Metal-Catalysed Reactions of Hydrocarbons / 14-Reactions of Higher Alkanes with Hydrogen

been confirmed with n-hexane.165 Perhaps the most interesting results have been obtained185,186 with ‘cluster-derived’ catalysts made either from (A) Chini complexes [Pt3(µ2CO)3(CO)3]n −2 (n = 2 − 5) or (B) from Pt3(µ2CO)3L4 (L = PPh3 or PEt3); they have been compared with those made by normal methods and have been characterised by TEM. Type A catalysts were less prone to deactivation than classical materials, and showed some differences in 2-methylpentane isomerisation; Type B catalysts showed very selective demethylation of MCP, attributed to the presence of residual phosphorus.

Very careful characterisation of catalysts having a variety of metal loadings is necessary if misleading conclusions are to be avoided. We have seen (Section 2.4) that two types of metal are possible; one very highly dispersed, formed first as the loading is increased, and reaching a limit of typically 1%, and a second, less well dispersed, the concentration of which increases with loading, and therefore predominating at high loadings. In the reaction of neopentane on Pt/γ-Al2O3, there was no change in TOF or activation energy at the changeover point,187 but stepwise hydrogenolysis was associated with the first type and isomerisation with the second. Si therefore increased with metal loading or sintering. These important differences are often overlooked, as they are not revealed by hydrogen chemisorption measurements and may even be missed by TEM (where attention is naturally focused on what is most easily seen). Decreasing particle size in Pt/Al2O3 favoured hydrogenolysis at the expense of aromatisation.188 The difficulty of obtaining an unequivocal connection between activity and particle size is underscored by the observation189 that mode of pre-treatment is in fact more important than size. HRTEM measurements on Pt/SiO2 catalysts suggested that particle shape per se did not affect product selectivity190 (see however results86 for Pt/Al2O3).

Values of TOF for hydrogenolysis of cyclopentane124,191 and MCP67,113,123 on Rh/Al2O3 at 353 K passed through maxima at 20–30% dispersion (H/Rh 0.4), lowest values occurring at high dispersion (Figure 14.9): MCP was much the less

Figure 14.9. Hydrogenolysis of n-hexane (•) and MCP (O) on Rh/Al2 O3 at 493 K: effect of dispersion (H/Rh) on TOF.67

Figure 14.10. Hydrogenolysis of 2,2,3,3-tetramethylbutane on Rh/Al2 O3 : effect of hydrogen pressure O, H/Rh = 0.08; , H/Rh = 1.7.

reactive. [Note: the use of a logarithmic scale minimises the scale of the effect; for MCP the change in TOF was about × 7]. With Rh/SiO2 the effect was less regular, although the trend was similar.67,123 n-Pentane hydrogenolysis on rhodium on various supports indicated192 that activity correlated with surface roughness and the concentration of B5 sites. With Rh/Al2O3 n-hexane showed somewhat irregular treads in the opposite sense, (Figure 14.9) as did 2,2,3,3-tetramethylbutane: values of TOF for the total reaction of this latter reactant are shown as a function of hydrogen pressure for Rh/Al2O3 of low and high dispersion (H/Rh = 1.7 and 0.08) in Figure 14.10. The order in hydrogen is slightly more negative on the larger particles, consistent with the greater integral heat of adsorption determined separately; activation energies (shown only for the αβ-fission process, Section 14.2.4) increase with hydrogen pressure in the expected way. The effect of particle size on the rates of the two bond-breaking modes in this molecule has already been considered (Section 14.2.4). These results confirm the unusual nature of the ring-opening reaction of MCP, compared to C––C bond breaking in acyclic alkanes.

At 453 K, chloride-free Ru/Al2O3 (unlike Rh/Al2O3) gave TOFs that increased 40-fold as dispersion decreased, as did the depth of hydrogenolysis measured by the ζ factor89,193 (Figure 14.11); a few percent of isomerised products were seen at high H/Ru. High dispersions and low TOFs were also obtained with other precursors.90 TOFs for MCP (and other branched alkanes) also decreased as dispersion rose.89 Arrhenius parameters for the reactions of several alkanes on variously dispersed Ru/Al2O3 were shown in Figure 14.8, and the effects of

Figure 14.11. Hydrogenolysis of n-hexane on Ru/Al2 O3 at 458 K: TOF and fragmentation factor ζ as a function of dispersion (H/Ru).89

dispersion of the ways of breaking 2,2,3,3-tetramethylbutane were also discussed in Section 14.2.4. For the reaction of MCP on Ru/Al2O3, see Section 14.2.5.

