H SV/ h- ' time on stream/h conversion'"'( Yo) selectivity' " ) oxygenates: methanol ethanol ethanal C, oxygenates others total oxygenates hydrocarbons: methane ethane ethene c, others total hydrocarbons 5 2920 0- 144 7 16.0 16.3 4.3 0.6 4.5 41.7 43 .O 2.3 0.9 4.3 6.2 56.7 30 17 960 144-216 3 45.2 5.6 7.8 4.3 0.1 63.0 35.5 0.2 0.4 0.9 37.0 - 30h 5 19 320 3000 216-300 0-36 S 6 42.8 9.7 13.1 29.3 3.5 9.1 1.5 1.6 0.5 0.4 61.7 so. 1 25.7 20.4 2.1 4.4 0.3 4.5 3.1 8.1 3.1 12.5 38.3 49.9 5 S 2559 3 000 0- 240 144- 200 11 6 - 3.5 - 3.1 18.6 2.8 8.2" 0.3 30.3 9.7 3.5" 8.0 40.4 57.3 16.0 13.4 0.2 4.4 9.5 9.5 3.6 5.7 69.7 90.3 I' Conversions and selectivities are averages over the times indicated.under standard conditions. ' Entirely propanal. '' Butanal and acetic acid. '' T = 623 K. Following re-reduction, material, which was almost inactive at 523 K. At 5 bar pressure, selectivity to higher oxygenates follows the sequence: V203 > Moo3 > Al2O3 > Crz03. The vanadia catalyst had the lowest selectivity to hydrocarbons at this pressure.The trend in selectivity to methanol was: A1203 > V203 > Cr203 > MOO,. The alumina supported catalyst was tested at both 5 and 30 bar pressure, with the space velocity adjusted to maintain constant residence time. The results show that the activity of this catalyst decayed with time, but that it could be regenerated by reduction in hydrogen at 673 K. At the higher pressure there is a marked increase in selectivity to methanol, largely at the expense of hydrocarbons and in particular methane. Catalyst Characterisation The alumina-supported catalysts have been studied in greatest detail and shall be considered first. Our preliminary study showed that after reduction in hydrogen at 473 K these are highly dispersed materials, with a nearest-neighbour coordination number of 4.8 * 0.5, corresponding to a mean particle size of ca.10 atoms. The scanning transmission electron micrograph shown in plate 1 confirms the highly dispersed nature of these materials and indicates considerable uniformity in particle size. These small particles are markedly disrupted by exposure to carbon monoxide at room temperature, as observed previously by Prins and co-workers.' Detailed analysis of the EXAFS spectrum, shown in fig. 2 in both reciprocal and real space, indicates the presence of three main shells of neighbours, carbon, chlorine and oxygen and some weak, residual Rh-Rh bonding. The best-fit parameters are given in table 3(6). Because of the obvious contribution of chlorine to this spectrum we have reanalysed that of the reduced catalyst, which is shown in fig.3, with the fitting parameters given in table 3 ( a ) . The result of exposure to synthesis gas (H,/CO ratio = 2/ 1) at 373 K and 1 bar pressure,1 2 r, I > z o Y v * - 2 - 4 0.30 h CI v) ..- c 4 0.20 m x c 0) c v Y v) .- CI .- 0.10 0.00 c t t 2.0 3.0 4.0 5.0 6.0 7.0 R I A Fig. 2. ( a ) The EXAFS spectrum of the Rh/AI,O, catalysts after reduction at 473 K and exposure to carbon monoxide, ( 1 bar), at 298 K: (-) experiment, weighted by multiplication by k 3 ; ( - - - ) calculated using the parameters given in table 3 ( a ) . ( b ) Fourier transform of the spectrum shown in fig. 2 ( a ) : (-) experiment, weighted by multiplication by k 3 ; (- - -1 calculated using the parameters given in table 3( b ) .P.Johnston et al. 97 Table 3. Best-fit parameters obtained from analysis of the EXAFS spectra of the 1% Rh/A1203 catalyst neighbour coordination number interatomic distance/A ( a ) reduced in hydrogen (1 bar) at 473 K oxygen 1.4 f 0.5 1.79 * 0.04 chlorine 1.5*0.5 2.28 * 0.03 rhodium 5.5 * 0.4 2.64 * 0.01 ( b ) exposed to carbon monoxide (1 bar, 298 K), after reduction carbon 1.6 f 0.4 1.8 