GENERAL DISCUSSION Prof. G. C . Bond (Brunel University) said: In addition to those items listed by Prof. Sachtler as requirements for a selective reaction, one should surely add the further stipulation that those factors associated with the catalyst itself should not be time-dependent. Prof. Sachtler knows better than I the strict standards of chemical and mechanical stability which new industrial catalysts must now meet, and research workers should be cautioned that a perfectly selective catalyst of only limited life will remain at best an academic curiosity. One of the principal problems I see in the design of selective metal catalysts to operate at high temperatures is the likelihood of quite rapid migration of atoms in the surface plane, and to a lesser extent normal to this plane: so that for example a well-dispersed bimetallic catalyst at 500 "C may well expose a semi-fl uid surface having no well-defined surface composition or structure.Some attention should be paid to the problem of making catalysts whose surfaces will retain the planned configuration under reaction conditions. Prof. W. M. H. SacMler (Shell, Amsterdam) said: After having dealt with the question " what makes a catalyst selective?" it is certainly of interest to turn to the question " how do we keep our catalysts active and selective?" Dr. 0. Zahraa (L'niuersity of Carbondale) said: While it is obvious that the elec- tronic properties of the faces, edges and corner atoms in a metal crystallite differ, I would like to ask if one considers a catalyst at high dispersion with an average crystal- lite size (20-25 A), which kind of atom could form the " ensembles " ? Are the atoms from the faces or edges? Or perhaps a combination of both? Prof.W. M. H. Sachtler (Shell, Amsterdam) said: For well-reduced transition metals with every atom in the zerovalent state it can safely be assumed that the energy of interaction of the atoms with SOz, or Al,03 support surfaces is smaller than the energy of interaction of these atoms with each other. Consequently the agglomerates will tend to form three-dimensional structures with overall shapes in agreement with the Gibbs-Wulff principle. Such clusters or crystallites can be described as polyhedra with cut-off corners, sometimes approximated by cubo-octahedra as we did in ref. (1).All exposed atoms of these agglomerates can, in principle, be part of ensembles used in chemisorption, but at high dispersion many of the ensembles will have a considerably wider spacing than those on the surface of a close-packed crystal face. W. M. H. Sachtler et al., J. Chim. Phys., 1954, 51, 1954. Dr. R. Burch (Uniuersity of Reading) said : In his lecture Prof. Sachtler discussed the importance of ensemble size in determining the selectivity of alloy catalysts. In the particular case of Ni/Cu alloys it has recently been reported1 that ethane hydro- genolysis requires an ensemble of 12 Ni atoms. It has also been reported2 for these same catalysts that the heat of adsorption of hydrogen at low coverage is decreased when Cu is added to Ni. Since it is difficult to envisage a mechanism in which an ethane molecule is attached to 12 Ni atoms, and since Cu appears to affect the strength80 GENERAL DISCUSSION of chemisorption bonds to Ni, is it possible that in alloys of this type there is a sig- nificant electronic ligand effect operating in addition to an ensemble effect? M.F. Guilleux, J. A. Dalmon and G. A. Martin, J. Catal., 1980, 62, 235, J. J. Prinsloo and P. C. Gravelle, J. Chem. SOC., Faraday Trans. I , 1980, 76, 2221. Prof. W. M. H. Sachtler (Shell, Amsterdam) said: We have addressed ourselves to the problem of the ligand effect in adsorption by alloys in theoretical studies1 and experimental work, including i.r. spectroscopy and temperature-programmed desorption of hydrogen or CO from alloy films.3 While I must refer to the original publications for an extensive analysis of the problem, I conclude from the data at present available that the selectivity by alloy catalysis is much less affected by changes in bond strength due to the electronic ligand effect than by the relative population of different adsorption states due to the geometric ensemble effect.With respect to the model of 12 nickel atoms required for the adsorption of ethane I refer to my remark in the written text of my lecture. R. A. van Santen, W. M. H. Sachtler, Surf. Sci., 1977,63, 358; W. M. H. Sachtler and R. A. van Santen, Adv. Catal., 1977, 26, 69. Y . Soma-Noto and W. M. H. Sachtler, J. Catal., 1974, 32, 315. J. J. Stephan, V. Ponec and W. M. H. Sachtler, Surf. Sci., 1975, 47, 403; J. J. Stephan, P.L. Franke and V. Ponec, J. Catal., 1976, 44, 359. Prof. J. J. Rooney (Qzreen’s Uniuersity, Belfast) said : In order to understand selec- tivities in catalysis we must understand mechanisms. This is quite clear from Prof. Sachtler’s lecture. However, mechanisms are extremely difficult to elucidate. especially as the catalytic centres are often far fewer in number than hitherto generally believed. For hydrocarbon conversions in hydrogen at elevated temperatures on transition-metal surfaces there is the additional complication that many reactions may simultaneously occur. Very often because of the breadth of our ignorance we accept postulated mechanisms for which there is really very little evidence. After these have been written into the literature for a decade or so they acquire the respecta- bility of dogma.I would therefore make an earnest plea to workers in the particular area mentioned to re-examine most of the mechanisms of bond shift and cyclizations, etc., because in the words of Daniel O’Connell one could “ drive a coach and horses ” through most of them. My second point is that most of the non-destructive reactions on metal surfaces also seem to be due to intermediates essentially bonded to only one metal atom. By the term ‘‘ non-destructive ” I am excluding drastic bond fission reactions such as Fischer-Tropsch synthesis and hydrocarbon methanation. The emerging role of a central mononuclear site vindicates a point made almost 20 years ago,‘ that catalysis on surfaces is often