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

reactions of cyclopentane and of n-heptane) and sometimes with the group having moderate activities (e. g. for neopentane). There is no obvious reason for this oscillation.

14.2.2. Effect of Varying Conversion

The extraction of meaningful information on the products formed in complex reactions is fraught with difficulties. It is of interest to know what are the initially formed products, i.e. before they are transformed by sequential reaction and before the catalyst has had time to acquire its equilibrium amount of ‘carbon’. There are several ways in which conversion can be systematically varied: (1) by changing the flow-rate and hence the contact time, (2) by altering the concentration of reactants, (3) by altering the temperature, and (4) by allowing the catalyst to deactivate spontaneously, noting the changes in rate and products with time- on-stream. Method (1) is undoubtedly the most satisfactory, because the other methods necessarily cause changes in the amount of ‘carbon’ deposited, in the concentrations of adsorbed reactants, and in the H/C ratio of adsorbed species, but even with method (1) the composition of the surface may not stay constant, so it is necessary to know whether the catalyst has been cleaned and reactivated between measurements. With a catalyst or under conditions where hydrogenolysis is the only reaction, it is sometimes possible to monitor how product selectivities change with conversion, and to deduce not only their initial values but also their reactivities as they move towards their final value, which is inevitably 100% methane if there is enough hydrogen. In such cases, changes of selectivities with conversion are often slow, so that initial values can be obtained very accurately;29,33−38 but when multiple types of product are formed, as for example in the reaction of n-hexane with hydrogen over a platinum catalyst,23,39 the changes occur very quickly, and even recording product compositions at conversions of only a few percent40 or less39 is not adequate to define initial selectivities.

We may illustrate these two extreme situations by reference to selected examples from the literature. A simple case is the hydrogenolysis of 2,2-dimethylbutane over a cobalt catalyst at 518 K;34 products were mainly methane and neopentane, selectivities for other molecules being less than 0.1 (Figure 14.2A); but even in the more complex case of n-hexane on Ru/Al2O3 at 422 K,34 rates of change were low and initial selectivities easily determined (Figure 14.2B). Publications by Pa´al and his associates provide numerous examples of the complexity of the early stages of the reactions of n-hexane23,39 and other alkanes13,16 on platinum catalysts. The extent of the changes depended on the hydrogen/alkane ratio used: when this was high (12 to 48) it increased rapidly with conversion but ‘carbon’ formation was limited, and each adsorbed species was at all times largely surrounded by hydrogen atoms. When it was low (e.g. 3), however, ‘carbon’ formation was important, and its change materially altered the composition of each reacting

Figure 14.2. Hydrogenolysis of (A) 2,2-dimethylbutane on Co/MgO-SiC at 518 K and (B) n-hexane on Ru/Al2 O3 at 422 K : product selectivities as a function of conversion.34 1 = CH4 ; 2 = C2 H6 ; 3 = C3 H8 ; 4 = n-C4 H10 ; 5 = n-C5 H12.

centre. In such cases, smaller alkanes were the main or perhaps the exclusive initial product, but at high ratios isomers and cyclic products also appeared early on. With this system it is next to impossible to disentangle the effects of formation of surface ‘carbon’ and changing reactant ratio from the natural progression of the reaction as determined by the reactivity of the products. Rhodium catalysts on the other hand, causing only hydrogenolysis with a little cyclisation, showed (as with Ru and Co) comparatively slow changes of selectivities up to high conversion ( 60%).35,36

14.2.3. Reactions of Linear Alkanes with Hydrogen

Of the reactions of the various types of alkane with hydrogen, those of the n-alkanes, especially n-hexane have been the focus of attention because of their

prime importance in petroleum reforming: they have been ably and extensively reviewed.11,13,14,19,41−45 In this section, attention is concentrated on the effects of varying hydrogen pressure, and the related effects of temperature, on rates of product formation and on selectivities. The alternative modes of presenting the results have already been noted (Section 14.1.4): expressing them as rates or TOFs emphasises the importance of the variable (e.g. hydrogen pressure), but so does the total rate, and the depiction of a rate versus hydrogen pressure at several temperatures can sometimes lead to negative apparent activation energies,46 which are hard to explain. On the whole, the use of selectivities is to be preferred,47 although this is also not without its problems, Treating each product as being formed by an independent reaction implies that the reactant at its initial chemisorption is destined to give ultimately a single defined product, and the options available for the interconversion of intermediates are thereby neglected. There is much convincing evidence to show that the structure of the site on which the alkane first chemisorbs (or to which it later moves), together with the ambient hydrogen atom concentration, determines its subsequent fate, but this does not mean that each product necessarily stems from a site of unique and specific geometry.