1 * 0.04 chlorine 1.8*0.3 2.33 * 0.02 rhodium 0.3 f 0.2 2.77 * 0.04 oxygen 4.2'"' 3.08 f 0.03 ( c ) after exposure to CO/H2 (ratio 1/2, at 323 K, 1 bar) carbon chlorine rhodium 2.0 * 0.4 1.9 * 0.3 2.6 * 0.3 1.87 * 0.04 2.25 * 0.03 2.73 * 0.02 0.25 0.20 0.15 0.10 - 0.05 ? Y, 0.00 4 * 0.05 0.10 0.15 0.20 0.25 '' Values not accurate owing to multiple scattering. k l k ' Fig. 3. The EXAFS spectrum of the Rh/A1203 catalysts after reduction at 473 K in hydrogen, ( 1 bar): (-) experiment, weighted by multiplication by k ; ( - - - ) calculated using the parameters given in table 3( a ) and including a rhodium-chlorine shell.98 0.15-- 0.10 In situ Studies of Supported Rhodium Catalysts P, I \ I \ -- 0.30 0.20 Fig.4. The EXAFS spectrum of the Rh/AI2O3 catalysts after reduction at 473 K in hydrogen, (1 bar), and exposure to synthesis gas at 323 K, (CO/H,: 1/2, 1 bar): (-) experiment, weighted by multiplication by k ; (- - - ) calculated using the parameters given in table 3(c). Table 4. XPS studies of Rh/AI2O3 catalysts pretreatment rhodium 3d,,, binding energy/eV Cl/Rh ratio as received 308.8 f 0.1 reduced in hydrogen, 1 bar, 473 K 307.6 f 0.1 exposed to carbon monoxide, 1 bar, 298 K 308.3 f 0.2 exposed to H,/CO 308.5 f 0.2 1.2 f 0.1 1.7 f 0.1 1.0*0.1 1.1 fO.1 after previous exposure to carbon monoxide, is shown in fig.4; fitting parameters can be found in table 3(c). An analogous series of experiments has been performed monitoring the catalysts by X-ray photoelectron spectroscopy (XPS). Of interest is the rhodium 3d5,, binding energy and the presence of chlorine, which had a constant binding energy of 198.8+0.2 eV. The results are listed in table 4. To parallel the catalytic studies, a Rh/AI2O3 catalyst disc has been pre-reduced at 673 K and examined by EXAFS, the structural parameters determined are listed in table 5. This catalyst has then been exposed to synthesis gas for 4 h at 523 K, and the EXAFS spectrum measured after cooling to room temperature in CO/H2 is shown in fig.5. Subsequently the catalyst was re-reduced at 523 K and then exposed to carbon monoxide at the same temperature. The EXAFS spectrum was measured at each stage and the results are also listed in table 5.P. Johnston et al. 99 Table 5. Best-fit parameters obtained from EXAFS analysis of Rh/AI2O3 under catalytically relevant conditions neighbour coordination number interatomic distance/ 8, ( a ) reduced in hydrogen 1 bar, 673 K oxygen 1.4 * 0.6 1.77 f 0.03 chlorine 1.2 * 0.6 2.27 f 0.03 rhodium 4.1 f0.5 2.66 f 0.01 ( b ) Exposed to synthesis gas (1 bar, H2/CO:2/1, 523 K), after reduction oxygen" 2.1 *0.5 1.88 f 0.03 chlorine 1.3 k 0.6 2.19f0.03 rhodium 5.0 * 0.5 2.74 * 0.02 ( c ) as ( b ) , after evacuation and exposure to hydrogen at 523 K o x y g e n ' I 1.8k0.5 1.91 *0.03 chlorine 1.5f0.6 2.25 * 0.03 rhodium 5.0 * 0.5 2.69 * 0.02 ( d ) as ( c ) , but after exposure to carbon monoxide at 523 K ( i ) fit with a carbon nearest neighbour carbon 2.8 * 0.5 1.92 f 0.03 chlorine 1.5k0.6 2.25 * 0.03 rhodium 5.1 k0.5 2.72 k 0.01 oxygen 1.5 * 0.05 1.75 * 0.03 rhodium 5.12k0.5 2.73 f 0.01 (ii) fit with an oxygen nearest neighbour chlorine 1.5 * 0.6 2.20 * 0.03 " This shell may represent both oxygen and carbon neighbours, see discussion in text.X-ray absorption studies have also been performed for the vanadia- and chromia- supported materials and the results are given in tables 6 and 7. A theory-experiment comparison for a vanadia-supported catalyst is given in fig.6. Discussion Structural Aspects of Alumina-supported Catalysts We consider first the information on structure of the catalysts derived from the X-ray absorption measurements and the attempt to relate these to the activity