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

nature of the products of hydrogenolysis, and (3) the structures of intermediate species and their reaction mechanisms. Most emphasis has been placed on the effect of surface structure (crystal orientation, particle size) and composition (surface state, presence of modifiers) on rates and the above three themes. These are developed in Sections 14.4, 14.2.6 and 14.5. There has been comparatively little straightforward kinetic work to explore the effects of experimental variables, and what there has been is not especially helpful. Positions of equilibria between the C6 isomers, benzene and methylcyclopentane have been calculated and compared with experimental values.2,70

The multiplicity of bond-breaking for 3-methylpentane at low conversion for various metal blacks was much as expected: rhodium, palladium, iridium and platinum give predominantly two fragments,91 the first formed new species thus desorbing quickly, while osmium in particular gave mainly methane, with other metals showing intermediate behaviour. The environment of the metal does matter, however; with 2-methylpentane, Co/Al2O3 gave chiefly deep hydrogenolysis, but Co/NaY zeolite was said to be an excellent isomerisation catalyst.92 As usual the most plentiful (and interesting) results have been obtained with platinum catalysts.

The derivation of reaction networks for deriving selectivity equations relevant to hydrogenolysis of branch alkanes has been extended from isobutane to 2-methylbutane33,37 and the dimethylbutanes.34,38 Arrhenius parameters for skeletal isomerisation and hydrogenolysis of 2-methylbutane and of n-pentane on 10% Pt/Al2O3 are shown93 as a compensation plot in Figure 14.7. Values for the movement of a 13C label within the structure but without change to its chemical identity are included; activation energy for this ‘self-isomerisation’ of n-pentane is notably high (300 kJ mol−1), and only the data for its demethylation are seriously wide of the mark. As we now expect, there was a distinct trend of the order of reaction in

Figure 14.7. Reactions of 2-methylbutane and of n-hexane on 10% Pt/Al2 O3 : Arrhenius parameters as a compensation plot.93 Hydrogenolysis (open); isomerisation (half-filled); self-isomerisation (filled).

hydrogen becoming more negative with increase in activation energy. The core of this important paper is the mechanism of skeletal isomerisation of labelled pentanes, and it will be revisited on Section 14.3. The order in hydrogen for isomerisation of 2-methylbutane was much more negative than that for hydrogenolysis on 0.2% Pt/Al2O3 heated by microwave radiation,94 so that isomerisation selectivity Si was greatest at low hydrogen pressure, as found for the reactions of n-butane (Section 13.4) and in qualitative agreement with what was found with 10% Pt/Al2O3.93

The manner of variation of rate with hydrogen pressure/concentration depends somewhat in the mode of reaction used. In the cyclisation of 3-methylpentane on platinum black, sharper maxima were seen when a recirculation reactor was used than when a pulse mode was employed, and inhibition at low hydrogen pressure due to ‘carbon’ deposition, which was marked 633 K in the former case, was absent in the latter.95 In both cases, the hydrogen pressure giving rate maxima increased with temperature.

While with linear and singly-branched alkanes there is clear but not extensive evidence that on platinum catalysts the intermediates for isomerisation and for hydrogenolysis differ in their extents of dehydrogenation, with doubly-branched alkanes as exemplified by neopentane (2,2-dimethylpropane) the situation appears not the same. In an extensive review of Arrhenius parameters for its reactions,96 activation energies for the two reactions were found to be of the same order,85,97 as were orders of reaction (for Pt/KL and Pt/KY zeolites98,99). On EUROPT-1 and on ‘oriented’ model platinum catalysts, activation energies for total reaction increased markedly with hydrogen pressure, as indeed they should.85 The two reaction paths thus seem to go via the same intermediate, which might be the αγ -diadsorbed species.100

The palladium-catalysed reactions present a different picture, however:63,95,98,99,101−104 activation energies for hydrogenolysis were uniformly higher ( 300–370 kJ mol−1) than for isomerisation ( 200–250 kJ mol−1), and orders in hydrogen much more negative (−4 compared to −1.9 for Pd/KL;98 −3.6 compared to −0.6 for Pd/SiO299). On palladium, therefore, the hydrogenolysis intermediate must be the more dehydrogenated, being perhaps an ααγ -species. This difference does not however prevent the data points for the two metals from sharing a common compensation line.

