Metal-Catalysed Reactions of Hydrocarbons / 13-Reactions of the Lower Alkanes with Hydrogen

Scheme 13.5. Reactions of n-butane showing suggested structures for intermediate species.

carbon atoms produced the desired effect (Scheme 13.5). Re-formation of a C C bond under hydrogenolysis conditions does however seem unlikely. The old simple-minded model111 of the projection of molecular orbitals from singlecrystal surfaces has been deployed91 to explain the various isomerisation efficiencies, and indicated that the best surfaces had sites at which a metallocyclobutane structure could be formed (Scheme 13.4). Scheme 13.5 shows an example of the kinds of structure that could be responsible and that are consistent with kinetic requirements.

13.6.HYDROGENOLYSIS OF ALKANES ON RUTHENIUM CATALYSTS

Ruthenium merits a separate section because of the extreme sensitivity of Ru/TiO2 and Ru/Al2O3 to conditions of pre-treatment, a sensitivity that incidentally is not shared by Ru/SiO2.101,103 It shares with osmium (which has been little studied102) the honour of being the most active metal for hydrogenolysis, and can be studied well below 373 K. The somewhat variable results shown for ruthenium in Table 13.2 and some of the later tables may find the basics for their explanation in the following paragraphs. The effects to be reported for Ru/TiO2 involve SMSI but are also additional to it; results relating purely to SMSI with other metals are considered in Section 13.7.

The reduction by hydrogen of the precursor RuCl3/TiO2 at 758 K (HTR1) gave ruthenium particles of moderate size ( 3 nm), but they were partially poisoned by chloride ion that adhered strongly to the support,112 and so transition to the SMSI state was incomplete, but the activity for alkane hydrogenolysis was low.99−101,103,113,114 Hydrogen chemisorption isotherms failed to show flat plateaux, and were useless for size estimation. Higher temperature (893 K) was needed to remove all the chloride ion, but a mild oxidation (623 K) (which did not give any RuO4) also removed it, and led after mild reduction (LTR) to higher

dispersions (H/Ru ≈ 0.5–1) and very much higher activities (typically by a factor of 10–70), with little change in activation energy. The effect was shown by ethane, propane and both butanes, by several types of titania, and was usually more marked at low metal loadings (0.1 and 0.5%). A second high-temperature reduction (HTR2) caused the H/Ru ratio to decrease, probably because of entry into the SMSI state (Section 13.7); activities were then very low. The O/LTR treatment (mild oxidation followed by mid reduction) clearly increased the number of active centres more than in proportion to the increase in area, since the Arrhenius parameters gave a compensation line that was well above that for the less active catalysts. Initial reduction at 893 K produced the SMSI state, and the rate was then increased by O/LTR by more than 200 times.101

Very significant changes in product distributions also occurred (Table 13.11). The effect of the O/LTR treatment was to decrease S2 and F, without greatly affecting T3; the rates at which selectivities changed with temperature also altered, becoming somewhat greater after the O/LTR.99 Following HTR2, the selectivity parameters (except T3) moved towards their original values, without quite achieving them. It was thought likely that the oxidation transformed ruthenium metal into the oxide, and that because of the similarity of the titania and ruthenium oxide structures (RuO2 has the rutile structure) Ru4+ ions could easily migrate over the surface to form a well-dispersed layer, so that gentle reduction would create small metal particles.

