Support Effects on Selectivity over Rhodium Bimetallic Cat a1 y st s BY GARY L. HALLER, DANIEL E. RESASCO AND ARMANDO J. Rouco Department of Engineering and Applied Science, Yale University, P.O. Box 2159, Yale Station, New Haven, Connecticut 06520, U.S.A. Received 19th May, 1981 Rhodium catalysts are about one order of magnitude more active for ethane hydrogenolysis when supported on alumina or titania than when supported on silica when reduction is carried out at low temperature. However, the difference in activity for ethane hydrogenolysis between low-(573 K) and high-(773 K) temperature reduced Rh on titania is much greater. When the selectivity between dehydrogenation and hydrogenolysis of cyclohexane is compared at low and high temperature, a modest increase in selectivity is found to accompany the increased reduction temperature.A more substantial effect on selectivity is evident when Rh-Ag on silica and titania, both reduced at low temperature, are compared. The direct effect of rhodium-titania interaction (varied reduction temperature) and the indirect effect of support (changed Rh-Ag interaction) appear to have a com- mon origin. It is proposed that in both the direct and indirect support interactions there may be preferential interaction with the smallest particles in the distribution. The effect of metal-support interaction on the selectivity between dehydrogenation and hydrogenolysis of cyclohexane was first reported by Nehring and Dreyer.' At 773 K, the Pt catalysed selective dehydrogenation to benzene was found to decrease in the order titanium oxide, aluminium oxide, magnesium oxide, silicon oxide, while zinc-oxide-supported Pt was found to be inactive.It was subsequently discovered that Group VIII noble metals supported on titanium oxide and reduced at low temperature (473 K) have distinctly different properties compared with catalysts where the reduction is carried out at high temperature (773 K).2 The high-temperature reduction decreases the hydrogen and carbon monoxide chemisorption to near zero. This effect has been designated by Tauster et al. as a strong metal-support interaction and shown to be a reversible effect when the catalysts were oxidised at 673 K and re-reduced at low temperature (473 K). Meriaudeau et aL3 report that the rates of both benzene dehydrogenation at 273 K and cyclohexane dehydrogenation at 523 K on titania-supported Rh follow the trend in hydrogen chemisorptive capacity, i.e.the activity per unit weight of Rh decreases about one order of magnitude when the reduc- tion temperature of the catalyst is increased from 473 to 773 K. Results on selectivity were not mentioned, probably because the activity for hydrogenolysis is negligible relative to dehydrogenation of cyclohexane over Rh at 523 K. However, the selecti- vity between isomerisation and hydrogenolysis of neopentane over titania-supported Pt has been studied by Foger and Ander~on.~ The percentage neopentane reacting (in the temperature range 473-553 K) to C5 products increased from 21 % to 65% when the reduction temperature was increased from 573 to 723 K.Because there was no evidence from electron microscopy for a change in particle size or morphology with reduction temperature and the activation energy remained constant at ca. 150 kJ mol- l, the increased isomerisation selectivity was ascribed to the decreased concentra- tion of adsorbed hydrogen on the catalyst reduced at 723 K. The frequency factor110 SUPPORT EFFECTS ON SELECTIVITY (rate per surface Pt atom) was lower by a factor of lo2 on the high-temperature reduced catalyst, indicating a much lower activity for both isomerisation and hydro- genolysis. In this work we have investigated two kinds of metal-support interactions. The direct effects of metal-support interaction on the selectivity between dehydrogenation and hydrogenolysis of cyclohexane were studied by comparing silica- and titania- supported Rh reduced at high and low temperature. An indirect effect of metal-support interaction, the influence of the support on the kind of metal-metal interaction between Rh and Ag, was probed by the change in selectivity that accom- panies a change in the support.Previous work has demonstrated two kinds of behaviour of Group VIII-Group IB immiscible bimetallics with regard to hydrogeno- lysis. One kind is exemplified by Ru-Cu (or 0s-Cu) supported on ~ilica.