3176
Langmuir 1991, 7, 3176-3182
The Flexible Surface. Correlation between Reactivity and Restructuring Ability Gabor A. Somorjai Department of Chemistry, University of California, Berkeley, and Materials Sciences Division, Lawrence Berkeley Laboratory, Berkeley, California 94720 Received September 3, 1991. In Final Form: October 29, 1991
Studies of the surface structure of metals on the atomic scale indicate the restructuring the clean surfaces with respect to their bulk atomic structure. Inward relaxation and reconstruction are more pronounced at steps, at open surfaces, and at low coordination sites. Chemisorption causes adsorbateinduced restructuring which drastically changes the surface structure. Such restructuring events occur on the time scale of chemisorption,on the time scale of catalytic reactions, and on a longer time scale that leads to redispersion and sintering. Thermodynamic considerations indicate more facile adsorptioninduced restructuring for more open surfaces where atoms have fewer nearest neighbors (flexiblesurfaces) while close packed surfaces are less likely to restructure (rigid surfaces). The dynamical restructuring of surfaces on the atomic scale during chemisorption and catalytic reactions explains several puzzles of surface chemistry and heterogeneous catalysis. In surface chemistry,these include (1)the transition from physisorption to chemisorption, (2) why rough surfaces are more active in chemical bond breaking and in catalysis, (3) epitaxy,and (4) unusual coverage-dependentchangesof bonding. In heterogeneouscatalysis the dynamic model of surfaces may also shed light on why certain reactions are structure insensitive and on the unique activity of bimetallic and oxide-metal interface systems.
Introduction This paper is dedicated to Art Adamson, the surface scientist, the photochemist, and, first and foremost, the educator. Among the modern surface science techniques, lowenergy electron diffraction (LEED) crystallography, field ion microscopy (FIM), scanning tunneling microscopy (STM), and the atomic force microscope (AFM), along with X-ray photoelectron diffraction (XPD),are well suited and have been utilized for studies of the atomic surface structure at the solid-vacuum and solid-gas interfaces. As a result, our views of surfaces have markedly changed over the past few years. When we consider surface phenomena, we have a model of the surface in mind that was developed based on experimental observations. Historically, the first model of the surface was one of smooth, structureless discontinuity developed in the 1930s (Figure l).l The electrons spill out by tunneling because there are less or no repelling negative charges outside the solid or on the vacuum side, leaving behind a smeared out positive charge. This jellium model explained well the properties of the surface space charge, why the work function varied from crystal face to crystal face, and how adsorbed atoms or molecules modify the work function. Such a homogeneous surface model was adequate to explain the Langmuir adsorption isotherm and dissociative chemisorption and the behavior of electrode surfaces in electrochemistry. In the 19505, the electron microscopy and FIM studies revealed the presence of steps and kinks at surfaces and the smooth surface model was replaced by the heterogeneous rigid lattice model' shown in Figure 2. The surface is heterogeneous; the various sites are distinguishable by the number of their nearest neighbors. The surface atoms are located at the same sites where atoms in the bulk are. Therefore, from knowledge of the bulk crystal structure, the location of the surface atoms can be predicted. This heterogeneous surface model served us well for at least 2 decades. I t explained the kinetics of ( 1 ) Somorjai, G. A. In Chemistry in Two Dimensions: Surfaces, Corne11 University Press: Ithaca, NY, 1981; Library of Congress Catalog Card NO. 80-21443.
Friedel electronic chorge density, nh) positive J background I
I
-4i
-2%
0
2%
x
Figure 1. Homogeneous surface model. TERRACE
/
-
STEP ADATOM
Figure 2. Heterogeneous rigid surface model. crystal growth and evaporation, the site dependence of chemisorption and variations in the work function and space charge with step density. Surface crystallography studies of recent years, both on clean surfaces and in the presence of chemisorbed monolayers, clearly indicate that the heterogeneous rigid lattice model is incorrect. When the surface is clean the atoms in the topmost layer relax inward to produce a shorter interlayer spacing.2 The rougher the surface (roughness is defined as l/packing density), the larger is the inward relaxation.3 This is shown in Figure 3. The relaxation is especially large at steps and kinks where the atoms move inward4to smooth out the surface irregularity as much as possible (Figure 4). (2) Wang, S. W.; Ogletree, D. F.; Van Hove, M. A.; Somorjai, G. A. Adu. Quantum Chem. 1989,20, 2-147. ( 3 ) Jona, F.; Markus, P. M. In The Structure of Surfaces; SpringerVerlag: Berlin, 1988; p 80. (4) Zhang,X.-G.; Van Hove, M. A.; Rous, P. J.; Tobin, D.; Gonia, A.; MacLaren, J. M.; Heinz, K.; Michl, M.; Lindner, H.; Meller, K.; Eheasi, M.; Block, J. H.; Somorjai, G. A. Submitted for publication in Phys. Rev. Lett.