a consequence of interconversions of various intermediates as reactive ligands of the same metal atom or ion.If this is true the idea of ensembles, as described, may be a barrier to progress since we should be thinking instead of a complex where the contiguous atoms or ions, metal or metalloid, act as permanent ligands of the one at the centre of the active site. Their number, disposition and electronic properties are then very important considerations as Dr. Burch points out. Mechanisms can then be discussed in the same language as that of the coordination and organometallic chemist such that the geometric and electronic factors are not seen as independent and separate foundations of any theory. I am not disputing that ensembles or clusters exist. I am merely stressing that jargon can be a great hindrance to progress and to the realization of where catalysis theoretically has its true place in chemistry in general.J. J. Rooney and G. Webb, J. Cafal., 1964, 3, 488.GENERAL DISCUSSION 81 Prof. B. C. Gates (University of Delaware) said: It is a common textbook assertion that high selectivities are characteristic of homogeneous catalysis, in contrast to surface catalysis. Some of Prof. Sachtler’s examples appear to be inspired by homo- geneous (molecular) catalysis. May I ask him to be an advocate of solids and surfaces ? What unique opportunities do they present for selectivity in catalysis? Prof. W. M. H. Sachtler (Shell, Amsterdam) said: Three facts of relevance to selectivity might be quoted to illustrate the opportunities of heterogeneous catalysts.(1) Many biocatalysts, renowned for high selectivity, contain active groups attached to protein bodies, while reactants are dissolved in the surrounding liquid. Such biphasic systems are therefore examples of heterogeneous catalysts. (2) Crystal faces and the pore system of zeolites can provide rigid templates, essential for stereoselectivity. Besides shape-selective zeolite catalysts, I would like to mention Ziegler-Natta catalysts, where different crystal modifications of the same compound, TiCl,, yield different (isotactic or atactic) isomers of the same polymer. (3) Numerous catalysts, including biocatalysts, are bi- or multi-functional. By fixing each functional group on a surface one can prevent undesired interactions between these groups, which would occur if they were dissolved in a common solution.Prof. V. M. Gryaznov (Lumumba University, Moscow) said: I would like to ask (1) Does he have an explanation why alloying with a Group Ib metal should (2) Can he explain why Pd is so easily self-poisoned and does he think dissolved Prof. Ponec the following questions. invalidate the valley position? hydrogen could be responsible ? Prof. V. Ponec (Rijksuniversiteit Leiden) said : (1) I think a theoretician could answer Prof. Gryaznov’s first question better than I can. My simple ideas in this respect are as follows. The fully occupied d-orbitals of Cu are spatially more contracted (higher atomic number 2) than the d-orbitals of Ni, which are directed into the same surface hollow (valley position). Both factors -occupation and contraction-diminish the binding strength of Cu atoms and may finally cause the CO molecules to feel more comfortable, on top of the Ni atoms than in the surface holes partially surrounded by Cu.This also means that the neigh- bouring Ni atoms are again covered so that the whole domain of the Ni-Cu alloy surface is finally occupied “ on top ”. (2) Pd is possibly more poisoned because the carbon layer or the layer of carbon- aceous residues (once it is formed) is less reactive than on Ni. In this case dissolved hydrogen may possibly be involved although this has not yet been proved in skeletal rearrangements. Prof. L. Guczi (Institute of Isotopes, Budapest) said : The importance of geometric arrangement could be proved by simple geometric arguments. Assuming the valence angle to be ca.109 O, the 1-3 complex can be accommodated on the Pt surface of Pt(ll1) without serious strain, whereas on Ni(ll1) it can not. This fact may be an explanation for the experimental finding that on Ni, isomerisation has not been ob- served. Concerning the similarity in the role of carbon and sulphur, this can indeed be found in most cases. However, for smaller molecules such as n-butane, we observed that although in both cases the activity decreased, in the case of sulphur isomerisation82 GENERAL DISCUSSION increased, whereas with deposited carbon CH,-formation increased when a Pt-Re/ A1203 catalyst was used.' L. Guczi, Bull. SOC. Chim. Belg., 1978, 88,497. Prof. V. Ponec (Rijksuniversiteit Leiden) said: If one can apply exactly the same " stereo rules " both to free molecules with localized bonds and to adsorbed species where the metal-carbon bond can be a multicentre (delocalized) bond, then interesting conclusions can indeed be made concerning the geometry of the adsorbed species and the consequences of that geometry.However, in that favourable case one should not forget that even if the overlap of the metal and carbon orbitals is not optimal, there is usually some overlap possible and binding cannot be excluded; the same is true for situations which bring about some strain in the adsorbed species. Prof. Bond once made a point that actually " less than optimal " fit is essential for making the adsorbed species an active intermediate and not only a poisoning blocking material.By the way, well-poisoned Ni reveals low activity in the isomerisation of hexane (not neohexane). Prof. Z. Pail (Institute of Isotopes, Budapest) said: (1) In his interesting and excellent paper Prof. Ponec put forward the idea of a " valley position " deactivation by C, S or other impurities. I should like to draw the attention to a special case where valley-position adsorption may enhance an unusual sort of activity, thus causing significant selectivity changes. This is the case with hydrogen adsorbed in valley positions. Hydrogen can thus block the most active sites where multiple hydrocarbon-metal interactions (important for hydrogenolysis and coking) occur. By doing so, the possibility is created that active sites are on the top of metal atoms. These sites form single metal-carbon bonds and we believe that such intermediates are important for C,-cyclization and ring opening.' This is illustrated by fig.1. It is also clear that after the formation of a single metal- FIG. 1.