Effects of Structure on Physical and Chemical Properties of Two-Dimensional Systems

Chapter 1 2

Effects of Structure on Physical and Chemical Properties of TwoDimensional Systems Materials science has long been capitalizing on the numerous practical applications of the relationship between the atomic structure of a material and its various physical and chemical properties. To make efficient use of a material in any construction or process, one has to know as much as possible about its structure and phase diagram. This chapter aims to demonstrate the influence of the structure of 2D systems, including phase transitions therein, on the behavior of these systems. This will help the reader understand why, from the practical viewpoint, one should study the two-dimensional crystals, whose beauty, in contrast with 3D crystals, cannot be admired with the naked eye. As in the previous chapters, we shall consider mainly surface systems.

12.1. Surface Diffusion of Adsorbed Particles Numerous experiments have shown a strong dependence of the diffusion coefficient of adsorbed particles D on particle concentration (Gomer, 1961; Butz and Wagner, 1977; Ehrlich and Stolt, 1980; Naumovets and Vedula 349

350

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

1984; Gomer, 1986). This dependence—as well as the concentration de­ pendences of parameters underlying the coefficient of diffusion, such as the activation energy Ea and the preexponential factor D0—are of interest both as explicit data on the mobility of particles at different coverages and as a reflection of adparticle interactions and hence of phase transitions in a layer. Indeed, the diffusion flux of interacting particles can be approximated by the formula (see, for example, Landau and Lifshitz, 1981c) J=-L(n)£,

(12.1)

where L(n) is the transport coefficient (dependent on the concentration), ζ is the chemical potential, and χ is the coordinate (for simplicity, a ID case is considered). Having separated the concentration gradient in this equation, one can rewrite it as J(n)=-D(n)^-

(12.2)

Here, the diffusion coefficient is

. «„m. .«,

m m

n

.3)

where the self-diffusion coefficient D{(ri) characterizes the diffusion in the absence of concentration gradient and is related to the adatom mobility B(n) by the Einstein equation D{(n) = B(n)T. As is obvious from the above expressions, the diffusion characteristics reflect all the peculiarities of the variation of chemical potential with concentration, which, on the other hand, determine the phase diagram of the layer. As a result, the overall pattern of phase transitions in overlayers is reflected in the behavior of the diffusion rate. This relationship in the interior of solids has long been under study (the so-called polyphase diffusion), but concerning solid surfaces, similar research has only been performed since the development of technique for registration of diffusion profiles with high concentration and distance resolution. Detailed information has been obtained on the diffusion of electropositive adsorbates, palladium, gold, and some gases on refractory metals, and of gold and silver on Ge and Si (see reviews by Naumovets and Vedula (1984) and Gomer (1990)). Figures 12.1 and 12.2 exemplify the findings for barium on the (110) plane of molybdenum (Vedula et a/., 1980). Clearly, virtually all phase transitions observed independently by the LEED method are revealed in the

(12

SURFACE DIFFUSION OF ADSORBED PARTICLES

FIGURE

351

12.1.

Coverage profile θ(χ) in diffusion of barium out of a multilayer step on M o (110); Τ = 400 Κ. Small circles in the structure models represent M o atoms, large circles Ba atoms. (Lyuksyutov et al, 1986.)

352

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

configuration of the diffusion profile and the variation of diffusion parameters with the concentration. In fact, the diffusion profile can be regarded as a sequence of 2D phases replacing each other and formed directly in the process of diffusion. Each phase has its own diffusion parameters and, in general, its own diffusion mechanism. Let us consider the above results (Fig. 12.2) in more detail. The rapid decrease in the diffusion activation energy, and consequent increase in the diffusion coefficient, with increasing degree of coverage θ for small θ can be naturally attributed to the repulsive interactions of adatoms (Fig. 12.3).

C FIGURE

12.3.

Effect of (a) attraction and (b) repulsion of adatoms o n the activation energy ΕΛ of surface diffusion; (c) surface potential corrugation V(x) for noninteracting adatoms. (Naumovets and Vedula, 1984.)

SURFACE DIFFUSION OF ADSORBED PARTICLES

353

Obviously, for this range of coverages, the conventional one-particle approach to the characterization of surface diffusion is well justified. Transitions to dense layers must reflect the increased importance of collective effects. When the concentration of adatoms approaches the critical value at which the first-order phase transition begins in the layer, one can observe an increase in the activation energy and a decline in the diffusion coefficient. This can be attributed to the formation of precritical nuclei with a lower mobility than that of individual atoms, as well as to the decrease of the concentration derivative of the chemical potential because of the proximity to the degree of coverage at which the lateral repulsion gives way to attraction (see Section 3.5.2). Within the first-order phase-transition range itself, the activation energy should be relatively high, since in this case the mass transport also incorporates the stage of adatom detachment from the islands of the twodimensional condensate (Shrednik and Odishariya, 1970; Golubev et al, 1971). For the system B a - M o (110) the data on the variation of D in the firstorder phase-transition range were not obtained for technical reasons. The behavior described was found experimentally for layers of lithium and barium on the (110) plane of tungsten (Loburets et al, 1982; Naumovets et al, 1988). Using the system B a - M o (110) as an example, one may study the correlation between diffusion-parameter variations and the transition from the commensurate to the incommensurate structure. The diffusion coefficient for the region of this transition has an absolute maximum (Fig. 12.2) corresponding to the plateau section in the concentration profile (Fig. 12.1), where the concentration gradient is very small. This peculiarity is so well defined that one can confidently assert that diffusion results in the growth of the c(2 χ 2) barium phase on the surface, after which the transition to an incommensurate lattice occurs in the layer. The c(2 χ 2) phase is a submonolayer one; as is attested by direct structural study (Fedorus et al, 1972), the buildup of the second layer over this phase is energetically unfavorable, so Ba atoms invade the first layer under the action of strong attraction to the substrate. Here, they shift adjacent adatoms from their potential wells, forming interstitial configura­ tions (crowdions). These configurations were hypothesized to form on the line of contact of the commensurate phase, spreading over the surface with the initial deposit of adsorbate, and it was suggested that the mass transport occurs due to the rapid movement of interstitial configurations to the "free" layer edge by a "relay race" mechanism (Vedula et al, 1980; see Fig. 12.4). This model was further developed in the work by Lyuksyutov and Pokrovsky (1981), where the diffusion mechanism has been treated with allowance for the fact that the pointlike interstitial configurations find it energetically advantageous to

354

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

SOLITON

FIGURE

12.4.

