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

Complex product distributions were observed with 1,2,4-trimethylcyclohexane using Ir/Al2O3. Those obtained with Pd/Al2O3 at 573 K depended neither on temperature or particle size, S being about 40%.153 Thus it appears that only with platinum and rhodium is there clear evidence for a particle-size effect on the direction of ring-opening; however, with platinum the effect only appeared below about 2 nm and with rhodium below 1.2 nm. On cobalt films, where complete degradation was minimised by use of low temperatures, reaction mechanisms have been studied with the help of deuterium labelling.19 Extension of the alkyl side-chain to C5 lowered ring-opening selectivity,15 but it remained above 70% for Pt/Al2O3, Rh/Al2O3 and even Ru/Al2O3, while for Ir/Al2O3 it was still 92%.

Ring-opening of alkyl-substituted cyclohexanes is slower and much less selective; selectivity was only about 5% with platinum, although Ir/Al2O3 gave 87% C7 alkanes. Lower values were found with n-butylcyclohexane.15

Larger and more complex ring systems undergo other types of transformation, especially on platinum catalysts, but they have not been subjected to quantitative treatment. With spiro(4,4)nonane, one of the two C5 rings was preferentially opened in all possible ways, but surprisingly one of the bonds adjacent to the quaternary carbon atom was the most reactive, so that n-butylcyclopentane was the main product. Isomerisation of the reactant also led via indane to o-ethyltoluene. Spiro[4,5]decane gave naphthalene and n-butylbenzene, spiro[5,5]undecane mainly n-pentylbenzene, and spiro[5,6]dodecane gave biphenyl.13 Cycloheptane reacted to give toluene, and benzene by demethylation. Rings containing eight or more carbon atoms undergo intra-annular dehydrocyclisation: cyclooctane gave Z -pentalene (bicyclo[3.3.0]octane), and cyclononane gave bicyclo[3.4.0]nonane. In these and other similar reactions the hydrogen atoms removed were those closest to each other in the stablest conformation of the ring, which is presumably maintained in the adsorbed state. On Ir/Al2O3, perhydroindan (bicyclo(4.2.0)nonane) was reduced to alkylcyclohexanes much faster than decalin, and bicyclo(3.3.0)octane gave 72% ring-opening selectivity at high conversion.

14.2.6. The Environment of the Active Site: Effect of ‘Carbon’17,42,154

It is desirable at this point to try to draw together a few of the threads that have permeated the previous discussion, in order to give them the prominence they deserve. We have seen that the product distributions and rates of reaction of the higher alkanes with hydrogen are dependent upon operational variables, especially temperature, reactant pressures, time-on-stream and the state of the surface, as well as on the nature of the metal, its support (if any) and its dispersion. Unfortunately the variables that are controllable, namely temperature and reactant pressures, do not give results that are immediately suitable for modelling, because these variables, and others, also affect the coverage of the surface by unreactive

carbonaceous residues (‘carbon’ for short), which are hydrogen-deficient species derived from the hydrocarbon reactant.155−157 This of course is not a new problem; we have mentioned it before (e.g. Section 14.2.2), but it is more prominent with the higher alkanes and has received more explicit attention in this context. The use of short reaction pulses, beneficially used with smaller alkanes (Section 13.1.3), has not been much used here, nor is it always clear that random alteration of the variables and adequate back-checking has always been employed. It therefore seems to be accepted that in most cases one has to be prepared to live with this situation, and to make the best of it. Antal S´ark´any56,158 has proposed a qualitative classification of various states of the surface as (i) Pt-H, where carbonaceous species are absent but the surface is hydrogen-covered, (ii) Pt-HC, where adsorbed species are not too much dehydrogenated, and (iii) Pt-C, where the H:C ratio in the adsorbed species is low. The character of reaction in each of these states can then be considered. Four factors may be at work: (1) lowering of the number of available sites results in loss of activity; (2) reduction in the mean size of the remaining free sites may render certain modes of reaction inoperable, and may facilitate others not formerly possible (this may be a consequence of a reduced availability of hydrogen atoms89), (3) the electronic character of the free sites may be influenced by the adjacent adsorbed species;159,160 and (4) hydrogen associated with the ‘unreactive’ species may participate in the continuing reaction. One is therefore left wondering whether it will ever be possible to obtain a measure of the true catalytic character of a metal uninfluenced by these factors, and whether important parameters such as particle size can ever be truly evaluated because their importance is itself dependent on the nature of the metal.

