The Role of the Support in CO Hydrogenation Selectivity of Supported Rhodium BY JAMES R. KATZER, ARTHUR W. SLEIGHT,? PATRICIO GAJARDO, JOHN B. MICHEL, EDWARD F. GLEASON AND SCOTT MCMILLAN Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Delaware 1971 1, U.S.A. Received 29th June, 1981 Catalytic hydrogenation of CO offers tremendous challenges in catalyst design to control selec- tivity because of the large variety of hydrocarbons and oxygen-containing species that can be formed; Rh can produce the full range of potential products. Rh on SO2, A1203, Ti02, CeOz and MgO has been characterized by a complement of techniques and by CO hydrogenation activity. Catalytic activity varied over 200-fold, dependent on the support; Rh/TiOz was the most active.Selectivity to alcohols us. hydrocarbons and to methanol us. ethanol varied over 50-fold. CO hydrogenation selectivity to alcohols varies with the basicity of the support; MgO, the most basic support, ex- hibited 90% selectivity to methanol. Mononuclear transition-metal complexes, when used as catalysts, can show widely varied activity and selectivity dependent on the ligands associated with the supported metal.' The control of activity and selectivity with the appropriate ligands is due to both steric and electronic factors. The electronic factors can be significant because a mononuclear transition-metal complex involves only a small number of orbitals and e!ectrons, and large gaps exist between energy levels. In this case, small shifts in electron density can have an important effect upon the catalytic properties of the transition-metal atom.On the other hand with bulk metals, which for practical purposes also include small supported metal crystallites (a 25 8, crystallite contains ca. 500 atoms), there are a large number of electrons delocalized over the system, and energy level gaps are very small.' Imposing significant shifts in the electron distribu- tion by ligand effects in such a system appears difficult, and recent alloy studies clearly confirm that most effects on catalytic behaviour are geometric and not elec- tronic in Thus, truly new catalytic chemistry seems unlikely with classical supported metals. However, for ultra-dispersed supported metals with no more than a few metal atoms in a cluster, all of which are in close contact with the support, the support surface or other atoms located thereupon can play the role of ligands, and since we are now dealing with a very small number of metal atoms, the support can take on an important role in determining catalytic activity and selectivity of the metal.This support ligand could play a role similar to that of the ligands in homogeneous catalysts. This allows for the potential of activity-selectivity control similar to that for soluble homogeneous catalysts but without all the problems of corrosion and catalyst separation and recovery associated with soluble catalysts. t Present address: Experimental Station, E. I. DuPont deNemours & Co, Wilmington, Delaware 19898, U.S.A.122 CO HYDROGENATION OVER SUPPORTED Rh Rh has shown some very unique catalytic behaviour in homogeneous catalytic ~ h e m i s t r y , ~ ' ~ ~ as a soluble cluster compound in CO hydrogenation," and as a sup- ported metal.l2-I6 I~hikawa'~-'~ has shown that the selectivity of supported Rh in CO hydrogenation depends markedly on the chemical nature of the support.These observations led us to study CO hydrogenation catalysed by Rh on a series of sup- ports and to characterize these catalysts fully by a complement of spectroscopic techniques to provide a basis for interpretation of their activity and selectivity behaviour. EXPERIMENTAL MATERIALS AND CATALYST PREPARATION Alumina was Conoco Catapal calcined at 550 "C to convert it into y-A1203 (194 m2 g-I); Si02 was Davison Grade 57 Si02 gel (210 mz g-').MgO was Baker Analytical grade (50 m2 g-I). All other supports were synthesized by hydrolysis of the metal compounds or salts to obtain materials of sufficient surface area and purity, all supports had surface areas in excess of 100 m2 g-l. Details are given e1sewhere.l' In order to achieve the highest possible extent of dispersion, the Rh was ion-exchanged onto all the supports; for MgO the pore volume saturation technique was applied. Thus, initially atomic dispersion of Rh was achieved. Sulphate-free Rh(N03)3 (10 wt % solution), diluted to give the desired uptake, was used. Catalysts were reduced in situ, in flowing dry H2 at 473 K for 4 h, and the desired measurements were carried out. A summary of the catalysts studied is given in table 1.TABLE 1 .