Palladium differs from the other metals in that there are only negligible or irregular effects of dispersion on TOF for cyclopentane,123 MCP123,194 or branched alkanes102,153 with alumina or silica as support. Pd/Al2O3 made using Pd(NO3)2 did not give lower alkanes from MCP,194 and selectivities for ring-opening were 50–60%. Reduction of chloride-free Pd/Al2O3 at 873 –1073 K resulted in a PdAlx bimetallic phase (xmax = 0.1), which prevented the formation of carbide and was reversed by oxidation. It caused Si in the reaction of neopentane at 527 K to increase from 20 to 80%, and the activation energy to decrease from 243 to only 92 kJ mol−1. Chloride ion inhibited formation of the bimetallic phase.

The above results emphasis the need to appreciate that structure-sensitivity is a property of the whole system, not simply of the reaction. We have had examples of where the effect of dispersion varies with hydrocarbon (compare MCP and n-hexane on Rh/Al2O389), with the support (compare Rh/Al2O3 with Rh/SiO2 etc67,123) and of course with the metal. TOFs of hydrogenolysis usually decrease as dispersion rises,71 (platinum, ruthenium), but not with Rh/Al2O3 (Figure 14.9) or palladium catalysts. MCP behaves as an ordinary alkane on platinum, ruthenium and rhodium (at dispersions above 30%, Figure 14.9); it is only the behaviour of n-hexane on Rh/Al2O3 that provides a contrast with that of MCP.67 The best efforts—and there have been very many—to find a geometric basis for structuresensitivity have been at best only partially successful. Variable extents of ‘carbon’ deposits and of strengths of hydrogen chemisorption have also been thought responsible,147 but positive identification of the culprit has not yet been possible. While the experimental observations are not in doubt, it remains uncertain whether geometric or electronic factors are the main cause, or whether some other factors that derive from the basic ones are not even more significant.

14.5. MODIFICATION OF THE ACTIVE CENTRE

14.5.1. Introduction

We have already considered how ‘carbon’ formed by excessive dehydrogenation of a reactant alkane can modify the characteristics of the active centre, by restricting its size, altering its electronic structure and possibly by providing a source of hydrogen atoms (Section 14.2.6). This chapter concludes with a review of deliberately introduced modifications. These may be classified as follows: (1) alteration of the environment of the metal particle by placing it within zeolite framework, which controls the movement of molecules around it (Section 14.5.2),

(2) modification by addition of rhenium (chiefly to platinum, Section 14.5.3) or elements of Group 14 (Section 14.5.4), (3) effects produced by admixing platinum with other inactive metals (mainly those of Group 11) or metals of different activity (Ru, Ir etc.) (Section 14.5.5, (4) effects due to other toxins either deliberately or accidentally added (S, Cl, H, etc.), and finally (5) effects attributable to metal-support interactions, especially of the SMSI kind.

The motivation for these studies is fairly obvious: it is to limit parasitic reactions such as hydrogenolysis and ‘carbon’ deposition when the target reactions are skeletal isomerisation, cyclisation or aromatisation. What is to be described will amplify the information contained in Sections 12.5.3, 13.7, and 13.8.

14.5.2. Metal Particles in Zeolites

Attention is confined to neutral or basic zeolites: it is unnecessary to discuss the structures of those that have been used, so we simply note the designations of those used most often. These include L, Y (see Further Reading section 4), β,195 mordenite,195,202 and ZSM-5;203 the nature of the balancing cation is denoted by the prefix. The complexity of zeolite structures has unfortunately obstructed any attempt to devise a suitable way of naming them so as to reveal what the structure is.