The possible reaction paths available to neohexane (2,2-dimethylbutane) were shown in Scheme 14.3. On various supported platinum catalysts, values of Si (to all isomers) were between 36 and 74%, depending no doubt on factors such as dispersion and surface composition, but only 9% on platinum black.17,106 Detailed product analysis revealed that most of the products (50–70%) were formed through the αγ route and most of the remainder through the αγ route. 2.2- Dimethylbutane reacted similarly on Pt/SiO2 at 568 K.107 This is striking evidence of the ease with which such species can be found on platinum surfaces, to

Scheme 14.4. Cyclic intermediates in the skeletal isomerisation of C7 alkanes. Routes only feasible by bond-shift are shown by dark arrows.

the almost total exclusion of the αβ route. Other doubly branched alkanes have been examined,17,55,107,108 and rates on platinum black as a function of hydrogen pressure sometimes passed through extraordinarily sharp maxima, usually at very low hydrogen pressure.52 Poorly-dispersed Pt/Al2O3 (H/Pt = 0.09) hydrogenolysed 2,2,3,3-tetramethylbutane almost exclusively by the αδ mode, as expected,109 but this continued the major way up to H/Pt = 0.99; the predicted correlation of reaction mode with particle size is clearly not straightforwardly obeyed.

Reactions of the C7 isomeric alkanes (Scheme 14.4) have been studied on platinum black and on EUROPT-1,16,52,110 and rates of formation of each product followed as a function of hydrogen pressure. C5 cyclisation products peaked at moderate pressures (20–30 kPa), while rates for isomerisation, aromatisation and hydrogenolysis usually rose continuously. From such results it is not easy to divine any general principle concerning the optimum degree of dehydrogenation for each process, although C5 cyclisation clearly requires a moderately dehydrogenated species.

The importance of the metal’s identity and degree of dispersion is dramatically illustrated by the reactions of neohexane over iridium catalysts. Ir/SiO2 (unlike Pt/SiO2) gave only hydrogenolysis, 94% of which occurred in the αβ mode (see Scheme 14.4), but iridium black (like platinum black) also gave much hydrogenolysis, but chiefly by the αγ route. 2-Methyl- and 2,2-dimethylbutane reacted with hydrogen over Ir/Al2O3 in the range 423–495 K giving mainly methane

and the corresponding branched alkane;51 variation of hydrogen pressures gave rate maxima that decreased with the number of branches. The small amounts of isomers decreased as hydrogen pressure was raised. It was noted that mechanisms for isomerisation starting with a mono-σ-bonded species are not consistent with the observed dependence of rate on hydrogen pressure. Neopentane was not isomerised on iridium film,42 and 2-methylpentane isomerised to 3-methylpentane on Ir/Al2O3 and iridium sponge by the cyclic mechanism.111

The effects of hydrogen pressure variation on rates and product selectivities in the reaction of 2-methylpentane have been reported36 for Rh/Al2O3 catalysts having 10 and 0.3% metal, and for the less active 5% Rh/SiO2. At 483 K hydrogenolysis predominated, although up to about 20% isomerisation also occurred, mostly by the bond-shift mechanism (Section 14.4). Si was greatest on 10% Rh/Al2O3, where it was independent of hydrogen pressure, but on 0.3% Rh/Al2O3 it decreased, and on the Rh/SiO2 it increased, with hydrogen pressure. Small amounts of cyclisation also happened, their selectivity decreasing with rising hydrogen pressure in all cases. Unlike the situation with n-butane, no maximal TOFs were observed.

Highly dispersed rhodium on various supports favoured hydrogenolysis of 2,2,3,3-tetramethylbutane by the αγ mode, giving methane and trimethylbutane; only the poorly active Rh/MgO gave isobutane as the chief product.67 A kinetic study88 of Rh/Al2O3 in states of high and low dispersion (H/Rh respectively 1.17 and 0.08) showed (i) a more negative order in hydrogen on the latter, i.e. stronger adsorption of hydrogen, (ii) mainly fission by the αδ mode on the latter, its selectivity decreasing as hydrogen pressure rose, (iii) mainly fission by the αβ mode on the former, its selectivity decreasing at low hydrogen pressures, and (iv) activation energies that were higher for αδ mode, although for both modes they increased (as expected) with hydrogen pressure. These observations suggest that the αδ intermediate is more highly dissociated than that for the αβ mode. The contrast in the hydrogen orders between this reactant and 2-methylpentane was very marked. Similar results have been obtained112 with Rh/Al2O3 catalysts at 453 K using 2,2,3-trimethylbutane as reactant.