~ For these catalysts the turnover frequency for ethane hydrogenolysis decreases precipita- tely as the Cu/Ru (or Cu/Os) ratio increases. In marked contrast, the turnover frequency of ethane hydrogenolysis on silica-supported Rh-Ag was observed to be nearly constant as the Ag/Rh ratio was increased.6 These catalyst systems exhibit a comparable change in hydrogen chemisorption as the Group IB/Group VIII ratio is increased which suggests that the extent of metal-metal interaction is comparable also.What appears to be different is the manner in which the Group IB metal is distributed on the surface of the Group VIII metal. It has recently been shown that Cu at sub- monolayer coverage on (0001) Ru is uniformly distributed and that the enthalpy change associated with the Ru-Cu interaction is ca. 20 kJ mol-' greater than the Cu- Cu intera~tion.~ However, the Rh-Ag interaction is apparently weaker than the Ag-Ag interaction.s If this is the case, then it may be expected that Ag will form islands or patches on the surface of Rh at submonolayer coverage when the support is silica, a fact that would rationalise the observation that turnover frequency of ethane hydrogenolysis (based on hydrogen chemisorption) is approximately constant on the Ag/Rh catalysts.Our goal was to investigate the Ag-Rh interaction on titanium oxide, a support with which metals are known to interact more strongly than on silica. EXPERIMENTAL MATERIALS USED AND CATALYST PREPARATION The catalysts were prepared by impregnating the support (silica, alumina and titania) with a solution of Rh(N03)3 obtained from Alfa Ventron Corp. The percentage of Rh by weight was 2% in most catalysts. The silica support used was provided by Cabot Corp. (grade HS-5, non-porous, 300 m2 g-I). Two titania supports were used, one provided by Degussa Inc.(grade P-25, SO% anatase/20% rutile, 50 If 15 m2 g-l) and one provided by Cabot Corp. (Cab-0-Ti M-85, 85% anatase/l5% rutile, 50-70 m2 g- I). Additional physical properties of these titanium oxides are given in ref. (2). The alumina support was Cabot Alon C, 90% gamma, 100 m2 g-I. The ratio of impregnating solution to weight of support adjusted to have similar degree of wetness was 3.75 cm3 g-I for SiOl-supported catalysts and 4 cm3 g- for Ti02-supported catalysts. After impregnation the catalysts were dried for two days at room temperature and then 12 h in air at 393 K. A similar prepara- tion was used for the alumina-supported catalyst. The pretreatment was exactly the same for silica-, alumina- and titania-supported catalysts.HYDROGEN CHEMISORPTION MEASUREMENTS Hydrogen adsorption measurements were carried out in a conventional Pyrex volu- An oil diffusion pump provided a dynamic vacuum of lo-' metric adsorption apparatus.G. L , HALLER, D . E. RESASCO AND A. J . ROUCO 111 Torr. Pressures were measured with an absolute pressure gauge (MKS Instruments). After the initial pretreatment the catalysts were reduced in situ at 573 K (LTR) and 773 K (HTR) in a hydrogen flow. Evacuation after reduction was carried out at 573 K for 5 h. Adsorption isotherms at room temperature were measured by admitting a known quantity of gas to the adsorption cell and waiting overnight before reading the equilibrium pressure for the first point. In determining the Rh dispersion, defined as the fraction of Rh atoms present in the surface of the metal crystallites, the isotherms were extrapolated to zero pres- sure.The value of H/Rh obtained in this way is taken as the degree of metal dispersion, i.e. the percentage exposed. ACTIVITY MEASUREMENTS ETHANE HYDROGENOLYSIS The activity measurements for the ethane hydrogenolysis reaction were performed in a microcatalytic pulse reactor. The reactors are Pyrex 6 mm 0.d. glass tubing connected to stainless-steel piping by a Cajon Ultra-torr union with Viton O-ring. The reactor is sus- pended in an electric furnace