These difficulties may be exemplified by reference to some of the publications already cited, which will also illustrate additional refinements. The extent of ‘carbon’ deposition (this term covers both Pt-HC and Pt-C states) increases with molar mass of the alkane, but not smoothly;161 we saw before that ethane and propane cause few problems, but with higher alkanes they are unavoidable. There is no direct link between hydrogenolysis to smaller alkanes and ‘carbon’ deposition. On relatively clean surfaces, the former is favoured by high hydrogen pressures, but the depth of the process goes oppositely, methane usually being a major product at low hydrogen pressures.36 Deposition of ‘carbon’ occurs also most readily at low hydrogen pressures,43 the H/C ratio decreasing as the hydrogen/alkane pressure ratio falls. This suggests that (i) isomerisation needs less fully dehydrogenated species than hydrogenolysis, (ii) C––C bond breaking needs fewer hydrogen atoms than reactive desorption of intermediates (conclusions already reached in Chapter 13), and (iii) ‘carbon’ comes mainly from the reactant alkane rather than fragments, although polymerisation of C1 species may occur.162 The state of the surface affects the types of product formed: thus on surfaces largely ‘carbon’-covered, selectivity towards hydrogenolysis is lowered, and is

less affected by varying hydrogen pressure.43 The view has been advanced that the way hydrogen pressure affects selectivities is through change in the ‘carbon’ coverage and not simply to variation in coverage by hydrogen, although as noted above the two are not easily separated.51,128

The occurrence of ‘carbon’ formation is most directly sensed by changes in reaction parameters with time-on-stream (TOS):163 so for example with MCP and Rh/Al2O3 the selectivity to fragments decreased with TOS and increased with temperature, while at and below 468 K the proportions of the C6 products also changed with TOS.36,146 Effects due to ‘carbon’ were also responsible for the different results obtained with the form of reactor used (pulse vs. continuous recirculation95). It also has to be remembered that large particles may be deactivated faster than small ones, so that effects of TOS etc. may be due to gradual elimination of certain classes of active site. The sense of variation of hydrogenolysis selectivity also appears to depend on particle size; with platinum black preheated to 633 K, selectivity decreased with increasing hydrogen pressure for several C7 alkanes, but if preheating was only to 433 K, and the particles therefore being smaller, it increased.164

Variation in conditions of pre-treatment can give major alterations in product selectivities: variables applied include reduction temperature119,146 (or temperature of hydrogen treatment39), oxidation,147,165 and manner of storage.86 Heating by microwave radiation during reduction or use also has had major effects.166,167 The precise effects of these changes are not always clear, but they probably affect surface contamination, particle size or roughness.

It would be nice to find a simple explanation for the outstanding activity of platinum for skeletal isomerisation and other desirable reactions. It is due in part to its inactivity for hydrogenolysis, which in turn follows from its inability to break C––M bonds: thus ‘carbon’ layers are relatively stable under reaction conditions, and carbon contamination is not easily removed by hydrogen.62,147 Thus only small ensembles of free atoms remain in the steady state, and these mercifully are capable of doing what is wanted. In my end is my beginning. With rhodium, however, its much greater activity for hydrogenolysis limits its utility for other reactions;147 the same goes for iridium. Iron was converted into a mixture of carbide phases at high isopentane/hydrogen ratios at 600 K, and there were changes in selectivities favouring intermediate products.116

This section has focused on self-generated effects that can be limited but not eliminated by appropriate choice of conditions. There are of course many other ways in which the environment of the active site can be influenced by deliberate alteration to the design of the catalyst. These include the use of zeolitic supports, and modification by other elements (Re, Sn, S) and inactive metals, adventitious poisons, and the Strong Metal-Support Interaction. These items will be discussed in Section 14.5.