The Pt/KL system has attracted much attention, as it is very effective for the aromatisation of n-hexane,206 but its disadvantage is a high sensitivity to sulfur, which is inevitably present in at least trace amounts in hydrocarbon feedstocks. The presence of the potassium appears essential, since it increases aromatisation and suppresses isomerisation,195,197 but at the same time decreases thiotolerance. Aromatisation selectivity has been correlated with terminal C––C bond breaking,198 and sulfur has been thought to poison the reaction of n-hexene.199 Pt/KL has been shown to be superior to Pt/KY for aromatisation,201 although activation energies for this and for MCP formation are the same on both: benzene and MCP were made without any intermediate molecules being detected. High electron concentration on the platinum particles (which are necessarily very small) seemed to favour

aromatisation,195 and the heat of hydrogen chemisorption was greater on Pt/KL than on Pt/SiO2 or the acidic Pt/H-LTL.130 This agrees with the rate maxima in the reactions of n-hexane and n-butane (Section 13.23) which occur at small hydrogen pressures: in the latter case, the rate was strongly inhibited by higher pressures. These properties are unique to platinum; Ir/KL did not give an unusual amount of aromatisation.200 neoPentane reacted with hydrogen only by hydrogenolysis on Pd/L containing either lithium, potassium or calcium; Arrhenius parameters were reported.103

The mechanism of aromatisation has been investigated in detail by 13C MASNMR using n-hexane-1-13C as reactant;22 basic catalysts (Pt/MgO-Al2O3, Pd/MgO-Al2O3, Pt/KL) were shown to function both by C6 cyclisation (via hexadienes and hexatriene, the chief route on platinum catalysts) and via C5 cyclisation on palladium. Mechanisms of MCP formation, isomerisation and hydrogenolysis were also examined: platinum catalysts were much more active than the palladium catalyst, this being explained by the need through π-alkenic species on the latter but not the former. Part of the benefit of using the KL zeolite arose through steric restriction on forming the C5 cyclic intermediate. It was concluded that small metal particles were electron-deficient on acidic supports but electron-rich on basic supports. Calcination of Pt/KL is however liable to move metal particles from within to the surface.196

14.5.3. Platinum-Rhenium Catalysts114,135,207

The combination of rhenium with platinum has provided one of the two outstandingly successful catalysts for petroleum reforming, the other being platinumiridium (see Section 14.5.5). PtRe/Al2O3 catalysts have therefore been subjected to intensive academic study, and a vast patent literature also exists. Commercial catalysts normally contain equal amounts of the two components (either 0.3 or 0.6% of each), although fundamental work has explored the whole composition range. For some period through the 1970s and 1980s there was much debate about the modus operandi of the bimetallic catalyst: the main question concerned the state of reduction of the rhenium in the working catalyst, i.e. whether it was all in the zero-valent state and associated with the platinum, or whether some if not all was on the alumina support in a positive oxidation state. Work on the recovery of values from spent catalyst at a very early stage revealed that the second alternative was at least partly true. Rhenium is, as we have seen (Chapter 13), very active for hydrogenolysis, and the performance of the binary catalyst in the unsulfided state shows that separate rhenium particles, or even large ensembles of rhenium atoms in a bimetallic particle, cannot be present, because it is not notably more active for hydrogenolysis than Pt/Al2O3. We noted in Section 13.8.4 that surface energy considerations ought to lead to segregation of platinum to the surface, with

some possible modification in their properties due to electronic interaction with neighbouring rhenium atoms; although this model has been proposed and used there is clear evidence208,209 that the reverse is true, with rhenium atoms preferentially occupying edge or step sites; mutual electronic interaction is plainly indicated by XANES measurements.208 It is also evident that no large assemblies of rhenium atoms are normally present,210 because activity for hydrogenolysis is limited. Although a number of significant studies have been made using unsulfided catalysts,70,211 industrial practice demands as complete elimination as possible of parasitic reactions, and sulfided catalyst are therefore used.212,213 It is now clear that sulfur atoms latch onto surface rhenium atoms, so that the working surface comprises small platinum ensembles separated by ReS species.214,215 In industrial use, the catalyst performs in a bifunctional manner (see Scheme 14.1) because the support is rendered acidic by chloride ion; a contribution to its role by some rhenium cations is not however impossible.