controlled within 0.5 K by a temperature programme-con- troller (Research Inc. models 7321 1 and 6391 1). Catalysts were activated in situ as described for hydrogen chemisorption. The amount of catalyst used for ethane hydrogenolysis was 50 mg.The 0.5 cm long catalyst bed was preceded by a preheater section of 3.5 g of 60 mesh glass beads that had been washed with sulphuric acid + chromic acid solution and heated to a high temperature in air. The temperature was monitored by an iron-constantan thermo- couple in contact with the catalyst bed. Blank runs indicated that up to 673 K the glass beads, the thermocouple, the supports and supported Ag are not detectably active for ethane hydrogenolysis or cyclohexane dehydrogenation/hydrogenolysis. Purified hydrogen was used as the carrier gas and the hydrocarbon pulse injection was accomplished by a Carle 2015 sampling valve. Each hydrocarbon pulse was 0.05 cm3. The carrier flow was meas- ured with a Hasting mass flowmeter at 30 cm3 min- Analysis was performed by an on-line gas chromatograph (Varian model 3700) by flame ionization detection coupled to Varian CDS 111 electronic integrator following separation on a Chromosorb 104 2m column operated at 348 K.The conversion levels were maintained below 5%. Thus we calculate the differential rate of reaction based on the conversion and a residence time assumed to be the reactor (catalyst bed) volume ratioed to the carrier flowrate and converted to units of molecules converted per surface Rh atom per min based on H/Rh = 1 for hydrogen chemi- sorption. In the case of titanium oxide the percentage exposed after a high-temperature reduction is assumed to be equal to that measured after a low-temperature reduction because of the known depression of hydrogen chemisorption that accompanies the strong metal- support interaction. CYCLOHEXANE REACTIONS The cyclohexane dehydrogenation and hydrogenolysis activity measurements were obtained in the same system used for pulse experiments but operated in steady-state flow mode.The reactant gas was passed over the catalyst for 12 min prior to sampling product for analysis by using a second Carle 2015 sampling valve. The carrier gas was helium. The reactant gas was a mixture of cyclohexane and hydrogen obtained by joining a stream of hydrogen (60 cm3 min-') with another hydrogen stream (2 cm3 min-') passed through a saturator and a condenser thermostatted at 268.7 K. The hydrogen/cyclohexane ratio was 1070. This high hydrogen/cyclohexane ratio was necessary in order to obtain an appreciable amount of hydrogenolysis products which made it possible to measure a change in the selectivity.Initial results performed in the pulse mode showed that the Rh-Ag/SiOz catalysts cracked the whole pulse to methane but Rh-Ag/Ti02 catalysts produced a large amount of benzene. By working at very low cyclohexane partial pressure we could approxi- mate this situation in the flow reactor. The high hydrogen/cyclohexane ratio allowed us to work at higher temperatures with low conversion far from the cyclohexane-benzene equili-112 SUPPORT EFFECTS ON SELECTIVITY brium so that dehydrogenation and hydrogenolysis could be measured simultaneously around 573 K. The differential rates were calculated in the same manner as described for the pulse reactor and the analysis apparatus used was the same as described for ethane hydrogenolysis. RESULTS The rate of ethane hydrogenolysis at 523 K measured in the pulse reactor is Note that the rate is expressed compared for three kinds of oxide supports in table 1.TABLE ET ETHANE HYDROGENOLYSIS AT 523 K ratec /molecule E a catalyst treatment ' CO/Rh (total Rh /kJ mol- atom min)-l 3% Rh/Si02 LTR - 0.40 230 f 10 HTR 0.79 0.33 230 2% Rh/A1203 LTR 0.70 3.52 1 70 HTR 0.72 2.65 180 2% Rh/Ti02 LTR 0.18 0.36 180 HTR 0.007 <0.001 - LTR 0.85 1.9 180 ~~ 11 LTR implies an in situ low-temperature reduction at 473 K following a 723 K reduction and exposure to air at room temperature; HTR is an in situ reduction at 773 K. This is the ratio of chemisorbed CO to total Rh atoms measured by a dynamic pulse method at room temperature.Because hydrogen chemisorption was not measured, the rate is expressed per total Rh atoms; for comparison to results in the figures and other tables the rates can be estimated by dividing by the CO/Rh ratio. This LTR was preceded by an oxygen treatment at 673 K. in molecules converted per total Rh atoms per min because the amount of hydrogen chemisorption was not measured. It should be observed that the low-temperature reduction was the same as used in ref. (2) (473 K), i.e. below the reaction temperature. Tables 2 and 3 compare various silica- and titania-supported catalysts which had TABLE 2.