-PROPERTIES OF SUPPORTED Rh CATALYSTS ~~ ~ B.E.T. surface (wt %) H,b cot support area" /m2 g-' Rh Rh Rh Ti02 (anatase) 105 0.95 0.98 2.17 A1203 194 2.4 1.15 NAd Ce02 99 1.6 3.21 2.68 Si02 210 0.89 0.99 NAd MgO 50 3 .O 0.68" NAd ~ ~~~ ~ ~~ " Measured by N2 adsorption using standard B.E.T. methods; * Determined by extrapolation of Determined by pulse the H2 and CO adsorption isotherms at 298 K to the zero pressure intercept; adsorption at 298 K; NA = not available. APPARATUS AND PROCEDURE Table 2 summarizes the characterization techniques applied to the supported Rh systems. Since the methods and results of application of each technique applied cannot be discussed here, only selected results will be presented. TABLE 2.--CHARACTERIZATION TECHNIQUES APPLIED TO SUPPORTED Rh CATALYSTS CO hydrogenation, activity and selectivity ethylene hydrogenation ethane hydrogenolysis H2 chemisorption (static) CO chemisorption (static) f.t.i.r.of chemisorbed CO proton n.m.r. t.p.d.-Hz e.p.r. electron microscopy X-ray photoelectron spectroscopy EXAFS isotopic tracer X-ray diffraction u.v.-visible laser Raman spectroscopyKATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 123 INFRARED MEASUREMENTS 1.r. measurements were carried out in a quartz cell with water-cooled NaCl windows using a Nicolet 7199 FTIR spectrometer. The cells were attached to a conventional high- vacuum system equipped with a manifold for gas flow which permitted us to perform all pre- treatments and measurements in situ. The cells could be heated to 873 K with flowing gases for pretreatments. Calcined samples were pressed into wafers (16 mg cm-2) and placed in a quartz sample holder.CHEMISORPTION Hydrogen was measured in a standard volumetric apparatus; CO chemisorption was The total chemisorption capacity was determined by extrapola- carried out gravimetrically. ting the total chemisorption us. pressure data to zero pressure. X-RAY PHOTOELECTRON SPECTROSCOPY Calcined catalyst was pressed into a wafer and attached to the end of a probe. Pretreat- ments were done in a prechamber attached to the Physical Electronics 550 X-ray photoelec- tron spectrometer. Samples were translated into the main vacuum chamber and X.P.S. spectra were taken, then retracted into the pretreatment chamber and reduced in flowing H1 at the desired temperature (473 K), reexamined by X.P.S., withdrawn for chemisorption of CO, and again examined by X.P.S.co HYDROGENATION CO hydrogenation studies were carried out in a steady-state differential plug-flow reactor interfaced with two gas chromatographs. A manifold equipped with traps for removal of oxygen, water and carbonyls allowed metering of high-purity gases to the reactor at 1-10 atm. Concentrations and flow rates were set by the feed rates to the manifold; pressure was controlled by a backpressure regulator. Everything downstream from the reactor was heated to prevent condensation, allowing direct injection of the whole product stream into each g.c. Hydrocarbons were separated on an activated A1203 column; alcohols were separated on a THEED on chromosorb column.RESULTS AND DISCUSSION Table 1 summarizes the H2 and CO chemisorption results for the catalysts studied. In all cases total hydrogen chemisorption capacity indicates high degrees of dispersion of the Rh. The Rh/MgO is not as highly dispersed because of the alternate prepara- tion method. For Rh/CeO, the total uptake of H2 was surprisingly high, higher than could be attributed solely to adsorption on Rh. The H2 adsorption on Rh/Ce02 involved a rapid process followed by a slow continued uptake which must have in- volved hydrogen uptake by the support.17 CO adsorption provides additional insight into the dispersion of the Rh. Ratios on the supported Rh catalysts approached or surpassed 2.0 CO molecules per Rh atom. Rh/Ce02 