0743-7463/91/2407-3176$02.50/0 0 1991 American Chemical Society
Langmuir, Vol. 7,No. 12,1991 3177
Flexible Surfaces
%C(lll)
flat
-
most rigid
ICC
open
-
I1101
less rigid
110
+
5
1 2 .
’
3*
l
4-
1
*5
I
*6
1
.7 1
J
ROUGHNESS
Figure 3. Experimental and theoretical first-layer relaxation (in % ) as a function of roughness (=l/packing density) for several bcc and fcc surfaces.
Icc (331)
stepped
- less flexible
cluster
-
most flexible
Figure 5. Four surfaces distinguishable by the coordination number of surface atoms. Atoms in the fcc (111)surface have the largest and atoms in the cluster have the smallest number of nearest neighbors. “Cracks” open close to the step edges
Each atom attempts to optimize its coordination
Figure 4. Restructuring at a step site on a clean metal surface.
Thus, the surface atoms are not rigid and are not located at the sites that are predicted from knowledge of the bulk structure. The surface is flexible and its inward relaxation is larger the fewer number of neighbors the surface atom has. We may divide surfaces according to their “flexibility” as shown in Figure 5. Those close packed, low Miller index surfaces with large coordination numbers are the most rigid. The more open rough or stepped surfaces are less rigid or more flexible, and small clusters are likely to be the most flexible. In the past, several structural configurations of small clusters have been known to have the same thermodynamic stabilities. The relaxation of surface atoms often leads to reconstruction because of the directionality of their chemical bonds. The surface assumes a long-range structure that produces surface unit cells that are very different from the projection of the bulk unit cell to that surface. Reconstructions of many insulator and metal surfaceshave been reported including those in Si, Ge, GaAs, Pt, Au, and Ir crystal surfaces.5~6
Chemisorption-InducedRestructuring of Surfaces It was believed that once chemisorption occurs, the inwardly relaxed surface atoms will be displaced outward to assume their bulklike rigid lattice positions. This does not happen, however. The substrate atoms relocate when chemisorption occurs, but they move into positions to optimize their bonding to the adsorbate atoms or mole~~
(5)Van Hove, M. A.; Weinberg, W. M.; Chan, C. M. Low Energy Electron Diffraction; Springer-Verlag: Berlin, 1986. (6) Van Hove, M. A.; Somorjai, G. A. Prog. Surf. Sci. 1989, 30, 201.