-ETHANE HYDROGENOLYSIS AND CYCLOHEXANE DEHYDROGENATION/HYDROGENOLYSIS AT 573 K FOLLOWING 773 K REDUCTION--673 K OXIDATION-573 K REDUCTION c ycl o hexane ethane rate dehydrogenation 2% Rh on Ag/Rh a H/Rh /molecule rate S " support (surface Rh /molecule atom min) - (surface Rh atom min)- SiO, 0 0.73 6.9 0.60 6 SiOz 1 0.13 0.26 1 .oo 15 TiO, 0 0.33 27.2 3.60 14 Ti02 1 0.22 0.02 6.90 250 ' Atomic ratio of Ag to Rh.Rate of benzene formation. S is the selectivity, the rate of cyclohexane conversion to benzene ratioed to the rate of cyclohexane disappearance to all other hydrogenolysis products.G . L . HALLER, D . E . RESASCO A N D A . J . ROUCO 113 TABLE 3 .-ETHANE HYDROGENOLYSIS AND CYCLOHEXANE DEHYDROGENATION/HYDROGENOLYSIS AT 573 K FOLLOWING 773 K REDUCTION--673 K OXIDATION-773 K REDUCTION cyclohexane ethane rate dehydrogenation 2% Rh on H/Rh /molecule rate a S b support (surface /molecule Rh atom min)- (surface Rh atom min)- SiOz 0.73 6.9 1.15 14 TiOz 0.01 2.7 x lo-” 2.82 60 a Rate of benzene formation.S is the selectivity, the rate of cyclohexane conversion to benzene ratioed to the rate of cyclohexane disappearance to all other hydrogenolysis products. This rate is not based on the H/Rh = 0.01 but on the H/Rh = 0.33 following a low-temperature reduction, see table 2. been reduced just prior to reaction at 573 and 773 K, respectively. Prior to the final reduction the treatment was identical for all catalysts shown in tables 2 and 3, i.e. reduction at 773 K in flowing hydrogen followed by oxidation at 673 K. The final low-temperature reduction used for the results given in table 2 was increased to 573 K (instead of 473 K as in table 1) so that all reactions could be carried out at or below the final reduction temperature.The ethane hydrogenolysis was performed in the pulse reactor while the cyclohexane dehydrogenation/hydrogenolysis was measured in the same reactor in the steady-state flow mode. Fig. 1 presents Arrhenius plots of the rate of benzene formation from cyclohexane at 573 K. All data were collected at times 12 min after cyclohexane was added to the hydrogen stream. This time period was chosen as a compromise, long enough to assure steady-state reaction but short enough so that there was no appreciable decrease in rate due to self-poisoning. The rate decreased by about a factor of 1.5 in the period between 12 and 20 min. This effect is entirely reversible by flowing hydrogen at the reaction temperature, i.e.it is presumably a slow accumulation of less-reactive hydrocarbon intermediates removed as hydrogenolysis products in the presence of pure hydrogen. Fig. 2 is a graphical representation of the dramatic effect of Ag on ethane hydro- genolysis activity. The cyclohexane hydrogenolysis activity is apparently similar but because of the extremely low conversion obtained at 573 K with Ag-Rh/Ti02, e.g. As can be seen in fig. 3, there is a clear correlation between activity and dispersion (percentage exposed), the rate of ethane hydrogenolysis increases by about a factor of 30 as the fraction of Rh atoms at the surface (as measured by hydrogen chemisorption) was increased from 0.3 to 0.8. There is probably an effect of the kind of titania on the dispersion, i.e.the four catalysts of lowest dispersion are on Degussa oxide while the three of highest dispersion are supported on Cab-o-Ti. It may be significant that the activity for the titania supports appears asymptotically to approach that of silica as the dispersion is decreased. it cannot be measured very accurately. DISCUSSION Before turning to the principal question, selectivity of bimetallic Ag-Rh catalysts influenced by the support, it is useful to consider support and preparation effects on pure Rh for the ethane hydrogenolysis reaction. In a previous investigation of114 SUPPORT EFFECTS ON SELECTIVITY silica-supported Rh where the texture of the silica, anion of the impregnating solution and degree of dehydration before reduction were varied, we observed a range of rates at 523 K varying from 0.36 to 4.9 molecules per surface Rh atom per min with no correlation between activity and disper~ion.