The early work on this system has been reviewed by Charcosset,216 and theoretical ideas to account for its success have been succinctly explained.217 As is often the case with catalysis, it is likely that a number of effects work in concert to produce the final result.218 Many of the early studies focussed on the dependence of rates of contributing processes on composition. With n-hexane217 and n-heptane219 and unsulfided catalyst, hydrogenolysis selectivity rose with rhenium content, but the TOF for total rate fell linearly.219 Cyclopentane hydrogenolysis at 773 K showed a marked maximum rate at 75% rhenium; the rate of this reaction acted as a measure of the number of binary sites, the PtRe unit being some 40 times more active than platinum itself.131,220,221 Arrhenius parameters for this reaction on catalysts of different composition showed a convincing compensation effect.28 However, in general, activation energies have not been measured, and kinetic measurements have been confined to more “realistic” conditions. Deposition of rhenium onto Pt(111) naturally accelerated hydrogenolysis of n-hexane, while addition of sulfur suppressed it and allowed cyclisation to remain:222 the maximum effect was produced by a monolayer of rhenium atoms. However on PtRe/Al2O3, sulfidation encouraged isomerisation rather than cyclisation.217 A small increase in methanation of 2-methylpentane on performing the reaction under microwave radiation has been reported.166

A particular feature of the PtRe/Al2O3 system is the extreme sensitivity of the surface composition to the conditions of pre-treatment applied. The need to reactivate used catalyst by removal of ‘carbon’ and to re-constitute it in its active configuration has generated extensive studies of the effects of oxidation, reduction and oxychlorination: when properly performed, these procedures are eminently successful. Recent publications180,209−211 by Anderson, Rochester and their associates treat these matters in detail, and cite many relevant references. The presence of chlorine, needed to provide acidity, modifies the surface composition, decreasing the surface segregation of the rhenium.180 Excess rhenium, not employed in forming

bimetallic particles, is located on the alumina support.205 Structural changes my also occur during use,180 as well as ‘carbon’ deposition.

The palladium-rhenium system has been little studied,223 and rhenium’s influence on other metals hardly at all.224 It rapidly negated the isomerisation activity of palladium, and the reaction of neopentane with hydrogen had a maximum rate (mainly hydrogenolysis) at 50% rhenium.225

The improved selectivity of sulfided PtRe/Al2O3 for non-parasitic reactions requires a short discussion. Although the PtRe unit has been implicated in greater rates of hydrogenolysis (perhaps because the rhenium atom can more easily accommodate a multiple C––M bond), the higher selectivity for isomerisation and cyclisation must be a feature of a small platinum ensemble. This conforms to the effect of dispersion, where small particles also show high isomerisation selectivity.

14.5.4. Modification by Elements of Groups 14 and 15 and Some Others

The platinum-tin system, the use of which was found to be most beneficial for dehydrogenations, has also been widely investigated for the characteristic reactions of the higher n-alkanes. Previous sections (e.g. 5.5 and 12.3.3) have described the bimetallic structures that can be formed, and likely configurations in aluminasupported catalysts; alumina is the support of choice for practical purposes. A brief r´esum´ is therefore all that is needed. There was much discussion in the earlier publications concerning the extent of reduction of the tin, and the form of its interaction with the platinum. It is now clear that at low Sn/Pt ratios (i.e. ≤ 1) most if not all the tin is reduced, and bimetallic particles are formed:226,227 at higher ratios, much of the tin remains as SnII on the support.228 The homogeneity of the metallic phase and other features depend very much on the method of preparation;188 procedures used include impregnation with H2PtCl6 and SnCl2,229,230 prior application of the tin component,226 coprecipitation,231 and the use of bimetallic complexes.229,230 Reaction of SnR4 (R = n-C4H9) with hydrogen covered platinum particles is also effective.109 Structural studies have been greatly helped by the use of Mossbauer¨ spectroscopy on the 119Sn nucleus.229,230