cules. Below we describe the adsorption-induced restructuring of several chemisorption systems.7 In the presence of a quarter of a monolayer of carbon, the (100) face of nickel is restructured in such a way that the four nickel atoms surrounding the carbon atom move outward and rotate by a small angle with respect to the underlying layer.8 This occurs to maximize the bonding to the carbon in the 4-fold site which can bond better this way to the nickel atom in the second layer underneath in addition to bonding to the four nearest neighbor surface nickel atoms. The restructured surface is shown in Figure 6. Oxygen chemisorption on the Cu(ll0) faceg produces a unit cell in which the 0 and Cu atoms are mixed in an ordered arrangement (Figure 7). STM studies could follow the growth of the domains of the new surface structure at the expense of the clean copper domains.1° Sulfur chemisorbed on the (110) iron surface creates its own 4-fold site [ 111by causing the restructuring of the surface to periodic four iron atom-one sulfur atom clusters (Figure 8). Hydrogen chemisorption restructures most transitionmetal surfaces.12 One of these structures, H on Rh(ll0) is shown in Figure 9, as a function of coverage. Ethylene chemisorption on the Rh( 100) crystal face causes restructuring even on this closest packed surface13 in a manner shown in Figure 10. Ethylene is converted to ethylidyne (C2H3) at 300 K that adsorbs with its C-C (7) Somorjai, G. A. In Adsorbate Induced Restructuring of Surfaces. The Surface Thermodynamic Puzzle. Bonding Energies and the Thermodynamics of Organometallic Reactions; American Chemical Society: Washington, DC, 1990; Chapter 15. (8) Overferko, J. H.; Woodruff, D. P.; Holland, B. W. Surf. Sci. 1979, 87, 357. (9) Coulman, D. J.; Wintherlin, J.; Behm, R. J.; Ertl, G. Phys. Rev. Lett. 1990, 64, 1761. (10) Woell, C.; Wilson, R.J.; Chang, S.; Zend, H. C.; Mitchel, K. A. R. Phys. Rev. B: Condens. Matter 1990,42,11926. (11) Shih, H. E.; Jona, F.; Marcus, P. M. Phys. Rev. Lett. 1981,46,731. (12) Nichtl-Pecher,W.; Oed, W.; Landskron, H.; Heinz, K.; Miiller, K. Vacuum 1990,41,297. (13) Wander, A.; Van Hove, M. A.; Somorjai, G. A. Phys. Rev. Lett., in press.
3178 Langmuir, Vol. 7,No. 12,1991
Figure 6. Carbon chemisorption induced restructuring of the nickel (100)crystal face.
bond perpendicular to the metal surface and with the C atoms bound to three Rh atoms.14 As a result of the formation of the chemisorption bond, the Rh-Rh distance increases around the C atoms which pushes the next nearest neighbor Rh atom more into the surface (by 0.15 A). Thus the surface becomes corrugated. At the same time the Rh atom in the second layer directly under the molecule moves upward, presumably to bond better to the adsorbed molecule. This restructuring is similar to that observed for carbon on the Ni(100) surface except in this circumstance, the molecular adsorbate occupies a 3-fold binding site for carbon instead of the 4-fold site on nickel.
Catalytic Reaction Induced Restructuring of Surfaces The Pt(100) surface is reconstructed to produce a hexagonal arrangement of metal atoms when clean or when covered with chemisorbed oxygen.15 When CO chemisorbs on this crystal face, the surface restructures and exhibits a square unit cell.15 During CO oxidation in certain ranges of partial pressures and temperatures, the reaction rate shows periodic oscillation.16 It has been shown that this is due to periodic alternation between two branchesof the reaction: one that takes place on the mostly oxygen-covered surface and the other on the mostly COcovered Pt surface. Since the work function changes when the surface is CO covered as compared to when it is 0 covered, photoemission electron microscopy can monitor the relative population of the 0-and CO-covered patches of the Pt surface a t low reactant pressures Torr).16 It was found that the surface restructures with a rate that is equal to the catalytic reaction turnover rate. The restructured areas expand and contract as the reaction moves in a wavelike fashion along the surface. Thus, the surface restructuring drives the catalytic reaction or, conversely, the catalytic surface reaction drives the rate of the surface restructuring.17 The same oscillatory reaction rates can also be observed at higher CO and 0 2 pressures, near 1atm or higher, for CO oxidation and on the Pt(ll1) or other metal crystal surfaces, none of which show surface reconstruction when clean.18 In these circumstances the reaction seems to be driven by the partial oxidation and reduction of the surface (14)Koestner, R.J.; Van Hove, M. A.; Somorjai, G. A. Surf. Sci. 1982, 121,321. (15)Van Hove, M.A.; Koestner, R. J.; Stair, P. C.; Biberian, J. P.; Kesmodel, L. L.;Somorjai, G. A. Surf. Sci. 1981,103,189. (16)Ertl, G. Ber. Bunsen-Ges. Phys. Chem. 1986,98,284. (17)Ertl, G. Catal. Lett. 1991,9,219.