in particular gave a ratio markedly larger than 2.0.In this case a second CO adsorption isotherm was measured after evacuation at 298 K for half an hour, and from this the amount of irreversibly adsorbed CO was determined to be 2.11 CO per Rh. These high values are apparently associated with adsorption on the support as appeared to be the case for H2 adsorption. In comparison, CO adsorption on single-crystal surfaces or on large supported metal crystallites will give no more than ca. 0.7 CO molecules per surface Rh atom at saturation.18-20 This is the steric limit for CO on a two-dimensional metallic surface; on small metal crystal- lites the CO to surface Rh atom ratio can approach 1.0.21,22 Only for very small clusters of Rh atoms having essentially all edge atoms could the CO to Rh ratios be124 CO HYDROGENATION OVER SUPPORTED Rh 2.0.These results support the H2 chemisorption studies and suggest a very high degree of Rh dispersion. Infrared studies of adsorbed CO allow further characterization of supported Rh. The three structures shown below represent the adsorbed species identified on the surface of supported Rh catalysts : dicarbonyl 1 i near bridged In the limit of very high dispersion involving either atomic dispersion or very small clusters containing mainly edge atoms the dicarbonyl species should predominate. As the Rh dispersion decreases the proportion of linear and bridged carbonyl species increases. These are the only species present on larger supported metal crystallites.18 In this work spectra of CO adsorbed on all supported Rh catalysts showed very strong dicarbonyl peaks.Fig. 1 shows the carbonyl spectrum for 2.4% Rh/A1203 34.40 2 3.40 I 2.40 I .40 2201 2111 202 I I931 1841 wavenumberlcm - FIG. 1 .-Infrared spectrum of CO adsorbed on Rh/A1203. Conditions : catalyst prepared by ion exchange of 2.4 wt % Rh on to Al2O3, calcined in O2 at 623 K, reduced in situ at 473 K in flowing dry H2, CO adsorption at 298 K and 120 Torr, evacuated just prior to taking the spectrum. reduced at 473 K. There are two principal bands in the CO region of the spectrum; these are assigned to the symmetric and asymmetric stretches of the dicarbonyl species. There is a very weak band assigned to the bridging CO species. Absorp- tion due to the linear species is completely absent from the spectrum.These results confirm that the ion-exchange preparation technique has given a uniquely high degree of Rh dispersion, i.e. atomically dispersed or very small clusters of Rh atoms which show no two-dimensional metal surface-like properties for CO adsorption. For example, the i.r. frequencies of adsorbed CO were all independent of CO cover- age from Torr up, in contrast to expected behaviour for two-dimensional metal-like surfaces, where large shifts in CO frequency are observed.18 These con-KATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 125 clusions are supported by the H2 chemisorption results. Clearly the ion-exchange preparation technique resulted in ultra-dispersed Rh/A1203. All previous prepara- tions of Rh/A1203 employed impregnation, and the resultant catalysts showed a band for linear CO that was almost as intense as that for the dicarbonyl species, indicating the presence of small metal crystallites or large clusters having two-dimensional metal surface-like proper tie^.^^'^^ The dicarbonyl assignment was initially made based on the dirhodium dichloro- tetracarbonyl compound shown in fig.2.26*27 In this case the Rh atoms have a + 1 OC cO dicarbonyl A , 2090 cm-l; B 2030 cm-' 2073 2026 1800 2068 2042 1883 FIG. 2.