Analogous but not identical effects have been seen with Ru/Al2O3(see Further Reading section). Use of a chloride-free precursor (e.g. Ru(NO)(NO3)3103 or the acetylacetonate115) removed any possible interference it might have caused; with only 1% metal, the first HTR gave extremely small particles (H/Ru = 0.88;76 mean size (TEM) = 1.2 nm,115 by EXAFS 12 atoms per particle116,117), but the TOFs obtained using 0.71 atm hydrogen were low. The reason became clear when the variation of rate with hydrogen pressure was determined.74 It increased markedly as hydrogen pressure was lowered, giving a sharp maximum at 0.03 atm. (Figure 13.19); product selectivities and derived parameters were virtually constant (Figure 13.20) and changed only slowly with temperature; this behaviour implied very strong hydrogen chemisorption. On applying the O/LTR treatment, TOFs at 0.71 atm hydrogen increased markedly, but inspection of the effect of varying hydrogen pressure showed that this was because hydrogen inhibition had been lowered, and the rate maximum occurred at higher pressures74 (Figure 13.19). Product selectivities then changed (Table 13.11) and became hydrogen-pressure-sensitive (Figure 13.20) and more temperature-dependent. This change in character was analysed in terms of the ES5B equation (Table 13.4), and this showed76 the way in which the constants changed varied with the alkane (Table 13.6): the decrease in bH became larger, and the decrease in KA became smaller, as chain-length increased, and the number of hydrogen atoms lost in forming the reactive species fell. It was then established that the O/LTR treatment caused a decrease in dispersion,75,103 and

Figure 13.19. Hydrogenolysis of n-butane on 1% Ru/Al2 O3 : variation of rate at 398 K with hydrogen pressure after HTR1 and O/LTR pre-treatments.75

formation of aggregates of larger particles.116 A second high-temperature reduction (HTR2) partly neutralised the effect of the O/LTR; there was no further important change in particle size and activity was much lower, but the original high values of KA and bH were not restored (see Table 13.12). These somewhat complex changes, involving the chemisorption of both reactants, were therefore effected by both a change in particle size and some changes in surface character. With a higher metal loading (4%), HTR1 produced larger particles (H/Ru = 0.25), and behaviour that more resembled the 1% Ru/Al2O3 after O/LTR, but even here O/LTR led to higher

Figure 13.20. Hydrogenolysis of n-butane on 1% Ru/Al2 O3 : selectivity parameters at 398 K as a function of hydrogen pressure after HTR1 and O/LTR.75

assumption of the ES5B equation) the composition of the C3 species should have been C3H5. Similar methodology gave the composition of the C2 species in the reaction of propane as C2H2.78 The method does not depend on the validity of the ES5B equation, but only on the Kempling-Anderson method of analysing selectivities. Fuller discussions of these results have been presented.76 The dependence of the rates of reaction of the first three linear alkanes on hydrogen pressure have been determined at various temperatures for 1% Ru/Al2O3 subjected to each of the three pretreatments.76 The form was generally similarly to that shown by Pt/Al2O3; apparent activation energies increased with hydrogen pressure,76,118 and values of bH were small and not temperature-dependent in a consistent way. The effect of particle-size on strength of hydrogen chemisorption cannot therefore be substantiated in this way, but isosteric measurements of the 1 and 4% Ru/Al2O3 catalysts after HTR1 do confirm the greater adsorption strength on the former.76 A selection of values of Et and of HA are shown in Table 13.12; they are of similar magnitude to those found for platinum catalysts at much higher temperatures. Since the concentrations of hydrogen atoms and the reactive hydrocarbon species must be equal at the point of maximum rate, whatever the temperature and their absolute values, the apparent activation energy should approximate to Et ; for 1% Ru/Al2O3 after HRT1 it was 62 kJ mol−1.

13.7.EFFECTS OF ADDITIVES AND THE STRONG METAL-SUPPORT INTERACTION ON ALKANE HYDROGENOLYSIS

The so-called ‘Strong Metal-Support Interaction’ (SMSI) was first revealed by a loss of capacity for hydrogen chemisorption by the metals of Groups 8 to 10 supported on oxides of the metals of Group 5, and on titania and manganous oxide, after first heating them in hydrogen to temperatures above about 573 K (see Section 3.3.5): oxides of sp metals and those of Group 3 and 4 did not respond in this way. This early work at the Exxon laboratories led to a torrent of further publications from elsewhere, and the following main conclusions were established. (1) While all the metals of Groups 8 to 10 were affected,119,120 there have been clear indications that all did not suffer equally,121 and that the extent of the effect might depend on the exact conditions of catalyst preparation (e.g. in the case of Ru99). (2) This loss of chemisorption capacity was not primarily due to decrease in metal-particle size, but rather to a partial or almost complete coverage or encapsulation of the metal by species emanating from the partially-reduced support (e.g. TiO in the case of TiO2).