Ruthenium is noted for its tendency to give multiple hydrogenolysis at low conversions,91 although this was less at high dispersion,89,90 and for its inability to show isomerisation except under special circumstances: small amounts of the latter, and some cyclisation,8,90 has however been reported with 2- methylpentane on Ru/Al2O3 of moderate to high dispersion (H/Ru > 0.35). On 0.5% Ru/Al2O3 2-methylbutane reacted with the same activation energy as neopentane (182 kJ mol−1), but some 200 times faster,33 suggesting that somewhat special sites are needed to allow αγ diadsorption to take place. Demethanation of 2,2-dimethylbutane to neopentane was the almost exclusive at 473 K, very similar results being obtained with supported nickel and cobalt catalysts:34 with 2,3-dimethylbutane, demethanation again predominated, but isobutane (from double-demethantion) as well as ethane and propane were also initial products.34

Figure 14.8. Hydrogenolysis of alkanes (n-hexane, 2-methylpentane and 2,2,3,3-tetramethylbutane, the last shown as half filled points) on Ru/Al2 O3 of various dispersions.89 Line A covers low dispersions (Sponge, O; H/Ru = 0.07, ); line B covers high dispersions (H/Ru = O.37, ; H/Ru = 1.1, ); line C covers the tetramethylbutane on the high-dispersion catalysts where the αβ mode predominates. Note that TOF for this reactant is more sensitive to changes in dispersion than the others.

2,2,3,3-Tetramethylbutane reacted much more slowly than either n-hexane or 2- methylpentane, due to notably higher activation energies; a selection of the Arrhenius parameters for these reactants on Ru/Al2O3 of various dispersions89 is shown in Figure 14.8. Activation energies did not however change smoothly with dispersion. Only the data points for the tetramethylbutane on poorly-dispersed catalysts lie well below the lines embracing the others; this is because the expected but slower αδ mode is more significant in these cases. Variation of hydrogen pressure revealed113,114 that on small (1 nm) particles the rate maximum for the αγ process occurred at a higher pressure than for the αδ process, the αγ /αδ ratio increasing continuously with hydrogen pressures above 40 Torr. On large (4 nm) particles the αδ rate was maximal at a very low hydrogen pressure and showed a large negative order (−2.9). Activation energies were independent of hydrogen pressure above 25 kPa (αγ mode, 139; αδ mode, 159 kJ mol−1) for 1 nm particles, but were variable between 110 and 210 kJ mol−1 for 4 nm particles. These results imply that (i) the αδ intermediate was the more dehydrogenated, and (ii) hydrogen chemisorption was strong, so that its coverage was high over most of the pressure range on both types of catalyst. Comparison between them is not easy because of the variable contributions of the two modes; it appears that it was the TOF for the αδ mode that was chiefly sensitive to dispersion. A parallel study with propane and n-butane115 concluded that hydrogen chemisorption was stronger on the smaller particles.

The base metals of Groups 8 to 10 resemble ruthenium quite closely in their preference for demethanation and easy multiple hydrogenolysis. Results are available for nickel,34,38,81,89,91,107,108 cobalt34,38 and iron34,38,116. With

2,2-dimethylbutane, splitting by the αβ mode between C3 and C4 was the principal route on Ni/MgO-SiC, but with 2-methylbutane37 it was the αβ C1––C2 fission that was preferred, and with 2,3-dimethylbutane38 double demethanation also occurred at low conversion. A detailed and systematic study81 of a number of branched alkanes on 20% Ni/SiO2 confirmed the preference for demethanation, the reactivity of the C––C bond decreasing as the multiplicity of the atom attached to the terminal atom increased (i.e. CI––CII > CI––CIII > CI––CIV, where CI is primary, CII secondary etc.). It was concluded that αγ - and αδ-species adequately explained the preference shown by nickel for demethanation.