-Schematic representation of hydrogen and carbon effects over the Pt (111) surface. Large empty circles denote Pt atoms, small shaded circles C atoms (to scale). Hydrogen adsorbed in " valleys " and on top of atoms is denoted by " H ". Top adsorption active for C5-cyclic reactions (shown for both 3-methylpentane and methylcyclopentane) is impossible on Pt atoms (light shading) where a C-atom occupies a valley position.GENERAL DISCUSSION 83 carbon bond (preferentially on the tertiary carbon atom having lowest C-H dis- sociation energy) the site of ring opening/closure will be determined by the distance between metal atoms.The active species shown in fig. 1 may give selective ring opening in positions b and c with respect to the substituent. Hydrogen atoms on top of metal atoms may also participate in ring closure/opening, which thus may proceed via an associatively adsorbed surface species such as that proposed by Liberman.’ Such a geometrical agreement between catalyst and reactant is apparently a unique feature of C,-cyclic reactions. We think that this is another important in- herent difference between the so-called “ bond shift ” and “ C,-cyclic ” mechanisms of skeletal rearrangement.3 Selective deactivation experiments support the above mechanistic as~umption.~ For example, the case of a radiotracer study provides evidence that ca.20% of the surface Pt atoms are covered by carbon atoms, the activity drop of C,-cyclization and skeletal isomerization of 3-methylpentane was 38%, whereas only a 3% decrease of aromatization-dehydrogenation activity was observed. Obviously one carbon in a valley position excludes more than one Pt atom from C5-cyclic reactions (shaded atoms in fig. 1) whereas these metal atoms still retain their activity in dehydrogena- tion-aromatization. (2) Prof. Ponec’s finding that Pt and Ir are more liable to catalyse internal C-C bond fission than Pd and Ni is in agreement with our previous results obtained with metal black^.^ Could he agree with an explanation whereby metals in lower rows of the Periodic Table having a second maximum in their photoelectron spectra possess a higher density of delocalized electrons.This facilitates less strong interactions which are more n-type in character.6 The experimental facts indicate that such interactions must favour internal fission, although at present we do not see the reason why this empirical correlation is true. Z . Paiil, Adu. Catul., 1980, 29, 273. A. L. Liberman, Kinet. Katul., 1964, 5, 128. H. C. De Jongste and V. Ponec, Bull. SOC. Chim. Belg., 1979, 88, 453. * Z. Paiil, M. Dobrovolszky and P. TCttnyi, J. Catal., 1977, 46, 65. Z. Padl and P. Ttttnyi, React. Kinet. Catul. Lett., 1979, 12, 131. Z. Knor, Kinet. KutuL, 1980, 21, 17. Prof. V. Ponec (Rijksuniuersiteit Leiden) said : (1) I sympathize with Prof. Pad’s idea, indeed, that hydrogen might also be causing an enforced shift of species from multibound types in the valley to those singly and doubly bound on the summits. I certainly would not exclude the possibility that carbon atoms could be double-bonded to metal atoms on the summits.I think it is conceivable that a double bond reacts with other groups as it should upon ring closure/ opening, but a conversion of two single metal-carbon bonds into a new carbon-carbon bond seems less probable to me; a single bond is much more localized in space, and its electrons do not interact so easily with other groups. Prof. PaXs information on the influence of carbon deposition on the selectivity is very interesting. This finding confirms the idea that 1-6 ring closure and aromatisation is a reaction run- ning via stepwise dehydrogenation (which can proceed even on isolated atoms), as he and his colleagues suggested some years ago.I think that the results of Mr Davis (from Prof. Somorjai’s laboratory) obtained with Pt monocrystals are similar. (2) I do not follow two points of Prof. PaQl’s suggestion. (i) What is meant by “ density of delocalized electrons ”: the number of sp elec- trons per atom (or spd electrons) or does it concern the spatial distribution of elec- trons and density in certain points? Which one, the valley centre? The existence84 GENERAL DISCUSSION of the two X.P.S. maxima with e.g. Pt or Ir is usually explained as a consequence of the more pronounced relativistic behaviour of valence electrons in heavy atoms (spin- orbit interaction), which fact leads to a splitting of the valence band into two sub- bands. Heavy atoms have the (n - 1) d-electron orbitals more contracted so that the density of these (n - 1) d more localized electrons is higher around the nuclei than with light elements.Its relation to the occurrence of two maxima in the X.P.S. spectra is just a coincidence, I think. (ii) The n-type metal-carbon bonding is generally believed to occur more easily when there are binding (but rather localized) metal d-orbitals able to form such a bond. However, I do not see any relation between the propensity of a metal to form n-bonds and the occurrence of two X.P.S. maxima, or a relation to any variation in the density of delocalized conductivity sp electrons. I would prefer an explanation of the exceptional behaviour of Pt and Ir along the lines outlined in our paper: these two metals have an intrinsic preference for binding carbon atoms on top of the surface atoms and this tendency is strengthened by the presence of deposited carbon.Neither Pt nor Ir are good methanation catalysts; once carbon is formed it stays longer than on Ni. If one accepts that isomerisation runs better in species bound on summits, the problem is solved. Dr. R. A. van Santen (Shell, Amsterdam) said: It can be argued that alloying causes a shift in adsorption from hollow sites to top sites for electronic-structure reasons. The electronic structure of a transition metal can be fairly well described by a constant electron density