Hydrogenolysis of Cyclopent ane and Hydrogenation of Benzene on Palladium Catalysts of Widely Varying Dispersion BY SERGIO FUENTES AND FRAN YOIS FIGUERAS* Institut de Recherches sur la Catalyse du C.N.R.S., 79, boulevard du 11 Novembre 1918, 69626 Villeurbanne Cedex, France Received 26th January, 1977 Adsorption of hydrogen and oxygen, titration of preadsorbed oxygen, benzene hydrogenation and cyclopentane hydrogenolysis were performed on palladium catalysts of widely varying dispersion. A change in the stoichiometry of oxygen adsorption is found when increasing the metallic dispersion. On clean supports, free of sulphur and iron, the turn over number for both reactions of hydrocarbons is constant and independant of, the dispersion of Pd. After a suitable reduction, sulphur may preferentially inhibit hydrogenolysis, while iron preferentially inhibits hydrogenation.On an industrial silica support the turn over for hydrogenolysis changes with dispersion ; this effect is attributed to contamination of Pd by iron from the support. Much recent work on catalysis by metals has been concerned with the question of how the degree of dispersion, or crystallite size, of a metal influences the specific catalytic activity. An effect of the dispersion has been found on hydrogenolysis of ethane on nickel and Rh/Si02,2 on hydrogenolysis of neopentane on platinum and of cyclohexane on Ru/S~O,.~ Maurel et aL5 recently demonstrated that trace impurities from the support could noticeably alter the selectivity of platinum, and the notion of selective poisoning was thus introduced.Only a few metals have been studied from this point of view, and the data are rather scarce, in particular for palladium. The present work is devoted to the study of the influence of metallic dispersion on the catalytic activity of palladium. We compared the rates for benzene hydrogenation, which is known to be structure insensitive, and for ring opening of cyclopentane which could be structure sensitive on platinum and which can be considered as a parent reaction of the hydrogenolysis of cyclohexane used recently by Lam and S i ~ ~ f e l t . ~ EXPERIMENTAL The catalysts were prepared by ion exchange using Pd(NH3)2C12 in basic solution with silica, and PdC12 in acidic medium with alumina. The solids were left overnight in contact with the solution, then filtered and dried at 110°C.The supports were : Davison silica gel (grade 70), of surface area 350 m2 g-I, containing traces of iron (0.02 % Fe) and sulphur (0.17 %) in the form of sulphate SO, ; Degussa alumina (llOC), a non porous alumina of surface area 180 m2 8-l containing only small traces of iron (< 200 p.p.m.) ; and y-alumina made in the laboratory with care taken to avoid introducing impurities, containing only weak traces of iron (< 100 p.p.m.) undetectable sulphur. The study of the possible influence of sulphur and iron impurities was undertaken on samples of y-alumina impregnated by known amounts of contaminants in the form of (NH4)2S04 and FeCl,. The chemical compositions of these solids are listed in table 1. A good dispersion (90-100 %) was obtained after calcination at 400°C in oxygen and reduction in dry hydrogen at 300°C.Changing the temperatures of calcination and reduction resulted in gradual sintering of the metallic phase, probably due to variation in the water content of the sample as underlined by Boudart.' 