(3) The importance of the effect increased with the reduction temperature and with decrease in metal loading, i.e. small particles succumbed more easily than large ones. (4) In the case of Pt/TiO2, the heat of hydrogen chemisorption was greatly lowered; there was a smaller effect with Pd/TiO2. This very brief summary (and that

in Section 3.3.5) does scant justice to the enormous volume of work performed to identify the cause of the effect; indeed the attention it commanded was surprising in view of its generally negative implications (except for the hydrogenation of carbon monoxide and carbonyl compounds, rates of which were enhanced by it).

The quantitative study of the SMSI was however made difficult because many of the oxides of interest did not readily lend themselves to become supports for metals. They had different surface areas and surface chemistry, so similar metal dispersions would not be obtainable, and with some their opacity to electrons would make TEM inapplicable. In a practical sense, their mechanical properties (e.g. hardness) did not encourage their large-scale use. An attractive alternative was therefore to mount the modifying oxide onto a strong conventional support (e.g. SiO2) and then to deposit the metal on top;122 or the modifier could be added to the silica-supported metal;103,123,124 or both could be introduced at the same time; or the modifier could be incorporated into the support, as with TiO2-SiO2.125 In this way it should be possible always to have the same metal-particle size, and to alter the modifier to metal ratio without having to change the pre-treatment temperature. These techniques were widely applied,126−129 and in general the same effects were observed as with the equivalent metal on modifier.

The effect of the SMSI on catalytic activity also aroused much interest. In the context of alkane hydrogenolysis, rates fell in parallel to ability for hydrogen chemisorption, and much of these reactions’ reputation for structure-sensitivity depends upon their sensitive response to the incursion of modifying species. This was qualitatively understandable in terms of a requirement for the reactive species to form several C M bonds (Scheme 13.5), the main consequence of the SMSI being to eliminate large ensembles of metal atoms. The conversion of this qualitative concept into reliable quantitative estimates of size of active centre has not in general been possible, and it is unfortunate that most of the studies, which involved mainly ethane130 and n-butane (see Table 13.13 for a selection of examples), reported only rates under one set of experimental conditions. A comprehensive study120 of ethane hydrogenolysis on all the metals of Groups 8 to 10 at 478 K showed that specific rates for titania-supported metals were lower than those for silica-supported metals by factors that were generally about 102 to 103, although much larger for iron and very much smaller for ruthenium. This, and much other early work, informed more on the conditions for and extent of the modification, and revealed little on the changes wrought on the reaction mechanisms. For Rh/TiO2, activation energies for the reactions of the four C2 to C4 alkanes increased progressively with reduction temperature, and exhibited compensation, but the temperature ranges had necessarily to be raised as activity was lost. In another study using Rh/TiO2, rates for both ethane and n-butane at 623 K decreased with increasing dispersion following high-temperature reduction, the opposite of what was found at 523 K after low-temperature;131 there was no change in the activation energy for the ethane reaction (197 kJ mol−1).

TABLE 13.13. Selected References to the Modification of Groups 8 to 10 Metals for Alkane Hydrogenolysis by Oxides Causing the SMSI

a When the support is put in brackets, it was optionally used to support the modifier.

Apparent SMSI effects have also been reported with non-Transition Metal oxide supports.119 Those obtained with magnesia were almost certainly due to evolution of traces of hydrogen sulfide, or of iron ions, from the bulk at the high temperatures used,132 but real effects have been seen with silica when very high pre-treatment temperatures were employed:41,133,134 the rate of the ethane reaction decreased more quickly than that of hydrogen chemisorption, pointing clearly to the need for multiatomic sites in hydrogenolysis.

