Metal-Catalysed Reactions of Hydrocarbons / 06-Exchange of Alkanes with Deuterium

Figure 6.6. Inversion of chemisorbed cyclopentane molecule via αα-diadsorbed species: (A) the attached carbon atom is sp2 hybridised; (B) it is sp3 .

effects, and (iv) use of bimetallic catalysts. By far the most work has been done with (i), so this will be reviewed first.

The cyclopentane molecule is almost planar, and the C H bonds are held in eclipsed configuration: it is therefore quite reasonably deemed impossible to form a 1, 2-diadsorbed state by removal of hydrogen atoms from opposite sides of the ring. The cyclohexane ring is more flexible, but in its exchange at low temperatures a maximum at -d6 is still observed, although qualitatively it is less marked than the -d5 maximum in cyclopentane: inversion to allow exchange of the second set is therefore easier, but it is not until the ring is enlarged to seven or eight carbon atoms2,73 that the barrier disappears and uninhibited total exchange occurs. One means of securing inversion of an adsorbed cyclopentyl radical is to assume that an αα-diadsorbed species is formed; if this is planar about the attached carbon atom, as has been assumed, then reversion to cyclopentyl may occur by addition of deuterium at either side, opening the way to total exchange (Figure 6.6). Difficulty of forming the αα species accounts for the higher activation energy shown by two-set exchange, and this of course parallels the behaviour of methane (Section 6.2.). The argument remains unaffected if, as seems probable on the basis of information presented in Chapter 4, the attached carbon atom retains its sp3 state, forming bonds to two different metal atoms or sites (Figure 6.6).

At least three other modes of inversion have received consideration. (1) The diadsorbed alkene may just desorb before responding to the thermodynamic urge to re-adsorb, which after flipping over will allow second-set exchange to proceed: the high activation energy might reflect the endothermic nature of the desorption.

(2) The diadsorbed alkene in the di-σ form might ‘roll over’ to another pair of sites without actually desorbing (Figure 6.7). (3) The adsorbed alkene in the π -form might lose a further hydrogen atom, becoming a π -alkenyl species (see Table 4.2); topside addition of a deuterium atom by an Eley-Rideal step would then lead to inversion and complete exchange as shown in Figure 6.8.

The trouble with all these four proposals is that each can be (and has been) defended on chemical grounds, but can also be criticised. We may note that

Figure 6.9. Transfer of exchange to the second set in cyclopentane via the methyl group: the σ -diadsorbed form is made by loss of the two indicated hydrogen atoms.

could extend to the methyl group through the formation of an eclipsed diadsorbed structure (Figure 6.9), giving maxima at the d8-species. However this latter process was harder than the one-side exchange, and had a higher activation energy. Ring inversion was easier for molecules that first adsorbed on the unhindered side, because interference between the methyl group and the surface encouraged formation of the intermediate responsible for inversion. Understanding of the subtleties of exchange in polymethylcyclopentanes may be tested by attempting to predict the maxima in product distributions formed from the structures shown in Table 6.8 over palladium film at 313 to 353 K.4,76

There is a further reaction that proceeds under exchange conditions and requires a mechanism similar to that for cycloalkane inversion: this is epimerisation, exemplified by conversion of Z -1,2-dimethylcyclohexane to the E -isomer72,77 (Figure 6.10). As with racemisation of branched alkanes, this also demands a planar intermediate, or some other mode of reconfiguration of a substituted carbon atom, but clearly an αα species cannot be formed at this point, and one of the other mechanisms must apply. These reactions can of course be observed using hydrogen, but more information is available when deuterium is used. With Z -1,2- dimethylcyclopentane, eleven of the fourteen hydrogen atoms were exchangeable, but only seven in the E -isomer (Figure 6.10). Epimerised molecules were however mainly perdeuterated, showing that the change of configuration involves the same intermediate as inversion. The common intermediate for obvious stereochemical reasons is more likely to form the E - than the Z -isomer. Over Pt/Al2O3

TABLE 6.8. Methyl-Substituted Cyclopentanes

Exchanged with Deuterium over Palladium Films

1,1,3-Trimethyl - E-1,1,3,4-Tetramethyl - Z-1,1,3,4-Tetramethyl - 1,1,3,3-Tetramethyl - 1,1,3,3,4-Pentamethyl -

Figure 6.10. Epimerisation of Z- to E-1, 2-dimethylcyclopentane.