1 74S. FUENTES AND F. FIGUERAS 175 The dispersion of palladium was measured by H2-02 titration, using volumetry at 0.1 Torr and a temperature of 70°C for H2 adsorption, to avoid H2 dissolution, as in a similar procedure used by other authors.8* The dispersion was defined in the usual way by Pd,/Pd, Pd, being the iiumber of superficial palladium atoms which adsorb hydrogen and oxygen and Pd the total number of palladium atoms of the sample. Catalytic activities were determined on an aliquot of the sample used for dispersion measurements.The sample was reactivated for 1 h at 300°C under hydrogen. The rates were measured in a conventional flow reactor at low conversion (<2 %) to avoid heat and mass transfer limitations. For benzene the conditions were : temperature 14O"C, partial pressure of hydrocarbon 56 Torr, partial pressure of H2 704 Torr ; under these conditions the reaction order relative to benzene is zero. TABLE 1 .-CHEMICAL COMPOSITION OF THE SAMPLES CONTAMINATED WITH SULPHUR AND IRON catalyst % Pd wt % S wt % Fe - 200 1 0.1 310 0.35 0.1 - 320 0.35 0.2 - 330 0.25 0.4 - 340 1 I 0.035 350 1 I 0.056 360 1 - 0.078 For cyclopentane the reaction temperature was 290"C, pressure of hydrocarbon 100 Torr and pressure of hydrogen 660 Torr.The reaction order relative to cyclopentane was found to be close to zero. The sole product of hydrogenolysis was n-pentane on Pd/alumina. On Pd/Si02 n-pentane was the major product (selectivity 75-80 % initially) but small amounts of methane,butane and isopentane were detected during a short initial period ; they decreased rapidly with time. The catalysts suffer deactivation in the reaction of conversion of cyclopentane. A better correlation between activity and metallic dispersion may be expected when using the initial activity. Therefore an effort was undertaken to determine the law of deactivation. The hyperbolic law first proposed by Germain and Maurel lo fits the data well as illustrated in fig. 1. The equation is a particular form of a more general rate law proposed by Szepe and Levenspiel.ll Levenspiel l2 recently discussed the means of testing several models of deactivation by a proper choice of the reactor and pointed out the usefulness of the well-mixed reactor for that purpose.A flow reactor operated at low conversion can be assimilated to a well mixed reactor. If we assume that the reaction proceeds by a parallel scheme : n-pentane I' 4 cyclopent ane coke and that resistance to diffusion is small, the theory developed in ref. (12) predicts for a second order of deactivation, the following law : where C A ~ is the inlet concentration of reactant, CA is the outlet concentration of reactant, k is the zero order rate constant of the reaction, ki is the deactivation constant, with k'd = k d ( C ~ ) ~ and z' = WCA,/FA, with W the mass of catalyst and FA^ the flow rate of reactant.176 HYDROGENOLYSIS OF CYCLOPENTANE timejmin 0, Pd+ S/Al203-334 ; A, Pd/Si02-47 ; + , Pd/A1203-21 1.FIG, 1 .-CataIyst deactivation in cycIopentane hydrogenolysis. Plots of the reciprocal conversion as a function of time for 3 samples. Therefore the plot of the reciprocal conversion CAJCA~- CA against time gives a straight line which allows the determination of the initial activity and of the rate constant of deactivation. A detailed discussion of the data gathered on deactivation will be given elsewhere, but we can note here that, a good representation of the kinetics can be obtained by this type of analysis after making some reasonable assumptions. RESULTS AND DISCUSSION DETERMINATION OF THE PALLADIUM AREA Benson et al.have extended to palladium the technique of titration of preadsorbed oxygen previously applied to ~1atinum.l~ On a sample characterized by a moderate dispersion of palladium ( D N 0.2), the stoichiometry observed for the ratios : hydrogen adsorption/oxygen adsorption/titration of preadsorbed oxygen was 1-1 -3. The reactions which represent the processes of adsorption are : Pd++H2 -+ PdH Pd++02 + PdO PdO + +H2 --+ PdH + H20. The results of the present work, summarized in