(a) Diffusion in the second monolayer over an immobile first monolayer ("unrolling-carpet mechanism"), (b) Diffusion of a domain wall (soliton) in a commensurate adlayer ("soliton mechanism"). Positions of adatoms in the substrate potential relief are shown schematically.

integrate into linear domain walls representing solitons in a commensurate lattice. Such a model, referred to as the soliton mechanism, has been treated in detail in Section 10.4 (see also Lyuksyutov et al, 1986). It is to be emphasized once more that the mass transport proceeds in the submonolayer phase. A high transport rate is provided by the strong lateral interaction, due to which the diffusion process acquires a collective character with solitons acting as mass carriers. Note that the above model of surface diffusion is similar to the crowdion mechanism of volume diffusion, the migration of domain walls in magnetic materials and ferroelectrics, and charge density waves in the interior of solids. So far, it is uncertain how universal mechanisms of this kind are. So far, most of the evidence attesting its existence has been obtained while in­ vestigating the diffusion of electropositive elements on metals. Thus, direct experiments performed with lithium layers on tungsten (Loburets et al, 1982) have demonstrated that at low temperatures the adatom mobility in the second layer was considerably below that in the first. Some interesting facts were reported in contributions by Dubrowski and Kleint (1982), Blaszczyszyn and Kleint (1986), and Beben et al (1987). By means of field-emission microscopy, these researchers have studied density fluctuations in 2D potassium overlayers on tungsten surface (see Section 2.4.2). They determined the cross-correlation function of the fluctuations of

SURFACE DIFFUSION OF ADSORBED PARTICLES

355

the number of particles within two areas with a diameter of «100 A separated by a spacing of «300 A. These fluctuations become possible only when there is adatom mobility. The cross-correlation function has the form of a curve with a maximum. The time corresponding to the maximum determines the average period of the propagation of the "signal" from one probed area to another. This implies that there is a period of time during which a decline in the number of particles in one area is followed by registration of an increased number of particles in the other. The experiments revealed very short propagation times for the "signal," which is hard to explain in terms of the conventional (one-particle) surface diffusion mechan­ isms. It is our belief that the effect in question may be produced by the soliton mechanism of the diffusion. The rapid spreading of submonolayer phases, the transitions between which are manifested rather clearly in the concentration profile, is also observed when noble metals (Au, Ag) diffuse on surfaces of Ge and Si (Suliga and Henzler, 1983; Gavrilyuk and Lifshits, 1983; see Fig. 12.5). The so-called unrolling-carpet mechanism can be regarded as an alter­ native to the above mechanism (Gomer, 1961). It is characteristic of systems where the first monolayer has a much stronger bond with the substrate and is

FIGURE

12.5.

Concentration profile and 2 D structures in surface diffusion of Au on Si (111) (solid line with open circles). I: initial deposit with boundary marked by a dashed line; ΠΙ: (*Jl χ y/3)R30°

+ (5 χ 1); IV: (5 χ 1). (Gavrilyuk and Lifshits, 1983.)

U:(y/3xy/3)R30°;

356

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

therefore less mobile than the second and subsequent layers. This occurs, for example, when chemically reactive gases are adsorbed on metals, in which case the first layer is chemisorbed and the rest of the layers are physisorbed. Here, spreading over the surface from the initial multilayer coverage is a densely packed monolayer, but this is due to the diffusion of physisorbed atoms within the second layer (Fig. 12.4). The diffusion of chemisorbed atoms can be observed (naturally, at a higher temperature) only when the initial coverage does not exceed a monolayer. In this case one can also observe the correlation between variations of diffusion parameters and the phase trans­ itions in the submonolayer film. Thus in submonolayer oxygen films on the (110) plane of tungsten, the diffusion coefficient is largest at a concentration preceding the occupation of surface by the continuous ( 2 x 1 ) lattice (Butz and Wagner, 1977). The results of Monte Carlo simulation with allowance for data on oxygen-adatom interactions on this face permitted the researchers to infer that this maximum was related to abrupt variation of the chemical potential in this coverage range (Bowker and King, 1978). The unrolling-carpet mechanism is supposed to act in the diffusion of both gases and metals. For example, for small bond polarities of adparticles and (hence) weak lateral repulsions, the mobility in the first layer may prove lower than in the subsequent layers of a metal film, thus satisfying the condition for the implementation of the carpet mechanism. Butz and Wagner (1979) concluded on the basis of their experimental data that this effect takes place in the diffusion of palladium and gold on the (110) plane of tungsten. Since diffusion in the first and the subsequent monolayers depends on the parameters £ a and D0, one can naturally foresee the possibility of a transition from one mechanism to another. However, it is to be borne in mind that there exists an upper temperature limit related to the initiation of desorption: above that temperature, the lifetime of adatoms on a surface becomes so short that they can migrate only over a very small distance prior to evaporation (see, for example, Naumovets et al, 1988). Obviously, the variation of temperature can lead to changes in diffusion mechanisms even at submonolayer coverages. Testifying to this fact are the jumps in the parameters ΕΛ and D0 revealed in the order-disorder transitions in adsorbed overlayers (Loburets et al, 1982; Naumovets et al, 1988). So far, we have been considering only the translational mobility of adsorbed particles. However, in the examination of adsorption of molecules and their reactions on the surface, other types of motion of molecules are of great interest. In particular, the study of quasielastic neutron scattering allows one to obtain data on the so-called rotational diffusion. In this case the molecule is engaged in free rotation (which either is spherically isotropic or proceeds about some axis), occasionally colliding with adjacent molecules and thus changing its orientation or the direction of its rotation (see, for

SURFACE DIFFUSION OF ADSORBED PARTICLES

357

example, Coulomb and Bienfait, 1986; Bienfait, 1987a). The nature of the movements performed depends on the structure of the layer (degree of coverage) and the temperature, as has been shown for ethane (C 2H 6) layers on graphite (Fig. 12.6). These molecules are of a rodlike shape and form ordered Si lattices (small coverages) or S 3 lattices (large coverages) at low temperatures. In the former case, the molecules lie on the surface, and at Τ = 10-53.5 Κ they begin rotating about their axes. In the S3 phase, the axes of molecules are normal to the surface, and rotational diffusion sets in at Τ = 57-67 Κ. At higher temperatures, of liquid-crystal and liquid phases form in which the rotational diffusion is supplemented by translational diffusion. The importance of information on the relation between diffusion and the

Coverage

AW /

ΦΦΦ

ΦΦΦ

s3

1-3

s 2?

L a?

///

la

W

Li

Temperature FIGURE

12.6.

Schematic configurations of ethane films on graphite in the submonolayer range. At low T, the molecules, depending on coverage, lie down (S^ or stand up (S 3). At medium T, molecules are reorienting about their C - C axis. At high T, the melted solids are 2 D liquid crystals or lattice liquids ( L t and L 3) . The structures 5 2 and L 2 are unknown. (Bienfait, 1987a.)