The reactions of linear alkanes catalysed by EUROPT-1 (6.3% Pt/SiO2) have been intensively studied by Pa´al and his associates(see Further Reading sections 1 and 2 at the end of the chapter). Most of these studies have used n-hexane, but n-heptane46,49 and n-nonane50 have also been employed. Detailed measurements have been made on the rates and selectivities of product formation over a range of hydrogen/n-hexane pressures and temperatures, and more recently the results have been subjected to kinetic analysis.25,26 In looking at these results, we must try to imagine that the surface coverage by the over-dehydrogenated species we refer to as ‘carbon’ will be variable,51 but may play a deciding role in determining what products are formed (Section 14.2.6) especially at low hydrogen: alkane ratios. It is evident, but not surprising, that at low ratios alkenes (i.e. hexenes) are the main products. Their formation is a complication not usually encountered with the lower alkanes, because of the less favourable thermochemistry, and especially at higher temperatures some further dehydrogenation to alkadienes may occur: these may either cyclise into benzene by a C6 dehydrocyclisation42,48 or may form unreactive ‘carbon’ (see Scheme 14.2). A typical form of dependence of selectivities on hydrogen pressure is shown in Figure 14.3.26 Results such as these, together with those found by varying the conversion (which will also affect the reactant ratio), and temperature, enable us to envisage the H/C ratio in the key intermediate leading to each product, at least in a qualitative way, and so to deduce the kind of reaction scheme shown in Scheme 14.2. The methylcyclopentane (MCP) selectivity passes through a maximum, showing that its formation requires hydrogen atoms when they are scarce, but is inhibited by higher hydrogen pressures due to suppression of the alkene from which it is formed. Skeletal isomerisation is also dependent on hydrogen availability, which is understandable if isomers are formed

Scheme 14.2. Simplified scheme for metal-catalysed reactions of n-hexane.

mainly through MCP. The yield of fragments (<C6) by hydrogenolysis decreased as temperature increased, unlike the situation with n-butane, but with n-hexane there is a greater variety of routes by which they may be formed. A comprehensive and quantitative model for the dependence of products of the n-hexane reaction on process variables seems to have eluded the efforts of the best minds to have attempted it.

A substantial amount of work has also appeared on the reactions of n-hexane catalysed by platinum black (see Further Reading section 2). Use of the unsupported metal avoids any possible complications due to support effects, but the particle size was large; ‘carbon’ formation was more noticeable,53 but it was more active for hydrogenolysis than the highly-dispersed EUROPT-1. The manner of variation of the other products with hydrogen pressure was however generally similar,52 but in the reaction of n-heptane49 the formation of toluene and of 3-methylhexane was more suppressed by high hydrogen pressure. In this reaction Pt/Al2O3 also gave more alkane fragments than EUROPT-1.49 Product distributions in the n-hexane reaction were generally similar on Pt/Al2O3 and Pt/SiO2, although activities (and therefore temperatures) differed; Pt/C gave mainly terminal fission.54,55 Other work on Pt/Al2O3 catalysts has been directed more towards the formation and removal of ‘carbon’.57 Platinum in KL zeolite is renowned for its efficient aromatisation of n-alkanes, and it has been extensively studied:50,58 it will receive further mention in Section 14.5. Results for various single-crystal surfaces have been compared with those for Pt/SiO2.49,52,59 They will be reviewed in the context of particle-size and surface geometry effects (Section 14.4). A Pt/silicalite catalyst also effected DHC of n-hexane.60

Figure 14.3. Hydrogenolysis of n-hexane over EUROPT-1 (Pt/SiO2 ): product selectivities as a function of hydrogen pressure at 603 K.26

Product selectivities observed with platinum catalysts vary markedly with the chain length of the alkane (Table 14.1).16,52 Products of hydrogenolysis, always greater on platinum black than on Pt/SiO2 (markedly so in the case of n-heptane),61 were a minimum with n-hexane, and skeletal isomerisation decreased with chain length as other options such as C5 cyclisation became available. We should note that the existence of five carbon atoms in a chain is not enough to ensure efficient cyclisation, and the progressively greater flexibility provided us chain-length is increased allows more opportunities for cyclisation to occur, the more so on the smaller platinum particles of EUROPT-1.62 All C––C bonds in n-alkanes have comparable probabilities of breaking, but there is a tendency for the chance to decrease on moving towards the centre of the molecule.19,51