14.3. MECHANISMS OF ALKANE TRANSFORMATIONS

14.3.1. A General Overview168,169

Much has been written about the mechanisms of the reactions undergone by alkanes having five or more carbon atoms (see Further Reading section 3), but little of note has been added to our understanding for the past 25 years or so. It is helpful to start by listing the types of input that have led to these mechanistic statements, and to assess the weight that should be given to them. (1) First and most importantly, there is simply the nature of the products formed, and their relative amounts. (2) Of equal importance is the way in which the amounts vary with experimental conditions, especially hydrogen pressure and temperature, and with the nature and form of the catalyst: the significance of these last variables means that ‘mechanism’ needs to be related to a particular catalyst, there being few if any general statements that can be valid. (3) Routes by which skeletal isomerisations proceed have been elucidated in depth by using isotopically labelled reactants, in a way that would otherwise have been impossible.

Concerning the interpretation, there have been two approaches: (i) input from the general body of organic chemistry and in particular reactions of ligands attached to organometallic complexes, and (ii) quantitative modelling of the reaction kinetics. While undoubtedly some heterogeneously-catalysed reactions have their counterparts in metal-complex-mediated processes, these analogies have led to the thesis170 that single metal atoms are sufficient to bear the key intermediates of the former; this of course is not impossible, and there is some evidence for an important role for atoms at steps and edges, but some of the proposed adsorbed states cannot be reconciled with the observed kinetics.19 Indeed it has generally been the case that mechanisms have been advanced in ignorance of the kinetics, and only in a few cases have orders of reaction been used as criteria.19,93,113 There is in fact remarkably little firm ground on which to build: even the interpretation of orders of reaction is debatable, largely because of the unknown relevance of ‘carbon’ and its possible dependence on reactant pressures, and as we saw in Chapter 13 activation energies are very frail reeds on which to lean. It is however unfortunate that the extensive results13,41 on effects of hydrogen pressure on rates and selectivities collected by the Budapest group have not proved susceptible to quantitative modelling.

What if anything can we then be certain of? The routes whereby skeletal isomerisations proceed are very clearly indicated by isotopic labelling experiments (Section 14.3.2), but even here the structures devised to explain them have relied heavily on organic chemical intuition, and have often involved multiple carbonmetal bonds (i.e. carbenes, CM, and carbynes, CM) that stretch the imagination to near-breaking point. Formal multiple C––M bonds of this type, i.e. having a π- component, are now thought unlikely, and representation of, for example, CM

as a di-σ CM2 is more probably correct. This reformulation places great strain on the availability of the necessary metal orbitals, so that the manner of bonding of α3ε3-hexa-adsorbed n-pentane were it to occur, would need careful thought. Isomerisation mechanisms will be considered further in the next section.

There is little more that can be usefully said about the mechanism of hydrogenolysis than was set down in Chapter 13. What is new concerns the particular form of hydrogenolysis that is responsible for the ring-opening of cyclopentane and its derivatives. On platinum, and to a lesser extent on other metals this process is easier than that of breaking C––C bonds in acyclic species, and requires a less dehydrogenated species13 (Section 14.25). It may be that the inflexibility of the C5 ring imposes additional strain in the C––C bond in the αβ-diadsorbed state, causing it to break easily. It is this that justifies the supposition that cyclic C5 species can be intermediates in skeletal isomerisation (see below). The two modelling studies24,130 already mentioned neither confirm nor deny this possibility, and there are no kinetic studies of C6 ring-opening to help us further.