358

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

structure of a layer for understanding the kinetics of many surface pheno­ mena is entirely evident. No less evident is the fact that ignoring layer structures precludes any valid inference on diffusion mechanisms. Making use of the difference of diffusion parameters in various two-dimensional phases, one can, to a certain degree, control the diffusion process by temperature variation, promoting preferential spreading of one or another adsorbate phase on the surface. The existing correlations between the rate of diffusion and the structure of a layer permit utilization of the surface diffusion for research on phase transitions in layers in order to locate the boundaries between different phases in phase diagrams and for other experimental purposes (see, for example, the publications by Bowker and King, 1978; Ehrlich and Stolt, 1980; Naumovets and Vedula, 1984; Zhdanov, 1989a).

12.2. Effect of Substrate Structure on Wetting The diffusion processes examined in the previous section are closely related to layer growth mechanisms, wetting phenomena, and spreading. The thermodynamical and statistical background of wetting were analyzed in Chapter 11. Here, we shall only briefly deal with the structural aspects of this phenomenon. According to Young's criterion (11.3), the stronger the adsorbate-tosubstrate interaction, the better the conditions for complete wetting. Thorough experimental studies of wetting processes in a large number of systems have nevertheless demonstrated that if the growing layer is in the solid state, the above criterion needs refinement (Seguin et a/., 1983; Bienfait, 1985; Ebner, 1986). Let us remind the reader that complete wetting in this case corresponds to the layer-by-layer growth of a film, while incomplete wetting involves the formation of 3D crystals on a surface covered by a certain small number of monolayers (see Section 11.1.1). Of course, complete wetting is absent in the case of the so-called weak substrates (the strength of a substrate is defined as the ratio of the heat of adsorption to the heat of sublimation of a solid adsorbate at Τ = 0). At the same time, tomplete wetting was also found not to exist on substrates that were too strong (the case of the so-called reentrant incomplete wetting). As a result, complete wetting can occur only within a fairly narrow range of substrate strengths (concerning this issue, see the review by Bienfait (1985)). From the structural viewpoint, this can be attributed to the fact that an excessively strong substrate imposes a specific structure on the layer. A structure of this kind would be obtained even if the surface were ideally smooth, because the attraction of adatoms to the substrate results in strong

ELECTRONIC PROPERTIES

359

compression of the monolayer. Conservation of this lattice under layergrowth conditions requires great energy expenditure for elastic deformation. At a certain critical thickness this becomes impossible, since the forces of attraction to the substrate weaken at a faster rate than does the mechanical stress in the strained layer. However, it is possible that layer growth is controlled not only by the substrate strength, but also by the amplitude of potential corrugation of the substrate. If the corrugation is sufficiently smooth, then upon occupation of the second and the subsequent monolayers, the structure of the first monolayer can rearrange with a certain adaptation to the structure of the adsorbate. It is believed that this effect underlies the virtually complete wetting of graphite by xenon, argon, and krypton, at low temperatures (layerby-layer growth was found to occur up to a thickness of at least 8-10 monolayers) (Bienfait, 1985). Typical of smooth substrates are the orienta­ tional transitions in the incommensurate lattices of the first monolayer (see Section 6.6). The ease with which these lattices rotate, also facilitates the alleviation of the structural misfit between the substrate and the growing solid layer. Effects of this kind have been observed for xenon layers on the (111) plane of platinum. Due to their action, this substrate, which is much stronger for xenon than graphite is, lends itself to complete wetting; the helium-atom scattering method made it possible to observe layer-by-layer film growth up to a thickness of at least 25 monolayers (Kern et al, 1986a). In this section we have dealt with the influence of the substrate's structure on wetting at low temperatures when the wetting process is determined by energy factors only and the entropy contribution is neglected. In Chapter 11 we discussed the wetting transitions at higher temperatures, resulting from to the formation of structures with domain-wall-type defects or of dislocations, or from the melting of a layer, where the structural effects are totally absent.

12.3. Electronic Properties The theory of and experimental research on two-dimensional electron systems have progressed rapidly over recent years. It suffices to mention in this respect the studies of two-dimensional conductivity and Wigner crystalli­ zation. The most impressive achievement in this area is the discovery of the quantum Hall effect in 1980 (von Klitzing et al, 1980). The 2D electron systems are abundantly dealt with in the literature (see Ando et al, 1982). This subject transcends the scope of our book, so we shall focus only on the relation between electronic properties and surface structures, including clean and adsorbate-covered surfaces.

360

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

12.3.1· Work F u n c t i o n The work function is an important characteristic of a surface. The results of investigations into clean surfaces of metals testify to the substantial de­ pendence of the work function on the atomic structure of a surface (the work functions of different crystal planes may differ by as much as 20-25%). Therefore, it is natural to expect that the phase transitions in overlayers should be reflected in peculiarities of work functions of adsorbate-covered surfaces. However, this problem has proved more intricate than was believed at first (Gerlach and Rhodin, 1970). Special experiments have been conducted for the accurate measurement of the variation of the work function due to an order-disorder transition in a homogeneous adsorbed layer (layers were selected for which the order is destroyed within a narrow temperature range). The resulting variation of the work function is usually a few hundredths of an eV, constituting ~ 1% of the total work-function variation caused by the adsorbed layer (Fedorus and Naumovets, 1980). It was concluded, on the basis of these facts, that the work function depends on the short-range rather than on the long-range order. The critical temperature 7^ marks the destruction of only the long-range order in a layer, while the short-range order persists in a broad temperature range above Tc, as attested, in particular, by the presence of diffuse reflections on diffraction patterns. Owing to the action of the potential corrugation of the substrate, the correlation function characterizing this short-range order bears the imprint of the substrate's symmetry. This means that the nearest neighbors are, in most cases, arranged about each other in a way similar to that in the lattice with long-range order. Adlayer-induced variation of the work function is determined by the dipole moment, which, in its turn, depends on the interaction of each adatom with the rest. As can be easily demonstrated, the effect of remote neighbors is only weakly dependent on their arrangement on the substrate. For closer neighbors the effect, naturally, is different, although their arrangement changes only insignificantly with the destruction of the long-range order. It can be assumed that the effects under discussion could be effectively explored by a method enabling the deter­ mination of the so-called local work function, i.e., the electrostatic potential close to the surface (Wandelt, 1987; see also Section 2.2.2). The dependence of the work function on the adatom concentration reflects the electronic state of adatoms and, first of all, the decline in the dipole moment of the adsorption bond. These variations take place as a result of lateral interactions or because the adparticles begin occupation of adsorption sites of another type on the substrate. As far as the metal atom layers are concerned, at a definite surface occupation stage metallization of the layer occurs, during which the layer rapidly acquires the properties of a bulk metal (Lang, 1971; Wojciechowski, 1976).