patterns observed. As noted already, small rhodium particles are susceptible to attack by carbon monoxide, with severe disruption of the metallic structure. The resulting entity has previously been analysed by Koningsberger and co-workers,' who concluded that the species formed ds (CO)?Rh03, where the oxygen is from the support and the Rh-0 distance is 2.12 A. We believe that this is not correct. The data are more accurately described by the proximity of chlorine rather than oxygen and the deduced RhzCI distance is very similar to that in the crystalline compound (CO),RhCI, at 2.33 A.The coordination numbers observed in the presence of carbon monoxide also show that the environment adopted by rhodium atoms is very similar to that in the cry$talline gem-dicarbonyl species. Not surprisingly, only a single Rh-C distance (1.81 A), is required to fit the data, while packing ic the crystalline state requires two rather different Rh-C separ- ations, ( 1.77 and 1.85 A ). The presence of chlorine on the catalyst surface is demonstrated by XPS, which indicates an average Cl/ Rh ratio of 1.3. Thus, almost half of the chlorine from the impregnation step is retained on the alumina-supported material. The import- ance of residual halogen in catalysis is widely recognised but has been emphasised by100 In situ Studies of Supported Rhodium Catalysts 0.25 0.30 t Fig.5. The EXAFS spectrum of the Rh/AI2O3 catalysts after reduction at 673 K in hydrogen ( 1 bar), and exposure to synthesis gas at 523 K, (CO/H2: 1/2, 1 bar): (-) experiment, weighted by multiplication by k ; (- - -) calculated using the parameters given in table 3 ( c ) . Bond et aZ.15 The results in table 2 spotlight the role of chlorine ions in stabilising the rhodium-carbon-monoxide species on the catalyst surface. Previous studies have emphasised the role of surface hydroxyl species in oxidising rhodium to the + 1 oxidation While this may be correct, the EXAFS results suggest that the presence of state.3. I k l 7 the chlorine ion may be necessary to stabilise Rh’ entities. Where the surface mono- rhodium dicarbonyl has been prepared in halogen-free conditions, Frederick et al. have shown that it is unstable at 298 K in uacu0.I’ It is surprising that the structural role of surface chlorine has not been recognised previously.In all cases where the disruption of rhodium particles has been established by structural techniques, rhodium chloride has been used as the catalyst precursor. The EXAFS spectra reported by Koningsberger and coworkers are essentially similar to those presented here and show the peak in the Fourier transform which we assign to Rh-CI bonding [see e.g. ref. ( l c ) , fig. 3 , peak B ) . These workers also detected the presence of considerable residual chlorine on the catalyst surface.’ ’ After reduction in hydrogen at 473 K, and before exposure to carbon monoxide, EXAFS indicates that the rhodium is present as small metallic particles.The pre- domiaant bonding is rhodium-rhodium, with an interatomic distance contracted by 0.05 A compared with the clean metal. Such contractions are common in small metal particles, and may be significantly larger than that reported here.’” The Rh-Rh coordi- nation number indicates that the average particle contains ca. 10 atoms, in other words somewhat smaller than the smallest quasi-spherical particle, which contains 13 atoms. Because the analysis of the catalyst in carbon monoxide highlighted the presence of chlorine, the spectrum of the reduced catalyst has been reanalysed. A statisticallyP. Johnston et al. 101 Table 6. Best-fit parameters obtained from analysis of the EXAFS spectra of the 1% Rh/V,03 catalyst.neighbour coordination number