of 1 electron per atom in a broad s-band and a narrower d-band with varying electron density.Alloying with a Ib metal does not appreciably change the s-band electron density and mainly affects the d-electron band density. We simplify the description of chemical bonding by assuming that the interaction of the adsorbate with the metal is additive for s and d electrons. We limit ourselves further to the interaction with d-electrons, which we suspect are responsible for the electronic effect. Note that the s-band contributes significantly to the chemical bond strength; changes in d-band structure (ligand effect) have a small, but certainly definite effect. In the bulk an atom has 12 neighbours, at the (100) face 4 atoms are removed from the first coordination shell. The z-axis is chosen perpendicular to the (100) plane.In this geometry one observes that dzz orbitals have no mutual 0 overlap. The same holds for the ~ , Z - ~ Z orbitals. On the other hand d,,, d,,, d,, have a 0 overlap. As a consequence the d-electron structure can be schematically described as a broad dxy,yz,zx band over- lapping a narrow d , ~ , , 2 - ~ ~ band. Hence for transition metals at the end of the Group VIII period, the d,z and ~ , Z - ~ Z orbitals will be filled, as this sub-band will be under the Fermi level. Upon surface formation the dzz orbital loses overlap with four dx2-y~ orbitals in the upper plane and the d,Z-,z orbital loses overlap with 4dzz orbitals. In this way the d,~, d,2-,~ band narrows upon surface formation (see fig. 2). There is no change in the d,, band, but the dyz and dxz orbitals lose 2 nearest neigh- bour overlaps with dy2, dzz, and dxz-yz and dxz, d , ~ and dx2-yz respectively.Since they maintain their n overlap in the xy plane only a small narrowing of the bands is expected (see fig. 2). So when one moves from left to the right in Group VIII of the periodic system one may expect that at the surface first the d,,, and d,, orbitals will become completely filled and further in the period the d,, band. Of course the validity of this description depends on the assumption that the Consider an atom at the (100) face of a f.c.c. crystal.GENERAL DISCUSSION 85 effect of increased electron-electron repulsion of a completely filled sub-band is small compared with the difference between the maxima of the sub-band energies. How- ever, the order of band filling in the bulk and at the surface that we deduced agrees with that computed by Fassaert and van der Avoirdl for Ni.In Ni the d,, and d,, orbitals turn out to be completely filled. We will illustrate the ligand effect upon the change in the relative bond strengths of an atom in a hollow and top position using adsorption of a hydrogen atom. A hydrogen atom adsorbed on a top position can only interact with the completely filled d,z orbital. So band filling or alloying has little effect. Bonding of an H atom r n(E) E 11 FIG. 2.-Schematic sketch of the electron distribution of a f.c.c. Group VIII metal and its (100) surface : (a) bulk electronic energy distribution ; (b) surface electronic energy distribution. in a bridging position requires overlap with a bonding combination of dz2, dxz and dyz orbitals, Obviously filling of some of the dxz and dyz orbitals will have a negative effect on the band strength of the bridge position. So a change in the electron density affects the bridged position more than the top position.Band narrowing cannot only lead to an increase in electron density, but in addition affects the localization energy. This can lead to a decrease in bond strength of bridge-coordinated hydrogen atoms2 even when there is no direct coordination to an alloying Ib atom. The essential argument for a hydrogen atom is that the ensemble of metal atoms involved in bonding with the adsorbate, if isolated from the rest of the lattice, is found to have electrons distributed over bonding as well as antibonding orbitals, when the metal atoms in the cluster are able to form direct bonds between each other.This occurs even when each of the atomic orbitals contributes only one electron.86 GENERAL DISCUSSION Embedding of the ensemble in a metal can lead to relaxation of the electrons from the antibonding orbitals into states near the Fermi level, when the antibonding orbitals remain higher than the Fermi level after embedding. This favours the adsorption energy of bridged adsorbates in large particles compared with that of particles of the size of the ensembles itself. There is an additional effect of relevance to alloying. Embedding of the ensemble into the metal leads to a decrease between bonding and antibonding levels of the ensemble with adsorbate because of delocaliza- t i ~ n .~ ’ ~ In the case that alloying leads to localization of electrons, this difference between bonding and antibonding levels will increase. As long as an antibonding level before alloying was below the Fermi level, this can lead to a decrease in bond energy of bridge-adsorbed species relative to that of top-adsorbed species. An alternative way to visualize this is to consider bridge bonding as an interaction of the symmetrical H atom orbital with symmetrical bonding combination of the ensemble metal atoms usually of lower energy. Because of symmetry there is no interaction with the antibonding ensemble metal atoms. When the ensemble is embedded in the metal, there will be a broaden- ing of the orbitals. Consequently when bonding of adsorbate to metal is not too strong, the bonding energy will increase since ensemble orbitals higher than in the isolated case become available.In the case of alloying, with an increased localiza- tion of the ensemble orbitals the maximum energy of the available binding ensemble orbitals will decrease, resulting in a lowering of the bonding energy of adsorbate to alloy surface. D. J. M. Fassaert arid A. van der Avoird, Surf. Sci., 1976, 55, 297. R. A. van Santen and W. M. H. Sachtler, Surf. Sci., 1977,63, 358. D. M. Newns, Phys. Rev., 1969, 178, 1123. J. R. Schrieffer, J. Vac, Sci. Technol., 1972, 9, 561. Prof. V. Ponec (Rijksuniversiteit Leiden, The Netherlands) said : In the simplified picture I