The extensive literature necessitates a short summary that catches only the main highlights; fortunately, most of the reported studies show similar or related conclusions, so the task is simplified. An early study232 using films revealed a trend that has frequently observed since,230,233−235 namely, a decrease in hydrogenolysis selectivity with increasing tin content. An additional effect was seen,232 not often specifically noted more recently: increasing tin content raised the C6 cyclic yield and decreased the C5 cyclic yield. A related effect was reported235 with PtSn/Al2O3 and with PtPb/Al2O3, i.e. increased yields of hexadienes from n-hexane; these are the necessary precursors to C6 cycles. As always, the beneficial effects are bought

at the expense of activity, although in the reaction of MCP a rate maximum was seen at a low Sn/Pt ratio.235 This is accounted for by the tin blocking hyperactive sites at which disruption of the alkane and formation of ‘carbon’ occurs; sulfur also acted in the same way as tin when deposited onto platinum foil.234 The presence of tin also raised the ratio of 2-methylpentane to 3-methylpentane in the reaction of n-hexane;234 the isomer composition is not however always reported.229 Catalysts prepared from bimetallic complexes have been used for the n-hexane reaction;229 they showed much higher isomer selectivities relative to those for benzene and MCP than those made by simultaneous impregnation of the salts. The latter behaved much as Pt/Al2O3, suggesting that bimetallic particles were not created by this method. The DHC of n-octane has also been followed on PtSn/Al2O3 catalysts;218,236 sodium and other basic additives increased aromatisation selectivity with n-heptane.237 PtSn/Al2O3 subjected to oxychlorination lost much of its activity for aromatisation.188

Divergent views have been expressed on the way in which the tin acts. Its role in limiting the size of platinum ensembles is not in question; what is at issue is whether there is any electronic modification of the active centre. It was claimed that, in the reaction of MCP with hydrogen, tin produced positive effects on dehydrogenation and aromatisation that were not shown by either ‘carbon’ or sulfur; they were attributed to an electronic action,226,227 for which Mossbauer¨ spectroscopy provided some evidence. In view of the proposal interpretation of the effect of sulfur on butadiene hydrogenation (Section 8.3) it would not be surprising if tin also influenced the platinum ensembles to some degree.

Platinum-germanium catalysts have also been used for the reactions of higher alkanes.161,238,239 Those prepared by co-impregnation of salts onto alumina gave higher aromatisation selectivities than those made by sequential impregnation,240 and bimetallic particles (PtGe, Pt3Ge2 and Pt3Ge) were detected in them; catalysts made by latter method were not however much superior to straight Pt/Al2O3. Oxidation of Pt97Si3 alloy produced an active catalyst for reactions of 2-methylpentane.241

The modifying effects of molybdenum242−244 and chromium245−247 have been looked at. The latter, introduced to Pt/Al2O3 as chromyl chloride or potassium dichromate, changed the direction of the reaction of n-pentane with hydrogen towards cyclisation rather than isomerisation or hydrogenolysis, due it was thought to electron transfer from reduced chromium ions to platinum.

Coq, Figu´eras and their associates have conducted wide-ranging investigations of surface modification of supported platinum,109 rhodium,170,248 and especially ruthenium89,113,114,248−250 catalysts by treating them when hydrided with alkyl compounds of aluminium, zinc, antimony, germanium, tin or lead. The purpose of this work was to explore the locations of the modifying atoms on the surface of the active metal particles, and to see whether in any case there was evidence for the selective blocking of sites on either low co-ordination number

(CN) sites on planar parts or high (CN) sites at edges and corners. The effect being sought was termed topological segregation; the influences at work would be (i) the relative surface energies, which in most cases would be expected to affect edge and corners sites, and (ii) the size of the modifier. To obtain selective effects, it was necessary to use particles of medium size (3 to 4 nm) displaying comparable amounts of the two classes of site; on small or large particles, one class only would predominate.

With the aid of a suitably sensitive reactant, namely, 2,2,3,3-

tetramethylbutane, topological segregation was observed with a number of systems. On Rh/Al2O3170 and Ru/Al2O3,89,250 tin and lead selectively decorated (or

occupied) low CN sites, and the reaction mode then changed to give more isobutane, characteristic of large particles; the same effect was seen with aluminium, zinc and tin on Pt/Al2O3.109 Germanium on the other hand showed no topological effects, decorating both classes of site indifferently. Addition of germanium to Rh/Al2O3 favoured demethylation over central C––C bond breaking;251 this simulated the behaviour of small rhodium particles, and FTIR of chemisorbed carbon monoxide showed that at sub-monolayer loading the atoms preferred to occupy low Miller index microfacets. Other molecules such as n-hexane and MCP however were not much affected by modifiers, suggesting that site environment is not greatly important in these cases: addition of tin to Pt/Al2O3 (H/Pt = 0.99) did however decrease the selectivity of MCP ring-opening to the point where it was perfectly non-selective.109 In the case of Ru/Al2O3, neither tin nor germanium had much effect on hydrogen orders or activation energies,114 although naturally TOFs were much smaller.248

