Surface Science Studies of Catalysed Reactions on Platinum Surfaces
Surface Science Studies of Catalysed Reactions on Platinum Surfaces
The Role of the Atomic Structure of Platinum Single Crystals in Determining Selectivity
Advances in platinum catalyst technology applied to hydrocarbon conversion have resulted in substantial improvements in the selectivity and durability of the processes. However, our understanding of the mechanism of the complex reactions involved, and in particular of their relationship to the physical and chemical properties of the catalyst, remains incomplete. This paper reviews important work carried out over the last ten years on the role of atomic sites on platinum surfaces and also presents for the first time a molecular model of the working platinum catalyst surface.
Platinum occupies a unique place in chemistry and in chemical technology. Its chemical stability at high temperatures and in hostile environments makes it an outstanding material for a wide range of applications but probably its most important use is as a catalyst to accelerate chemical reactions and, when several reactions occur simultaneously, to catalyse the most desirable one selectively. A considerable proportion of the annual output of platinum is used in the manufacture of catalysts for two major applications: to convert the mixture of hydrocarbons found in petroleum to high octane number gasoline, and in the catalytic converter that is attached to an automobile to aid the conversion of carbon monoxide and unburned hydrocarbons in the exhaust gas to non-polluting carbon dioxide and water vapour. These two applications which greatly affect our well-being demonstrate that platinum is an outstanding catalyst in both oxidising and reducing environments. Indeed, this metal comes close to being the all purpose catalyst that will perform with high activity and selectivity when used under very diverse circumstances.
Platinum, when suitably prepared, selectively converts the straight-chain saturated hydrocarbons (alkanes) to branched and cyclic organic molecules (1). These types of reactions are exemplified by the conversion of n-hexane, as shown in Figure 1. The reactions leading to branched isomers (isomerisation) and cyclic molecules (dehydrocyclisation) are especially desirable for producing high octane gasoline from petroleum naphtha. This mixture of organic molecules containing five to ten carbon atoms is rich in low octane number alkanes, and the reaction is often called “reforming” to underscore its importance. The third type of hydrogenolysis reaction that involves the scission of carbon-carbon bonds yields low molecular weight gaseous products and is, therefore, undesirable when producing gasoline. All the reactions proceed in excess hydrogen at high pressures of 1 to 50 atmospheres and at temperatures in the range of 550 to 800 K. Since both dehydrocyclisation and isomerisation have relatively high activation energies (in the range of 25 to 45 kcal/mole), they proceed at a higher rate as the temperature increases (1). This consideration is counterbalanced by the higher rate of deactivation or “cokeing” that occurs with increasing temperature, due to the formation of an unreactive carbonaceous deposit that necessitates the frequent regeneration of the catalysts, an expensive process indeed. Platinum, an excellent “reforming” catalyst, is employed in large reforming units, each containing up to 3000 pounds of platinum. Each unit may convert up to 5 × 104 barrels (42 gallons/barrel) of naphtha into 100 research octane number (RON) gasoline daily. Efficient utilisation of the expensive platinum catalyst is achieved by depositing the metal from solution onto a high surface area oxide support such as silica or alumina. While catalysts prepared in this manner may contain only about 0.3 per cent platinum by weight, the effective surface area of the dispersed platinum particles, which average about 10 to 20 Ångströms in diameter, often approaches one square metre per gram of catalyst; which means that essentially every metal atom is exposed at the surface.
How does platinum work as a selective catalyst? The answer has been obtained by the atomic scale scrutiny of the metal surface by several techniques that were developed over the past 10 to 15 years by modern surface science. In response to the demands of the electronic and computer technologies where devices were developed with an ever-increasing surface-to-volume ratio, a very large number of techniques has been developed for studies of the outermost layer of atoms at the surfaces of solids and of the near monolayer surface region (1). As the devices became smaller in size their performance was controlled more and more by charge transport in the surface region which in turn depended on the structure and composition of the surface. Thus, the analysis of these surface properties and the ability to tailor them became essential. A partial listing of the surface analysis methods that have been developed is provided in the Table. The composition of the surface and the oxidation states of surface atoms can be determined with a sensitivity of 1 per cent or better and since the outermost surface layer of most solids contains about 1015 atoms/cm2, the presence of less than 1013 impurity atoms/cm2 or of other types of atoms is readily detectable. The locations of surface atoms or of molecules adsorbed on the surface can be determined together with their bond distances and bond angles to their neighbours. Also the binding energies and energy distribution of electrons in the surface and in the chemical bonds among the atoms in the surface can be determined.