Arrhenius parameters calculated26 for n-hexane removal at various hydrogen pressures for EUROPT-1 two types of platinum black give an excellent compensation plot (Figure 14.4); this is not unexpected, as the same behaviour had been found with the lower alkanes (Section 13.2.2; Figure 13.12 and 13.13) and is explicable by the Temkin equation (Section 5.2.5) and the general model presented in the last chapter. The points obtained at or above the rate maximum agree well with the line for the total reaction of lower linear alkanes on EUROPT-1 (Figure 13.4); those obtained below the rate maximum, although not inconsistent with those for higher pressures, lie somewhat below the standard line. This may reflect the difficulty of getting perfectly clean surfaces at low hydrogen pressures. Parameter

Figure 14.4. Compensation plot of Arrhenius parameters for the total reaction of n-hexane on two kinds of platinum black (Pt-N, ; Pt-HCO, ) and EUROPT-1 (Pt/SiO2, O) measured at various hydrogen pressures.26 The parallel lines delineate a band within which the points fall; the broken line is that for EUROPT-1 taken from Figure 13.4. Note that Pt-HCHO is somewhat less active than the other catalysts.

values estimated25 for the major contributing processes on EUROPT-1 also show compensation (Figure 14.5), and with care and a little imagination separate lines for each product can be identified. The propriety of applying the Arrhenius equation to reactions that are not rate-limiting has already been questioned, and it is doubtful whether this exercise greatly advances our understanding of the system. Interpretation of the isokinetic parameters presents many pitfalls for the unwary.26

Much less work has been done on palladium catalysts, which show comparable activity to platinum for the n-alkanes, but with very different behaviour.63 Tremendous differences were shown between unsupported palladium (foil and (111) surface)64 and Pd/Al2O3 treated in various ways65,66 (Table 14.2). The former gave dehydrogenation as one of the major routes, hydrogenolysis being the other, with small amounts of cyclic alkanes and benzene. Of the lower alkanes, methane was the major component, its production being increased with the hydrogen/n- hexane ratio. On Pd/Al2O3 reduced at 573 K (LTR) hydrogenolysis was the chief route,66 but isomerisation and cyclisation to MCP also occurred, in amounts that tended to rise with metal loading. Reduction at 873 K (HTR) produced dramatic changes: activity at low metal loading (0.3%) was increased some 300 times and activation energy lowered (Figure 14.6), and isomerisation then became the major route ( 90%). The rise in rate was less at higher metal loadings, although the

Figure 14.6. Arrhenius parameters for hydrogenolysis of n-hexane on Pd/Al2 O3 shown as a compensation plot: effect of various metal loadings and pre-treatments (see text). The parallel lines delineate the zone that contains the points for lower alkanes on catalysts in Figure 13.3; the broken line is that for EUROPT-1.66 O, 0.3% Pd; , 0.6% Pd; , 2.8% Pd. Open points, LTR; half-filled points, HTR; filled points, regenerated.

mainly hydrogenolysis (Table 14.3) The small amount of cyclisation that occurred decreased with increase in hydrogen pressure;36 it did not happen at all on rhodium film. With Rh/Al2O3 and Rh/SiO2 reduced at either 603 or 1253 K, high hydrogen coverages (i.e. high pressure, low temperature) gave random single C––C bond fission, while the opposite conditions encouraged multiple breaking and methane formation.71 There was a trend from terminal to internal C––C fission as hydrogen pressure was increased. Rh/SiO2 resembled Rh(111) more than Rh(100) in showing a high activation energy (197 kJ mol−1) and a preference for internal C–C bond breaking with n-pentane hydrogenolysis:70 pre-oxidation gave higher rates with lower activation energy, and no products other than lower alkanes were noted.

n-Pentane reacted with hydrogen at 423 K on Ir/Al2O3 to give mainly ethane and propane, with traces of cyclisation and homologation.51 On iridium film,51 n-hexane gave all the lower alkanes but no benzene below 544 K, its yield increasing above 598 K with methane as the other chief product. Hydrogenolysis was also the principal route on iridium single crystal faces72 and on iridium foil.73