interatomic distance/ A ( a ) reduced in hydrogen (1 bar) at 473 K oxygen 1.2f0.6 1.98 * 0.04 rhodium 5.9 f 0.4 2.68 f 0.02 rhodium 2.0 f 0.5 3.69 f 0.03 rhodium 4.8 f 1 .O 4.67 f 0.04 rhodium 6.2" 5.29 f 0.1 ( b ) reduced in hydrogen, (1 bar) at 673 K oxygen 1.4 f 0.6 1.93 f 0.04 rhodium 6.8 f 0.4 2.68 f 0.02 rhodium 2.0 f 0.5 3.72 f 0.03 rhodium 4.4 f 1 .o 4.64 f 0.04 rhodium 7.3" 5.26 f 0.1 ( c ) exposed to synthesis gas, ( H2/CO: 2/1) at 523 K, 1 bar oxygen 0.7 f 0.4 1.99 f 0.04 rhodium 6.3 f 0.4 2.69 f 0.02 rhodium 2.0 f 0.05 3.72 f 0.03 rhodium 4.8 * 1.0 4.67 * 0.04 rhodium 5.7" 5.28 f 0.1 '' Values not accurate due t o multiple scattering. Table 7. Best-fit parameters to EXAFS spectra of 1% Rh/Cr203 catalysts reduced in hydrogen (1 bar) at 673 K ~~ ~ neighbour coordination number interatomic distance/ A rhodium rhodium rhodium rhodium 10.1 f0.5 5.5 * 1.0 1 1 .O * 3.0 12.7" 2.69 f 0.01 3.74f0.05 4.67 f 0.04 5.28" No change in parameters was observed after exposure to synthesis gas at 523 K.'' Subject to errors due to multiple scattering. significant improvement in the fit is obtained if two non-rhodium shells are included, as shown in fig. 2. The Rh-0 and Rh-CI distances calculatedoare both shorter than those observed in the bulk compounds, in each case by ca. 0.1 A. This contraction is thought to be real and not due to systematic errors in the phase shifts. The phase shifts employed give a good description of bonding in bulk rhodium oxide and rhodium chloride, with nearest-neighbour distances accurate to k0.02 A.Reduction at 473 K is therefore ineffective in removing residual chlorine from the proximity of the catalyst particles. The continued presence of chlorine on the catalyst is confirmed by XPS (table 4). Fig. 4 and table 3(c) indicate that exposure of the Rh/A1203 to synthesis gas at a temperature as low as 323 K is sufficient to start the restoration of the metallic structure. Rhodium-rhodium bonding with a coordination number about half of that in the reduced catalyst is observed. The table also indicates a general feature of the rhodium particloes in synthesis gas, irrespective of the support, namely a significant expansion (ca. 0.07 A) of the Rh-Rh distance with respect to the reduced particle.This is not observed in0.30 "'it ( a ) 2 2 p \ r . /-_,-. 0.0 - I 10 2.0 3.0 4.0 5.0 6.0 7.0 R / A Fig. 6. ( a ) The EXAFS spectrum of the Rh/V203 catalysts after reduction at 673 K, ( 1 bar): (-) experiment, weighted by multiplication by k (an experimental glitch has been removed from the spectrum at ca. 10.2 A I ) ; ( - - -) calculated using the parameters given in table 6. ( b ) Fourier transform of the spectrum shown in fig. 6 ( a ) : (-) experiment, weighted by multiplication by k ; ( - - - ) calculated using the parameters given in table 3 ( b ) .P. Johnston et al. 103 the presence of hydrogen alone and is therefore a result of carbon monoxide chemisorp- tion. Although not shown in the figure, significant EXAFS oscillations were observed out to k > 14 k’ (750 eV above the absorption edge), and the full data range was used in determining nearest neighbour Rh-Rh distances.A more restricted data range was used in the analysis where low Z scatterers such as oxygen, carbon and chlorine were of interest. At this point it is appropriate to comment on a problem involving fitting of the EXAFS data where coordination to both oxygen of the support and carbon monoxide is suspected. Because of the low scattering cross-section of carbon and oxygen, there is insufficient information contained in the spectra to allow sensible fitting of both Rh-C and Rh-0 shells with interatomic distances <2 A. Table 5 ( d ) shows alternative best fits with carbon