table 2, are in agreement with this stoichiometry when the dispersion is low ( D < 0.4). However at higher disper- sion ( D fi 0.9), the stoichiometry shifts to 1-0.5-2 as illustrated in table 2 by the samples 176 and 211. In the intermediate range (sample 1702) the stoichiometry is not a simple one.This situation is quite similar to that described by Wilson and or Dalla Betta and Boudart for well dispersed platinum catalysts. The interpretation now accepted in the case of platinum is a change in the stoichiometry of adsorption of oxygen which give PdO on large particles and Pd20 on small particles. The reaction of titration of oxygen on small particles is then : Pd2O + 2H2 --+ 2PdH + HZO.S . FUENTES AND F . FIGUERAS 177 The shift from PtO to PtzO was attributed in ref. (7) to the fact that small metallic particles are electron deficient and cannot therefore supply enough electrons to complete a monolayer of oxygen. We can suppose, since the behaviour of palladium is the same as that of platinum as a function of the dispersion, that the assumptions made previously for platinum are valid for palladium.The change in stoichiornetry of oxygen adsorption with particle size is thus a general phenomenon and not restricted to platinum. TABLE 2.-sTOICHIOMETRIC RATIO FOR THE ADSORPTION OF HYDROGEN, ADSORPTION OF OXYGEN AND HYDROGEN TITRATION OF PREADSORBED OXYGEN ON DIFFERENT SAMPLES OF SUPPORTED PALLADIUM H/Pd O/Pd H/Pd Pd/A1203 176 1 .o 0.48 1.85 Pd/A1203 21 1-1 1.03 0.59 2.13 Pd/A1203 1702 0.73 0.51 1.62 Pd/Al2Oj 185 0.31 0.28 0.82 catalyst adsorption adsorption titration Pd/A1203 211-2 0.96 0.49 1.99 Pd/Si02 1 0.35 0.34 1 .oo Pd/Si02 66 0.47 0.39 1.22 CATALYTIC ACTIVITIES OF CLEAN SAMPLES The catalytic activities of some samples, supported on y-alumina or silica are reported on table 3 in terms of turn over number N (number of molecules reacting per hour and per surface palladium atom).On a clean support like y-alumina, the turn over numbers N for both reactions are constant and independent of the dispersion. The values for benzene hydrogenation are in good agreement with previous determinations on palladium catalysts. 9 l6 This behaviour is character- istic of structure insensitive reactions, and of the absence of any support effect. In contrast, Davison silica supported palladium gives a low but constant turn over for benzene hydrogenation and a striking increase in the turn over for hydrogenolysis with dispersion. The results obtained with Pd/Si02, if they were taken separately, could lead to the conclusion that the hydrogenolysis of cyclopentane is structure TABLE 3.-TURN OVER NUMBER FOR THE HYDROGENATION OF BENZENE AND HYDROGENOLYSIS OF CYCLOPENTANE AS A FUNCTION OF THE DEGREE OF DISPERSION, ON TWO SUPPORTS N2 cyclopentane at 290°C N1 benzene at 140°C samples wt %Pd Tnd/OC % D / molecules h- 1 / molecules h-1 R = N I I N ~ y-aluminas 2202 0.25 21 1 0.5 1702 1 .o 176 1 .o 184 1 .o 185 1 .o 122 1.2 252 8.0 83 0.7 41 1.3 4 1.3 2 1.3 72 1.3 400 300 400 300 400 500 300 500 650 500 400 300 400 100 305 100 286 55 243 80 270 40 264 28 220 12 282 12 230 Davison silica gel 4 115 26 104 33 101 48 94 32 140 33 24 30 24 28 21 30 22 8 15 19 34 50 9.2 12 9.7 11.2 9.4 10.5 9.4 10.5 14.4 6.9 5.3 2.5 2.5178 HYDROGENOLYSIS OF CYCLOPENTANE sensitive.Indeed, Lam and Sinfelt interpreted their results obtained on the hydrogenolysis of cyclohexane on Ru/SiO, in this way.However, such an effect should appear on alumina as well as on silica and this is not the case. Therefore we cannot conclude that there is a particle size effect on the catalytic properties of palladium, but rather may suspect possible influence of an impurity. Sulphur and iron being common impurities of silica, we investigated