~ These results were obtained in the same pulse reactor used in the present work and, considering the different reactor used, are in fair agreement with the range of rates reported by Yates and Sinfelt'' when their results are extrapolated to the same temperature and converted to the same units, 4.0 2.0 ' \ 0.\ 0 a '. 1.75 1.80 1.85 103 KIT FIG. 1.-Arrhenius plot of the rate of benzene formation from cyclohexane. The numbers given in parentheses are approximate activation energies in kJ mol-'. Silica support is represented by circles and titania support by squares. Filled symbols indicate pure Rh, 2% by weight; open symbols indicate catalysts with Ag/Rh atomic ratio of one. 0.28-9.7.The rate of ethane hydrogenolysis given in table 1 for silica-supported Rh falls within the range of previous values. However, the rate given in table 2 extra- polated to 523 K with the measured activation energy of 200 kJ mo1-1 is only 0.13 molecules per surface Rh per min. The difference is attributed to the oxidation treatment at 673 K used for all results reported in tables 2 and 3. While it is not possible to rationalise these results at the present time, the observed facts are that oxidation at relatively low temperature, 673 K, and high temperature, 1073 K,1° decreases the rate of ethane hydrogenolysis while an intermediate oxidation tempera- ture, 873-923 K,9 has a relatively small effect on activity.We have previously argued that the effect of oxidation cannot be a particle-size effect alone and should probably be interpreted as a support interaction induced by the ~xidation.~*l~ Two points must be borne in mind in the light of the above discussion: (i) a comparison of supportG . L . HALLER, D. E. RESASCO A N D A . J . ROUCO 0 c, 0 1 . 1 I I 0.5 1 .o 115 FIG. 2.-Rate of ethane hydrogenolysis at 573 K as a function of the AglRh atomic ratio on titania support. I I116 SUPPORT EFFECTS ON SELECTIVITY effects requires an identical preparation and pretreatment, and (ii) an order-of- magnitude difference in activity between two different supports is comparable to observed preparation/pretreatment effects and may not be significant, The rates of ethane hydrogenolysis on silica, alumina and titania are compared in table 1.Alumina and titania, when reduced at low temperature, have considerably greater activity than silica-supported Rh and a significantly lower activation energy. The support effect that arises between titania-supported Rh reduced at low and high temperature is more striking than is the comparison with the other oxides. After the high-temperature reduction, Tauster et al.2 suggested Rh on titania was bonded to Ti cations via electron donationfrom the metal to the cation by analogy to certain barium titanates which contain Group VIII metals. However, the titanates always contain the Group VIII metals as isolated cations instead of small clusters of metal atoms on the surface.A recent theoretical interpretation by Horsely l2 indicates a metal-metal bonding with donationfrom the Ti cation to metal atoms (Pt in Horsely’s calculation) located on surface oxygen-ion vacancies. This picture is consistent with the experi- mental observation of charge transfer from Ti3+ in the surface of (100) SrTiO, to Pt atoms or from Ti3+ in the surface of (110) TiOz to Ni atoms measured by photo- emission.13*14 It should be emphasised that the Ti3+ on the surface of the single crys- tals studied by Chung and co-workers was produced by Ar-ion bombardment and not by reduction in hydrogen at 773 K, the normal procedure for inducing a strong