Mechanisms of aromatisation of several branched alkanes have been investigated on nickel catalysts, with the aid of 13C labelling using 3-methyl -pentane.117 In addition to a C5 cyclisation plus ring enlargement route, two other routes participated: (i) an ethenyl-shift of 3-methyl -pentane to n-hexane, followed by 1,6- cyclisation, and (ii) an addition-abstraction route involving mono-carbon species (i.e. 3-methylpentane + C1 → toluene − C1 → benzene).

14.2.5. Reactions of Cyclic Alkanes with Hydrogen118

The reactions of cyclic alkanes with hydrogen have attracted enormous interest. Cyclic C5 molecules have been strongly implicated as intermediates in the skeletal isomerisation of alkanes (see Section 14.3), either as visible transitory products or as inferred adsorbed (virtual) species not vacating the surface:19,100 their role in aromatisation of alkanes through ring-enlargement on catalysts having a purely metallic function is however more debatable (see later). For these reasons, cyclic C5 molecules, especially methylcyclopentane, have been extensively studied; the variety of reactions that they undergo119 (Section 14.1.3), and in particular selectivities towards the various isomers that ring-opening generates, have proved irresistible magnets for those who have hoped to use catalytic reactions as means of characterising metal surfaces.120 Although some useful generalisations have emerged, the conclusions reached have not perhaps fully repaid the effort invested. Cyclopentane itself has not been much studied, except to compare the activities of various metals28,121 and to assess the particle-size dependence of its hydrogenolysis rate;122−124 its tendency to deactivate rhodium and palladium catalysts has also been noted.125 Cyclohexane reacts mainly be dehydrogenation and aromatisation,13,126 and its hydrogenolysis has been neglected.127 Larger ring systems undergo reactions of considerable interest,13,19 although these too have not received detailed attention.

Although the C––C––C bond angles in the C5 ring are close to that for tetrahedral carbon, the fission of a C––C bond by hydrogenolysis occurs more readily than that of linear alkanes.2 Reaction takes place at lower temperatures and with lower activation energies, and the order of reaction in hydrogen (for methylcyclopentane) is positive, where for an acyclic alkane it would be negative.128 This is

clearly shown at least for platinum catalysts, where the corresponding acyclic alkanes constitute the major if not the sole initial products.24128−131 These are rapidly desorbed, although with some other metals more active for hydrogenolysis further fragmentation occurs.113 The molecule is therefore more strongly chemisorbed than an ordinary alkane, possibly by reaction with a hydrogen atom to give a cyclopentyl radical in a rate-determining step. It certainly appears that the essential intermediate is more hydrogen-rich than those for all other transformations. Subsequent steps have not been clearly defined19,113 (see Section 14.3.1), but the suggestion that dehydrogenation to an α2β2-tetra-adsorbed species must precede C––C bond breaking is not in accord with the kinetics. The process has been aptly described as ‘a peculiar sort of C––C bond rupture”.13

The introduction of a methyl group differentiates the C––C bonds of the C5 ring into three types according to whether the product is n-hexane, 2- or 3- methylpentane. Although discussion of the effects of particle size and surface geometry on rates is to be deferred to Section 14.4, it is impossible to consider the reactions of methylcyclopentane (MCP) without reference to particle-size and related effects. The literature13,19 recognises two extremes of mechanisms according to the extent that the substituent shields the adjacent C––C bonds. By the so-called selective mechanism this shielding is complete and no n-hexane is made; this situation has been found with 10% Pt/Al2O3,132 but only at a single temperature (506 K); at higher temperatures, n-hexane became a significant product.132,133 It has also been said to be absent from the products on Pt(100) and Pt (111) at 540 to 650 K,134 but this observation was not confirmed on Pt(111) or Pt(557) (or Pt foil) at 623 K.133 When it is certainly observed, however, it is with large particles109,132 or extended surfaces.134 With the non-selective mechanism this shielding effect is absent, and all three types of bond are reactive: if they were equally so, the products n-hexane, 2-and 3-methylpentane would be as 2:2:1. This type of distribution has been most often closely approached by catalysts containing small platinum particles21,109,129,135 (Table 14.3), but extremely small platinum particles have inexplicably given only 10% n-hexane.130 With Pt/Al2O3, the mechanism moved towards the non-selective form as dispersion and hydrogen pressure were decreased and as temperature increased.19,113 While in the great majority of cases the observed selectivities lay between the expected limits, the ratio of 2-methylpentane 3-methylpentane has been not infrequently less than two (e.g. with Pt/KL zeolite130) and sometimes a little greater.129,135 Fact other than the purely statistical can clearly affect the case of breaking of the C2––C3 and C3––C4 bonds. A third partially selective mechanism has been proposed, but it is hard to see what this might entail. It is probably safer to believe that the condition of the surface and the nature of the sites available determine the extent to which the methyl group interferes with the process of chemisorption or the stability the adsorbed state, so that a gradual transition between the two limits might be envisaged. It is not however clear what circumstance could prevent the chemisorption