Somorjai
as the reaction rates oscillate between the O-rich surface branch and the CO-rich surface branch.lg The oscillation in this case is monitored by the periodic temperature changes that accompany this exothermic reaction and the surface composition by Auger electron spectroscopy (AES). Let us assume that adsorbate-induced surface restructuring is associated with catalytic reaction turnovers even when such reactions are in their steady state. The rate of restructuring as the reactant adsorbs or the product desorbs controls then the reaction turnover rate. The faster the rate of periodic restructuring, the higher the activity of these sites. It is likely that low coordination sites that are more “flexible” can restructure a t higher rates and are the active sites for catalytic reactions. Rapid periodic restructuring may cause structural instability, however. The challenge of catalyst design is to produce active clusters of great structural flexibility but well anchored and stabilized against self-destruction. In order to strengthen the experimental evidence for surface restructuring driven catalytic reactions, we must develop time-resolved spectroscopic techniques that operate under high-pressure reaction conditions and can monitor the periodic desorption of adsorbates along with the periodic alteration of the local atomic structure around the binding site.
Puzzles of Surface Chemistry That Can Be Explained by the Flexible Surface Model There are many interesting and puzzling phenomena that are associated in a unique way with the behavior of surfaces and interfaces. Some of these will be listed below. They can all be rationalized or explained by the flexible surface model. Thermal Activation. The adsorption of reactive molecules on reactive surfaces at low enough temperatures does not result in bond breaking or in molecular rearrangements. As the temperature is increased, however, the breaking of a selected bond will take place at a welldefined temperature for a given adsorbate-substrate system. This was first observed in the 1930s when molecular adsorbates, 0 2 or N2, were found to undergo dissociative chemisorptionat certain temperatures, which was called the physisorption to chemisorption transition. For hydrocarbons, sequential loss of hydrogen and simultaneous molecular rearrangement occurs at wellcharacterized temperatures on a transition-metal surface with a given surface structure.20 These bond scission processes occur at much lower temperatures at surface irregularities, steps, and kinks; that is, on more flexible surfaces as compared to close packed rigid surfaces. Recentlya molecular mechanism has been proposed that considers surface restructuring as the driving force for dissociative chemisorption. This model also predicts a coverage dependence for the phenomenon as the chemisorption of more molecules should lower the activation energy for the surface restructuring.21 Rough Surfaces Do Chemistry. Surface irregularities, steps, and kinks are very effective for breaking adsorbate chemical bonds and in catalysis as well. This is best shown by the temperature programmed desorption (TPD) of Ha from flat, stepped, and kinked surfaces of Pt (Figure 11). H2 desorbs a t maximum rates at the highest temperature1 from kink sites, then a t somewhat lower temperatures from step sites, and at even lower tem(18)Yeates, R.C.;Turner, J. E.; Gellman, A. J.; Somorjai, G. A. Surf. Sci. 1985,149 (l),175. (19)Somorjai, G. A. Catal. Lett. 1990,7, 169. (20)Salmeron, M.;Somorjai, G. A. J. Phys. Chem. 1982,86,341. (21)Levine, R. D.;Somorjai, G. A. Surf. Sci. 1990,232,407.
Langmuir, Vol. 7,No. 12,1991 3179
Flexible Surfaces (a)
(b)
lo011
I
100 A
b -,
[iiol
100A
(2x1)-O/Cu( 110)
Figure 7. (a, left) Oxygen chemisorptioninduced restructuring of the copper (110)crystal face. (b,right) Scanningtunneling microscopy picture of the oxygen adsorption induced restructuring process.
1x3-H 8-03
e - os 1x2- H
-
1~2-3H 8 1.5 Figure 8. Sulfur chemisorption induced restructuring of the iron (110) crystal face.
peratures from the flat (111) terraces. This indicates higher heats of adsorption of the H atom at these defect sites. Thus, the thermodynamic driving force for dissociation is certainly greater at these sites, which can explain their enhanced bond-breaking activity. It is difficult to understand, however, that these same strongly adsorbing sites are also very active sites for catalysis. This is shown in Table I. The reaction probability of H2/D2 exchange on stepped surfaces is near unity at low pressures on a single scattering while the reaction probability is below the detection limit on the flat (111)crystal face as shown by molecular beam surface scattering studies.22 How is it possible that the strongly adsorbingstep sites, where H has a long residence time because of its high binding energy, are also the sites of rapid reaction turnover? One possible explanation is that the strongly adsorbed hydrogen restructures the surface near the step, thereby creating the active site for the catalytic exchange process. At the low pressures of these molecular beam scattering experiments, the low coverages keep the structure of the flat part of the surface unaltered. (22) Salmeron, M.; Gale, R. J.; Somorjai, G. A. J. Chem. Phys. 1979, 70 (6), 2807.