361

ELECTRONIC PROPERTIES

From the physical viewpoint it is clear that the abovementioned variations of electronic properties must be linked to changes in the atomic structure, and this relation can be revealed by sufficiently accurate experiments. Experiments allow a distinct correlation of the work function with firstorder phase transitions to be observed. Naturally, the condensation of atoms into a denser phase is normally accompanied by significant variations of the dipole moment, due to which the concentration curve of the work function shows a break (Fig. 12.7). If the method employed determines the surfaceΦ, q, eV

FIGURE

12.7.

Curve 1: Heat of adsorption of La on W (110) as a function of the degree of coverage; curve 2: the same for La on M o (110); curve 3: work function of the L a - W ( l l O ) system versus degree of coverage; curve 4: the same for L a - M o ( l l O ) ; curve 5: the intensity (in arbitrary units) of the Auger peak of La (78 eV) on W(110) versus degree of coverage. (Vedula et aly 1977a.)

362

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

averaged work function, then from the region of the first-order phase transition one can extract the difference in this parameter between values characteristic of coexisting phases, which is linear in concentration (Fedorus et a/., 1972; Vedula et al, 1977a). The first-order phase transition in which the adsorbate evaporates from 2D condensate islands and passes over to the 2D gas phase is also accompanied by significant variations of the work function (Fig. 12.8). The surface-averaged concentration of adatoms in this case remains the same. However, the area occupied by the condensate decreases (to zero in the limiting case), whereas the area covered by the two-dimensional gas increases, and so does the density of this phase. Naturally, this involves variation in the short-range order and the dipole moment of adatoms, which is reflected in the change in the work function. The use of these effects underlies one of the methods for studying two-dimensional sublimation and plotting phast diagrams of adlayers (Kolaczkiewicz and Bauer, 1984a). Let us review once more the concentration dependences of the work

-0.5 1

I

I

600

I

J

800

TEMPERATURE

Τ (Κ)

FIGURE

I

1

1000

L— 1200

—

12.8.

Temperature change of the work function of an Au-covered W(110) surface in the first-order phase transition. (Kolaczkiewicz and Bauer, 1984a.)

ELECTRONIC PROPERTIES

363

function. When a transition between ordered lattices occurs through a disordered phase or as a second-order phase transition with increasing concentration, the dipole moment of adatoms varies smoothly, especially in the presence of long-range interaction between the lattices. This leaves no pronounced features in the concentration curve of the work function. In experiments at submonolayer coverages one can often observe a fairly pronounced peculiarity in the form of the work-function minimum. Its emergence must be attributed to different reasons inherent in systems of different nature. The deepest minima of the work function are typical of layers of electropositive metals. Calculations demonstrate that in that case the minimum appears in the intermediate area between "ion" and "metal" states of the layer (Lang, 1989b; Norskov, 1989). It reflects the progressive depolarization of the adsorption bond with the rise in the degree of coverage. Given the sufficiently close approach of the atoms, this process results in the metallization of the layer, which implies that it acquires a set of properties similar to those of a bulk adsorbate. In most cases the appearance of the minimum correlates with a certain phase transformation, as was found for many overlayers on faces with a smooth or furrowed corrugation in crystals of tungsten, molybdenum, rhenium, and other metals (see, for example, Bolshov et al, 1977; Naumovets, 1984). Completion of the formation of the first densely packed monolayer is also reflected in the concentration dependence of the work function (Fig. 12.7). On adsorption of alkali and alkaline-earth metals, one sees a small maximum or a transition to the saturation section, while the occupation of the second and the subsequent layers causes a relatively insignificant variation of the work function. However, with sufficiently high accuracy of measurements (~lmeV) the work-function minima corresponding to the beginning of formation of subsequent monolayers can be clearly identified (Vedula and Poplavsky, 1987; see Fig. 12.9). The emergence of the above minima is due to the fact that the nucleation of each new layer leads to increased roughness of the surface. The adatoms protruding above the surface assume a certain positive dipole moment owing to the tendency of the electron gas to smooth out the contour of its boundary—the Smoluchowski effect (Smoluchowski, 1941). By calculating the number of work-function minima, the number of the deposited layers can be found in the same way as when observing the oscillations of diffraction-beam intensity (Fig. 2.2). High sensitivity of the work function to the surface structure was revealed also by Kolaczkiewicz and Bauer (1984b) during experiments with gold and silver overlayers on the (112) plane of tungsten. In this case, a decline in the work function by «30meV was recorded, accompanying the initiation of uniaxial buckling in the compressed first monolayer of these adsorbates. This solitonlike buckling occurs because the atomic diameters of Ag and Au

364

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

FIGURE

12.9.

Work-function change during the growth of a lithium film on M o (110). The top coverage scale is 2 in monolayers (the interval from 4 to 10 monolayers is shown); the bottom scale is in a t o m s / c m . (Vedula and Poplavsky, 1987.)

exceed the diameter of W by approximately 5%. The structural transitions in these layers also show up in the variation of the work function. The reported data indicate that obtaining the desired value of the work function for a surface with a definite atomic structure virtually presupposes the formation of a certain optimal two-dimensional phase on the surface. It is to be emphasized that work-function optimization problems present them­ selves in electronics when developing cathodes of various types (see, for example, Hatsopoulos and Gyftopoulos, 1979). Moreover, similar problems are encountered in the formation of adsorbed layers on metalsemiconductor (Spicer et a/., 1984) and electrode-electrolyte interfaces (Schultze and Dickertmann, 1976; Takayanagi et a/., 1980; Ross, 1982; Kolb and Scherson, 1983). 12.3.2. Electronic S t r u c t u r e The variation of the spectrum of surface electron states induced by surface reconstruction constitutes another illustrative example of the interrelation between atomic and electron structures of a surface. The (111) plane of silicon has been studied in most detail. Its surface states were investigated by means

365

ELECTRONIC PROPERTIES

of photoelectron and optical spectroscopy, and also by the high-resolution LEELS technique (Himpsel and Fauster, 1984; Lieske, 1984; Himpsel, 1986). What was observed experimentally was the sharp restructuring of the spectrum of these states on the irreversible transition from the metastable (2 χ 1) structure to the (7 χ 7) structure stable at low temperatures (Fig. 12.10), as well as upon the reversible transition from the (7 χ 7) structure to the (1 χ 1) structure, which is stable at elevated temperatures. These variations cause abrupt changes in the adsorption rate on a silicon surface (see Section 12.4). The surface diffusion coefficient was found to increase by several orders of magnitude on transition from a metastable structure of the silicon surface obtained by evaporation in a high electrical field to a stable structure formed by the annealing of a crystal (Borziak and Dadykin, 1984). Such a smoothing out of the potential corrugation is interpreted as a consequence of partial saturation of broken valence links of surface atoms under reconstruction.