14.5.5. Other Bimetallic Catalysts252–256

It is hardly surprising that the reactions of the higher alkanes with hydrogen on bimetallic catalysts have attracted interest (see Further Reading section 5), because the scope for variation in product selectivities is so great, and understanding of the factors affecting their structure is now well advanced. It might be thought, in view of what has been said about bimetallic systems in earlier chapters, and about other means of modification already in this chapter, that it should be possible to predict with fair accuracy what effects to expect. To a certain extent this is so, but there are still some surprises and puzzles. It appears to be difficult to make generalisations: especially with the Groups 8 to 10 and Group 11 systems, what happens seems to depend on the nature of the alkane reactant, and on the manner in which the catalyst has been constructed. The various methods used affect behaviour in ways not yet well understood.

We may start with the platinum-Group 11 systems. Deposition of gold onto Pt(111) gave epitaxial islands of the former, and activity for the reaction of n-hexane decreased linearly with significant changes to selectivities to reach a

minimum value when a gold monolayer had been formed.257 Heating a multilayer of gold gave surface alloys and with these the selectivities at 573 K for hydrogenolysis and aromatisation fell, and that for isomerisation rose, as gold concentration increased. Contrary results were however obtained with a ‘bimetallic molecular cluster precursor’-derived PtAu/SiO2catalyst,258 which gave with n-hexane at the higher temperature of 675 K more hydrogenolysis, at the expense of isomerisation and 1,6-cyclisation than Pt/SiO2,but it deactivated less quickly. These results are not easily harmonised without knowing the temperature coefficients of the component processes. Single platinum atoms in a very dilute PtAu/SiO2 catalyst are reported to catalyse skeletal isomerisation of alkanes.259 With PtCu/SiO2,made either conventionally260 or via a cluster compound,258 hydrogenolysis selectivity was increased, due it was thought to the PtCu sites. Little use has been made of bimetallic single crystal surfaces for these reactions, but the DHC of 4-phenyl- 1-butene to naphthalene has been followed on PtCu3(111) below 500 K.261 This surface also catalysed the cyclisation of 1-hexene to benzene at 405 K,262 the rate being limited by C––C bond formation rather than loss of hydrogen atoms. With platinum-gold films, the rate of neopentane decreased with increasing gold concentration;263 both activation energy and pre-exponential factor rose, the one partially compensates the other (Figure 14.12A). The mechanism of n-hexane isomerisation on nickel-platinum crystals has been followed using 13C-labeled molecules.264

Palladium-Group11 systems provide some interesting contrasts to the platinum-Group 11 systems. With PdAu/SiO2 catalysts made by Barbier’s direct redox method (Section 2.32), activation energies for the neopentane reaction decreased with rising gold content,265 and were again partially compensated by changes in pre-exponential factor (Figure 14.12B): isomerisation selectivities (Si ) were not much altered, their values being between 20 and 40%. The precise trend is not easily seen, because of the need to use different temperatures in order to keep conversions low. The PdAu/SiO2 system needs careful characterisation, because the size distribution may be binodal, each fraction having a different composition.266 The astronomically high activation energy for palladium (324 kJ mol−1) compared to the more modest value for platinum (116 kJ mol−1) has been noted before (Section 14.24) and must reflect the much greater difficulty of forming the αγ species on palladium in cases where the π-alkenyl route is note available. Exactly why this process becomes easier with the bimetallics is not at all clear; similar observations have also been reported for palladium-gold powders267 and films.268 Significant differences between the effects of copper and silver in palladium-based films have also been demonstrated.269 In the neopentane reaction, values of Si were maximal at about Pd85Cu15, rates at about 573 K falling to low values at a copper content about 40%, but with silver as the inert component, rates fell precipitately at silver contents of only a few per cent.270 This may also suggest that binary sites (e.g. PdCu) can show activity.