CHAPTER 1
Catalysis by Single-crystal Surfaces
BY R.W. JOYNER
1 Introduction
Judged by two important criteria, the field of catalysis by single-crystal surfaces has prospered significantly since the earlier reviews in this series, the later of which covered the period to 1977. The number of studies has increased to an undeniable extent and these form almost the whole matter of this report. Little room has been left to detail other areas of 'Surface Science' that might be of interest to the catalytic practitioner. This Reporter also believes that the quality and sophistication of information in the area of single-crystal catalysis has improved markedly since 1977 and that a more detailed understanding of many important catalytic processes is being achieved. The remainder of this report attempts to put a gloss onto these statements, each section dealing with an individual reaction or class of reactions. Sections are designed, if the reader desires, to be read independently. Each opens with an account of advances in our knowledge of adsorption, structure, theory, or other topics of relevance. Investigations of the reaction in question are then outlined. No attempt is in general made to describe the experimental techniques used or to discuss detailed interpretation of results. A short glossary of acronyms is given in Table 1. Nearly all of the reactions considered are of industrial or technological significance and most of the catalysts described are metallic. This reflects the much greater ease of preparation of metal single crystals vis à vis large oxide or sulphide single crystals. Catalysis by zeolites, now of immense significance in catalytic cracking, has, however, been excluded by reason of space and lack of competence of the reviewer.
There are three reasons why catalysis by single crystals can usefully be reviewed as a sphere of activity apart from catalysis in general. The first is that crystal structure is itself an influence on catalytic rate. However, as Boudart has reminded us changing the surface structure usually changes the rate of reaction by less than a factor of twenty, while changing the metal may alter the rate by many orders of magnitude. A second point is that single crystals are somewhat easier to characterize than supported catalysts, where the role of surface and bulk phase transformation and the influence of impurities may be more insidious. The main justification for isolating single-crystal catalysis for study is that it is often performed in conjunction with some of the formidable array of surface science techniques listed in Table 1. The information obtained can also be related to that obtained in studies of adsorption and structure of single crystals. A more detailed and fundamental understanding of the reaction of interest may therefore be expected.
The reactions of hydrocarbons are considered in three groups, the first involving synthesis, and the second dealing with isomerization and hydro- genolysis. The third is involved largely with oxidation reactions and reactions of oxidized hydrocarbons. CO and NO oxidation are then examined, followed by NH3 synthesis and oxidation.
2 Hydrogen Reactions
Synthesis – Successive oil crises have focused attention on forthcoming shortages of liquid hydrocarbons. A consequence of this has been a revived interest in the Fischer–Tropsch (FT) synthesis [equation (l)], of which methanation, [equation (2)] is a special case. Fe, doubly promoted by K2O and Al2O3
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
CO + 3H2 [right arrow] CH4 + H2O (2)
The adsorption of both reactants on transition metals has been much studied. H2 adsorption on Pt, as evidenced by the rate of H2-D2 exchange, is structure sensitive. H2 adsorption has been shown to cause metal surface reconstruction, rather than simple 'on top' chemisorption, of the (001) face of W, a surface which itself has been shown to reconstruct at low temperature.
Of major significance for our understanding of the FT reaction has been the recognition that CO adsorption may be dissociative on many surfaces. First suggested by King et al. in 1972, the unequivocal demonstration of CO dissociation was an early success for photoelectron spectroscopy studies of surfaces." The involvement of dissociated CO in FT synthesis on supported CO catalysts has been firmly established by the ratio-tracer studies of Biloen et al. and on Fe by the work of Perrichon et al.
Those metals which dissociate CO have been identified by Rhodin et al. as shown in Figure 1, and the relation of dissociation to the existence of stable carbides has been noted." The existence of dissociative CO adsorption on Ru has not been established by XPS or UPS. On the close-packed (001) plane Menzel et al. suggest that electron bombardment is required. However Singh and Grenga have shown C build-up resulting from CO dissociation at edge sites on a spherical single crystal. McCarty and Wise have demonstrated that isotopic mixing occurs in CO above a supported Ru catalyst at 350K.
There is clear evidence for structural sensitivity in CO dissociation on some metals. On Ni dissociation was observed on evaporated films and ion-bombarded single crystals, but not on a well annealed (111) surface. Detailed confirmation of these early results comes from a study of CO adsorption on Ni(111) and the stepped plane [5(111) x (110)]. Temperature-programmed desorption from the stepped plane showed a CO peak at ~ 820 K in addition to that observed from the low-index surface, at 430 K. Erley and Wagner attribute this higher temperature peak to activated recombination of C and 0 adatoms formed by CO dissociation. Erley et al. give further proof of dissociation in an HREELS study. Figure 2 shows the loss spectrum resulting from the adsorption of 0.4 Langmuir of CO at 180 K. The peak at 1520 cm-1 appears before those at 1890 and 2010 cm-1, and is assigned to CO adsorbed at the steps. 1520 cm-1 represents a very low C=O stretching frequency and is in the range associated by Nguyen and Sheppard with CO triply bonded to the surface. Erley et al. suggest that this species is the precursor of the dissociatively adsorbed species indicated in curve (b) of Figure 2, observed after heating the surface to 430 K. This spectrum shows no peak assignable to C=O stretching, but does indicate a peak, at 550 cm-1 which is assigned to Ni–O and/or Ni–C vibration. Auger electron spectroscopy (AES) showed that both C and O remained on the metal surface at this stage.