-1.r. frequencies of rhodium carbonyl cluster compounds. charge, and the dicarbonyl stretching frequencies are 2090 and 2030 cm-'. These values agree with those for Rh/A1,03 to within 9 cm-'. Also given in fig. 2 are the frequencies observed for Rh,(CO),, and Rh4(C0)', cluster compounds. The di- carbonyl frequencies of Rh/A1203 do not match those of the larger clusters.How- ever, the bridged CO peak of Rh/A1203 is composed of two components, one at ca. 1869 cm-l and the other at ca. 1818 cm-'. These values are similar (within 20 cm- ') to frequencies observed in Rh4(C0)12 and Rh6(C0)',, respectively, where edge and face bridging carbonyls occur. These results clearly show that all the Rh cannot be present as atomically dispersed or dimeric species but is present in clusters having only small numbers of Rh atoms. Fig. 3 shows the carbonyl spectra of Rh/TiO, after 473 K reduction, after 673 K reduction and then following recalcination (of the 673 K reduced sample) and re- reduction at 473 K. Clearly the 473 K reduction results in ultra-dispersed Rh/Ti02 as only the CO stretches of the dicarbonyl species are evident in the infrared spectrum.There is no bridged species evident in this spectrum. Reduction at 673 K results in significant cluster growth as evidenced by the marked reduction in the dicarbonyl peak intensities, the appearance of an intense linear carbonyl peak at ca. 2071 cm-' and a weak bridging carbonyl peak at ca. 1880 cm-l. Similar behaviour was ob- served upon 673 K reduction for the other supported Rh systems studied. The total CO chemisorption on Rh/Ti02 was reduced from 2.17 to 1.71 CO/Rh (Tred = 673 K), consistent with the marked reduction in the intensity of dicarbonyl peaks in the spectra (fig. 3). The total H2 chemisorption capacity was reduced from 0.98 H/Rh to 0.24 H/Rh (Tred = 673 K).The more substantial reduction in the H2 chemisorption appears to be related to the metal-support interaction phenomena reported for the third-row Group VIII metals on reducible oxide^.,^-^* In contrast, the H, chemi-126 CO HYDROGENATION OVER SUPPORTED Rh sorption capacity of Rh/A1,0, remains constant, independent of the reduction temperature, although the carbonyl spectra showed evidence of clustering similar to that observed for Rh/Ti0,.31 Following the 673 K reduction, calcination and rereduction of Rh/Ti02 at 473 K resulted in significant redispersion as indicated by the increased intensity of the dicarbonyl peaks and reduced intensity of the linear carbonyl peak. 0 0 ..4 *g c E Y 2155 2110 2065 2020 1975 1930 wavenumberlcm - FIG. 3.-Infrared spectra of CO adsorbed on Rh/Ti02.Conditions: catalyst prepared by ion exchange of 0.95 wt % Rh on to Ti02 (anatase), calcined in O2 at 623 K, reduced in situ at 473 K in flowing dry H2, CO adsorption at 298 K and 120 Torr, evacuated just prior to taking the spectrum; sample then reduced in situ at 673 K in flowing dry H2, CO spectrum taken as before; sample then calcined in situ in flowing O2 at 623 K, followed by rereduction in situ in flowing dry H2 at 473 K, spectrum taken as before. Temperature of H2 reduction as follows: (a) 473 K, (b) 673 K, (c) 473 K (after calcination of 673 K reduced sample in O2 at 623 K). The infrared and chemisorption data indicate the Rh is present in the form of small clusters, even when reduced at 473 IS; reduction at 673 K increases the size of these clusters.Similar information has been interpreted by others to indicate the presence of atomically dispersed Rh, in particular concluding that the dicarbonyl species formed on atomically dispersed Rh.23-25932 The question of dispersion can be clarified by determining the coordination of Rh atoms to adjacent atoms by EXAFS. Determination of the coordination number can give an indication of the number of Rh atoms in the cluster. EXAFS results are an ensemble average, and without addi- tional information it is difficult to distinguish between a uniform population having a single coordination number and a population having two different properties, i.e. coordination numbers. EXAFS measurements of Rh/A1203 and Rh/Ti02 reduced at 473 K show that these samples, which exhibit CO/Rh E 2.0 and which show onlyKATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 127 dicarbonyl i.r.bands, are not atomically dispersed but exhibit a Rh-Rh coordina- tion number of 4 & 1.33 These results further support our speculation that the Rh is not atomically dispersed but is present as small clusters of Rh atoms having less than ca. 10 atoms per cluster. The EXAFS results particularly do not support earlier speculations about atomic dispersion of even a significant fraction of the Rh.23-25*32 Reduction at 673 K resulted in an increase in the coordination number determined by EXAFS, supporting the suggestion of agglomeration of the small clusters of metal atoms to larger clusters. These clusters are still much smaller than units which could warrant the terminology " small metal crystallites ".TABLE 3.