[OOl I
Figure 9. Hydrogen chemisorption induced restructuring of the rhodium (110)crystal face. Note the hydrogen coverage dependent changes of surface structure.
Epitaxy. When a monolayer of metal is deposited from the vapor phase onto the ordered surface of another metal, the condensed metal layer assumes the periodicity of the underlying substrate. This phenomenon is called epit a ~ y Epitaxial .~~ monolayers form for most metal on metal (23) Nix, R. M.; Somorjai, G. A. In Concepts in Surface Science and Heterogeneous Catalysis; Perspectives in Quantum Chemistry;Jortner/ Pullman: 1989.
Somorjai
3180 Langmuir, Vol. 7, No. 12, 1991
Different ethylidyne species: bond distances and angles (rc =carbon covalent radius; rM = bulk metal atomic radius)
0.12A
o.ioA
i
l u
-
1
-
1
-
C [AI
m
rM
rc
uI"l
C03 KO), CCH3
1.53 (3)
1.90(2)
1.25
0.65
131.3
H3 Ru3 KO), CCH3
1.51 (2)
2.08 (1)
1.34
0.74
128.1
H3 Or3 (COIg CCHj
1.51 (2)
2.08 (1)
1.35
0.73
128.1
Pt (111) + ( 2 X 2) CCH3
1.50
2.00
1.39
0.61
127.0
Rh (111) + (2 X 2) CCH3
1.45 (10)
2.03 (7)
1.34
0.69
130.2
H3C CH3
-
1.54
0.77
109.5
H2C = CH2
1.33
0.68
122.3
HC I CH
1.20
0.60
180.0
Rh (1 11) - (2x2) - C2H3
Figure 10. (a, left) Ethylene chemisorption induced restructuring of the rhodium (111) crystal face. (b, right) Structure of ethylidyne chemisorbed on the rhodium (111)and platinum (111)crystal faces and in metal cluster compounds.
t
01 0
Potassium Heat of Adsorption vs. Coverage, Y = I O ~ ~ S ~ C - ~
'
'
'
'
'
! A
1.2" Bulk K Coverage (monolayers)
0.4
0.8
e
Figure 12. Heat of adsorption of potassium on Rh(ll1) as a function of coverage. 0
20 0 Temperature
400
("C)
Figure 11. Thermal desorption spectra of hydrogen from a flat ( l l l ) ,a stepped (557),and a kinked (12,9,8) surface of platinum. Table I. Structure Sensitivity of Hz/D2 Exchange at Low Pressures (=lo* Torr) reaction probability 0.9 stepped Pt(332) =lo-' flat Pt(111) 110-3 "detect free" Pt(ll1)
systems in spite of the sometimes large mismatch of interatomic spacings in the two bulk metal systems (differences in atomic sizes and bulk crystal structure). The flexible surface model would predict restructuring of both sides of the metal-metal interface to optimize bonding.
This restructuring will lead to epitaxy, a predominant feature of most metal-metal. interface systems. It is hoped that surface crystallography studies in the near future will explore the nature of restructuring of these interfaces. Unusual Coverage-Dependent Changes of Surface Bonding. When the coverage of a chemisorbed atom or molecule is increased, its heat of adsorption per molecule decreases rather markedly. This is shown for an adsorbed layer of potassium24on the (111)crystal face of rhodium (Figure 12). The decline in binding energy is due to an adsorbate-adsorbate interaction which is repulsive,leading to a weakening of the adsorbate-substrate bonds as the coverage is increased. (24) Garfunkel, E. L.; Somorjai, G. A. In Alkali Metals as Structure and BondingModifiersof Transition Metal Catalysts. Alkali Adsorption on Metals and Semiconductors; Bonzel,H. P., Bradshaw, A. M., Ertl, G., Eds.; Elsevier: Amsterdam, 1989; p 319.