Studies of n-butane have been somewhat more informative (Table 13.13). It has been thoroughly established that the effect of the SMSI (whatever it is) can be partially or even completely restored by oxidation, and a sequence of reduction (HTR1) . . . oxidation and low-temperature reduction (O/LTR) . . . reduction (HTR2) applied to supported and modified ruthenium catalysts has produced results that extend our understanding both of the SMSI and of those described in the last section.103 Application of the oxides of the first row Transition Metals (Ca through Mn) to Ru/SiO2 led124 to the results shown in Figure 13.21. HTR1 and HTR2 produced a particularly severe depression of the n-butane hydrogenolysis rate with vanadia, with smaller but still very significant effects with chromia and manganic oxide, not fully restored by oxidation, due perhaps to the formation of stable mixed oxides. However, in each case the rate was enhanced by O/LTR and

Figure 13.21. Hydrogenolysis of n-butane on 1% Ru/SiO2 modified by oxides of Groups 2 to 7: rates (mmol gRu−1 h−1 ) at 433 K.123

decreased by HTR2 (the effects with Ru-Ti/SiO2 were smaller than those reported before because the precursor was not calcined before HTR1). Arrhenius parameters for reactions after O/LTR fell on a separate higher line than those for the HTRs, as before. Various generally minor changes to product selectivities were seen, so that evidence for electronic or geometric alteration to the active centre was absent (except perhaps with Ru-V2O5/SiO2 after HTR1), and it seemed possible that residual activity was due to a few particles that had escaped unscathed from the assault of the SMSI.

Pre-coating supports (SiO2, Al2O3 and TiO2) with various amounts of V2O5, followed by ruthenium from RuC13, gave similar but somewhat clearer results.122 In the n-butane reaction, high values of F and T3 were found after the HTRs, but were lowered by O/LTRs and partially restored by HTR2s. It appears that this method of composing the catalysts gave a closer interaction between metal and modifier than that obtained when the reverse sequence was used. Analogous but less detailed results have been found with Rh/V2O3 and Rh-V2O3/SiO2, and corresponding platinum catalysts.135,136 HTR of Pt/CeO2 caused an increase in activation energy, a decrease in rate and a significant rise in methane selectivity.137 HTR of Rh/TiO2 and of Ir/TiO2 virtually eliminated activity for hydrogenolysis, but allowed isomerisation and dehydrogenation to be seen at 688 K.90,138 Deposition of titania onto Pt/SiO2 (EUROPT-1) decreased rates of both hydrogenolysis and isomerisation of n-butane progressively,139 the latter more than the former, and both more quickly than H/Pt, so that activity was almost killed while H/Pt had fallen only to 0.75. The formation of ‘carbon’ and of titania modifiers (e.g. TiO) on the metal had complementary effects, both suppressing isomerisation

selectivity and to a smaller extent increasing S2 and F. These effects are not explicable in terms of hydrogen atom availability (Schemes 13.3 and 5), but require the idea of a site of different structure at which the intermediates for isomerisation can be accommodated. It therefore seems that some aspects of particle-size and ensemble-size effects can be equally well or perhaps better explained by specific site requirements than by adsorbed hydrogen concentration (Section 13.5). However, reaction of n-butane on Pt/TiO2 affords a lower S2 than does Pt/SiO2, even after reduction at only 573 K, and, with Pt/TiO2-SiO2 containing a large amount of titania, reduction at 773 K almost eliminates central C C bond-breaking, due to restriction of site size.125 There will be further news of the effects of SMSI and modifiers in Chapter 14, when reactions of larger alkanes are considered, and when in consequence a wider range of mechanistic options becomes available.