-CO HYDROGENATION OVER SUPPORTED Rh (wt %> product formation rate" support Rh CH4 CHjOH CzH50H Ti02 0.95 39.6 7.0 3.4 A1203 2.4 1.6 1.1 0.04 Ce02 1.6 0.40 1 .o 0.060 Si02 0.89 0.16 0.30 0.001 8 MgO 3 .O 0.40 4.69 0.0 Product formation rate (turnover number) x lo4, (g mol product)/(g atom surface Rh s) turn- over number based on Rh dispersion in table 1 , all dispersions assumed to be 1 .O except for Rh/MgO. Reaction conditions: 473 K, 10 atm total pressure, 4% CO in HZ, differential plug-flow reactor operation, conversion < ca. 7%. The oxidation state of Rh in the Rh(CO), species observed for the supported Rh catalysts has also been c o n t r ~ v e r s i a l . ~ ~ ~ ~ ~ The close similarity of the dicarbonyl stretching frequencies of supported Rh to those of the (Rh)2(C0)4(C1)2 model com- pound (where Rh has a +1 charge) suggests a monovalent Rh'(C0)2 species occurs for supported Rh catalysts.The question of the oxidation state of Rh can be answered by X-ray photoelectron spectroscopy (X.P.S.). Furthermore, X.P.S. can clarify the question of electron transfer between Rh and the support. The Rh 3d5/2 electron binding energy of reduced Rh foil is 306.8 & 0.1 eV. Rh/Ti02 reduced at 473 K has a binding energy of 307.0 & 0.1 eV; Rh/A1203 reduced at 473 K has a binding energy of 307.4 5 0.1 eV.35 These results show that the Rh is reduced to the metallic state on A1203 and on TiO,; it is not present as Rh1.36 They strongly suggest, however, that Rh undergoes a small amount of electron transfer to the A1203, relative to Rh/Ti02.This is consistent with X-ray absorption edge studies for Pt/A1203 relative to Pt/TiO,, which show that significant electron transfer occurs from Pt to Al,03, but that there is a small amount of electron transfer from the TiO, to the Pt.37*38 Addition of CO resulted in a +0.4 eV shift in the Rh 3dS/, electron binding energy for Rh/Ti02. This shift is far short of the shift required for a Rh valence change to +1.36 The shift observed upon adsorption of CO is simply due to the electron-withdrawing properties of the CO ligand. Therefore, we find no evidence for the assertion that dicarbonyl is associated with atomically dispersed Rh in the + 1 oxidation state; the Rh is present as clusters containing very small numbers of Rh atoms in a reduced metallic-like state.Table 3 summarizes the CO hydrogenation rate behaviour of the supported Rh catalysts. The activation energy for methane formation was 134 & 4 kJ mo1-I for Rh supported on SiO,, A1203 and Ti02. The >200-fold rate variation between Rh/Ti02 and Rh/Si02 appears to be due to a variation in the number of active sites128 CO HYDROGENATION OVER SUPPORTED Rh or other geometric factors which would appear in the pre-exponential term, and not due to changes in the energetics of the reaction which should appear as changes in the observed activation energy. CO and H2 chemisorption capacity changes as well as changes in the infrared spectra of chemisorbed CO (fig. 3) as a function of reduction temperature (473 K us.673 K) led us to expect similar dependence of CO hydrogenation activity and selec- tivity on reduction temperature. However, the rate of CO hydrogenation (gmol product per g of catalyst per s) and the product selectivities for Rh/Ti02 reduced at 673 K were essentially the same as for Rh/Ti02 reduced at 473 K. The effects referred to as strong metal-support interactions (SMSI) for Group VIII third-row transition metals supported on reducible oxides such as Ti028-30 are not observed here. The nature of the support rather than the reduction temperature is the most important factor ; the effects of reduction temperature are secondary. TABLE 4.-CO HYDROGENATION SELECTIVITIES OF SUPPORTED Rh product selectivity" total (wt %) c2+ CF+ alcohols C2HSOH support Rh CH4 CH4 CH4 CH30H TiOz 0.95 0.023 0.1 1 0.26 0.49 A1203 2.4 0.034 0.0 0.71 0.046 Ce02 1.6 0.039 0.0016 2.72 0.058 Si02 0.89 0.050 0.0 1.95 0.062 MgO 3.0 0.0 0.0 12.0 0.0 rate of formation of species i (mol/g cat s) rate of formation of species j (mol/g cat s).