or oxygen as the nearest neighbour.The interatomic distances deduced from the carbon neighbour appear more realistic and are preferred; both sets of parameters give similar values of the fitting index. It is nonetheless unlikely that this shell represents bonding both from adsorbed carbon monoxide and from oxygen of the support. We now consider the results of EXAFS analysis where the Rh/AI2O3 catalyst has been examined under conditions of catalytic relevance. The EXAFS results are sum- marised in table 5. The first thing to note is that reduction at the higher temperature, 673 K compared to 473 K, does not appear to cause sintering. If anything, EXAFS shows that the particles are smaller as a result of the higher temperature reduction. The most significant result is that on exposure to synthesis gas at 523 K there is no disruption of the small particle structure, even though there is evidence for some residual chlorine.We therefore conclude that rhodium metal and not Rh’ is responsible for all of the catalytic activity observed. There appears to be an increase both in the Rh-0 and Rh-Rh coordination numbers. Although this is within the absolute error bars in both cases, it is believed that the relative trends are real. We believe that the increases are the signature of chemisorbed carbon monoxide on the catalyst particles. As commented above, there is insufficient information in the spectra to allow reliable fitting of both carbon and oxygen near-neighbour2hells. It seems, however, to be significant that the ‘Rh-0’ distance increases by 0.1 A in the presence of synthesis gas, and we suggest that this is due to carbon monoxide chemisorbed on the small particles.It does not mean that the Rh-C distance is 1.9 A, because this distance is calculated with Rh-0 and not Rh-C phase shifts. The suggestions that the changes reflect chemisorbed CO are reinforced by the results in table 5 ( d ) , where the largest ‘Rh-0’ coordination number is noted in the presence of CO. The absence of a major Rh-0 contribution requires comment, especially since this shell is strong in the gem-dicarbonyl spectrum. The weakness of this feature probably reflects a combination of static and dynamic disorder in the adsorbed layer. It is known that the wag mode of CO adsorbed on metal surfaces is quite soft and this will significantly weaken the Rh-0 contribution to the spectrum.It is disappointing that carbon monoxide adsorbed on these very small particles cannot be studied with greater precision. The behaviour of this catalyst on exposure to hydrogen and subsequently to carbon monoxide at 523 K has been studied and the results are summarised in tables 5 ( c) and ( d ) . The only changes noted are those thought to reflect increased CO chemisorption, as discussed above. Structural Aspects of Other Supports The EXAFS data obtained on vanadia- and chromia-supported catalysts are more limited and also less interesting than obtained from the alumina-supported catalysts. Two features of the vanadia-supported materials are worthy of note; the absence of any evidence for chlorine, and the presence of non-nearest-neighbour rhodium shells.The average particle size on vanadia is similar to that on alumina, so that chlorine should104 In situ Studies of Supported Rhodium Catalysts be observed if present. The catalysts show evidence only for some bonding to oxygen of the support. The presence of higher Rh-Rh shells can be seen most clearly from the peaks at R> 3.5 8, in the Fourier transform [fig. 6( 6)], and suggests that the particle size distribution is less uniform than on the alumina support. Greegor and Lytle'" have calculated the way in which coordination numbers vary for