FIGURE 12.10. Surface states (tick marks) on different Si surfaces, clean and covered with hydrogen and oxygen. The data were obtained by photoemission and inverse photoemission methods (left-hand and right-hand side, respectively). (Himpsel and Fauster, 1984.)

366

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

The change in semiconductor surface structure can be accompanied by intense variations of surface conductivity (Aristov et al, 1982; Zavaritskaya, 1984; Katrich and Moiseev, 1985). When the bulk conductivity of a specimen is small, a 2D conductor can be created. The formation of surface superlattices permits controlling the behavior of such systems. The superlattices can be implemented by tilting the surface by a small angle from a direction corresponding to low Miller indices. Then the regularly arranged atomic steps serve as electron scatterers, leading to the formation of small gaps in the energy spectrum of a two-dimensional conductor (Volkov et al, 1980; Κ von et al, 1981). Pronounced correlation between variations of electron and atomic structures is also observed during the reconstruction of metal surfaces (Campuzano et al, 1981; Inglesfield and Holland, 1981; King, 1983). Let us examine in brief the variation of electron structures of adsorbed overlayers induced by changes in their phase state. Concerning this issue, many data are available, especially in connection with investigations into the mechanisms of catalytic reactions (see Section 12.5). The adsorption of alkali elements on refractory metals, semiconductors, and graphite is a good example of the strong effect of the substrate on the electronic behavior of adsorbed particles. At small coverages, these atoms are almost completely ionized (Sesselmann et al, 1983; Woratschek et al, 1985, 1987, Lang, 1989b; Norskov, 1989), or, according to another interpretation, strongly polarized (Soukiassian et al, 1985). This results in repulsive interactions between adatoms, rapid decline in the work function on adsorption, and the promotion of catalytic reactions (Bonzel, 1987). With the gradual compaction of an adlayer, the adatom degree of ionization decreases. The variations of band structure correlate with the changes in 2D lattices that are usually seen with the simultaneous employment of LEED and electron spectroscopies. Thus, at the transition from the commensurate to the incommensurate lattice one clearly observes significant changes of the electron state density (Katrich et al, 1988; see Fig. 12.11). The spectra of characteristic losses of electron energy demonstrate that at a certain critical adatom concentration of the order of \ monolayer a new peak appears. This can be interpreted as a peak of plasma losses caused by the metallization of the adsorbed layer (MacRae et al, 1969; Gorodetsky and Gorchinsky, 1979; Lindgren and Wallden, 1980; Tochihara, 1983; Ishida et al, 1985). It goes without saying that the adsorption of alkali metals, with their small ionization potentials, represents a limiting case. At the same time, even when there is no similar drastic difference between the properties of the adsorbate and the substrate, the adsorbed layers possess a specific electron structure, which is also controlled by the atomic structure of the 2D lattice.

367

ELECTRONIC PROPERTIES

FIGURE 12.11. Energy spectra of electrons photoemitted from the S r - M o ( 1 1 2 ) system at various Sr adatom concentrations, hv = 3.38 eV. N o t e the sharp change in the spectra near the Fermi level EF in the region of commensurate-to-incommensurate ( C - I ) transition. The transition is one-dimensional compression of the first monolayer. SL denotes the second layer filling. (Katrich et al., 1988.)

Most data on this problem were obtained by the photoelectron spec­ troscopy method. Let us consider, as an example, the data related to palladium layers on the (110) plane of niobium (El-Batanouny et al, 1983). The photoelectron spectra indicate the displacement of the center of the d band in monolayer of Pd on Nb (110) by 2eV downward with respect to its position on the (111) plane of bulk palladium, whose atomic structure is very close to the structure of the above monolayer. The resulting states in the monolayer turn out to be located below the Fermi level, which means that the layer possesses an electron structure typical of noble rather than transition metals. These experimental findings agree with the results of analyses performed by Kumar and Bennemann (1983). The variations of electron structure of palladium cause the adsorption rate on the palladium monolayer to be radically different from that on the surface of bulk palladium. Thus, the sticking probability for hydrogen molecules on bulk palladium is ~ 10" \ the molecules dissociate, and Pd acts as an efficient catalytic agent in reactions with the participation of hydrogen. In contrast, the sticking probability on a Pd monolayer structure tends to zero (as on copper), dissociation is absent,

368

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

and no catalytic activity is displayed. A pronounced difference between adsorption properties of the surfaces of bulk crystals and on pseudomorphous monolayers on foreign substrates was also revealed for Ni, Pd, and Fe on W (110) and (100), and for Cu on Ru (0001) (see the work by Berlowitz and Goodman (1988) and references therein). It is of interest to compare these data with the results of the examination of palladium monolayers on the (110) plane of aluminum, a metal in which the unoccupied d band is absent (Xu and Smith, 1988). Unlike niobium, the aluminum substrate exerts no sensible effect on the behavior of palladium, which is virtually identical to the behavior of free monolayer as predicted by theory. The formation of electronic structure typical of the surface of bulk palladium occurs when the film becomes 3 monolayers thick. These data demonstrate the feasibility of controlling the electron pro­ perties of surfaces using both the specificity of the electron state of adsorbed particles caused by the action of the substrate, and the specific features of different two-dimensional phase states of the layer itself. We note, in conclusion, the new opportunities in the study of surface electron states opened up by scanning tunnelling microscopy. As was shown in the work by Becker et al. (1985), the application of this apparatus as a tunneling spectrometer allows one to obtain unique information on the spatial distribution of the density of electron states on a surface that can be reliably bonded to the atomic corrugation of a surface. It is hoped that these data will allow a deeper understanding of the electronic behavior of surfaces on the atomic scale (for this prospect see also the proceedings of conferences on the applications of the scanning tunnelling microscopy: Burstein et al. (1988); Feenstra (1988)).

12.4. Kinetics and Energetics of Adsorption The adsorption rate is exceedingly sensitive to the structure of a clean substrate as well as to the structure of two-dimensional phases formed in the process of adsorption. Thus, for the (111) surface of Si with the (7 χ 7) lattice, the oxygen sticking probability is 10 times higher than on the same face with a (2 χ 1) lattice (Su et a/., 1981). The latter, however, shows higher activity in the adsorption of hydrogen and chlorine (Lieske, 1984). The reconstruction of the silicon surface exerts a significant effect on the initial stages of the process of epitaxy (Gossman, 1985; Akinci, et a/., 1988). Sticking probabilities for molecules on different planes of the same metal crystal may differ by approximately one order of magnitude (see, for example, Somorjai, 1984; Yates, 1985).