the influence of these contaminants. INFLUENCE OF CONTAMINANTS The relative ratio poison/palladium was varied by changing the chemical composi- tion (table 1) and the temperature of reduction. The contaminant content was kept low in order to simulate a catalyst supported by a commercial carrier. The palladium area of contaminated samples is difficult to determine by H2-02 titration owing to the presence of impurities, the behaviour of which is unknown in the process of adsorption.In this case a useful procedure is that proposed by Maurel which consists in the comparison of the rates of hydrogenation and of hydrogenolysis on the same sample. The hydrogenation of benzene is known to be insensitive to structure and the activity for this reaction can be a measure of the palladium area. If the hydrogenolysis of cyclopentane proceeds on the same sites as hydrogenation, the selectivity ratio, rate of hydrogenation/rate of hydrogenolysis must be constant ; a variation of this selectivity ratio must reflect that hydrogenolysis and hydrogenation proceed on different sites, therefore that hydrogenolysis is structure sensitive.INFLUENCE OF SULPHATES The results are summarized in table 4. The blank runs on clean samples give the activities in the absence of sulphates. The temperature of reduction has a strong influence on activity and selectivity, TABLE 4.-cATALYTIC ACTIVITIES ( X lo5 mol S-l 8-l Pd) OF Pd/Al203 SAMPLES, CONTA- MINATED WITH SULPHATE AND REDUCED AT 300 OR 400°C catalyst 176 2201 1702 201 204 311 321 331 202 205 312 332 loading activity A 1 for temperature of cyclopentane %Pd % S reductionl'c at 290°C 1 0 300 (2 h) 5.3 0.25 0 300 (2 h) 9.3 1 0 400 (2 h) 4.2 1 0.1 300 (1 h) 2.9 1 0.1 300 (1 h) 2.9 0.35 0.1 300 (1 h) 6.7 0.3 0.2 300 (1 h) 8.0 0.25 0.4 300 (1 h) 4.32 1 0.1 400 (12h) 0.85 1 0.1 400 (2 h) 1.08 0.35 0.1 400 (12h) 0.52 0.25 0.4 400 (2 h) 0.08 activity A2 for benzene at 140°C S = A21Ai 62 11.7 100 10.7 50 11.9 average 11.4 45 16 35 12 90 13 87 11 60 14 average 13.3 20 24 22 20 13 25 2.5 31 average 25 -S .FUENTES AND F. FIGUERAS 179 The samples contaminated with sulphate and reduced at 300°C have activities close to the blank, the maximum decrease is by a factor of 2. The selectivity of these catalysts is slightly higher than that of clean samples, but the difference is not large. In contrast, after reduction at 400°C, the activities for hydrogenation and hydrogenolysis decrease sharply, but the selectivity for hydrogenation increases : this increase means that sulphur preferentially represses hydrogenolysis. These results obtained on palladium are similar to those published for platinum : 5 on both metals a selective poisoning of hydrogenolysis can be observed in the presence of sulphate after reduction at 400°C, while reduction at 300°C yields a non-selective poisoning.It is gratifying that palladium has the same behaviour as platinum in this respect, and the same interpretations may be valid in both cases. Maurel et aL5 concluded that hydrogenolysis proceeds on metal sites with specific geometrical requirements as suggested by Boudart,l while hydrogenation would proceed on all superficial metallic atoms. Elementary sulphur produced by the reduction of sulphates at 400°C would be preferentially adsorbed on these sites of low coordination and could specifically poison hydrogenolysis. The reduction at 300°C would yield H2S or SO2 which are non-selective poisons of platinum.The influence of sulphate poisoning on palladium is thus in good agreement with literature data concerning platinum. However it seems clear that the results obtained in the present work with Pd/Si02 cannot be attributed to sulphur poisoning since the selectivity ratio is shifted in