Structure sensitivity is also observed in CO dissociation on Fe and Co. Absorption is dissociative at 300 K on the open Fe(100) and (111) planes, but is molecular on the close-packed (110) face. Even on the (110) face, dissociation occurs on heating the absorbed layer. Enhanced dissociation relative to the low-index face occurs on a stepped (111) Fe surface. Lambert et al. have shown that CO adsorption is dissociative at low pressures (< 10-4Torr) on the open Co(102) plane, but not on the basal (001) plane.
There have been numerous investigations of molecular CO adsorption on metals, and salient references have been reviewed by Bradshaw. Points that have been established include the orientation of the adsorbed molecule on Ni(100), the assignment of the valence orbital photoelectron spectrum, and the role of lateral interactions in the adlayer. These results are of greater relevance for CO oxidation and methanol synthesis catalysis than for the FT reaction. An intriguing result regarding CO–H2 coadsorption on Rh(111) is nevertheless worth mentioning. Williams et al. argue, from LEED data, that repulsive interactions between adsorbed CO and H2 cause segregation into islands of each species at 100K. This repulsive interaction is effective at distances up to 3 Å and can be expected to inhibit CO–H2 reactions on Rh, compared to metals such as Ir and Pt where a weak attractive interaction may occur. Controversy exists regarding the existence of CO dissociation on Rh. Somorjai et al. claim to have detected this on stepped surfaces whereas Williams et al. have demonstrated that the ratio of dissociation to desporption does not exceed 10-4.
Goodman et al. report an interesting CO-H2 coadsorption phenomenon on Ni(100). Coadsorption at exposures > 1L and T = 77 K produced new desorption peaks for each gas at about 200 K. This is interpreted as indicating the irreversible formation of an adsorbed CO–H complex. The limiting concentration ratio observed was 6H–CO, suggesting that the complex involves a CO molecule surrounded by adsorbed H atoms.
Extensive studies of CO hydrogenation over single-crystal metal catalysts have been carried out by two groups, at National Bureau of Standards, Washington DC, and at Kernforschungsanlage (KFA), Julich, W. Germany. In both cases a high-pressure (1 bar) reactor was connected directly to an u.h.v. chamber equipped for surface characterization. The NBS group of Goodman, Kelley, Madey, and Yates have studied CO hydrogenation over single-crystal Ni and Ru. On Ni their results are summarized in Figure 3. At Ptotal = 120 Torr and H2 CO = 4, linear Arrhenius plots are observed spanning 5 orders of magnitude in rate and for temperatures between 450–850 K. As can be seen, very similar specific rates, or turnover numbers, are observed on (100) and (111) planes. The speccific rates are in good agreement with those reported for supported Ni catalysts by Vannice.
Deviations from the Arrhenius law occur at lower total pressures. Where the Arrhenius law is obeyed, rate is reported to be independent of total pressure. Goodman et al. have also observed that the rate of reaction is the same on the (110) and (001) faces of Ru.
Goodman et al. have used AES to characterize the C-containing species present on the surfaces of interest. Unusually, in AES the C (KLL) peak shape shows marked chemical sensitivity. On a catalytically active Ni(100) crystal the C peak shape resembles that from Ni carbide, the steady-state concentration being typically 20% of a carbide monolayer. When graphitic carbon is deposited on the surface, the rate of methane formation is inhibited. It is suggested that the surface carbide may be hydrogenated, perhaps existing under reaction conditions as surface –CH groups.
Goodman et al. have studied the rate of C build-up on a Ni surface, by reaction with 24 Torr of CO. Their results are illustrated in Figure 4; similar observations were also made on Ru.
The detailed and systematic investigations of these workers are valuable, important, and intriguing, particularly in view of the structure sensitivity of CO dissociation mentioned above. They add further weight to the strong body of evidence that FT reaction proceeds largely through a mechanism involving dissociated CO on active catalysts. It is also gratifying that similar rates to those noted on supported catalysts can be observed. This is a tribute to the care taken by the NBS group in these very difficult experiments, although its significance for catalytic mechanism may be limited. Clearly strong interaction with the supports can be ruled out in those catalysts discussed by Vannice. We may, however, question the validity of the comparison since the pressure responses of the single-crystal catalysts and the supported catalysts are markedly different. Zero-order kinetics were observed on the single crystals, whereas on the supported Ni materials the kinetics have the form shown in equation (3), where 0.8 < x< 1.4 and -0.9 < y < -0.2.