suggested in my answer to Prof. Gryaznov, the hollow becomes less favour- able for binding CO or hydrocarbons, because Cu itself binds less, and a (1 11) hollow among, say, two Ni atoms and one Cu atom is less binding than the summits of the two Ni atoms.In my picture this is the “ maximum ” effect of the electronic structure difference between pure Ni and a Ni-Cu alloy. I did not recall any effect of Cu on the behaviour of Ni atoms. This is, of course, a simplification, and Dr. van Santen’s suggestion may be considered as a first-order correction to such a simple picture. The question now is how important such a correction is for the adsorption and cataly- sis by Ni atoms in Ni-Cu alloys, etc. My associates, Toolenaar and Stoop addressed themselves to this question. Using i.r. spectra, they investigated how much the summit position properties change with regard to CO adsorption, when Pt is alloyed with Cu.Namely, on alloys the frequency of a stretching CO vibration is lower than on pure Pt and this has al- ways been considered as evidence of the ligand effects of alloying, in the literature. However, Toolenaar and Stoop obtained results which showed that within the error of the method any effect of Cu on the Pt properties was negligible. By using the isotopic dilution (l2C0, 13CO) technique, Stoop and Toolenaar proved that the main effect causing the decrease of v(CO/Pt) was the dilution of the CO layer and the suppression of the CO-CO interaction. The results appeared in J . Chem. Soc., Chem. Commun., 1981, 1027. I think that with alloys like Pt-Cu and other endothermic and less exothermic ones, the ligand effects are not detectable by strong chemisorption. However, the shift in adsorbed species from the valIey to the atom tops caused by alloying a Group VIIIGENERAL DISCUSSION 87 metal with a Ib metal occurs widely and is easily observable.Therefore, I think that an explanation in terms of effects which do not assume changes on Group VIII metal atoms is preferable for the data known up to now. I am convinced that some type of (weak?) adsorption, sensitive to the ligand effect of, e.g., the type Dr. van Santen has suggested, will be found in the future. Prof. J. J. Rooney (Queen's University, Belfast) said : The " metathesis of carbene " mechanism for 172-bond-shift is not correct. There are many good reasons for this conclusion, some of which have been published.' The correct mechanism of neo- pentane rearrangement, etc.on Pt is as follows: or The metallacyclobutane or ay adsorbed species can undergo fission, as in meta- thesis, but fission only gives hydrogenolysis products. 'Pt ' The second mechanism has been published,2 but largely ignored although it is theoretically sound and also explains why there may be a common intermediate for both 1,2-bond-shift and hydrogenolysis. It involves only a simple extension of the m.0. theory for the first mechanism but can be very readily appreciated in terms of canonical forms, e.g. C C M I I - M The first mechanism has already been proven2 for a wide range of model com- pounds, but both may occur for simple alkanes. Indeed it has been found3 that methylcyclopentane undergoes 100% selective bond shift on Au-rich Pt alloys, a result completely in accord with expectation on the basis of the first mechanism.This could well be in general the only important mechanism. The second one via metallacyclobutanes remains to be proven, although there is some evidence for it in the field of metathesis.88 GENERAL DISCUSSION Selective demethylation and substantial hydrogenolysis without much cyclization is a feature of Ni catalysts, so it seems that phosphorus on the surface of Pt (" Chatt clusters ") makes Pt behave like Ni. Since preferential demethylation on Ni is due to preferential attacl; at primary C atoms and fission via map-type triadsorbed species, I would suggest that similar species are responsible for demethylation on the P- modified Pt.If the metallacycle theory for demethylation were correct (my+ m y , fig. 2 of Maire's paper) significant 1,2-bond shift giving ring enlargement and ulti- mately benzene formation should have been noted. Have the authors any comment on this aspect of their results? Finally, catalysts such as Co and Ni, which favour surface reactions involving mxp and ccc@ species, i.e. metallacarbenes and metallacarbynes, do not favour cycliza- t i ~ n . ~ On the other hand Pt is usually not very good at forming surface carbenes and is a poor methanation catalyst, but it is very good at cyclization reactions. There is therefore every reason to rethink all the popular mechanisms of cyclization where carbenes are freely postulated with little or no evidence.C. O'Donohoe, J. K. A. Clarke and J. J. Rooney, J. Chem. SOC., Faraday Trans. I , 1980, 76, 345. J. K. A. Clarke and J. J. Rooney, A h . Catal., 1970, 25, 125. A. F. Kane and J. K. A. Clarke, J. Chem. SOC., Faraday Trans. I , 1980,76, 1640. F. G. Gault and J. J. Rooney, J. Chem. SOC., Faraday Trans. I , 1979, 75, 1320. Prof. G. Maire (Universite' Louis Pasteur, Strasbourg) said: As underlined by Prof. Rooney it seems that the olefin metathesis mechanism is not correct for bond- shift. However, it is interesting to note that the mechanism proposed in our fig. 2 is the same as that proposed by Chauvin and Herrisson for metathesis.' In the temperature range >250 "C homologation is extensive on Pd, Rh and W. Some activity to a lesser extent was also observed with Pt by O'Donohoe et aL2 Olefin metathesis seems to occur on all metals but to different degrees. Dr.F. Luck3 observed in our laboratory: (i) homologation of 2-methylpentane on platinum black at 300 "C under 40 Torr of HZ; (ii) that the bond-shift and homologation reactions of alkanes are directly correlated to the particle sizes of platinum. On low dispersed supported platinum catalysts or platinum black the bond-shift and homologation reactions are favoured in agreement with table 4 of ref. (2) for sintered and unsintered Pt films. To answer more precisely Prof. Rooney's specific questions: (1) On " Chatt cluster-derived catalysts " no formation of benzene was detected from 2-methyl- pentane or methylcyclopentane. (2) Toluene led exclusively to benzene by selective demethylation excluding map species.Y . Chauvin and J. L. Herrisson, Makromol. Chem., 1971, 141, 161, C. O'Donohoe, J. K. A. Clarke and J. J. Rooney, J. Chem. SOC., Faraday Trans I , 1980, 76. 