metal- support interaction on dispersed titania-supported catalysts. The hydrogen reduction produces a surface concentration of Ti3+ which is too small to be detected by photo- emission.Moreover, to the extent that there is interaction in the dispersed catalysts, it is an interaction with metal clusters and not individual atoms. This suggests that the strong metal-support interaction of catalytic interest may be more subtle interaction than that described theoretically by Horsely or physically by the experiments of Chung et at. The recent n.m.r. results of Gajardo et al.15 for a 3.8% Rh on a 205 m2 8-l titanium oxide, comparable with the preparation used in this work, suggest an interaction similar to that described by Chung et al. for the single crystals. A sample with a fraction exposed equal to 0.6 (measured by hydrogen chemisorption) had a high-field resonance attributed to hydrogen on or near paramagnetic centres on the surface of titanium oxide when the reduction temperature was 573 K.These paramagnetic centres disappeared upon reduction at 673 K (and are not detected in the absence of Rh). This suggests that the strong metal-support interaction induced by high-temperature reduction may involve electron transfer from a surface Ti3+ to a Rh cluster. We have attempted to probe for the negative charge on Rh using N2 chemisorption.16 Rh has been reported to be unable to chemisorb nitrogen” and only slowly decomposes ammonia at high temperature.18 However, Rh supported on activated carbon or alumina and promoted by K has been shown to chemisorb nitrogen19 and to have a stable activity for ammonia synthesis comparable with Fe.20 The effect of alkali metals on transition metal activity for ammonia synthesis is generally ascribed to its electron- donating nature since chemisorption and activity increase with decreasing ionisation potential of the alkali, i.e.Cs > K > Na.21 Thus, it is reasoned that a high-tempera- ture reduction of Rh on titania should promote N2 chemisorption to some extent if negative charge is transferred to Rh. This has been verified experimentally and shown to be reversible following oxidation at 673 K.16 Most previous investigations have implied that titanium oxide is only quantitatively different from other oxide supports after a high-temperature reduction. Fig. 3, which plots the activity for ethane hydrogenolysis as a function of dispersion after low-temperature reduction, would suggest a support interaction that increases as theG.L . HALLER, D . E. RESASCO A N D A . J . ROUCO 117 particle size decreases, an effect which must be different from the strong metal-support interaction of Tauster et aL2 No comparable correlation between activity and particle size was evident for silica in our own work9 but Yates and Sinfeltlo have claimed a maximum activity for a particle size around 2 nm for ethane hydrogenolysis and Fuentes and Figueras 22 see a similar maximum for cyclopentane hydrogenolysis around 5 nm. However, the maximum selectivity between simple ring opening to pentane and hydrogenolysis to methane occurred at 2 nm. The comparison in table 2 of the effect of Ag on Rh supported on silica and titania also indicates that these two supports are quantitatively different even when reduced at low temperature.There is more than an order of magnitude depression of the ethane hydrogenolysis rate when Ag is added to Rh on a silica support, but the effect is at least three orders of magnitude on titania. Moreover, the effect on silica may have a different origin. In a more detailed investigation of Ag-Rh silica-supported catalysts where no oxida- tion treatment was used and the reduction temperature was 723 K, the specific rate did not vary by more than about a factor of 2 as the Ag/Rh atomic ratio was varied from zero to values >1.6 It may be that Rh-Ag catalysts which have been oxidised at 673 K are not completely reduced at 573 K although the fact that the rate of cyclo- hexane dehydrogenation was greater on the Rh-Ag catalysts than pure Rh catalysts independent of support suggests that incomplete reduction is not a factor for either support.The precipitous decrease in ethane hydrogenolysis activity on titania-supported Rh as a function of increasing Ag/Rh ratio is quite similar to that found on Ru-CuS where the Group IB metal is known to deposit uniformly over the ~urface.