catalysts having dispersions between 5 and 65%, the TOF for epimerisation of Z -1,2-dimethylcyclohexane at 453 K was essentially constant, i.e. the reaction was structure-insensitive.77

Further attempts to resolve mechanistic ambiguities required the use of polysubstituted cycloalkanes78 and polycyclic molecules of Byzantine complexity, the synthesis of which must have taxed the ingenuity of research students, and helped to fill many a PhD thesis. One example is 1,1,3,3-tetramethylcyclohexane which contains an isolated trimethylene group in which on palladium film only five of the six hydrogen atoms are exchangeable;65 this is good evidence for a π-alkene- π-alkenyl mechanism. The study of fused-ring cycloalkanes has also been a fertile field for mechanism discrimination. The argument here revolves around the ability of the alkene formed in the exchange to adopt a π structure that is almost coplanar, and thence a π-alkenyl or mono-adsorbed alkene, as against its retaining the di-σ form and tetrahedral stereo-chemistry, in which ‘roll over’ can operate. It has to be remembered that these debates were held before the spectroscopic and other structural studies mentioned in Chapter 4 were available: these would suggest that a π -alkene would be the preferred form at low coverage and low temperature, but only at high coverage at high temperature. This would suggest that under most of the conditions used in exchange experiments the di-σ form would be stabler than the π form.

It is however worth citing a few of the results obtained with fused-ring cycloalkanes. In norbornane (bicyclo[2.2.1]heptane) there are only two pairs of hydrogen atoms in eclipsed confirmation, but these are isolated by the bridge-head methyne groups, and so αβ exchange gives only the -d2 product (Figure 6.11A). With bicyclo[3.3.1]nonane, however, if one of the cyclohexane rings adopts the boat conformation, the bridgehead hydrogen atoms eclipses with an adjacent methylene group and hence the exchange of eight hydrogen atoms becomes possible72,79 (Figure 6.11B). Further maxima at -d10 and -d12 may be due to exchange of two atoms in each of the trimethylene groups. Only stepwise exchange is possible with adamantane, because the two cyclohexane rings are fused in the chair conformation (Figure 6.11C). With bicyclo[3.3.0]octane, however, there is no hindrance to inversion, and -d8 and -d14 maxima have been seen80 (Figure 6.12D); replacement of one of the bridge-head hydrogen

Figure 6.11. (A) Norbornane (bicyclo[2.2.1]heptane): one of the pairs of eclipsed and therefore exchangeable hydrogen atoms is shown.

(B)Bicyclo[3.3.1]nonane: one pair of eclipsed hydrogen atoms is shown, and exchange of eight atoms becomes possible.

(C)Adamantane: all methylene groups are isolated, and only stepwise exchange can take place.

(D)Bicyclo[3.3.0]octane: one hydrogen in each of the methylene groups adjacent to the bridge-head hydrogen is eclipsed with it, so all hydrogen atoms are exchangeable.

(E)Endo-trimethylenenorbornane: exchanges only five hydrogen atoms. Find them!

(F)Heptacyclotetradecane: shows only stepwise exchange.

(G)Tricyclo[2.2.0]decane: exchange ten hydrogen atoms.

atoms by a methyl group limits the extent of the exchange to seven and eleven atoms, a result which clearly favours the di-σ adsorbed intermediate. Endo- trimethylenenorbornane (Figure 6.11E) exchanges only five hydrogen atoms, not more. Further attempts to distinguish between di-σ -adsorbed alkene and π-alkene intermediates hinged on finding molecules that should have a pair (or pairs) of eclipsed C H bonds but which could not by any stretch of the imagination allow the relevant part to adopt the nearly flat configuration demanded of an adsorbed π-alkene. Such weird and wonderful molecules as heptacyclotetradecane and tricyclodecane (Figure 6.11E and F) have also been obtained and used. The former only showed stepwise (not αβ) exchange, perhaps because of steric interference by the methylene bridge; the latter exchanged up to ten hydrogen

atoms, but it is difficult to envisage any of the carbon atoms changing to the sp2 state or anything close to it. The unbiased observed might conclude that if there is really a conflict of interpretation (a point we come back to shortly), the idea of the diadsorbed alkane intermediate as presented by Bob Burwell has defeated by π-alkene intermediate mechanism favoured by John Rooney–but only on points.