nearest-neighbour shells with particle size. For particles containing 10-15 atoms we expect only a very weak contribution to the EXAFS from non-nearest shells. This is the result obtained from the alumina-supported materials, where the contribution from these shells is below statistical significance over the range analysed.The ability to detect these shells in the vanadia-supported catalyst data suggests that there is a fraction of larger particles present. This would of course imply that particles smaller than the 12 atom average must also be present, to yield the observed Rh- Rh nearest-neighbour coordination number. Electron microscopy studies are planned to probe this point. The results on the chromia-supported catalyst show that the mean particle size is much larger than on either of the other supports. EXAFS becomes much less useful at large particle size, but the observed coordination numbers suggest that the average particle diameter is 17 f ': A, containing 200-300 atoms.The observed relatively large non-nearest-neighbour coordination numbers support this estimate of particle size. Structure-Performance Relationships The results in table 2 show that the activity pattern of the catalysts is: MOO, > Al2O3 = Vz03 >> Cr203. No data are available on the dispersion of the molybdena-supported material, but the activities of the other catalysts are broadly in line with the EXAFS observations on particle size. The technique is not very sensitive for large particle sizes, such as are found in the chromia-supported catalysts, so we cannot be certain that the low activity of the chromia catalysts is entirely due to poor dispersion. Because of the poor activity of the chromia-supported catalyst, realistic selectivity comparisons can be made only for the other three materials.Of these, alumina and vanadia show rather similar patterns, but molybdena is very different, with much higher hydrocarbon yields, complete suppression of methanol and ethanol and the production of significant quantities of aldehydes. These changes may suggest the presence of stronger acid sites on molybdena than for the other two supports, perhaps due to some reduction of Mo"' local to the rhodium particles. The behaviour of the molybdena- supported material will be the subject of further investigation. The characterisation studies provide no evidence for the presence of Rh' species under catalytically relevant conditions. However, there are three interesting differences between the alumina- and vanadia-supported catalysts: that on alumina yields more methanol but much less ethanol than that on vanadia, but hydrocarbon chain growth appears more significant on vanadia.Before ascribing a role to the surface of the support, two factors relating to the rhodium particles must be considered. The presence of chlorine rather than oxygen, proximate to the metal particles on alumina, may have an influence on selectivity. Nearby chlorine could alter the electronic environment over much of the surface of the small particles present here. We have calculated that its range of operation as a catalyst poison or promoter is ca. 3.5 A'' very similar to the radius of the metal particles involved here. 4lternatively, if ethanol synthesis involves carbon monoxide bound to both metal and support, as envisaged by Sachtler et al.," the proximity of chlorine could change alcohol selectivity. I t should also be remembered that, although the average particle size is quite similar on alumina and vanadia, the distribution of sizes may be much broader on vanadia, as indicated by the strength of higher Rh-Rh shells in the EXAFS spectra.At the very small particle sizes involved, some structure sensitivity in the selectivity would not be unexpected. This is particularly the case where CO dissociation is involved, which is known to be structure sensitive.''P. Johnston et a]. 