14.3.2. Mechanisms of Skeletal Isomerisation2,13,19,42,113,135

There is now no doubt that the thermochemical advantages of increasing the degree of branching of alkanes can be realised in practice by metallic catalysts, and that an acidic function is not needed. The elegant and demanding work performed, with the assistance of his colleagues, by the late Fran¸cois Gault before his untimely decease in 1979, and prosecuted subsequently by Maire, Garin and others, has led to a deep understanding of the extent and subtlety of hydrocarbon transformations. The following short account does scant justice to this outstanding work; the several major reviews and the original papers will have to be studied to appreciate their major contribution to the science of catalysis.

There are two separate and distinct mechanisms by which skeletal isomerisation can occur: (i) the bond shift mechanism, and (ii) the C5 cyclic mechanism.

The first is clearly the only possibility when there are less than five carbon atoms in the chain: so the way of isomerisation of n-to isobutane has to be by bond-shift. Two somewhat different mechanisms with a number of minor variations have been proposed.19,42 The first involves an actual or virtual cyclopropanoid species formed by loss of four (or thereabouts) hydrogen atoms, with the bond-shift occurring by the subsequent breaking of a C––C bond other than that just made (Scheme 14.5). Depending on how the ring is formed, the reaction may lead to either a methyl or an ethyl shift. The second involves forming an αγ -diadsorbed species attached to a single metal atom (i.e. a metallocyclobutane), which then dissociates, and following rotation of the alkene part re-assembles and is released as the isomer (Scheme 14.6). The intermediate may be formed directly on platinum, as there is evidence from deuterium exchange (Chapter 6) that this is a favoured mode of

Scheme 14.7. A mechanism for 1,5-cyclisation.

for neopentane hydrogenolysis on palladium are uniformly higher than those for isomerisation, whereas for platinum they are similar, and orders in hydrogen support the view that on palladium hydrogenolysis (but not isomerisation) requires a substantially dehydrogenated species98 (Section 4.2.4). It is possible that because neopentane cannot immediately form a π-alkenyl species because of its quaternary carbon atom, it is driven faute de mieux to find an alternative route in which the two processes differ, and differ from those occurring on platinum. Activation energy values have also been used to argue for two bond-shift mechanisms,19 one applying to reactions in which the degree of branching is unchanged, and another in which it is increased or decreased.

The C5 cyclic mechanism (Scheme 14.7) is clearly established as a strong possibility where the necessary chain exists: it is preferred by palladium and small platinum particles, and is the sole means of isomerisation on iridium catalysts. However metals are further distinguished, on the basis of their behaviour in the hydrogenolysis of methylcyclopropane, as giving products by either selective or nonselective breaking of the C5 cycle: with platinum (small particles) and palladium, the non-selective route predominated, while on iridium it was non-existent.113 This startling difference between adjacent metals is one of the minor mysteries of catalysis, but there does not appear to have been any kinetic study of reactions on iridium to help explain what is happening. Where the selective mechanism of ring opening operates, its reverse is ruled out as a route in isomerisation; n-hexane cannot then isomerise to methylpentanes. Recent work has however shown74,75 that Ir/Al2O3 catalysts are able to effect the dehydrocyclisation of n-hexane to a limited extent in competition with the predominant hydrogenolysis; such catalysts ought then to show a degree of non-selectivity in MCP hydrogenolysis. Disagreements of this kind may be resolvable in terms of variables such as surface cleanliness or operating conditions; it has for example been recently shown172 that with Ir/SiO2 the routes followed in the reaction of 1,4-dimethylcyclohexane changes dramatically with time-on-stream, from mainly hydrogenolysis to mainly dehydrogenation, with small amounts of other products in between. The species involved in the cyclisation step has been considered;113 since for the selective mechanism two adjacent methylene groups are needed, it has been argued that an α3ε3-hexa-adsorbed species is required. Such a deeply dehydrogenated species

is however unlikely to possess the necessary reactivity, and other alternatives are more likely.