Langmuir, Vol. 7, No. 12,1991 3181
Flexible Surfaces
P. (OOOd
- (1.0- s
Figure 13. (a, far left) Scanning tunneling microscope picture of the ordered arrays of sulfur atoms on the Re(0001) crystal face and a model of the sulfur chemisorption induced restructured metal surface. (b, middle left) Scanning tunneling microscope pictures and a model ordered arrays of sulfur trimers on Re(0001) in the (3d3X31/3)R3Oosurface structure. (c, middle right) Scanning tunneling microscope picture and a model of arrays of sulfur tetramers in the (i surface structure. (d, far right) Scanning tunneling microscope pictures and a model of arrays of sulfur hexamers in the highest su fur coverage (21/3X21/3)R30°.
r'
The restructuring of the substrate induced by chemisorption can turn the repulsive adsorbate-adsorbate interaction attractive. This appears to be the case for S adsorbed on the (0001)crystal face of Re.25 At low coverages S occupies %fold sites on the metal surface. The substrate reconstructs as determined by LEED surface crystallography. As the S coverage is increased, S forms trimers, then tetramers, and finally hexamers at the highest coverages. These were detected by STM in ultrahigh vacuum (UHV) and the STM pictures are displayed in Figure 13. The S-S bond length is that of the metal interatomic distances and not that of the shorter covalent bonds in sulfur dimers or hexamers. The clusters are arranged in an ordered superstructure, a most striking indication of the attractive adsorbate-adsorbate interaction that persists in this circumstance to the highest coverages in the monolayer.
Puzzles of Heterogeneous Catalysis That Can Be Explained by the Flexible Surface Model Structure-Insensitive Reactions. One of the most successful classifications of heterogeneous catalysis reactions is their division into structure-sensitive and structure-insensitive gr0ups.~6 Structure-sensitive reactions (NH3synthesis, hydrogenolysis, dehydrocyclization, etc.) change their rate markedly with changing surface structure. Surface structure dependence is studied by changing the catalyst particle size or by varying the single crystal surface structure in model studies. Structure sensitivity is expected for surface reactions, since the reactant and product molecule bonds are altered as the structure of the surface is changed. In light of the flexible (25) Ogletree, D. F.; Hwang, R. Q.; Zeglinski, D. M.; Lopez Vazquezde-Parga, A.; Salmeron, M.; Somorjai, G. A. J. Vac Sci. Technol. B 1991, 9 (2), 886. Hwang, R. Q.; Zeglinski, D. M.; Lopez Vazquez-de-Parga,A.; Ocal, C.; Ogletree, D. F.; Salmeron, M.; Somorjai, G. A. Phys. Rev. Lett., in press. (26) Boudart, M. Adu. Catal. 1969,20,153. Carrazza, J.; Somorjai, G. A. Ind. Eng. Chem. Fundam. 1986,25,63-69.
surface model, all one has to invoke is that there is a correlation between the clean surface structure and the restructured surface that is active during the catalytic reaction. Structure insensitivity is much more difficult to understand. How can any surface reaction be structure insensitive? Using the flexible surface model, we argue that the surface prepared before the reaction commences is inactive. Chemisorption of the first monolayer of reactants restructures the surface and this process creates the active sites. In this circumstance, the concentration of catalytically active sites depends on adsorbate coverage or reactant pressure and not on the structure of the freshly prepared catalyst. Hydrogenation of olefins, CO, CO2, and hydroformylation on some metal surfaces (Pt, Pd) is a reaction that belongs to this class of catalytic process. High Activity of Oxide-Metal Interfaces. When an active metal is deposited on high surface area oxides of different types, its catalytic activity may vary by orders of magnitude for the same reaction. For example, nickel powder or nickel dispersed on silica as small particles are orders of magnitude less reactive for the formation of methane from CO and H2 than Ni dispersed on TiO2, a catalyst that produces very high rates of m e t h a n a t i ~ n . ~ ~ The oxide alone is inactive for this reaction. These results can be quantified by depositing Ti02 on a transition-metal surface and measuring the rate of the reaction as a function of oxide coverage. This is shown in Figure 14 for Ti02 deposited on a rhodium The system exhibits a maximum rate for the hydrogenation of CO2 or CO at about 50% of a monolayer oxide coverage of the metal. Since the oxide is inactive for this reaction and the metal alone is much less reactive, the oxide-metal interface has a very high catalytic activity. Since the oxide-metal (27) Vannice, A.; Garten, R. J. Catal. 1979,56,236. Bartholomew, C. H.; Pannell, R. B.; Butler, J. L. J. Catal. 1980, 65, 335. (28) Williams, K. J.; Boffa, A. B.; Levin, M. E.; Salmeron, M.; Bell, A. T.; Somorjai, G. A. Catal. Lett. 1990, 5, 385.