To utilise surface science techniques for studies of the problems of heterogeneous catalysis we had to develop a model catalyst configuration that permits us systematically to vary its atomic surface structure and its surface composition. Then a reaction chamber had to be constructed in which the catalyst could be subjected to the usual chemical reaction conditions of high pressures and temperatures in the range 400 to 900 K so that we could monitor the reaction rates and the distribution of the reaction products. Simultaneously, we must be able to analyse the atomic structure and composition of the working catalyst surface to relate these atomic scale catalytic properties to its reactivity and chemical selectivity (2, 3).
We found that the various crystal faces of platinum single crystals with surface areas of about 1 cm2 and 1 mm thick serve as excellent model catalysts (4, 5, 6). These samples can be prepared with quite uniform and ordered surface structures that can be analysed by the various surface science techniques. Low energy electron diffraction was particularly useful for the determination of the structure of single crystal surfaces as long as they were ordered. The flat surfaces, where each platinum atom is surrounded by six and four nearest neighbours, respectively, are two closest packed platinum crystal faces of the highest atomic density, (see Figure 2). Stepped crystal faces displaying close-packed terraces several atoms wide, and separated by atomic steps one atom high, can also be prepared easily. The lowered coordination of the step atoms is responsible for the unique chemical activity that is often displayed at these surface sites. There can be kinks in the steps, and atoms at these ledges have even lower co-ordination. The structure and concentration of steps and kinks, along with the structure and width of the terraces, can be varied by cutting the platinum single crystals along different crystal planes, and then by appropriately polishing and etching them to remove the surface damage introduced by the mechanics of surface preparation (7, 8, 9).
High Vacuum-High Pressure Chamber
Combined catalytic studies and chemical and structural analysis of the small area single crystal surfaces are carried out in a stainless steel chamber, shown in Figure 3, that can be evacuated to 10−9 torr or less. In these circumstances the surface remains free from unwanted impurities that may adsorb from the surroundings, for thousands of seconds. The chamber is equipped with viewports and with various surface analytical techniques to permit characterisation and cleaning of the platinum crystal surface. Then, by enclosing the sample in an isolation cell, various gases or liquids are introduced and the chemical reaction is induced by heating the catalyst to the desired reaction temperature. The reaction mixture is circulated by a pump and is periodically sampled by a gas chromatograph or other suitable chemical composition detector. Outside the isolation cell all of the surface analytical techniques are kept in vacuum since gas leakage from the cell is negligible. The reaction can be interrupted at anytime by evacuating and opening the isolation cell, and the catalyst, again in vacuum, can be readily analysed. After surface analysis, or appropriate modifications of the surface structure or composition, the isolation cell is closed and the reaction studies are restarted. Thus reaction rates and product distributions are related to the working structure and composition of the model catalyst surface (10).
One important question we may ask is: how long does it take for the reaction to “turn over”? In fact what distinguishes catalytic reactions from stoichiometric reactions is that many product molecules are produced on the surface as reaction sites are continually generated following the desorption of the products. The turnover rate for dehydrocyclisation at 700 K and at 1 atmosphere hydrogen/hydrocarbon reaction mixture is about 10−3 per surface site per second, if we assume that each platinum surface site is active, which is not a very good assumption. The surface is then producing one molecule per site in each 103 seconds, a slow but not atypical conversion rate. However, this may be viewed as a lower limit of activity that may be increased by an order of magnitude if only 10 per cent of the surface sites are active. Multiplying the turnover rate with the reaction time gives the turnover number and this must be greater than unity for a reaction to be called catalytic. The slow turnover rate indicates long residence times of the order of seconds for the molecules on the surface, and this has two important consequences:
 Within their residence time the molecules may visit many surface sites and may undergo subsequent rearrangement or chemical bond breaking at these sites. The surface species may diffuse over a distance of 10−5 to 10−4cm before leaving the surface.
 Since the surface is continually covered with a monolayer of reacting molecules, any incident reactant molecule has a very low probability of finding an empty surface site to adsorb on. The reaction probability that is obtained by dividing the turnover rate by the rate of incidence of the molecules, obtained from knowledge of the pressure, is of the order of 10−9 at 1 atmosphere (11). Thus high surface area catalysts are needed to achieve high conversion of reactants to products in reasonable times. Because turnover rates at platinum surface sites are of the order of 10−3 to 10−1 per second, even with the one square metre of platinum surface area per gram of commercial reforming catalyst, an acceptable conversion of 60 to 70 per cent needs the use of about 10 pounds of catalyst per daily barrel of naphtha.