* Selectivity r , j E Reaction conditions: 473 K, 1 atm total pressure, 4% CO in H2, differential plug flow reactor operation, overall conversion < ca. 7%. Essentially all Rh atoms are exposed, and there are no major differences in the dispersion of the Rh in terms of the number of Rh atoms in each cluster, as indicated by all the characterization techniques applied to these catalysts. Therefore, the rate differences must be due to metal-support interactions which change the number of active sites capable of catalysing the formation of CH4.The rates of methanol and ethanol formation also vary by a large factor; the ethanol formation rate is highest for Rh/Ti02, and the methanol formation rate is highest for Rh/MgO (table 3). The activation energy for methanol formation was typically 88 & 4 kJ mol-I; that for ethanol formation was significantly higher. The activation energies for methanol formation over Rh/MgO, Rh/Ti02 and Rh/ A1203 are indistinguishable, indicating that the reaction mechanism remains the same and that there are no electronic effects that change reaction energetics induced by the support. Therefore, the effect must be due to a change in the number of sites active in methanol formation; again, the metal-support interaction is directly affecting the number of active sites.Selectivity to alcohols us. total hydrocarbons varies by ca. 50-fold; selectivity to ethanol us. methanol varied greatly (table 4). The selectivity to alcohols is related to the acid-base properties of the support, Table 5 correlates the selectivity to al- cohols synthesis with the acidic properties reported in the literature for the supports.KATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 129 The selectivity to alcohols formation increases with the basic character of the support; MgO exhibits ca. 90% selectivity to methanol. The ordering is good except for the Ti02-A1203 pair for which the acidity-selectivity relation appears to be reversed. This could well be due to acidity differences between the oxides used here and those upon which the acidity measurements were made.Oxides used here were calcined TABLE 5 . 4 0 HYDROGENATION SELECTIVITY OF SUPPORTED Rh TO ALCOHOLS (wt %) total alcohols support Rh acidity, Hoa ref. CH4 Ti02 0.95 2 +1.5 39 0.26 A1203 2.4 2 - 3 40,41 0.71 CeOz 1.6 5 +3.3 43 2.72 SiOz 2.89 2 +4.8 44 1.95 MgO 3 .O 2 +14 45 12 H,, = Hammett acidity function. at 623 K following ion exchange, air equilibrated, reduced in situ at 473 K, and ex- posed to water, a reaction product, during reaction. The oxides for which acidity measurements are reported were typically calcined at high temperature, and water was rigorously excluded during acidity measurement. Therefore, tabulated acidities should be used as only a rough guide. Furthermore, it is not clear that titration methods are able to measure the appropriate acid property.V I VII Vlll IB /- ambient synthesis temperatures temperatures FIG. 4.-Regions of dissociative and non-dissociative adsorption of CO at room temperature and at CO hydrogenation reaction temperatures. The alcohol selectivity can be explained on the basis of the CO dissociation activity of transition metals. In fig. 4 the metals to the left-hand side of the heavy line designated " ambient temperature " spontaneously dissociate adsorbed CO at room temperature; those metals to the right of the line adsorb CO non-dissociatively at room temperature. The heavy line marked " synthesis temperature " separates those metals that dissociate adsorbed CO from those that do not dissociate adsorbed CO at temperatures of 473-573 K.Metals that dissociate CO catalyse hydrocarbon synthesis from CO and HZ; metals that adsorb CO non-dissociatively catalyse synthesis of methanol. Rabo et ~ 1 . ~ ~ showed that Pd, Ir and Pt synthesized methanol with 100% selectivity at higher pressures. Rh, which lies between Ru (which produces exclusively hydrocarbons) and Pd (which can produce exclusively methanol), catalyses130 co HYDROGENATION OVER SUPPORTED Rh formation of both alcohols and hydrocarbons from CO + HZ. Since it is capable of synthesizing both types of compounds small effects can markedly alter its selectivity. Thus Rh in the form of bulk