Having exhausted the inspiration obtainable by tinkering with molecular structure, we must turn for further insights to the use of other approaches. Little use has been made of kinetic analysis, as has been done so extensively (and profitably) with ethane and methane, orders of reaction when determined being expressed just as exponents of reactant pressures.50,81 Exchange distributions observed with cyclohexane have been interpreted as for ethane,4 but dependence of the variables on physical properties of the metals used has not been pursued. It has often been noted that above some critical temperature around 373 K there is hydrogenolysis of the cycloalkane, which above about 473 K leads to deactivation by ‘carbon’ deposition.43,68,69,81,82 This is more quickly suppressed by addition of an inert metal (Cu to Ni;68 Sn or Au to Pd30), and many (but not all) deliberately introduced poisons affect multiple exchange more than stepwise77 (the exceptions are SnCl2 and PbEt4). While the inert metals quickly depress the rates of exchange processes30,68 (although a small addition of gold to palladium can cause an increase, as is seen with other reactions63), there has been no attempt to quantify these results in terms of active ensemble size. One is however left with the impression that the needful size decreases as

‘carbon’ deposition > multiple exchange > stepwise exchange.

This sequence does not however point uniquely to any of the conflicting mechanisms.

The tendency to inversion (i.e. two-set exchange) with palladium is unaffected by addition of gold, but is actually increased by adding tin and is decreased by dissolved hydrogen:30 it was concluded that raising the energy of the palladium 4d states weakens the metal-alkene bond, while lowering it had no effect. Where the bimetallic pair comprises two active metals (Pt-Ir;29 Pd-Ni31), the results approximate to the weighted average of the two components: thus iridium favoured stepwise and pairwise exchange of cyclopentane to -d1 and -d2, and platinum preferred multiple exchange to -d4 and -d5, while nickel tended to give more pairwise and less total exchange than palladium. High -temperature pre-reduction (HTR) of Rh/TiO2 created the SMSI state (Section 3.53), but while this lowered the rate it did not affect the multiplicity of the exchange, which showed much cyclopentane-d1 and -d2, and little -d10. It did however reduce to multiplicity shown by Pt/TiO2.25

Cyclopentane exchange has been used to ‘count’ the number of exposed metal atoms in alumina-supported metals: rates at 373 K were about the same for Pt/A12O3 and Re/A12O3, but no exchange past -d4 was found with the latter.83

Rates were however about twice as fast with PtRe/A12O3, and the exchange pattern was intermediate between those of the pure metals. In the case of Pt/A12O3, specific (areal) rates of stepwise exchange of cyclopentane were almost independent of particle size, increasing only three-fold as it increase from 1.3 to 17.5 nm;77 rates of multiple exchange increased by a factor of about 20. The areal epimerisation rate for 1,2-dimethylcyclohexane was also independent of particle size with this catalyst.52 Similar results have been found with Pt/SiO269,84 and Ni/SiO2,50 although the type of pre-treatment has a significant effect with the former.

We have therefore only a very incomplete matrix of isolated sets of results from which to deduce anything about site requirements for exchange reactions. It seems that stepwise and perhaps the single αβ steps need no special site, a conclusion that does not rule out Frennet’s mechanism for stepwise methane exchange (Section 6.2.1), but that multiple exchange and inversion appear to need a larger site. It is however very difficult to disentangle the effects of support, particle size and reaction temperature arrive at any generally valid conclusions.

More recent work by Rooney’s group85,86 and others87 has however revealed further complexities to those which already exist in the behaviour of supported metal catalysts for exchange reactions. Low-temperature reduction (LTR) in the reactants after pre-oxidation completely removed inversion of cyclopentane on palladium film and on Pd/γ -A12O3, but HTR at 573 K restored it, at least in part; it was thought that inversion was somehow prevented by electron-deficient sites such as Pd2+ ions near palladium atoms, or low co-ordination number atoms (Table 6.9): but HTR of palladium on other aluminas or on titania did not show the -d10 products of inversion. Similar effects were shown by platinum on acidic supports, although Pt/MgO showed much more total exchange than Pt/SiO2 or Pt/A12O3: the rate was however much smaller. The difficulty with observations of this sort is that the effects cannot be definitively connected with any physical property of the catalysts (electron deficiency, presence of adventitious poisons, (S)MSI etc.) because of the lack of appropriate characterisation. They do however appear to operate with rhodium, as well as palladium and platinum; and unusual catalytic behaviour of metals on magnesia or influenced by other basic components will be encountered again in a later chapter.