How does the reaction rate depend on the atomic structure of the catalyst surface? To answer this question, reaction rate studies using flat, stepped and kinked single crystal surfaces with variable surface structure were very useful indeed. For the important aromatisation reactions of n-hexane to benzene and n-heptane to toluene, we discovered that the hexagonal platinum surface, where each surface atom is surrounded by six nearest neighbours, is three to seven times more active than the platinum surface with the square unit cell (12). Aromatisation reaction rates increase further on stepped and kinked platinum surfaces. Maximum aromatisation activity is achieved on stepped surfaces with terraces about five atoms wide with hexagonal orientation, as indicated by reaction rate studies over more than ten different crystal surfaces with varied terrace orientation and step and kink concentrations. These are represented in Figure 4.
The reactivity pattern displayed by platinum crystal surfaces for alkane isomerisation reactions is completely different from that for aromatisation. Our studies revealed that maximum rates and selectivity (rate of desired reaction/total rate) for butane isomerisation reactions are obtained on the flat crystal face with the square unit cell (13, 14). Isomerisation rates for this surface are four to seven times higher than those for the hexagonal surface and are increased only slightly by surface irregularities (steps and kinks) on the platinum surfaces, as is shown in Figure 5.
For the undesirable hydrogenolysis reactions that require C-C bond scission we found that the two flat surfaces with highest atomic density exhibit very similar reaction rates. However, the distribution of hydrogenolosis products varies sharply over these two surfaces. The hexagonal surface displays high selectivity for scission of the terminal C-C bond, whereas the surface with a square unit cell always prefers cleavage of C-C bonds located in the centre of the reactant molecule. The hydrogenolysis rates increase markedly (3 to 5-fold) when kinks are present in high concentrations on the platinum surfaces (12, 15).
Since different reactions are sensitive to different structural features of the catalyst surface, we must prepare the catalyst with an appropriate structure to obtain maximum activity and selectivity. The terrace structure, the step or kink concentrations, or a combination of these structural features, is needed to achieve optimum reaction rates for a given reaction. Our studies indicate that H-H and C-H bond-breaking processes are more facile on stepped surfaces than on the flat crystal faces, while C-C bond scission is aided by ledge sites that appear to be the most active for breaking any of the chemical bonds that are available during the hydrocarbon conversion reactions (16). Since molecular rearrangement must also occur in addition to bond breaking, it is not surprising that the terrace structure exerts such an important influence on the reaction path that the adsorbed molecules are likely to take (17, 18). The difference in chemical behaviour of terrace, step and ledge atoms arises not only from their different structural environment but also from their different electronic charge densities that result from variation of the local atomic structure. Electron spectroscopy studies reveal altered density of electronic states at the surface irregularities; there are higher probabilities of electron emission into vacuum at these sites (lower work function) indicating the redistribution of electrons (1).
One of the important attributes of transition metal surfaces is that they atomise large binding energy diatomic molecules (hydrogen or oxygen) by forming strong M-H or M-O bonds, and hold the atoms in high surface concentrations so that they are readily available during the surface reaction. The hydrogen atom surface concentration is especially important in permitting a catalysed hydrocarbon conversion reaction to proceed unimpeded. The presence of excess hydrogen facilitates removal of product molecules and also inhibits catalyst deactivation. For this reason the reforming reactions of organic molecules are always catalysed in the presence of excess hydrogen.
While many reactions of organic molecules are catalyst structure sensitive, there are some that show no structure sensitivity: the ring opening of cyclopropane to propane (4) and the hydrogenation of cyclohexene to cyclohexane (6) are two structure insensitive reactions when carried out on various metal surfaces.