It was surprising to find that Os/TiO2 did not behave in the same way as Ru/SiO2:102 O/LTR caused rates to decrease, and S2 to fall quite considerably in the reactions of both propane and n-butane. With Ir/TiO2,102 on the other hand, the O/LTR treatment did cause the rate of the n-butane reaction to increase, although the effect with different metal contents was not consistent, but there was no significant change in selectivity parameters throughout the various pre-treatments, suggesting that iridium once reduced could not be oxidised as was ruthenium. To obtain the effect, it was not necessary to use titania as the support; a small amount ( 1 wt.%) on silica or alumina produced equivalent changes as pre-treatments were altered.103

Two other forms of modification deserve to be noted. A series of platinumcontaining smectite-like clays have been exchanged with the divalent cations Ni2+, Co2+and Mg2+, and their behaviour in n-butane hydrogenolysis compared with that of Pt/SiO2;140 the results are summarised in Table 13.14. The first two were more active than the Pt/SiO2, and the activation energies were the exact inverse of the activities. Excellent compensation was shown, but the slope of the line was greater than that of the mean slope in Figure 13.3. As reaction temperature necessarily rose, so did both Si and S2, suggesting that decreasing hydrogen coverage was the main factor; this would also cause activation energies to fall (see Figure 13.16). It seems likely that the platinum catalysed the reduction of the Ni2+ and Co2+ions (but not Mg2+ions), which then contributed to the activity.

TABLE 13.14. Hydrogenolysis of n-Butane on Pt-containing Smectic-like Clays (SM) and on Pt/SiO2 140

The addition of potassium to Ru/Al2O3 has produced quite dramatic effects on propane hydrogenolysis.141 The changes produced by O/LTR and HTR2 were abolished, and the almost constant behaviour was characterised by larger values of KA and bH, and by smaller values of Et and − Ha , although S2 values changed marked with hydrogen pressure. A convincing explanation of these effects is not yet available, but resistance to the effects of oxidation may be the result of the formation of KRuO2. The effect of potassium on ethane hydrogenolysis catalysed by Ru/SiO2 was to decrease activation energy and increase hydrogen order.142 Inclusion of potassium into palladium143 and platinum on LTL zeolite, and onto Pt/SiO2,144 inhibited hydrogenolysis of propane, and it also interfered with the process of hydrogen spillover.

When pyridine was added to n-butane during its hydrogenolysis over Ru/SiO2, the rate was suppressed, but (as with sulfur poisoning123) isomerisation then became visible.145 Pyridine also scavenged hydrocarbon radicals, which were detected as their pyridine adducts.

13.8.HYDROGENOLYSIS OF ALKANES ON BIMETALLIC CATALYSTS24,146−148

13.8.1. Introduction

Of the modifiers considered in the previous section, most if not all were believed to exercise their influence while remaining in a positive oxidation state. With those about to be considered there is good evidence that in the main they are in the zero oxidation state, forming real bimetallic catalysts. It is possible, indeed likely, that in some cases not all of the modifier is metallic;149 some may remain on the support (e.g. in the Pt-Re system), but even with elements not easily fully reduced it appears that bimetallic particles may be formed (e.g. Pt-Mo,150,151 Pt-Zr152). In principle however the effects produced are expected to be similar,153 i.e. a lowering of the mean size of the active ensemble, with the possibility in some cases of electronic modification as well. The main additional feature is the chance that a bimetallic site will be of comparable activity to that composed only of ‘active’ atoms; this chance rises to near-certainty when the two metals are drawn from within Groups 8 to 10.

Its reputation for structure-sensitivity has made alkane hydrogenolysis attractive for investigation using bimetallic catalysts, and there is an extensive literature on the subject. Much of it however relates only to the ability of a pair of metals to form and retain bimetallic particles, and adds little to the understanding of reaction mechanisms. Sometimes for example the calcination of two precursor compounds gives a binary oxide that is easily reduced to a bimetal,126,154 while at other times reduction of the precursors gives the desired product, but oxidation undoes the