345. F. Luck, Thesis (Strasbourg, 1978). Prof. V. Ponec (Rijksuniversiteit Leiden) said: Prof. Maire and his colleagues men- tion in this and in some earlier papers a novel mechanism of isomerisation. In this mechanism occur-at a certain step-two fragments of the isomerising molecules completely separated (see fig. 2 of Maire's paper). This happens at a temperature when the separated fragments are thermodynamically more stable than the original molecule. My question is: is it then reasonable to assume that the mechanism suggested is really quite general ? One knows that sometimes isomerisation selec- tivity is higher than 90% (see my paper).Tt means that in those cases the reconstitu-GENERAL DISCUSSION 89 tion of carbene and olefine fragments back into the non-destroyed molecule would have been that complete. The mechanism suggested assumes a rather free rotation of n-bonded ethylene (olefin). My question is: is there any example known of such rotation for a Group VIII metal complex which would justify the assumption made for the situation at 600-700 K on the surface of a metal? Does Prof. Maire think this is really possible? Prof. G. Maire (Universite' Louis Pasteur, Strasbourg) said : Prof. Ponec asked first about the general validity of the mechanism of isomerization proposed in fig. 2 of our paper, supposing a metallocyclobutane species as precursor.Such a pro- posed species was chosen by Gault and Garin [see ref. (2) of our paper] because (i) the rotation of the adsorbed olefin allows the methyl migration; (ii) an adsorbed ethylidene formed via such a metallocyclobutane precursor is rapidly isom erized to an adsorbed o1efin;l isomerization is then replaced by hydrogenolysis of the C-C bond ; (iii) the internal fission has the same activation energy as methyl shift [see ref. (1) of our paper], 45 kcal mo1-l compared with 55 kcal mo1-l for chain lengthening or shortening. In our paper (see fig. 2) we suggested the metallocyclobutane species as precursor responsible for the methyl-migration and the selective demethylation on the " Chatt cluster-derived catalysts ". In this case the selectivity for isomerization was 12% leading to CH4, at first sight in agreement with the remark of Prof.Ponec. But to reply straightforwardly to the question we believe more in a carbenoid species as transition state as recently proposed by us to interpret our results on platinum single-crystal surfaces [see ref. (5) of our paper]. Furthermore the selectivity includes isomers formed via bond-shift mechanisms and via cyclic mechanisms which can be higher than 90% on high or low dispersed platinum catalysts. Concerning the second question of Prof. Ponec an example of concerted rotation in the rearrangement of platinacyclobutanes has been proposed by Casey et aL2 which is closely related to Puddephatt's propo~al.~ Chatt4 in 1953 proposed a detailed picture of the orbitals involved in the formation of olefin-platinum complexes corresponding to a back donation of electrons from Pt to the olefin, i.e.Zeise's salts. This back donation of electrons implies a rotation barrier of 15.3 kcal mol-1 in Mo(C,H,),(diphos), at 98 0C.5 C. P. Casey, Org. Chem., 1976, 33, 189. C. P. Casey et al., J. Am. Chem. SOC., 1979, 101, 4233. R. J. Puddephatt, J. Chem. SOC., Chem. Commun., 1976,626. J. Chatt, J. Chem. SOC., 1953, 2939. J. Byrne, H. Blaser and J. Osborn, J. Am. Chem. SOC., 1975, 97, 3871. Prof. L. Guczi (Institute of Zsotopes, Budapest) said : The importance of dispersion in the hydrogenolysis reaction was emphasized earlier where turnover number increases with dispersion for hydrogenolysis of ethane and n-butane. For the latter case, selec- tivity for ethane formation, i.e.the rupture of the middle C-C bond, also increases. This could well be interpreted by the formation of metallocyclopropane intermediate on Pt catalyst. The same is valid for n-pentane., This, however, cannot be confined only to platinum, but it is characteristic of highly dispersed metal. On highly dis- persed Ru the same phenomena are ob~erved.~ L. Guczi and B. S. Gudkov, React. Kinet. Catal., 1978, 9, 343. A. Sirkany et aZ., Proc. 7th Congr. CataZ., Part A, (Kondasha, Tokyo and Elsevier, Amsterdam, 198l), 291. L. Guczi et a!., Bull. Soc. Chem. Belg., 1979, 88, 497. Dr. P. B. Wells (University of Hull) said: Prof. Knozinger proposes breakdown of O S ~ ( C O ) ~ ~ on his supports to give Os(CO), and Os(CO), bound to oxide.We have90 GENERAL DISCUSSION impregnated O S ~ ( C O ) ~ ~ onto silica, alumina and titania and have observed three-band infrared spectra such as he reports. We have also obtained u.v.-visible reflectance spectra which, for OS~(CO)~~ in solution, contain two bands which have been assigned to electronic transitions in the Os,-framework of the cluster molecule.' These bands are retained when OS~(CO)'~ is impregnated onto the supports mentioned above, and when the impregnated materials are rendered catalytically active by thermal activation. Thus we have evidence for the retention or partial retention of cluster nuclearity when the same materials also provide the three-band infrared spectrum. Furthermore, our catalysts are not only active for CO-hydrogenation (as Prof.Knozinger reports) but also for ethane hydrogenolysis, for which the apparent activa- tion energy may be one-quarter of that exhibited by a conventional supported osmium catalyst. Ethane hydrogenolysis is normally considered to require sites consisting of a considerable number of metal atoms, and hence this observation supports our view that multinuclear clusters are present. In view of these observations, I wish to ask: could the three-band infrared spectrum be assigned to some clustered state (possibly involving the dimerisation or trimerisation of the Os,-unit) of high symmetry? Alternatively, may we have a situation here in which the species which is mostly responsible for the infrared spec- trum is not that which is responsible for CO-hydrogenation and ethane hydrogenolysis ? H.B. Gray, R. A. Levenson and D. R. Tyler, J. Am. Chern. Soc., 1978,100,7888. Prof. H. Knozinger (Uniuersitat Miinchen) said : (1) The observed three-band infrared spectrum can certainly not be attributed to a single surface species. Decarbonylation and recarbonylation cycles clearly demon- strated that the three bands must be due to two interconvertible surface carbonyl complexes which we consider as Os(CO), and OS(CO),.