What is the composition of the working platinum catalyst surface? When the surface is examined after carrying out any one of the hydrocarbon conversion reactions, it is always covered by a near monolayer amount of carbonaceous deposit (14). This is not surprising since we interrupted a hydrocarbon chemical reaction to determine the composition, and the adsorbed molecules could have stayed on the surface. In this case the adsorbate would stay only as long as the time needed for one turnover. Conversely, the surface may be partly covered by a more tenacious deposit that stays on the metal surface during many turnovers. In order to determine the surface residence time of the carbonaceous deposit, the platinum surface was dosed with 14C labelled organic molecules under the reaction conditions (19). Carbon-14 is a β particle emitter and using a β particle detector we monitored its surface concentration as a function of time during the catalytic reaction. Using another surface science technique, thermal desorption, we could determine the hydrogen content of the adsorbed organic layer by detecting the amount of desorbing hydrogen with a mass spectrometer. From these investigations we found that the residence time of the adsorbed carbonaceous layer depends on its hydrogen content, which in turn depends on the reaction temperature as shown in Figure 6. While the amount of deposit does not change much with temperature, its composition does; it becomes much poorer in hydrogen as the reaction temperature is increased (14, 20). The adsorption reversibility decreases markedly with increasing temperature as the carbonaceous deposit becomes more hydrogen deficient. As long as the composition is about CnH1.5n and the temperature below 450 K, the organic deposit can be readily removed in hydrogen and has residence times in the range of the turnover time. As this deposit loses hydrogen with increasing reaction temperatures (>450 K), it converts to an irreversibly adsorbed deposit with composition C2nHn that can no longer be removed readily (hydrogenated) in the presence of excess hydrogen.
Nevertheless, the catalytic reactions proceed readily in the presence of this active carbonaceous deposit. Above 750 K this active carbon layer is converted to a graphitic layer that deactivates the metal surface and all chemical activity for any hydrocarbon conversion reation ceases. Hydrogen exchange studies indicate rapid exchange between the hydrogen atoms in the adsorbing reactant molecules and the hydrogen in the active but irreversibly adsorbed deposit. Only the carbon atoms in this layer do not exchange. Thus one important property of the carbonaceous deposit is its ability to store and exchange hydrogen.
How is it possible that the hydrocarbon conversion reactions exhibit great sensitivity to the surface structure of platinum, while under the reaction conditions the metal surface is covered with a near monolayer of carbonaceous deposit? In fact, often more than a monolayer amount of carbon-containing deposit is present as indicated by surface science measurements. In order to determine how much of the platinum surface is exposed and remains uncovered, we utilised the adsorption and subsequent thermal desorption of carbon monoxide. This molecule, while readily adsorbed on the metal surface at 300 K at low pressures, does not adsorb on the carbonaceous deposit (14). Our results indicate that up to 10 to 15 per cent of the surface remains uncovered while the rest of the metal surface is covered by the organic deposit. The fraction of uncovered metal sites decreases slowly with increasing reaction temperature. The structure of these uncovered metal islands is not very different from the structure of the initially clean metal surface during some of the organic reactions, while thermal desorption studies indicate that the steps and ledges become preferentially covered in others (14).
As a result of our catalysed hydrocarbon conversion reaction studies on platinum crystal surfaces a model has been developed for the working platinum reforming catalyst and is shown in Figure 7. Between 80 and 95 per cent of the catalyst surface is covered with an irreversibly adsorbed carbonaceous deposit that stays on the surface for times much longer than the reaction turnover time. The structure of this carbonaceous deposit varies continuously from two-dimensional to three-dimensional with increasing reaction temperature and there are platinum patches that are not covered by this deposit. These metal sites can accept the reactant molecules and are responsible for the observed structure sensitivity and turnover rates. While there is evidence that the carbonaceous deposit participates in some of the reactions by hydrogen transfer by providing sites for rearrangement and desorption while remaining inactive in other reactions, its chemical role requires further exploration.
Building Improved Platinum and Other Metal Catalysts
The atomic scale ingredient of selective hydrocarbon catalysis by platinum has been identified and a model of the working catalyst constructed. Attention now turns toward building improved catalyst systems. Additives are being used to alter beneficially the surface structure, to reduce the amount of carbon deposit, or to slow down its conversion to the inactive graphitic form. Bimetallic or multi-metallic platinum catalyst systems have been developed by the addition of one or more other transition metals (palladium, iridium, rhenium or gold) that can be operated at higher reaction temperatures to obtain higher reaction rates. They show slower rates of deactivation (have longer lifetimes) and can also be more selective for a given chemical reaction (dehydrocyclisation or isomerisation) than the one-component catalyst.