the opposite way. INFLUENCE OF IRON The experimental data are listed in table 5. In the case of iron the temperature of reduction of the sample also influences the results. Moreover, an effect of the iron content can be noted. Iron contamination has only a small effect on the catalytic properties when the sample is reduced at 300°C. The selectivity ratio is slightly lowered but the difference is not large. This is consistent with the results obtained on Fe+Pt/A1,03 samples reduced at 300"C,18 which show that iron is a non selective and low toxic poison of platinum under such conditions.When the sample is reduced at 400"C, the activities for hydrogenation and hydrogenolysis decrease, and the selectivity ratio tends to decrease also. The decrease in this ratio is more pronounced at low iron loading than at high contents. With 0.035 % Fe the effect is clear since an average value of 6 is obtained, compared with 10 for Pd/A12Q3 and >20 for Pd+S/A120,. A reduction at 500°C yields a further decrease in the activities of both reactions, but a relative increase in the selectivity, since the activity ratio of pure Pd/A1203 catalysts is reobtained. From the literature we know that the degree of reduction of iron increases with temperature and with platinum loading on Pt + Fe/A1203 cata1ysts.l 9 9 2o After reduction at 500"C, Mossbauer spectroscopy detects PtFe clusters in a sample characterized by a low Fe/Pt atomic ratio of 0.2; when the iron content increases PtFe clusters are formed but some iron remains as ferrous ions.In a separate study of the reduction as a function of temperature it was observed that a ferrous ion spectrum was obtained at low reduction temperatures, namely 300°C. Similar results were reported by Garten 21 on PdFe/A1203 catalysts. The catalytic activity of well characterized PtFe clusters has been also measured. Bartholomew and Boudart 22 reported that PtFe clusters supported on carbon have a lower activity but the same selectivity for isomerisation of neo-pentane as180 HYDROGENOLYSIS OF CYCLOPENTANE Pt/C.Similarly, Vannice and Garten 2o observed that PtFe/A1203 have the same selectivity pattern for methanation of CO as Pt/A1203. Therefore, it may be concluded that metallic iron is not a selective poison of platinum. It is reasonable to suppose that the reduction of our Pd+Fe/Al,O, samples at 500°C yields PdFe clusters which have a low activity, but the same selectivity as Pd/A1203 . From what is known from the literature on platinum+iron catalysts, the possible interpretation of the results obtained by reducing the sample at 400°C could be a selective poisoning by Fe2+. This hypothesis could explain the influence of tempera- ture and iron content. It is supported by the results obtained by reducing the catalysts TABLE 5.-cATALYTIC ACTIVITIES (lo5 XI01 S-' 8-l Pd) OF IRON CONTAMINATED SAMPLES loading catalyst % Pd % Fe 176 1 0 1702 1 0 341 1.0 347 1.0 342 1.0 345 1.0 351 1.0 354 1.0 362 1.0 365 1.0 343 1.0 346 1.0 348 1.0 3401 1.0 352 1.0 3501 1.0 364 1.0 366 1.0 0.035 0.035 0.035 0.035 0.056 0.056 0.078 0.078 0.035 0.035 0.035 0.035 0.056 0.056 0.078 0.078 344 1 0.035 354 1 0.056 363 1 0.078 activity A 1 for activity A2 for conditions of reduction1OC cyclopentane at 290°C benzene at 140°C R = A 2 / A 1 300 (2 h) 5.3 62 11.7 400 (2 h) 4.2 50 11.9 300 (1 h) 300 (2 h) 300 (4 h) 300 (3 h) 300 (1 h) 300 (2 h) 300 (1 h) 300 (3 h) 400 (12 h) 400 (1 h) 0,400-H 2400 H2fH20 400 (12 h) 400 (12 h) 400 (12 h) 400 (2 h) H2+H20 400 (12 h) 5.0 3.6 6.6 3.6 6.2 3.6 4.8 4.1 3.4 3.2 1.5 0.94 2.9 1.5 1.26 3.05 43 34 56 35 62 30 38 42 17 16 10 23 12 24 5.4 5.0 8.6 9.4 8.5 10.5 10 8.4 7.9 10.2 average 9.1 5 .o 5.0 6.6 5.8 7.9 8 .O 4 7.9 500 (12 h) 0.9 11 12 500 (12 h) 0.5 