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
Bonzel et al., at KFA Julich, have studied CO hydrogenation on single-crystal Fe surfaces. They have concentrated on the analysis of carbon layers present on the crystals after reaction, and have applied XPS in addition to AES. After reaction at a total pressure of 1 bar (H2:CO ratio typically 20:1) for different times, three carbonaceous phases were identified from XPS peak positions and AES peak shapes. These were assigned as: Phase 1, -CHx, i.e., CH, CH2, CH3, or possibly a polymeric species; Phase 2, a carbidic carbon phase containing some H; Phase 3, a graphite carbon, possibly containing some intercalated H.
Studies on polycrystalline Fe show that the first two phases could rapidly by hydrogenated in 1 bar H2, at 560 K. These results demonstrate a similar pattern to those discussed above, on Ni crystals.
Bonzel et al. have investigated the influence of adsorbed K on the Fe(110) surface. K2O is a common promoter in commercial FT catalysts. At 298 K, LEED indicates the presence of an h.c.p. K layer, with a maximum density of 5.3 x 1014 atoms cm-2 K deposition, as would be expected, causes a fall in the work function, which reaches a minimum value of ~1.5 eV at a K coverage of 3 x 1014 atoms cm-2 and thereafter rises slowly with increasing K coverage. XPS indicates changes in the core level binding energies characteristic of charge transfer from K to Fe. UPS studies indicate that CO is more tightly bound to the Fe–K surface than to the clean Fe(110) plane. Figure 5 shows that no molecularly adsorbed CO remains on the Fe surface after heating to 390 K, while CO is still detectable on the Fe-K surface. The size of the O(2p) and C(2p) derived peaks indicates that the extent of dissociative CO chemisorption is also greater on the Fe(110)–K surface. Bonzel et al. discuss the hypothesis that electropositive promoters act by weakening the C–O bond in the adsorbed CO molecule and suggest, as the converse, that electronegative promoters should inhibit FT synthesis. This is true for the well known case of S poisoning, where, in agreement with many studies of supported catalysts, Goodman and Kiskinova have shown that synthesis is markedly inhibited by <0.1 monolayer of S. Bonzel et al. indicate, however, that adsorbed O does not inhibit FT synthesis on Fe. Indeed Dwyer and Somorjai suggest that O2 adsorption increases the rate of FT synthesis on Fe foils.
One element in the FT synthesis, which has so far not been much studied by surface science techniques, is the nature of the catenation or polymerization step in the formation of ethane, ethene, and higher hydrocarbons. Understanding this step may be of key importance in the design of selective FT catalysts of the type that will increasingly be required. Dwyer and Somorjai studied the effect of adding ethene or propene to the feed gas for the H2–CO reaction over Fe at 573 K, with Ptotal = 6 bar and H2 :CO = 3:1. Addition of <3% molar of these gases markedly reduced methane production and increased C4 or C5 synthesis rates. Dwyer and Somorjai argue that readsorption and secondary reaction of olefins may be an important route to higher hydrocarbons in FT synthesis. This view is in accord with the known importance of support pore-size distribution and reactant contact time in influencing the product distribution of commercial Fischer–Tropsch catalysts.
Hydrocarbon Adsorption and Reaction. – The adsorption and structure of simple hydrocarbon layers on metals such as Cu, Ni, Pt, and Pd has received a great deal of attention. In particular UPS and HREELS have demonstrated considerable potential in the examination of the molecular events involved. Although many of the arguments are not resolved, the area is of sufficient importance for hydrocarbon catalysis to justify a reasonably detailed progress report. We consider firstly the weakly chemisorbed layers known to form when these metals are exposed to monolayer doses of ethene or ethyne at 80 K. Figure 6, from work by Demuth, shows the U.V. photoelectron spectra (UPS) observed for ethene on Cu(111) taken with He I (21.2 eV) and He II (40.8 eV) radiation. The close correspondence between the spectrum of the gas phase and the adsorbed layer (allowing for differences in reference levels) shows that the molecule has not been severly perturbed, for example, by dissociation. Interest however centers on the extent to which the bonding of the ethene molecule is distorted by adsorption. Demuth has addressed this question by examining small changes in valence orbital binding energy occurring on adsorption. His previous studies' have considered the extent to which adsorption rehybridized the molecule. He now uses the relative changes in binding energy of the σ orbitals of the adsorbed species to examine its geometric distortion from its gas-phase equilibrium configuration; numerical values are listed in Table 2. As can be seen, very small changes result from adsorption on Cu(111), in contrast to Pt(111), where shifts in binding energy of 0.8 eV are noted (1 eV = 96 kJ mol-1). Demuth has used a self-consistent molecular orbital calculation (Gaussian-70, ab initio SCF–LCAO) to determine which changes of the ethene geometry best reproduce the shifts observed in the σ molecular orbitals, which are assumed not to interact directly with the metal substrate. Distortions considered are reduction of the C–C–H bond angle from the gas-phase value of 120° to 107°, H–C–H angle reduction from 120° to 106°, and expansion of the C–C bond length from 1.34 Å to 1.54 Å. The best fits obtained from the comparisons are shown in Table 2. Greatest distortions from equilibrium geometry are suggested for adsorption of ethene on Pt and Pd. By contrast, and perhaps surprisingly to the chemist, much smaller distortions are predicted for ethyne on all of the metals studied.