One of the major challenges in preparing scientifically tailored high technology metal catalysts is to deposit the metal particles with the specific surface structure needed to obtain optimum reaction selectivity. The structure of the support and its chemical interaction with the metal is utilised to achieve this goal: deposition of ordered platinum monolayers on sulphides or oxides with well-defined substrate structure is one important approach in this direction. Zeolites, aluminosilicates that are available with variable well-defined pore structure and aluminium : silicon ratios, could perhaps provide the structural definition that was obtained on the low surface area single crystal catalysts without sacrificing the availability of high surface area. Strong chemical interaction between the metal particles and the support induces charge transfer toward or away from the metal that again could beneficially alter its catalytic properties. Other additives are being investigated to increase catalytic activity by decreasing the surface residence times required for reaction and product desorption to take place, thereby reducing the amount of platinum required in conventional reforming catalysts.
As combined surface science and catalytic reaction studies develop working models for catalysts of other types, the building of new “high technology” catalysts, using this molecular level understanding, will become more frequent. This transition from art to catalysis science cannot come soon enough. The rapidly rising cost of petroleum necessitates the use of new fuel sources such as coal, shale and tar-sand, and the use of new feedstocks for chemicals including CO + H2 and coal liquids. The new fuel and chemical technologies based on these new feedstocks require the development of an entirely new generation of catalysts. Ultimately, our fuels and chemicals must be produced from the most stable and abundant molecules we live with on our planet, among them carbon dioxide, water, nitrogen and oxygen. To build the catalytic chemistry starting from these species is a considerable challenge that, we believe, will be met by catalysis science in the future.
G. A. Somorjai, “Chemistry in Two Dimensions: Surfaces”, Cornell University Press, 1981
D. W. Blakely,, E. I. Kozak,, B. A. Sexton and G. A. Somorjai, J. Vac. Sci. Technol., 1976, 13, ( 5 ), 1091
A. L. Cabrera,, N. D. Spencer,, E. Kozak,, P. W. Davies and G. A. Somorjai, Rev. Sci. Instrum., in press
D. R. Kahn,, E. E. Petersen and G. A. Somorjai, J. Catal., 1974, 34, ( 2 ), 294
B. A. exton and G. A. Somorjai, J. Catal., 1977, 46, ( 2 ), 167
S. M. Davis and G. A. Somorjai, J. Catal., 1980, 65, ( 1 ), 78
R. W. Joyner and G. A. Somorjai, The Chem. Soc. Publ., 1973, 2, 1
B. Lang,, R. W. Joyner and G. A. Somorjai, Surf. Sci., 1972, 30, ( 2 ), 440
D. W. Blakeley and G. A. Somorjai, Surf. Sci., 1977, 65, ( 2 ), 419
G. A. Somorjai, Adv. Catal., 1977, 26, 1
S. M. Davis, and G. A. Somorjai,, “The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, 4, Fundamental Studies of Heterogeneous Catalysis”, ed. D. A. King and D. P. Woodruff, Elsevier, 1982, 217 – 364 .
W. D. Gillespie,, R. K. Herz,, E. E. Petersen and G. A. Somorjai, J. Catal., 1981, 70, ( 1 ), 147
S. M. Davis,, F. Zaera and G. A. Somorjai, J. Am. Chem. Soc., 1982, 104, ( 26 ), 7453
S. M. Davis,, F. Zaera and G. A. Somorjai, J. Catal., 1982, 77, ( 2 ), 439
R. K. Herz,, W. D. Gillespie,, E. E. Petersen and G. A. Somorjai, J. Catal., 1981, 67, ( 2 ), 371
G. A. Somorjai, Surf. Sci., 1979, 89, 496
R. J. Gale,, M. Salmeron and G. A. Somorjai, Phys. Rev. Lett., 1977, 38, ( 18 ), 1027
M. Salmeron,, R. J. Gale and G. A. Somorjai, J. Chem. Phys., 1977, 67, ( 11 ), 5324
S. M. Davis,, B. E. Gordon,, M. Press and G. A. Somorjai, J. Vac. Sci. Technol., 1981, 19, ( 2 ), 231
G. A. Somorjai,, J. E. Crowell,, R. J Koestner,, L. M. Dubois and M. A. Van Hove, “Recent Advances in Analytical Spectroscopy”, Proc. 9th Intl. Conf. on Atomic Spectroscopy and 22nd Colloquium Spectroscopicum Internationale, Tokyo, 1981, ed. K. Fuwa, Pergamon, 1982, 211