It was explained in Section 13.5 that the increase in n-butane isomerisation selectivity with decreasing hydrogen pressure was due to the loss of a further hydrogen atom from the first-formed C4H7 species. In the case of n-heptane isomers, however, molecules formed by the bond-shift route increased progressively with hydrogen pressure, while those arising from a C5 cyclisation route passed through maxima. It is hardly to be expected that these alkanes would isomerise by the bond-shift route through different intermediates; a possible explanation might be that with the C7 isomers the final step becomes rate-limiting, this requiring several hydrogen atoms (see Scheme 13.5) and hence proceeding best at high hydrogen pressures. Precise structures, specifying H/C ratios, for intermediates in C5 cyclic and bond-shift isomerisation are rarely set down.

In the case of the isomerisation of 2- to-3-methylpentane, the switch from the non-selective to the selective C5 mechanism starts (in the case of platinum) as the particle size is decreased below about 2.5 nm, and progresses continuously: however, the bond-shift mechanism (where possible) is the major mechanism above 1 nm size. The extents of these changes depend on the structure of the reactant, and at low dispersions it contributes much more to the 2- to 3-methylpentane reaction than to the 2-methylpentane to n-hexane process. The above statements concerning critical sizes for apparent changes in mechanism rest on TEM studies of the catalysts rather than on mean dispersions obtained by hydrogen chemisorption.19,113 Single-crystal studies (see Section 4.4) show that stepped surfaces are the origin of both types of isomerisation,173 and it has been thought therefore that lowcoordination number atoms in small supported particles are responsible.19 Attempts to find a geometric explanation for the totality of the results have however failed, and opinion now favours the change in electronic structure that accompanies decrease in size (Section 2.5) as being the cause of the changes noted above.19,113 A detailed explanation of how this effect might operate is however still awaited. Activation energies for small platinum particles (in 0.2% Pt/Al2O3) are uniformly higher than those for the larger particles in 10% Pt/Al2O3; a possible explanation for this,19 following the argument developed in Section 13.2.4, is that hydrogen is more strongly chemisorbed on the smaller particles; but there are no kinetic measurements to bear this out.

14.3.3. Dehydrocyclisation2,57,135,174

The process of converting alkanes into cyclic compounds by loss of hydrogen and formation of new C––C bonds, i.e. dehydrocyclisation (DHC), is an important contributor to the matrix of reactions that comprise petroleum reforming (Section 14.1.2); the products thus made, i.e. alicyclic molecules and aromatics, give added value to the output as a fuel, although aromatics are now disliked because of their

carcinogenicity. We have already met DHC in the context of the reactions of linear alkanes (Section 14.2.3) and in the reversible formation of C5 rings as a route to skeletal isomerisation (Sections 14.2.3 and 14.3.2). It simply remains to comment briefly on the actual process of C––C bond formation.

The literature recognises the occurrence of both 1,5- and 1,6-cyclisation.2 The latter takes place through unsaturated linear species (dienes, trienes, Scheme 14.3); it is naturally most prevalent when the concentration of adsorbed hydrogen atoms is low. The process accounting for the former has been much debated;2,175 it differs in that a much less dehydrogenated species is adequate, but simple elimination of two hydrogen atoms at C1 and C5 seems unlikely. Alkanes and alkenes reacted at similar rates, suggesting that the common alkyl radical was involved.2 A mechanism based on an alkenyl radical appears more probable, however (Scheme 14.8): this may be regarded as alkyl-alkene insertion. Carbene-alkyl insertion has also been considered, but the carbon atom is an αα-bonded species is probably sp3 rather than sp2.

14.4. STRUCTURE–SENSITIVITY2,42

14.4.1. Reactions on Single-Crystal Surfaces

Frequent mention has already been made of the effects of surface structure as thought to be affected by changing particle size on product selectivities and reaction mechanisms. In this section, emphasis falls chiefly on effects on rates or TOFs, with cross-reference to mechanistic information where necessary.