KINETICS AND ENERGETICS OF ADSORPTION

369

Surface defects also strongly affect the rate of adsorption. This issue was dealt with qualitatively for surfaces that are vicinals of close-packed planes. These consist of regularly arranged terraces of close-packed planes separated by atom-high steps. By changing the vicinal's misfit angle relative to the initial ("ideal") plane, one can vary the width of terraces over a fairly broad range (the experiments were performed with the terrace widths of 3 to 20 lattice periods). Kasupke and Henzler (1980) have found that if one step corresponds to a terrace 7 lattice periods wide, the adsorption rate of oxygen on Si (111) is increased by about an order of magnitude over that for an "ideal" Si (111) plane. In their recent work, Akinci et al (1988) discovered an extremely strong influence of steps on the (111) planes of silicon, and of their orientation, on the growth of nickel silicide. Significant effects are also observed on the stepped faces of metals (see reviews by Wagner, 1979; Somorjai, 1984). The increase in chemisorption rates on stepped surfaces is usually related to the fact that the steps facilitate the dissociation of molecules followed by the migration of chemisorbed atoms on terraces, whereby the whole surface becomes occupied. Such a mechanism was found to exist for hydrogen on W(110): the hydrogen is virtually not chemisorbed on a defectless surface (Polizotti and Ehrlich, 1979). It is noteworthy, however, that the substrate's structure exerts a substantial effect only on the kinetics of formation of the first (chemisorbed) layer. The next (physisorbed) layers of gas molecules are filled up almost irrespective of the substrate's structure (see, for example, the data for oxygen on different tungsten planes in Chuikov et al9 1989). The concentration dependence of the sticking probability for gas molecules in the first monolayer correlates with the 2D phase transitions in this layer. This is due to the fact that intermediate-state adsorption usually precedes the transition to a chemisorbed state. A particle in the intermediate state is bonded to a surface by weak van der Waals forces, has a high mobility, and is characterized by a short average lifetime on the surface at not too low temperatures. During this lifetime the particle is either chemisorbed (if it spots a suitable adsorption site on the surface) or desorbed into the gas phase. Naturally, the relative probability of such transitions depends on the structure of the 2D lattices formed during the adsorption process itself. In other words, each two-dimensional lattice has its own sticking probability. It is to be borne in mind, however, that the above assertions relate to the adsorption of gases. Adsorption of metals at not too high temperatures (below the desorption temperature level) is characterized by 100% sticking probability, whatever the coverage and the substrate's structure. Let us discuss now the heat of adsorption, which is the basic parameter behind the thermal stability of adlayers. At present, there is much experi­ mental evidence on the relationship between the heat of adsorption, the

370

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

structure of adlayers, and the phase transitions therein. As the first example we consider the dependence of the heat of adsorption of xenon on the (111) plane of platinum on the degree of coverage (Fig. 12.12). It was calculated from adsorption isotherms observed with thermal-energy helium atom scattering, using the Clausius-Clapeyron formula (Kern et al, 1988a). In the range of 0 < θ < 0.33 the heat of adsorption grows by about 10%, reflec­ ting the action of lateral attraction forces in the xenon layer. Another manifestation of these forces is the formation of two-dimensional islands of o xenon with the commensurate (y/3 χ y/3)R30 structure, which coexist with a 2D gas at θ < 0.33. The transition to an incommensurate phase beginning at θ = 0.33 is accompanied by an abrupt drop in the heat of adsorption caused by the repulsion of domain walls (solitons) forming a system of parallel stripes. Kern et al also derived the differential entropy of the xenon layer as a function of coverage. This quantity varies oppositely to the heat of ad­ sorption ("compensation effect") and shows an upward jump at θ = 0.33. Estimates demonstrate that the jump is related mainly to the emergence of low-frequency modes in a weakly incommensurate (soliton) phase.

FIGURE 12.12. Isosteric heat of adsorption qst of Xe on Pt (111) as a function of coverage 0 X .e (Kern et a/., 1988a.)

KINETICS AND ENERGETICS OF ADSORPTION

co D

371

Coverage θ ( M L )

FIGURE

12.13.

Desorption energy Ε(Θ) and preexponential factor v(0) of Ag on 2W ( 1 1 0 ) as functions of the 14 coverage 0. Coverage Θ = 1 corresponds to 1 4 . 1 2 χ 1 0 atoms/cm . The data are derived from thermal desorption spectra. (Kolaczkiewicz and Bauer, 1 9 8 6 . )

Let us now consider the findings for systems of the other type, i.e., metals on metals. To begin with, we deal with systems for which the dipole moment of the adsorption bond is small. Figure 12.13 shows the dependences of the desorption energy and the frequency factor ν in the desorption rate equation (2.2) on the degree of coverage for silver on W(110) (Kolaczkiewicz and Bauer, 1986). These curves were derived by means of the TDS technique, and they are of different shape at low and elevated temperatures. In the latter case (lower curves) the whole layer is in the state of a 2D gas, and the increase in

372

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

the desorption energy with increasing coverage reflects the lateral attraction. Some features observed in this curve are interpreted by the authors as a consequence of the fact that the proportion of silver evaporated as individual atoms or dimers can vary as a function of coverage. Noteworthy as well is the strong variation of the frequency factor with the coverage. Its value can differ 1 3 -1 by several orders of magnitude from the value ~ 1 0 s predicted for desorption from the ideal 2D gas phase. Obviously, these differences are due to variations of entropy in the process of desorption (for detailed discussion of the reasons for the dependence of the frequency factor on the coverage, see the recent works by Seebauer et al (1988) and Zhdanov (1989b)). The variations of the desorption energy and frequency factor for silver on tungsten are correlated, so that their effects on the desorption rate partially cancel. At low temperatures, the lateral attraction results in a first-order phase transition, viz., the formation of 2D condensate islands against the background of the two-dimensional gas. So long as condensate islands exist on the surface, the desorption kinetics is controlled by parameters character­ istic of the islands. Therefore, the TDS method employed at low temperature yields the desorption energy and the frequency factor, which are actually independent of the degree of coverage (upper curves). In this case the parameter χ in Eq. (2.2) determining the desorption reaction order equals zero (see Kern et al, 1979). Paunov and Bauer (1987) have obtained data on the variation of desorption energy and frequency factor for the coverage range of 0 to 2.5 monolayers characterizing the kindred system of Ag-Mo (110). Within the first monolayer the variations are similar to those described above. The desorption-energy and frequency-factor minima correspond to the beginning of the buildup of the second and the third monolayer. It is also interesting that the difference in the kinetics of desorption from the same-type faces of different crystals (tungsten and molybdenum) is less pronounced than in desorption from different faces of the same crystal (see, for example, the works by Bauer et al (1974,1977) and Kolaczkiewicz and Bauer (1986)). Finally, let us dwell on the concentration dependences of the heat of adsorption of electropositive metals forming strongly polarized adsorption bonds. The difficulties arising in the comparison of these dependences with structural data are usually due to the fact that the desorption kinetics and overlayer structures are studied at significantly different temperatures. Apparently, at desorption temperatures, the long-range order in layers of this type is destroyed in the majority of cases. In the concentration range where the adlayer forms a series of structures with a large interatom spacing and remains homogeneous with increasing concentration, the heat of adsorption is monotonically decreasing (Fig. 12.7). Beyond doubt, the factor underlying correlations of this kind is the repulsive interaction of adatoms caused by the