TABLE 14.3. Product Selectivities for the Reaction of n-Hexane with Hydrogen over Rhodium Catalysts at 498 K.36

On Ir/Al2O3 containing various chlorine contents, n-hexane at 513 K afforded mainly the lower alkanes, the selectivity for which increased with temperature and hydrogen pressure;74,75 small amounts of other products varied with operating conditions much as with platinum catalysts. Depth of hydrogenolysis signalled by the ζ factor increased rapidly with rising temperature.76 DHC of n-heptane was more effective on Ir/Al2O3 than on Pt/Al2O3, but there was also more hydrogenolysis, which could not be suppressed by sulfiding.77

Ruthenium catalysts are also noted for their high activity for hydrogenolysis, but are capable of giving some surprises. An early study34 of the reaction of n- hexane with hydrogen on Ru/Al2O3 showed that at low conversion (422 K) the breaking of C––C bonds was more or less statistical; conversions were followed to 80%. Ru/TiO2 catalysts showed78 the same marked variations in activity for n-hexane hydrogenolysis as the pre-treatment was changed as were found with the lower alkanes (Section 13.6); exceptionally fast rates were observed with 5% Ru/TiO2 after LTR following reduction at 893 K. Catalysts having 0.1 or 0.5% ruthenium were however able to show skeletal isomerisation when in the SMSI state,79 after reduction at 758 or 893 K, a value of Si of 94% being found at 633 K with 0.1% Ru/TiO2 made by ion exchange using the [Ru(NH3)6]3+ion. Residual hydrogenolysis activity could be further lowered by treatment of 0.5% Ru/TiO2 with thiophene.78 These results nicely illustrate the priority given to reactions demanding large ensembles of atoms when such are available. When they are not, isomerisation becomes possible.

Ru/ZSM-5 catalysts modified by inclusion of either rhenium or rhodium or nickel have been examined80 for n-hexane hydrogenolysis at 403 and 433 K: methane was the main product in most cases, but Ru-Ni/ZSM-5 gave chiefly isoheptane from the reaction of n-heptane at 423 K.

Nickel resembles palladium in giving principally demethylation of alkanes,81−83 although C2 to C4 alkanes were also initial products of n-hexane with Ni/MgO-SiC at 528 K:34 less methane was formed from n-pentane on Ni/SiO2 as hydrogen pressure was increased.82 The possibility of reaction proceeding via a π-allylic intermediate again needs to be considered. Cobalt catalysts show greater tendency to multiple fragmentation to methane.34

Rhenium film gave minor amounts of benzene and isomers in the products of the reaction of n-hexane.84

14.2.4. Reactions of Branched Alkanes with Hydrogen

The introduction of single or double branches (i.e. of tertiary or quaternary carbon atoms) into alkane molecule immediately further differentiates the C––C bonds; thus for example 2-methylpentane has four. The presence of branches also allows a greater variety of modes of attachment to the surface by dissociation of C––H bonds: it is generally assumed that reactive species must be σ-diadsorbed,

Scheme 14.3. Reactions of neohexane.

although for skeletal isomerisation a single point bonding is sometimes preferred.85 So neohexane (2,2-dimethylbutane) can be diadsorbed in either the αβ-, the αγ or the αγ modes; those and the products to which they may give rise are shown in Scheme 14.3. This molecule has been the subject of intensive study.17,86 While all branched alkanes can undergo hydrogenolysis, possible alternative products are circumscribed by the molecule’s structure. Neopentane (2,2-dimethylpropane) is initially limited to isomerisation to 2-methylbutane, but molecules containing five or more carbon atoms in a straight line can also cyclise.2 The range of possible hydrogenolysis products is also much increased by the presence of branches, and there has been great interest shown in different reactivities of various types of C––C bond, although the results have not always been explained in terms of the preferred forms of chemisorption. Quite complex molecules can however show very simple reaction paths: thus both 2,2,3,3-tetramethylbutane and 2,2,4,4-tetramethylpentane87 contain only two distinguishable sorts of C––C bond. The former has been particularly widely used18,67,88−90 in the expectation that formation of the αδ-diadsorbed state, presumed to be needed for breaking the central C––C bond, might be difficult in very small particles, so that the product distribution would be sensitive to surface structure, i.e. particle size and composition.

The principal themes of work in branched alkanes have therefore been (1) selectivities for isomerisation and cyclisation as opposed to hydrogenolysis, (2) the