of MCP by loss of the hydrogen atoms at C1 and C2 on the unobstructed side. However this may be, we may assign a degree of selectivity S to the ring-opening by linear interpolation of the observed amount of n-hexane between the theoretical limits of zero and 40%.19 The non-selective route may proceed by an adlineation mechanism involving adjacent sites on the metal and the support.136,137 The evidence suggested the MCP was dissociatively chemisorbed on Pt/SiO2 at 623 K, since some -OD groups on the support were changed to -OH.138

Table 14.4 contains a small selection of the reported product distributions, but a fuller recapitulation is not warranted because in much of the earlier work (and some of the later) there is a lack of awareness of the importance of experimental variables in determining product selectivities. Relevant factors include (i) surface cleanliness (especially ‘carbon’) and type of pre-treatment,139 (ii) temperature,95 and (iii) the hydrogen/MCP ratio used.24,130,135 There is however some disagreement as to the importance of the latter.

TABLE 14.4. Hydrogenolysis of Methylcyclopentane on Platinum Catalysts: Selectivity Parameters

Column headings: D = fractional dispersion, d = size (nm): S subscripts; < 6 = smaller alkanes; 2 = 2-methylpentane; 3 = 3-methylpentane; n = n-hexane: S = mechanistic selectivity (%).

a EUROPT-1 (6% Pt/SiO2 ) :

b ‘model’ catalysts made by vacuum deposition : c [Pt] = 10%;

d [Pt] = 0.2%, e [Pt] = 1%.

Other products formed in the reaction of MCP with hydrogen are usually minor (Table 14.4). They include benzene, the formation of which increased with decreasing hydrogen pressure140 and with increasing temperature128 (which has the same effect on the concentration of adsorbed hydrogen). It is now thought to arise mainly by 1,6-dehydrocyclisation of n-hexane via linear unsaturated C6 species (Scheme 14.3), and is therefore sometimes included with it.130 Alkanes were very major products of the reactions of cyclopentane and MCP over platinum black and EUROPT-1 at 603 K,141,142 in amounts that understandably decreased with increasing hydrogen pressure. The same trend was observed in <C6 fragments, which must therefore have arisen from thoroughly dehydrogenated intermediates; their amounts have been analysed in detail.139

In the further consideration of the platinum-catalysed reaction, we focus on a few quite recent papers that contain kinetic information, and from which references other than those already cited can be gleaned. There have been two studies relating partly or wholly to EUROPT-1 (6.3% Pt/SiO2), but unfortunately they are in substantial disagreement, and different models were used to explain the results. The first24 gave the dependence of rate on hydrogen pressure at three temperatures; the pressure giving maximum rate increased with temperature, and the activation energy increased from 90 to 220 kJ mol−1 as the hydrogen: MCP ratio was increased from 6.7 to 83. The second130 gave the intermediate value of 138 kJ mol−1 at a ratio of 40. A basis for explaining this variation was suggested in Section 13.2.4. The first paper24 also showed that the order in MCP increased as the fixed hydrogen pressure and temperature were raised. This showed that MCP was competitively and exothermically adsorbed. This paper also reported that selectivities depended on MCP and hydrogen pressures, in the sense that the purely non-selective reaction only took place at high hydrogen and low MCP pressures. In the second paper130, selectivities were independent of hydrogen/MCP ratio over much of the range covered, but they were not expressed in terms of separate reactant pressures. This paper also reported minor benzene formation; the other did not see it. The implied participation of two ‘mechanisms’, the contributions of which vary with experimental conditions, makes the modelling exercise difficult. The analyses proposed in both these papers can be criticised on a number of grounds. One model suggested different modes of adsorption of MCP as being responsible for the two ‘mechanisms’, and a rate-controlling step involving molecular hydrogen.24 The slow step in the other model130 apparently involved several hydrogen atoms. Comparison of these two papers illustrates how the diversity of experimental and theoretical procedures so often used in this field renders arrival at agreed conclusions a hazardous business. For if the trumpet give an uncertain voice, who shall prepare himself for war?