Small-ring cycloalkanes have not escaped attention. Cyclopropane underwent exchange as well as hydrogenation on rhodium film at 173 K,4 as did cyclobutane82 and ethylcyclobutane4 on nickel catalysts at about 430 K, although in both these cases ring-opening occurred at the same time. All hydrogen atoms were exchanged on nickel, a fact having some mechanistic significance, as it shows that inversion is possible. On palladium it did not occur with methylcyclobutane, however, a maximum of six hydrogen atoms being exchangeable, and with 1,1- dimethylcyclobutane multiple exchange appeared not to occur at all;8 these last observations are not mutually consistent. Endocyclic alkenes are most unlikely to be formed in C4 rings, so other mechanisms of exchange must apply.

TABLE 6.9. Parameters of Cyclopentane Exchange with Deuterium over Palladium Catalysts Variously Reduced, and over Supported

Platinum Catalysts

Notes:

a Similar results were obtained by reduction at the reaction temperature.

bBoehmite but may transform to Al2 O3 under reaction conditions.

6.4.INTERALKANE EXCHANGE

There have been occasional reports that exchange can take place between normal and deuterated molecules in the absence of hydrogen or deuterium, thus:

The process took place on Ni/Cr2O3 between 373 and 525 K and must require the dissociative chemisorption of the methanes followed by random recombination: not surprisingly, deactivation was found at all temperatures.4 The reaction between normal and deuterated n-butane has also been studied on a number of supported platinum catalysts.4 Little or no exchange was seen between light methylcyclohexane and deuterated n-octane (or the reverse) over Pt/SiO2 at 755 K either in the presence or absence of hydrogen or deuterium because of rapid conversion to other molecules; reaction did however take place between them at lower temperatures, and on Pt/A12O3 it occurred both on the metal and the support.88

6.5. CONCLUSIONS

Almost all the work mentioned in this Chapter was performed before about 1975, so that either metal films or powders (‘blacks’) or supported metals were of necessity used. Thus attempts to assess the sensitivity of exchange processes

either to particle size, or to ensemble size in bimetallic systems, have yielded only results of qualitative significance, and measurement of particle size by TEM and the possible presence of accidental poisons have been largely ignored. Quantitative kinetic analysis has also been omitted (except for methane and ethane), so that it is difficult to connect exchange multiplicity and tendency to inversion to any independently measured property of the catalysing metal. The only study using a single crystal surface (Pt(111)) has been rewarding in the sense that distinct processes are shown to occur through different adsorbed structures on a uniform surface rather than through a variety of sites. It is strange that, for example, cyclopentane exchange has not yet been examined on the low-index faces or stepped surfaces of palladium or platinum (or any other metal): such a study would surely be most informative.

When two very intelligent people inspect the same body of experimental facts and proceed to draw quite different conclusions as to their cause, one might think that the truth lies somewhere in the middle or that both are partly right and partly wrong. We must therefore ask whether the alternative views expressed particularly mechanisms of cycloalkane exchange represent a genuine conflict or whether they are essentially synthetic. Rooney’s ideas were originally based on analogies with organometallic complexes, which as we have seen in Chapter 4 are indeed useful models for adsorbed states that are formed under certain conditions. However the fact is that a particular form, say the π-alkene, exists in mono-nuclear complexes such as Zeise’s salt does not necessarily mean it will also be formed on metal surfaces, where there are many atoms crying out to have the free valences neutralised. The regions of temperature and surface coverage in which this form are stable is restricted. We now realise too that the σ-diadsorbed and the π -adsorbed alkene are only extreme representations of a whole range of intermediate forms, the tendency towards one or the other depending on the chemistry of the surface and probably on the nature of the adsorbed molecules as well. Some of the recent very precisely determined structures of adsorbed species (Section 4.5) are not easily rationalised in terms of emergent molecule orbitals, and theoretical explanations for the structures adopted are still awaited. Applications of this knowledge to reaction mechanisms require these to be inspected on similar surfaces, and it is of little use to the traditional forms of metal catalyst mainly used for alkane exchange, with their evident proliferation of types of surface site.

In conclusion, we may feel that it has been an error to try to force all the observations of cycloalkane exchange into a single mould, and to require all molecules (and surfaces) to conform to a single type of mechanism. Many (but not quite all) the results are explicable by π-alkene, π-alkenyl and mono-adsorbed alkene structures; opposition to the view that σ -diadsorbed alkane structures predominate seems to have been based on intuition rather than on hard fact, and the lack of any attempt to identify adsorbed structures by spectroscopic methods (perhaps difficult when surface concentrations are low) has not helped to resolve the difference of