KINETICS AND ENERGETICS OF ADSORPTION

373

strong polarity of the adsorption bond. It is to be noted that on furrowed planes, where the adatom interactions are sharply anisotropic (see Section 3.6), the heat of adsorption decreases at a slower rate than on planes with isotropic repulsion (Kanash et a/., 1975; Medvedev and Yakivchuk, 1975a). In the regions of first-order phase transitions, where the adlayer is heterogeneous, the concentration curves of the heat of adsorption have plateaus (Fig. 12.7). The nature of this effect has been dealt with in this section. Finally, in the incommensurate phase the growth of the concentration is accompanied by a smooth decrease in the interatom spacings and the heat of adsorption, which reflects the repulsive character of the lateral interaction. Thus, the principal stages of layer buildup are displayed in the variation of the heat of adsorption. Obviously, the more delicate manifestations of the relation between the heat of adsorption and the substrate's structure can be revealed only through the more accurate measurements of the heat of adsorption and its determination at temperatures at which the structure is also known. Let us discuss, in passing, the issue of the substrate's effect on the heat of adsorption of electropositive metals and compare, to this end, the concen­ tration dependences for cesium on the (110) and (111) planes of tungsten (Fedorus and Naumovets, 1970; Medvedev and Yakivchuk, 1975b). The heat of adsorption of cesium on the (110) plane of tungsten at small coverages (n -> 0) is 0.5eV higher than on the (111) plane (see Fig. 12.14). At increased concentration, the heat of adsorption decreases on both planes due to the repulsive lateral interaction of adatoms, with the (110) plane featuring a much 14 2 faster decrease. As a result, at η = 5.2 χ 1 0 cm " , this plane shows a heat of adsorption of only 0.8 eV (equal to the cesium heat of sublimation), whereas the heat of adsorption on the (111) plane amounts to 1.3eV even at a 14 2 significantly greater concentration, η = 5.8 χ 1 0 c m " , corresponding to the commensurate (1 χ 1) lattice. These findings indicate that the (111) plane has an additional interaction among adatoms resulting in the increased heat of adsorption at large coverages. As was presumed by Medvedev and Yakivchuk (1975b), this can be due to the more favorable conditions for lateral interactions via the substrate, namely, through the strongly protrud­ ing substrate atoms with unsaturated orbitals. As in the treatment of a chain of adatoms on a furrowed plane as a peculiar linear molecule consisting of alternating adatoms and substrate atoms, the (1 χ 1) lattice on the (111) plane can be regarded as a binary crystal. Of course, an approach like this is only an approximation, but its application to overlayers on loose faces possesses a certain validity, since in this case the adatoms find themselves greatly immersed in the substrate. The details of the concentration de­ pendences of the heat of adsorption of electropositive metals on different

374

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

q,eV

FIGURE 12.14. Concentration dependence of the heat of adsorption of Cs on tungsten faces (110) (curve 1, Fedorus and Naumovets, 1970) and (111) (curve 2, Medvedev and Yakivchuk, 1975b).

substrates can be employed in the implementation of various adsorption systems with predetermined thermal stability, such as thermoemitters (Hatsopulos and Gyftopoulos, 1979) and catalysts (Bonzel, 1987). The heats of adsorption for electropositive metals on semiconductors are less well understood than in metals. However, some interesting data are available on the influence of the reconstruction of a silicon surface on the heat of adsorption of alkali metals at small coverages (Greene et a/., 1982, 1984). Using the surface ionization method, these authors demonstrated that the transition of the Si (111) plane from the high-temperature ( l x l ) lattice to the low-temperature (7 χ 7) lattice leads to an increase in the heat of adsorption of alkali metals by 30-40%. This is testimony to the pronounced difference between adsorption sites on these surface lattices.

CATALYTIC PROPERTIES

375

In conclusion, we give a few more examples illustrating the specificity of the state of a substance in 2D layers. Meyer et al. (1988) discovered that a monolayer of lead on the (111) plane of copper has a melting temperature 200 Κ above that of bulk lead. This is considered to be caused by a compression of the monolayer by 3.2% as compared to bulk lead. The work by Daum et al. (1988) contains a detailed characterization of surface phonons in iron layers on the (100) plane of copper. These authors established the correlation of the layers' properties with thickness and the variation of their structure on reversible reconstruction caused by the displacement of atoms. Although the discussion has been focused only on adlayers deposited from the gas phase, one has to bear in mind that the structure of a surface also exerts a strong effect on the adsorption of substances dissolved in the interior of a substrate, i.e., on the process of segregation. For example, Samanta and Unertl (1986) have found that the concentration of carbon segregated to the surface varies sharply in the phase transition in a tellurium layer on the (111) plane of nickel. An overview of research on surface segregation is given in the contribution by Buck (1982) and the collection of papers edited by Johnson and Blakely (1979). This problem has a number of practical applications, since the strength of structural materials is largely determined by the behavior of grain boundaries, which, in its turn, is strongly dependent on segregation (Briant, 1982). Considerable progress has been made over the recent years in the study of the atomic structure of grain boundaries, which are also quasi-two-dimensional objects (see, for instance, the monograph by Kopetsky et al. (1987)). Research on the relation between the structure of grain boundaries and the segregation processes thereon is undoubtedly a promising area of scientific effort.