metal produces mainly hydrocarbons;47 in the form of very small clusters on acidic supports Rh produces mainly hydrocarbons.On more basic supports the selectivity to alcohols is enhanced. This can be explained in terms of the more basic oxides having a higher Fermi level and thus having greater electron- donating capabilities; the supported Rh is thus forced to behave more like Pd (fig. Consideration of the mechanism of the synthesis reactions leads to better under- standing of the selectivity behaviour (fig. 5 and 6). Methanol formation on Rh 4)- 0 II C II Co + - * ADSORPTION 0 II H K C / Q H C I II II + 2 * - 3s HYDROGEN AT I ON w OH H-C’ H II + 2 l - = It w FIG. 5.-Mechanism of CO hydrogenation occurs by a non-dissociative mechanism such + CH30H to methanol on supported Rh. as shown in fig. 5. This has been demonstrated by 13C160 and 12C1s0 tracer CH30H is not an effective reactant for C2H50H formation under synthesis conditions over Rh/Ti0z.49 Further, the formation of high alcohols followed a Fischer-Tropsch type molecular-weight distribution as should result from the mechanism of fig. 6.Thus the synthesis of methanol is related to the amount of undissociated CO on the surface. This amount may be determined by the fraction of the catalytic sitesathat are unable to dissociate CO (sites behaving like Pd) or if all catalytic sites are equivalent, by the relatively slow CO dissociation reaction on Rh (i.e. hydrogenation of CO competes with CO dissociation). Although we are unable to distinguish between these two possibilities, we speculate that the methanol reaction involves different sites from the hydrocarbon and higher alcohols reactions.We speculate that the synthesis of hydrocarbons relates to the dissociation of CO;50 the synthesis of higher alcohols, mainly ethanol, relates to the relative concentration of surface carbene and undissociated CO. This explains the selectivity behaviour observed. The support is inferred to act through direct ligand bonding to the Rh. This bonding involves electronic interactions with the metal changing the relative numbers of the active sites involved in the various CO hydrogenation reactions.KATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 131 0 II II CO + - C ADSORPTION m 0 c o C II I I ll 2-H-% II C - % - 3W + H20 DISSOCIATION 11 m m C H CHx II + I - I I CARBIDE HYDROGENATION m m m CH3 CH2 I + I t - a € % CHAIN GROWTH, INITIAL STEP )Ic yc R I y 2 R CH2 I II - m m P R 0 PAGAT I 0 N * M R H 1 I CH2 + WC - a€ + R-CH3 TERMINATION: HYDROCARBON FORMATION I m 0 II R R C I I II - m C = O COINSERTION m m I m R I H OH C = O + 3 I - R - C ' TERMINATION: ALCOHOL FORMATION I )Ic I'H m H FIG.6.-Postulated mechanism of CO hydrogenation to hydrocarbons and higher alcohols.132 CO HYDROGENATION OVER SUPPORTED Rh SUMMARY AND CONCLUSIONS Ion-exchange preparation, initially giving atomically dispersed Rh, results upon H2 reduction in unique highly dispersed supported Rh catalysts. Chemisorption capacities, infrared spectra of chemisorbed CO, and EXAFS confirm the ultra-high dispersion achieved. EXAFS results show that after reduction the Rh is not atomic- ally dispersed but forms very small clusters containing only a few Rh atoms per cluster ; similar conclusions were inferred from the infrared results.The supported Rh is reduced to the metallic state, but there is evidence for electron transfer between the support and the Rh. CO hydrogenation activity varied over 200-fold dependent upon the support; selectivity to alcohols 0s. hydrocarbons varied over 50-fold. Activity and selectivity behaviour are inferred to be due to absolute and relative numbers of active sites as determined by the support. Methanol formation involves a non-dissociative mechanism on sites that can hydrogenate adsorbed CO. Hydro- carbon formation appears to involve CO dissociation. We infer that the formation of higher alcohols (ethanol) is related to the relative concentration of surface alkyl and CO species. This work was supported by a National Science Foundation grant from the Industry-University Cooperative Research Program and by the Industrial Sponsors of the Center for Catalytic Science and Technology.B. C. Gates, J. R. Katzer