3182 Langmuir, Vol. 7, No.12, 1991
Somorjai
an n
T = 553 K P = 1 atm H,:CO = 2:l
h
b
In a
2
0.10
I N
a Y
t 4
z
0
,"
a05
W
z
z u 0 -I
2 0 c
0
0.5
xA" Au-Pt ( I l l ]
periphery area is no more than 10% of the total surface area (metal inactive oxide), these sites have turnover rates that are more than 50 times that of the metal alone. One possible explanation for this unique "activation" of the metal by an inactive oxide is the roughening of the metal at the interface by the formation of sites of lower coordination. Partial reduction of Ti02 to Ti3+-oxide at the interface produces low coordination metal sites that can undergo adsorbate-induced restructuring at a much higher rate than the metal alone, and these sites appear to be stabilized by the oxide. High Activity Bimetallic Interfaces. Pd is a good catalyst for the oxidation of H2 to produce water. However, when gold is added to Pd, the reaction rate increases by about 50-f0ld.~~ Gold itself is a poor catalyst for this reaction. Similar effects were found for HC conversion reactions30when gold was added to platinum (Figure 15). These puzzling effects can also be rationalized by the flexible surface model. The lower coordination of active metal atoms at the bimetallic interface can undergo adsorbate-induced restructuring at higher rates leading to the formation of more active sites as a result of this dilution effect due to gold.
+
The Experimental Challenge Surface science studies are hampered by the limitations of the experimental techniques that are available. Most spectroscopic tools-infrared (IR), X-ray photoelectron spectroscopy (XPS),s u m frequency generation (SFG)-m only monitor the molecule side of the surface chemical bond and not the substrate side. Other techniques like electron microscopy can only monitor the substrate side of the bond. We need the development of techniques that can simultaneously monitor both sides of the surface chemical bond and can operate a t high coverages (high pressures). We also lack time resolution for most of our structural or spectroscopic techniques in surface science. We need techniques that can monitor the dynamics of surface processes on a time scale that is shorter than ~~
(29) Lam,Y. L.; Criado, J.; Boudart, M. Nouu.J. Chim. 1977,1, 461. (30)Sachtler, J. W. A.; Somorjai, G . A. J. Catal. 1983, 81,77.
15-
--
4.
0
0-0
0
I I-: l o ,\
.
I
0
1
Q
Alloys
+
Epitaxlal
1
Gold Coverage 1 Monolayers 1
6
-
7
05
I
I
3
2
+
1
Gold Surtace Concentration 0
Figure 15. (a,top) Rate of n-hexaneisomerization. (b, bottom) Rate of cyclohexanedehydrogenationover platinumas a function of gold coverage. catalytic turnover times. Time resolution in the range of 10"3-10-5 s would be most useful and again should be applicable at high pressures. Finally, we should aim to obtain atomic spatial resolution during our surface and interface studies whenever possible. There are surface science techniques whose development can provide some of the desired information. These include STM, AFM, nonlinear laser optics (SHG, SFG), solid-state NMR, and EXAFS. It is my hope that new developments in surface science instrumentation will permit us to carry out molecular level studies of the dynamic processes that occur at the flexible surface. Acknowledgment. This work was supported by the Director, Office of Energy Research, Office of Basic Energy Sciences, Material Sciences Division, US. Department of Energy under Contract No. DE-AC03-76SF00098.