6.5 13 500 (12 h) 0.7 10 14.3 by wet hydrogen (hydrogen saturated with water at room temperature) : the activities but not their ratio are changed.That treatment will probably not yield metallic iron, but could give Fe2+. In every case, it has been demonstrated here that the selectivity of Pd/A1,03 may suffer large variations in the presence of small amounts of contaminants like iron and sulphur. When an inhibition occurs, these two contaminants have opposite effects on selectivity. Since the phenomenon observed on Pd/Si02 is a relative increase in hydrogenolysis, it may be interpreted by a contamination of the support by iron, which is commonly present in industrial carriers or has been introduced accidentally.The larger effect obtained with silica could be explained by a different reducibility of Fe3+ on silica compared with alumina.23S. FUENTES AND F. FIGUERAS 181 CONCLUSIONS The hydrogenolysis of cyclopentane on palladium catalysts appears to be insensi- tive to metallic dispersion when the support is clean. In the presence of impurities like SO$- or Fe3+, the catalytic properties may be noticeably modified. Both contaminants behave as poisons, but their influence depends mainly on the temperature of reduction of the sample. When an inhibition occurs these two contaminants have opposite effects : sulphur preferentially poisons hydrogenolysis and iron preferentially represses hydrogenation.It is worth recalling here that all the methods which are used to change the disper- sion of a supported metal, like modification of the metal loading or reduction temperature, can also affect the chemical state of the contaminants. The partial pressure of water can also play an important role in this respect. In the case of an impure support it would not be easy to discriminate between the effect of particle size and that of contamination of the metallic surface ; therefore it is useful to compare the results for several supports to obtain safe conclusions on the influence of the particle size for a given reaction. S. F. thanks the " Consejo Nacional de Ciencia y Tecnologia de Mexico " for a scholarship. D. J. C. Yates, W. F. Taylor and J. H. Sinfelt, J. Amer. Chem. Soc., 1964, 86, 2996. D. J. C. Yates and J. H. Sinfelt, J. Catalysis, 1967, 8, 348. M. Boudart, A. W. Aldag, L. D. Ptak and J. E. Benson, J. Catalysis, 1968,11,35. Y . L. Lam and J. H. Sinfelt, J. Catalysis, 1976, 42, 319. R. Maurel, G. Leclercq and J. Barbier, J. CataZysis, 1975, 37, 324. R. Maurel and G. Leclercq, Bull. SOC. chim. France, 1971, 1234. R. A. Dalla Betta and M. Boudart, Vth Int. Congr. Catal., ed. J. W. Hightower (North Holland, Amsterdam, 1973), vol. 2, p. 1329. P. C. Aben, J. Catalysis, 1968, 10,224. J . E. Benson, H. S. Hwang and M. Boudart, J. Catalysis, 1973,30, 146. lo J. E. Germain and R. Maurel, Compt. rend., 1958,247,1854. l 1 S . Szepe and 0. Levenspiel, Proceedings of the Fourth European Symposium on Chemical l2 0. Levenspiel, J. Catalysis, 1972, 25,265. l3 J. E. Benson and M. Boudart, J. Catalysis, 1965,4,705. l4 G. R. Wilson and W. K. Hall, J. Catalysis, 1970,17, 190. l6 F. Figueras, R. Gomez and M. Primet, Ada Chem. Ser., 1973,121,480. l S J. Barbier, mesis (Poitiers, 1975). * O M. A. Vannice and R. L. Garten, J. Mol. CataZysis, 1975,1,201. 22 C . H. Bartholomew and M. Boudart, J. Caralysis, 1972,25, 173. 23 T. Yoshioka, J. Koezuka, H. Ikoma, J. Catalysis, 1970, 16, 264. Reaction Ewineering (Brussels, 1968, Pergamon, London, 1971), p. 265. P. C. Aben, J. C. Platteuw and B. Stouthamer, IVrh Ini. Congr. Catalysis (Moscow, 1968), paper 31. M. Boudart, Adv. Catalysis, 1969,20, 153. R. L. Garten and D. F. Ollis, J. Caralysis, 1974,35,232. R. L. Garten, J. Catalysis, 1976,43, 18. (PAPER 7/133)