There appears to be some disagreement as to whether ring-enlargement of MCP to cyclohexane and thence rapidly to benzene can occur on catalysts having only a metallic function.143 It does not take place on single-crystal

platinum surfaces,134,135 nor on supported catalysts that are strictly neutral;128,130 it does however appear to go readily on platinum black,13 although it may result from 1,6-cyclisation of n-hexane. Some benzene was also observed on EUROPT-1.139

Detailed studies have also been made on the dimethylcyclopentane isomers,141 using EUROPT-1 and platinum black at 603 K; these help to illuminate factors governing the adsorption of the C5 ring. In 1,1-dimethylcyclopentane, the C1––C2 bond was strongly deactivated, presumably by steric interference, but the other bonds were very reactive, giving mainly fragments at low hydrogen pressure, large amounts of aromatic products and only small amounts of alkene. The behaviours of the Z - and E-1,2-dimethyl isomers differed significantly: both gave large amounts of the cyclic 1-alkene even at high hydrogen pressures as the consequential flattening of the ring reduced strain in the adsorbed state. The Z -isomer was the more reactive, and only this gave n-heptane as one of the products, by breaking of the C1––C2 bond. Z-E isomerisation was observed, equilibrium being attained at high hydrogen pressure; the mechanism must involve breaking and reforming the C1––C2 bond. Demethanation also took place, but is importance decreased with increasing hydrogen pressure. Reaction mechanisms were exhaustively discussed. On rhodium, palladium and platinum films, 1,1,2- and 1,1,3-trimethlcyclopentanes gave either aromatisation or demethylation to the 1,1-dimethyl compound without ring-opening.19,144 The intervention of C7 cyclic species in the interconversion of the heptane isomers is illustrated in Scheme 14.4; certain transformations do however necessitate a bond-shift mechanism.

Methylcyclopentane reacted with hydrogen on Rh/SiO2 and Rh/Al2O3 at500 K to give quantities of smaller alkanes that fell as the hydrogen pressure was raised;35,36,145−148 C5 product selectivities (≥60 %36) showed little dependence on this variable or on conversion up to 70%.35 In another study with Rh/Al2O3, the selectivity S decreased (from 95 to 65%) as dispersion increased.18,67 The activation energy for fragmentation exceeded that for ring-opening, because the intermediates were more hydrogen-deficient; Arrhenius parameters for both reactions showed compensation.146 Over Ru/Al2O3 at 458 K, the reaction gave much fragmentation at low dispersion (H/Ru = 0.07)89, but the C6 selectivities were unaffected, and approached those expected for the selective mode (S ≥ 83%);149,150 on Ru/SiO2 a lower value (42%) was reported15 at 548 K. Ir/Al2O3, Ir/SiO2 and iridium sponge all gave extremely small amounts of n-hexane (S ≥ 99%),15,111,151 and ratios of 2-methylpentane/3-methylpentane greater than two, as is often the case. On 10% Pd/SiO2 at 496 K, ring opening was essentially non-selective.105

With various alkyl-substituted cyclopentanes, ring-opening selectivity with Ir/Al2O3 depended15 on the number of CH2––CH2 bonds, which argues for the dicarbene mechanism, in which the intermediate is a 1,1,2,2-σ4 species. The chemisorbed state of cyclopentene was reported152 to be more or less perpendicular to the Ir(111) surface by a NEXAFS study, in support of this mechanism.