Most of the work that has been reported on reactions of the higher alkanes on single-crystal surfaces concerns platinum.176,177 There have been detailed studies59,133,173 of the reaction of n-hexane and other alkanes on Pt(111), Pt(100) and several stepped and kinked surfaces: the effects of changing temperature and hydrogen pressure were observed.59 Arrhenius plots were non-linear due to selfpoisoning at higher temperatures, and activation energies tended to rise with hydrogen pressure. Product selectivities depended on surface structure in interesting ways: (100) terraces favoured internal C––C bond rupture, while (111) terraces promoted terminal breaking, and (111) microfacets also favoured formation of benzene. The bond-shift mechanism was more important on stepped surfaces than on low-index planes.133,178 The dependence of rates and selectivities on operational parameters was much as found with supported platinum. It did not however prove possible to identify the facets exposed on supported particles by comparing products formed on them with those appearing on single crystals because amounts of MCP on the latter often exceeded those on the former.59 MCP reacted faster on Pt(100) than on Pt(111), but the activity of stepped surfaces was like that of flat surfaces because the steps became inactivated by ‘carbon’.134 This is unexpected,

because it is generally thought that ‘carbon’ formation occurs preferentially on terraces. n-Hexane yields at 623 K were significant on Pt(111), Pt(557) (and platinum foil), in disagreement with a later observation that the ring-opening was wholly selective.133 Reactions of C6 hydrocarbons on Pt(111) have been compared with those on palladium foil, but not with other single-crystal surfaces,64 and n-pentane hydrogenolysis on Rh(100) and Rh(111) has been compared with the reaction on Rh/SiO2 treated in various ways.179 Ir(111) and (755) were less active than the same platinum faces for DHC of n-heptane at 423 and 523 K.72

Preparation of supported micro-crystals by vacuum-evaporation of platinum onto crystalline supports gives ‘model’ catalysts exposing specific planes.85,97,136,180,181 Deposition of the metal onto the (100) or (111) face of NaCl followed by evaporation of the support (Al2O3, SiO2 etc.) and dissolution of the NaCl in water has given metal particles exposing only the face originally in contact with the NaCl. Detailed studies have confirmed particle-size effects on the selectivity of MCP ring-opening, but effects on rates were small.136 The technique did however permit study of the mobility of the support at moderate temperatures, in harmony with the encapsulation seen with ordinary supported catalysts at much higher temperatures.

14.4.2. Particle-Size Effects with Supported Metals

There have been numerous studies of the reactions of the higher alkanes on supported metal catalysts, the particle size of which has been controlled by metal loading,36,92,182 calcination,181,183 sintering or some other method. Before attempting a brief summary, we should remind ourselves that mean size (as determined by hydrogen chemisorption or X-ray diffraction) is only a poor guide to the true state of a catalyst, unless the distribution happens to be narrow. Within a broad distribution there may be significant differences in activity, especially as the tendency to lose activity by ‘carbon’ deposition is itself variable. It has even been speculated that apparent particle-size dependence of reaction parameters may be induced by deposited ‘carbon’160 (see also Section 12.3). TEM is a much better indicator of what is in a catalyst.67 Furthermore, the simple measurement of a TOF under a single set of conditions, without determination of kinetics, is only of limited value. A short survey will therefore suffice, especially since some of the trends seen merely confirm those found with the smaller alkanes (Chapter 13).

The structure-sensitivity of alkane hydrogenolysis catalysed by platinum was first established by Oles Poltorak many years ago;184 since then, comparatively little work has been done on this system. The TOF for cyclopentane decreased about five-fold on Pt/Al2O3 catalysts of 7 to 65% dispersion at 573 K, this dependence being similar to that of the stepwise exchange of methane, but less than that for multiple exchange.122 The trends of isomerisation and hydrogenolysis selectivities with particle size observed with n-butane (Section 13.3) have