12.5. Catalytic Properties Since the kinetics of adsorption, surface diffusion, and desorption, as well as the electronic properties of clean and coated surfaces, are very dependent on the structure of the surface, one can reasonably assume that they will affect the rate of heterogeneous catalytic reactions too. The desire of researchers to examine the mechanisms of catalytic reactions and to improve the efficiency of catalysts is another motive for the study of surface phenomena (Ertl and Kuppers, 1979; Somorjai, 1981, 1984). Detailed research results are available on the series of model catalytic reactions of the carbon oxide hydrogenation type, on ammonia synthesis

376

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

reactions, and on the hydrocarbon conversion reactions proceeding on metal crystal surfaces and thoroughly evaluated by means of various diagnostic instruments. The general conclusion boils down to the fact that the catalytic reactions are very sensitive to the surface structure of a catalyst. By properly selecting a substrate with an optimal structure, one can appreciably step up the reaction rate and the selectivity of a catalyst. For example, the conversion of ^-hexane and ^-heptane to aromatic hydrocarbons on a platinum plane with a hexagonal lattice is 3 to 7 times more efficient than on a plane with a square lattice. On the other hand, the effect on reaction rates in the isobutaneto-^-butane transformation reaction on these substrates is the reverse. Under appropriate conditions, the ammonia synthesis rates on the (110), (100), and (111) planes of iron are related as 1:32:420 (Fig. 12.15). The reaction rates can be strongly affected by surface defects, including atom steps and atom-step kinks (the total effective concentration of defect adsorption sites in this case can be considerably higher than that of point

(Λ Ν

Ε ο w 500
Ε Γ

I

20 atm. 3:1, Η 2 : Ν 2 798 Κ

4001 300Η

CO

Ε 200 ο 5> 100 (Ο DC (110)

JZL

(100) (111) Crystal Face

FIGURE

12.15.

Dependence of the rate of iron-catalyzed ammonia synthesis on surface structure (Somorjai, 1984).

CATALYTIC PROPERTIES

377

defects). The rate of dissociation of hydrogen molecules on a stepped (111) plane of platinum is increased by about one order of magnitude. Also increased is the probability of breaking C - H and C - C bonds. There exists some optimal (reaction-specific) step density for which the desired reaction rate will be the highest (Somorjai, 1984). Another important problem is the sensitivity of catalysis to the structure of 2D phases formed by the adsorbates themselves. If the layer proves to be a heterophase one, there may occur a situation in which the reaction proceeds at the highest rate either in an individual phase or on the phase boundary. Naturally, this leads to different requirements for the optimization of an adlayer. It is also necessary to take into account the likelihood of the reciprocal action of adsorbates on the substrate, that is, of reconstructive adsorption, which changes the reactivity of the surface. All this can lead to self-sustained oscillations of the reaction rate. The nature of such an effect has been thoroughly studied for the CO oxidation reaction on the (100) plane of platinum (see the overview by Ertl (1987)). It is common knowledge that in a stable state this surface is reconstructed, which implies that its lattice has hexagonal, rather than square, symmetry we shall denote the latter by (1 χ 1)). However, upon adsorption of half a CO monolayer on the reconstructed (hexagonal) surface, the platinum (1 χ 1) lattice is restored. An 0 2 molecule filling the vacancy in the CO layer from the gas phase dissociates, and the oxygen atoms react with the neighboring molecules of CO, forming C 0 2 molecules, which are quickly desorbed (Fig. 12.16). As a result, the spot of uncovered surface, (1 χ 1), is expanded, with the arrival of new 0 2 molecules (the reaction proceeds when the pressure of 0 2 is an order of magnitude above that of CO) and the progressive removal of CO from the surface. However, if the coverage of CO drops to a certain critical value of θ « 0.3, there occurs a reconstruction of platinum surface into the hexagonal phase. The oxygen dissociative adsorption probability on it is two orders of magnitude lower than on the (1 χ 1) lattice. Therefore, the surface is again covered mainly by CO molecules, and the process is repeated. This example is a convincing illustration of the close relation between catalytic reactions and the phase state of the surface. Utilization of this relation opens up a new potential for controlling the characteristics of catalysts and their adjustment to a given chemical reaction. Recent years have seen vigorous exploration of the catalyst promotion effect with the use of up-to-date surface diagnostics. The subjects under study include the nature of interaction between promoters and the adsorbed substances, the effect of promoters at all stages of catalytic reactions, etc. For example, alkali-metal additives readily give off their electrons, thereby increasing the electron density on the surface, due to which the bond in the coadsorbed molecules is weakened and the dissociation of molecules becomes

378

EFFECTS OF STRUCTURE ON TWO-DIMENSIONAL SYSTEMS

Λ m>m22 χ

1*1

1X1

t

CO

0 2

0 0

c c

/

0 0 0

c c c

1X1 co2 0 \ CO c ^ o t o

c o

*

2

y

0

0

c c

0

1X1 0 0

CC

0

C — 0

0

c

0 0

c c

0

0

c

hex

'1*1

0 0 0 0 00 0O 0O 0O (O Q C C C CC C C C ( c c c c c t x

FIGURE

1χί 12.16.

Periodic structural transformations of a Pt ( 1 0 0 ) surface that occur in a mixed C O + 0 2 atmosphere and result in oscillations of the rate of catalytic oxidation of C O (see text). (Ertl, 1987.)

easier. Concerning this issue, see the reviews by Somorjai (1984), Norskov et al (1985), Goodman (1986), Bonzel (1987), Kiskinova (1988), and Bonzel and Krebs (1989). Another research trend encompasses the study of the mechanisms of poisoning of catalysts. In particular, the effect of carbon on the catalytic properties of the (111) plane of iridium was studied in the work by Zandberg et al (1972). The carbon in the 2D gas phase was found to poison the catalytic activity of iridium only slightly: at a coverage of θ = 0.3 the activity in relation to the dissociation of CsCl molecules is decreased less than twofold. At the same time, the activity in the condensed two-dimensional phase

CATALYTIC PROPERTIES

379

represented by a dense carbon layer with the structure of the basal (0001) 3 6 plane of graphite is reduced by a factor of 10 to 10 , a fact that the authors attribute to the formation of a surface with saturated valence links. The work function of an Ir (111) surface covered by the above carbon phases also shows a strong difference (by 30%). The same effect was observed for carbon on the surface of platinum, which served as a catalyst for hydrocarbon conversion reactions (Somorjai, 1984). Summarizing the results discussed in this chapter, it can be concluded that the diverse physicochemical properties of a surface are strongly dependent on its structure and on the phase transitions occuring on it. This awareness motivates further research into 2D crystals, including not only the surface systems treated here, but also all the other two-dimensional objects of which examples were cited in Chapter 1. Despite the diversity of these objects, it is hoped that they will show effects similar to those discussed above. Their study will allow comprehensive utilization of the properties of two-dimensional crystals.