~ Because of this similarity we believe that the titania support has altered the Ag-Rh energetics so that Ag deposits uniformly over the Rh surface in contrast to the silica support where the Ag is supposed to form islands or patches below a monolayer.6 If ethane hydrogenolysis requires several Rh atoms to constitute a site as is generally believed5 then the uniform deposition of the Group IB allows a disproportionate number of sites to be poisoned but when islands are formed, that fraction of the surface not covered behaves like normal Rh.Our interpretation requires a linear decrease in hydrogen chemisorption with added Ag up to a monolayer. In fact, this kind of behaviour has now been documented for deuterium on (100) Rh8 and, like the ethane hydrogenolysis, it is in marked contrast to hydrogen coverage dependence on Cu coverage on (0001) R u . ~ The nature of the titania-rhodium interaction that affects the way Ag deposits on small Rh particles is not obvious. It may not involve electron transfer from the support to the Rh since we are considering an effect observed at low-temperature reduction. Whatever the nature of the interaction, the results of fig.3 indicate that it increases as the particle size decreases. Moreover, since the activity for ethane hydrogenolysis apparently increases as the interaction grows on low-temperature reduced titania which was previously oxidised at 673 K and the activity for ethane hydrogenolysis always decreases greatly after a high-temperature reduction, it is conceivable that the low-temperature reduction (or more probably the oxidation that precedes it) involves charge transfer from Rh to titania. This would appear to be consistent with the n.m.r. results that a low-temperature reduction in the presence of small particles of Rh creates paramagnetic centres (probably Ti3+) not formed in the absence of Rh and that these centres disappear when the reduction temperature is increased.15 A simple comparison of the effect of added Ag on Rh activity for ethane hydro- genolysis and cyclohexane dehydrogenation would lead one to surmise that cyclo- hexane dehydrogenation/hydrogenolysis selectivity would be dramatically increased by the addition of Ag on either support.The selectivity does increase but rather118 SUPPORT EFFECTS ON SELECTIVITY modestly even on titania support. If one assumes that the sites for dehydrogenation require only a single Rh atom and ethane hydrogenolysis sites require several, then one would have to conclude that ring opening (and subsequent hydrogenolysis) sites are much more like the dehydrogenation sites than the hydrogenolysis sites. In the terminology of Ander~on,~ the critical reactant site for cyclohexane hydrogenolysis is smaller than for ethane hydrogenolysis.Moreover, these sites may also be distingui- shed by their activation energies. For pure Rh on titania, ethane hydrogenolysis has an activation energy of ca. 180 kJ mo1-I while cyclohexane hydrogenolysis has a lower activation energy, ca. 140 kJ mol-I. A comparison of low- and high-temperature reduced pure Rh on titania is even more striking. Ethane hydrogenolysis decreases by three orders of magnitude, dehydrogenation by ca. 20% but the selectivity only increases by a factor of 4 when low- and high-temperature reductions are compared. As reported by Anderson for Pt,4 the activation energies are apparently not affected as one moves from low- to high- temperature reduction, i.e. it is primarily the number of sites that is changed by the change in reduction. However, all sites cannot be uniformly attenuated or one would not observe any effect on selectivity.If we assume that the strong metal-support interaction (high-temperature reduction) beomes more effective as the particle size decreases, a fact that is demonstrated for the low-temperature reduction in fig. 3, then it may be that the high-temperature reduction selectively