105 Thus Lin et al. have demonstrated that the extent of hydrocarbon chain growth on small ruthenium particles supported on alumina depends on the particle size, with smaller particles favouring longer chains.24 At the present stage it is not fruitful to speculate further on the origins of the observed selectivity differences, especially since the change in the reaction pressure so markedly affects the selectivity of the alumina-supported catalyst.We are grateful to the S.E.R.C. for their support of this work, to Mr R. Billsborrow of the S.E.R.C. Daresbury Laboratory for experimental advice, to Johnson Matthey PLC for the loan of precious metals and to BP Research, Sunbury-on-Thames, for the use of their in situ cell. Mr R. Devenish (Dept of Materials Sciences and Engineering, University of Liverpool), took the beautiful micrograph which is plate 1 . R.W.J. acknowl- edges a helpful conversation with Professor D.C. Koningsberger. References 1 ( a ) H. F. T. Van’t Blik, J. B. A. D. Van Zon, T. Huizinga, J. C. Vis, D. C. Koningsberger and R. Prins, J. Phys. Chem., 1983, 87, 2264; ( h ) H. F. T. Van’t Blik, J. B. A. D. Van Zon, T. Huizinga, D. C. Koningsberger and R. Prins, J. Mol. Catal., 1984, 25, 379; ( c ) H. F. T. Van’t Blik, J. B. A. D. Van Zon, T. Huizinga, D. C. Koningsberger and R. Prins, J. Am. Chem. Soc., 1985, 107, 3139; ( d ) H. F. T. Van’t Blik, J. B. A. D. Van Zon, T. Huizinga, D. C. Koningsberger and D. E. Sayers, J. Chem. Phys., 1985, 82, 5742. 2 G. van der Lee, B. Schuller, H. Post, T. L. F. Favre and V. Ponec, J. Catal., 1986, 98, 522. 3 M. I. Zaki, G. Kunzmann, B. C . Gates and H. Knozinger, J. Phys. Chem., 1987, 91, 1486. 4 P. Johnston, R. W. Joyner and P. D. A. Pudney, J. Phys. Condensed Matter, 1989, 1, SB171. 5 R. W. Joyner, Vacuum, 1988, 38, 309. 6 A. Takeuchi and J. R. Katzer, J. Phys. Chem., 1981, 85, 937. 7 M. lchikawa and T. Fukushima, J. Chem. Soc., Chem. Commun., 1985, 321. 8 J. T. Yates, E. D. Williams and W. H. Weinberg, Surf: Sci., 1980, 91, 562; D. G. Castner, L. H . Dubois, B. A. Sexton and G. A. Somorjai, Surf Sci., 1981, 103, L134. 9 W. A. Dietz, J. Gas Chromatogr., 1967, pp. 68. 10 R. W. Joyner and P. Meehan, Vacuum, 1983, 33, 691. I I S. J. Gurman, N. Binstead and I . Ross, J. Phjqs. C, Solid State Phys., 1984, 17, 143. 12 R. W. Joyner, K. J. Martin and P. Meehan, J. Phys. C, Solid State Phys., 1987, 20, 4005. 13 L. F. Dahl, C. Martell and D. L. Wampler, J. Am. Chem. Soc., 1961, 83, 1761. 14 J. H. Scofield, J. Electron. Spectrosc., 1976, 8, 129. 15 G. C. Bond, R. R. Rajaram and R. Burch, Appl. Catal., 1986, 27, 379. 16 A. K. Smith, F. Hugues, A. Theolier, J. M. Basset, R. Ugo, G. M. Zanderighi, J. L. Bilhou, V. Bilhou- 17 P. Basu, D. Panayotov and J. T. Yates, J. Phys. Chem., 1987, 91, 91. 18 B. G. Frederick, G. Apai, and T. N. Rhodin, J. Am. Chem. Soc., 1987, 109, 4797. 19 E. S. Shpiro et a/., submitted. 20 R. B. Greegor and F. W. Lytle, J. Catal., 1980, 63, 476. 31 R. W. Joyner and J. B. Pendry, Catal. Lett., 1988, I , 1. 12 W. M. H . Sachtler, Proc. 8th Int. Congr. Catal. (Verlag Chemie, Weinheim, 19841, 1, pp. 151; W. M. H. Sachtler and M. Ichikawa, J. Phys. Chem., 1986, 90, 4752; M. Ichikawa, P. E. Hoffmann and A. Fukuoka, J. Chem. Soc., Chem. Commun., 1989, pp. 1395. Bougnal and W. F. Graydon, Inorg. Chem., 1979, 18, 3104. 23 W. Erley, H. Ibach and H. Wagner, Su~-/:f: Sci. 1979, 83, 585. 24 Z - Z . Lin, T. Okuhara, M. Misono, K. Tohji and Y. Udagawa, J. Chem. Soc. Chem. Commun., 1986, 1673. Paper 0/00324G; Receiced 22nd Januar?,, 1990