and G. C. A. Schuit, Chemistry of Catalytic Processes (McGraw-Hill, New York, 1979), chap. 2 and 3. V. Ponec, Cutal. Rev. Sci. Eng., 1975, 11, 41. J. R. H. Van Schaik, R. P. Dessing and V. Ponec, J. Cutul., 1975,38, 273. V. Ponec, in Chemistry and Chemical Engineering of Catalytic Processes, ed. R. Prins and G. C. A. Schuit (Sijthoff and Noordhoff, Alphen aan den Rijn, The Netherlands, 1980), pp. M. Araki and V. Ponec, J. Cutul., 1976, 44, 439. ' J. J. Stephen, V. Ponec and W. M. H. Sachtler, Surf. Sci., 1975,47,403.257-269. ' D. Evans, J. A. Osborn and G. Wilkinson, J. Chem. SOC. A, 1968,3133. * C. K. Brown and G. Wilkinson, J. Chem. SOC. A, 1970,2753. lo J. F. Roth, J. H. Craddock, A. Hershman and F. E. Paulik, Chem. Tech. (Leipzig), 1971, 600. l1 R. L. Pruett and W. E. Walker, U.S. Patent 3, 833,634. l2 P. C. Ellgen, W. J. Bartley, M. M. Bhasin and T. P. Wilson, Ado. Chem. Ser., 1979, 178, 147. l3 M. M. Bhasin, W. J. Bartley, P. C. Ellgen and T. P. Wilson, J. CutuZ., 1978, 54, 120. M. Ichikawa, Bull. Chem. SOC. Jpn, 1978, 51, 2273. l5 M. Ichikawa, Bull. Chem. SOC. Jpn, 1978, 51, 2268. l6 M. Ichikawa and K. Skikakura, Proc. 7th Int. Congr. Cutul., Tokyo, 1980, paper B-17, to be F. E. Paulik, A. Hershman, W. R. Knox and J. F. Roth, U.S. Patent 3 169 329. published. P. Gajardo, A.W. Sleight and J. R. Katzer, paper in preparation. L. H. DuBois and G. A. Somorjai, in A.C.S. Symposium Series (Am. Chem. SOC., Washington, D.C., 1980), no. 137, paper 9, pp. 163-190. l9 T. A. Dorling and R. L. Moss, J. Cutal., 1967, 7, 378. 2o D. Brennan and F. H. Hayes, Philos. Truns. R. SOC. London, 1965, 258, 347. 22 L. Spenadel and M. Boudart, J. Phys. Chem., 1960,64,204. 23 A. C. Yang and C. W. Garland, J. Phys. Chem., 1957,61, 1504. *' J. T. Yates, Jr., T. M. Duncan, S. D. Worley and R. W. Vaughan, J. Chem. Phys., 1979, 70, 25 J. T. Yates, Jr., T. M. Duncan and R. W. Vaughan, J. Chem. Phys., 1979,71,3908. 26 J. G. Bullitt and F. A. Cotton, Inorg. Chim. Actu, 1971, 5, 637. 27 E. R. Corey, L. F. Dahl and W. Beck, J. Am. Chem. SOC., 1963, 85, 1202; P.Chini and S. 28 S. J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. SOC., 1978,100, 170. J. Freel, J. Catul., 1972, 25, 149. 1219. Martinengo, Inorg. Chim. Actu, 1969, 3, 315.KATZER, SLEIGHT, GAJARDO, MICHEL, GLEASON, MCMILLAN 133 29 S. J. Tauster and S. C. Fung, J. Catal., 1978, 55, 29. 30 S. J. Tauster, S. C. Fung, R. T. K. Baker and J. A. Horsley, Science, 1981, 211, 1121. 31 E. F. Gleason, J. R. Katzer and A. W. Sleight, paper in preparation. 32 H. C. Yao and W. G. Rothschild, J. Chem. Phys., 1978, 68, 4774. 33 D. R. Short, D. E. Sayers and J. R. Katzer, paper in preparation. 34 M. Primet, J. Chem. SOC., Faraday Trans. I , 1978, 74, 2570. 35 J. B. Michel, J. R. Katzer and A. W. Sleight, paper in preparation. 36 C. D. Wagner, W. M. Riggs, L. E. Davis, J. F. Moulder and G. E. Mullenberg, Handbook of 37 D. R. Short, R. J. Emrich, A. Mansour, D. E. Sayers and J. R. Katzer, J. Catal., submitted for 38 D. R. Short, A. Mansour, D. E. Sayers and J. R. Katzer, J. Catal., submitted for publication. 39 K. Tanabe, C. Ishiya, I. Matsuzaki, I. Ichikawa and H. Hattori, Ann. Meeting Chem. SOC. Japan, 23rd, Tokyo, 1970, paper no. 03408. 40 T. Yamaguchi, H. Shimizu and K. Tanabe, unpublished data. See K. Tanabe, Solid Acids and Bases (Academic Press, New York, 1970), chap. 4, p. 45. 41 M. Ito, K. Matsuura, Y. Yoshii, K. Igarashi and A. Suzuki, Chem. SOC. of Japan, 1969, no. 06420. 42 K. Tanabe and M. Katayama, J. Res. Insf. Catal. Hokkaido Univ., 1959,7, 106. 43 K. Nishimura, Nippon Kagaku Zassani, 1960, 81, 1680. 44 K. Tanabe, Solid Acids and Bases (Academic Press, New York, 1970), chap. 4, pp. 45-101. 45 J. Take, N. Kikuchi and Y. Yoneda, Shokubai (23rd Symposium Catalysis), 1968, 10, 127. 46 M. L. Poutsma, L. F. Elek, P. A. Ibarbia, A. P. Risch and J. A. Rabo, J. Catal., 1978,52, 157. 47 B. A. Sexton and G. A. Somorjai, J. Catal., 1977, 46, 167. 48 A. Takeuchi and J. R. Katzer, J. Phys. Chem., 1981, 85,937. 49 A. Takeuchi and J. R. Katzer, paper in preparation. 50 P. Biloen and W. M. H. Sachtler, Adu. Catal., 1981, 30, in press. X-Ray Photoelectron Spectroscopy (Perkin-Elmer Corp., 1979), p, 109. publication.