deactivates the smallest particles in the distribution. This would lead to the conclusion that the larger parti- cles are more selective for benzene formation than for hydrogenolysis of cyclohexane. Our results would at first glance appear to be in conflict with those of ref. (3) where it is observed that rate of cyclohexane dehydrogenation at 523 K decreases by a factor of 7 when the reduction temperature is increased from 473 to 773 K.How- ever, no oxidation treatment was used in this study. In an experiment where we allow a room-temperature oxidation, the cyclohexane dehydrogenation at 573 K is a factor of 4 lower when the reduction temperature is increased from 573 to 773 K and, if both reductions are preceded by a 673 K oxidation, the increase in the reduction temperature only decreases the dehydrogenation rate by 20% (compare tables 2 and 3). Thus preoxidation of the titania-supported Rh catalysts has a levelling influence with respect to reduction temperature effect for a less demanding reaction like dehydro- genation, but for a demanding, site-structure-sensitive reaction like ethane hydro- genolysis, the effects of reduction temperature remain pronounced. In conclusion, we find that the order of increasing site-structure sensitivity or critical size of sites is cyclohexane dehydrogenation, cyclohexane hydrogenolysis and ethane hydrogenolysis on Rh catalysts and this is the order of increasing influence of support interaction with titanium oxide for the direct Rh-Ti02 interaction.Greater leverage on selectivity can be gained by an indirect support effect for the case of Rh-Ag bimetallic catalysts. In this case the nature of the support, silica or titania, changes the kind of metal-metal interaction but the magnitude of the effect follows the same ordering of reactions as for the direct effect. This suggests that the effect of Ag on Rh sites and the effect of strong metal-support interaction on Rh sites have a common mode of attack.The simplest hypothesis is that the smallest particles are being preferentially deactivated in both the direct and indirect metal-support inter- actions on titanium oxide. We thank the National Science Foundation for support of this research. A. J. R. thanks the Consejo Nacional de Investigaciones Cientificas y Tecnicas de Argentina ; D. E. R. thanks the Bolsa de Comercio de Mar del Plata, Argentina, for scholarships.G . L . HALLER, D . E. RESASCO AND A . J . ROUCO 119 D. Nehring and H. Dreyer, Chem. Tech. (Berlin), 1960,12,343. S . J. Tauster, S. C. Fung and R. L. Garten, J. Am. Chem. SOC., 1978,100,170. P. Meriaudeau, H. Ellestad and C. Naccache, Proc. 7th Int. Congr. Catal., Tokyo, 1980, ed. T. Seiyama and K. Tanabe (Kodansha, Tokyo and Elsevier, Amsterdam, 1981), part B, p. 1464. J. R. Anderson, Am. Chem. SOC., Div. Pet. Chem. Prepr., 1981, 26, 361. J. H. Sinfelt, J. Catal., 1973, 29, 308. ti A. J. Rouco and G. L. Haller, J. Catal., to be published. ’ K. Christmann, G. Ertl and H. Shimizu, J. Catal,, 1980,61, 397. J. M. White, personal communication. A. J. Rouco and G. L. Haller, in Proc. 7th Ibero-American Symposium on Catalysis, La Plata, Argentina, July, 1980. lo D. J. C. Yates and J. H. Sinfelt, J. Catal., 1967, 8, 348. l1 A. J. Rouco and G. L. Haller, J. Chim. Phys., to be published. l2 J. A. Horsely, J. Am. Chem. SOC., 1979, 101, 2870. l3 M. K. Bahl, C. S. Tsai and Y. W. Chung, Phys. Reo. B, 1980,21, 1344. l4 C. C. Kao, S. C. Tsai, M. K. Bahl, Y. W. Chung and W. J. Lo, Surf. Sci., 1980,95, 1. l5 Y. Gajardo, T. M. Apple and C. Dybowski, Chem. Phys. Lett., 1980,74, 306. D. Resasco and G. L. Haller, J. Chem. SOC., Chem. Commun., 1980, 1150. B. M. W. Trapnell, Proc. R. SOC. London, Ser. A , 1953,218, 566. G. C. Bond, Catalysis by Metals (Academic Press, New York, 1962), p. 380. l9 M. Oh-Kita, K-I. Aika, K. Urake and A. Ozaki, J. Catal., 1976, 44, 460. *O K-I. Aika, S. Yamaguchi and A. Ozaki, Chem. Lett., 1973, 161. I. R. Shannon, in Catalysis, senior reporters C. Kemball and D. 0. Dowden (Specialist Periodi- cal Report, The Chemical Society, London, 1978), vol. 2, p. 33. 22 S. Fuentes and F. Figueras, J. Catal., 1980, 61, 443.