~ Dimerisation and tri- merisation reactions seem to be extremely unlikely at the low metal loadings used in our catalysts. (2) We can certainly not absolutely exclude the possible presence of species other than those detected by infrared spectroscopy and TEM. These may indeed be responsible for the catalytic properties of the samples, a situation which can hardly ever be excluded in heterogeneous catalysis.H. Knozinger and Y. Zhao, J. Catal., 1981, 71, 337. Prof. L. Guczi (Institute of Isotopes, Budapest) said: R. C . Baetzoldl has stated In Prof. Knozinger's How can one that ca. 100 metal atoms are needed to form metallic properties. work the number of 0 s atoms which form clusters is between 4 and 6. comprehend the action of those supported metals as metal catalysts? R. C. Baetzold, Surf: Sci., 1981, 106, 243. Prof. H. Knozinger (Uniuersitat Miinchen) said: We have never said that the sup- ported osmium species would have metallic properties. They are to be described as surface carbonyl complexes, the properties of which will be determined by the nature of ligands, the coordination number and formal oxidation state of the 0 s atoms.Prof. J. Cunningharn-(University College, Cork) said: At one point of his paper Prof. Knozinger suggests for catalysts prepared from H,OsCl, on A1203, that " Lewis- acid centres on the surface of the A1,0, support could play a role in the catalysis, such as aiding in the dissociation of CO ',. However, at another point he advances the idea that smaller Os-0s distances (than in osmium tricarbonyl) may be importantGENERAL DISCUSSION 91 for CO dissociation upon ensembles of three osmium atom complexes on A1203. My question is whether it might not be more consistent to envisage upon all the sup- ported osmium catalysts a role in CO dissociation for Lewis-acid centres of the support.The proximity of such sites to osmium atoms seems assured, as also does the different Lewis-acid character towards the oxygen of CO. Prof. H. Knozinger (Universitat Miinchen) said: We have indeed suggested the two alternative possibilities mentioned by Prof. Cunningham to explain the CO dissociation. Although we do not yet have direct experimental evidence, we feel that the CO dissociation with the direct participation of Lewis-acid sites perhaps through intermediate structures such as c--0 / 0 s \ ~ 1 3 + might be the preferred route. The findings of Katzer et al.' regarding the selectivity control via acid-base properties of the support might perhaps be understood in a similar way. J. R. Katzer, A. V. Sleight, P. Gajardo, J. B. Michel, E. F. Gleason and S. McMillan, Faraday Discuss.Chem. Suc., 1981,72, 121. Dr. S. D. Jackson (University of Hull) said: Prof. Knozinger states that there is a slow loss of catalytic activity due to carbon deposition which affects all the products, then also states that electron microscopy gives results consistent with the carbon- aceous deposit being on the support. Could he please explain how carbon on the support affects the activity of the metal? Also, does he see any evidence from X.P.S. for metal carbide on the used catalyst? Prof. H. Knozinger (Universitlit Miinchen) said: The reasons for deactivation dur- ing high-pressure experiments have not been studied in detail. The suggestion of carbonaceous deposits being the reason, came from the observation by TEM that carbon had probably formed on the support.This statement, however, should not necessarily be interpreted in the sense that carbon on the support affects the activity of the metal. Carbonaceous deposits (not directly detected by TEM) might also occur on the metal species. Moreover, if the Lewis sites on the support surface participate actively in CO dissociation, these may be blocked by carbonaceous deposits as well. Photoelectron spectra of the used catalyst have not yet been meas- ured. Prof. M. W. Roberts (University College, Card@): Do the authors have in addi- tion to the Os(4f) spectra shown in fig. 1 of their paper any corresponding C(ls) or O( 1s) spectra ? The C( 1s) and O(ls) spectra would enable comment to be made on the nature of the CO-Ru interaction, since a correlation has been shown to exist' between AH and the O(ls) binding energy.Valence-level spectra might also be helpful to define the CO-0s system. In fig. 1 how did the authors arrive at the conclusion that the Os(4f) spectrum can be deconvoluted to give two peaks attributed to a single 0 s and two edge 0 s atoms ? R. W. Joyner and M. W. Roberts, Chem. Phys. Lett., 1974, 28, 246. Prof. H. Knozinger (Universitdt Miinchen) said: It would certainly be extremely interesting if the correlation between AH and O(1s) binding energy as suggested by92 GENERAL DISCUSSION Prof. Roberts could be applied. C(ls) and O(ls) spectra have indeed been measured. I am afraid, however, that they do not contain much information on the CO-0s interaction. The l'eason is the extremely low concentration of 0 s complexes in the oxide-supported samples; therefore, the O( 1s) signal is essentially produced by the oxygen of the support and relatively small contaminations would also dominate the C( 1s) signal.The conclusion that the Os(4f) spectrum of a silica-supported cluster indicated the superposition of two doublets of two types of chemically different 0 s atoms came from the comparison with the spectra of an authentic molecular compound: Valence-level spectra have not been measured. \I/ The spectra of the supported and unsupported cluster compounds were closely similar with respect to binding energies and full width at half maximum. Prof. J. M. Thomas (Cambridge University) said: The difficulties inherent in imaging small supported clusters of metals or alloys (5-50 A diameter) by conventional transmission, high-resolution electron microscopy (HREM) are well known and are again highlighted in Prof.Knozinger's micrographs. Yet, as he rightly emphasizes, with heavy atoms such as osmium dispersed on light supports such as silica or alumina the chances of success are improved. I would like to suggest that, instead of em- ploying phase contrast or absorption contrast, as is usually done in conventional HREM, one should adopt an approach that uses so-called atomic-number contrast. (With an electron beam of cross-section ca. 3 A sweeping across the surface it is possible to record images from both elastically and inelastically scattered beams. The elastically scattered electrons are deviated through larger angles than the inelastic ones.If a ratio of the elastic to the inelastic signal is recorded, the resulting intensity distribution is directly proportional to the atomic number.) From the results of Wall and others [see ref. (2) and (3) of this comment and plate 11 there is every indication that atomic-number contrast electron microscopy will directly reveal the nuclearity of 0 s clusters. The micrograph reproduced here clearly shows clusters of uranium consisting of two, three and seven atoms, the latter being disposed as in a fragment of a close-packed sheet. J. M. Thomas and D. A. Jefferson, Endeavour, New Ser., 1978, 2, 136. J. Wall, in Nobel Symposium No. 47 (Aug. 1979), published as Direct Imaging of Atoms in Crystals and Mofecules (Royal Swedish Academy, Stockholm, 1979), p.271. .T. M. Thomas, Nature (London), 1979, 281, 523. Prof. H. Knozinger (Universitdt Miinchen) said : In fact, as Prof. Thomas stresses very clearly, resolution was not a problem in our conventional TEM studies. The difficulties in interpretation of contrast arise from the fact that the material used is polydisperse. For this reason we did not want to over-interpret our micrographs. However, it appears to be safe enough to say that TEM results were not in contra- diction to the structural model derived from infrared spectra. The technique which Prof. Thomas suggests would certainly be of great help for elucidation of the structural details of 0s-atom ensembles in our catalysts. Un- fortunately, the technique has not hitherto been available to us.PLATE 1.-Atomic-number contrast image of uranium supported on carbon [from ref. (2)] showing individual atoms of uranium and clusters consisting of two, three and seven atoms.Each white spot represents an atom. [To face page 92GENERAL DISCUSSION 93 Dr. A. P. G. Kieboom (Delft University of Technology) said: I have two short questions for Prof. Gryasnov: (1) Has he tried membrane catalysts for the partial hydrogenation of benzene to cyclohexene ? High selectivity could be expected by regulation of the hydrogen transfer through the catalyst. (2) Please would he give in a nutshell the major advantages and disadvantages of the hydrogen-porous membrane catalysts? Prof. V. M. Gryaznov (Lumumba University, Moscow) said: In reply to Dr. Kieboom's first question I should like to cite the results of a benzene hydrogenation study' on a hydrogen-porous membrane catalyst made of an alloy of 94.1 wt % palladium and 5.9 wt % nickel. The rate of cyclohexene formation was ca.70% of the cyclohexane formation rate at the beginning of the experiments in the flow system. In connection with the second question of Dr. Kieboom I will try to summarize the major advantages of the hydrogen-porous membrane catalysts. They are as follows: (1) the control of atomic hydrogen concentration on the catalyst surface irrespective of partial pressures of hydrogen and other substances in the reaction zone; (2) the increase in hydrogenation rate and selectivity by diminishing the competition in adsorption of hydrogen and the hydrogenatable substance; (3) the enhancement of dehydrogenation rate and selectivity as a result of hydrogen removal through the catalyst ; and (4) the coupling of dehydrogenation and hydrogenation reactions, which are carried out on two surfaces of the membrane catalyst with higher efficiency than during separate courses. The disadvantages of the membrane catalyst are caused by the small fraction of palladium atoms on the surface in comparison with the total amount of palladium atoms in the catalyst. This fraction may be increased by using very thin films of palladium alloys on hydrogen-permeable supports.2 The precious metal losses which are inevitable in the case of other catalysts are eliminated when using membrane catalysts because of their high mechanical strength and corrosion resistivity. That is why the higher initial investments required for membrane catalysts are adequately compensated. V. M. Gryaznov et ul., in Mechanisms of Hydrocarbon Reactions, ed. F. Marta and D. Kallo (Akademiai Kiado, Budapest, 1975), p. 107. V. M. Gryaznov, V. S. Smirnov, V. M. Vdovin et al,, U S . Patent, 4, 132, 668. However, this correlation decreases with time. Prof. V. Ponec (Rijksuniuersiteit Leiden) said : I appreciated very much that Prof. Gryaznov so clearly formulated one important principle : catalysed reactions can be influenced in an important manner when one succeeds in shifting the conditions of the steady state of the working catalyst. He demonstrated what membranes can do in this respect. I would now like to turn attention to the work done at Delft by Prof. H. Van Bekkum. Van Bekkum and associates' use, e.g., zeolites to abstract water from running condensation reactions to achieve higher yields of desirable products, etc. Another example is to use zeolites to remove selectively lower alcohols from a reaction mixture upon an interchange reaction of esters and alcohols, etc. These are other examples of the same approach in manipulating the selectivity. D. P. Roelofsen, J. W. M. De Graaf, J. A. Hagendoorn, H. M. Verschoor and H. Van Bekkum, Rec. Trav. Chinl. Pays-Bas, 1970, 89, 193; D. P. Roelofsen, E. R. J. Wils and H. Van Bekkum, Rec. Trav. Chim. Pays-Bas, 1971, 90, 1141.