New Ways of Looking at Surfaces

E.WARD PLUMMER, TORGNY GUSTAFSSON, DONALD R.HAMANN, INGOLF LINDAU, DOUGLAS L.MILLS, CALVIN F.QUATE, and Y.RONALD SHEN

The tremendous progress in surface science in the last 10 to 15 years is due primarily to the development of new theoretical and experimental techniques capable of probing the surface of materials in microscopic detail. The scanning tunneling microscope now shows the surface in near-atomic detail, revealing the surface order, steps, and defects. In the near future this instrument will be advanced to the stage where high-resolution inelastic tunneling spectra can be obtained for each atomic site, giving chemical as well as structural information. Experiments based on scattering of electrons, photons, and ions to determine structure have shown significant progress in the last few years. It is nearly correct to say that the static structure of any ordered surface can now be determined, thus placing emphasis on understanding why a given structure occurs. Recent experiments have examined changes in surface structure as a function of temperature or the presence of foreign impurities. X-ray scattering from adsorbate layers is used to explore the nature of two-dimensional melting for both physisorbed and chemisorbed layers. Recent ion-scattering experiments have shed light on the nature of surface melting of a lead crystal. The use of insertion devices on high-energy synchrotron sources will produce orders of magnitude more intensity for surface x-ray scattering experiments, making them nearly routine.

This next generation of insertion devices also opens up the possibility of doing magnetic x-ray scattering experiments from surfaces. The development of the double alignment technique for medium-energy ion scattering shows great promise for routine surface structural analysis even on multicomponent systems.

Synchrotron radiation facilities have become the workhorse of all forms of surface electron spectroscopy. The coupling of this light source with angle-



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 283
Advancing Materials Research New Ways of Looking at Surfaces E.WARD PLUMMER, TORGNY GUSTAFSSON, DONALD R.HAMANN, INGOLF LINDAU, DOUGLAS L.MILLS, CALVIN F.QUATE, and Y.RONALD SHEN The tremendous progress in surface science in the last 10 to 15 years is due primarily to the development of new theoretical and experimental techniques capable of probing the surface of materials in microscopic detail. The scanning tunneling microscope now shows the surface in near-atomic detail, revealing the surface order, steps, and defects. In the near future this instrument will be advanced to the stage where high-resolution inelastic tunneling spectra can be obtained for each atomic site, giving chemical as well as structural information. Experiments based on scattering of electrons, photons, and ions to determine structure have shown significant progress in the last few years. It is nearly correct to say that the static structure of any ordered surface can now be determined, thus placing emphasis on understanding why a given structure occurs. Recent experiments have examined changes in surface structure as a function of temperature or the presence of foreign impurities. X-ray scattering from adsorbate layers is used to explore the nature of two-dimensional melting for both physisorbed and chemisorbed layers. Recent ion-scattering experiments have shed light on the nature of surface melting of a lead crystal. The use of insertion devices on high-energy synchrotron sources will produce orders of magnitude more intensity for surface x-ray scattering experiments, making them nearly routine. This next generation of insertion devices also opens up the possibility of doing magnetic x-ray scattering experiments from surfaces. The development of the double alignment technique for medium-energy ion scattering shows great promise for routine surface structural analysis even on multicomponent systems. Synchrotron radiation facilities have become the workhorse of all forms of surface electron spectroscopy. The coupling of this light source with angle-

OCR for page 283
Advancing Materials Research resolved photoemission has given the community a detailed picture of the electronic states at a surface. For clean surfaces, almost every form of surface state imaginable has been identified and understood theoretically. Dangling-bond or back-bond surface states on semiconductor surfaces, free-electron surface states on metals as well as magnetic surface states have been documented. Studies of adsorbed atoms or molecules, or impurities segregated to the surface, have revealed the nature of the adsorbate-adsorbate and adsorbate-substrate interaction. Simple symmetry rules that exploit the polarized nature of the light source are used routinely to determine bonding geometry. The new insertion devices on the synchrotron, coupled with much more efficient detectors, will produce a sufficiently strong signal that spin-polarized angle-resolved photoemission will become common, enabling the experimentalist to observe with unprecedented detail the magnetic properties of the surface. The new sources will also enable experiments to be conducted with much higher energy and momentum resolution. One of the most exciting applications of synchrotron radiation sources to surface science has been to the area of core-level spectroscopy. Conventional core-level photoelectron spectroscopy measures the binding energy shifts of a specific element in different environments. The intensity from undulator sources, coupled with the resolution from specially designed monochromators, will give a hundred to a thousand times more signal with less than 0.1 eV resolution. For the first time the inherent line shape of core excitations from chemisorbed atoms or molecules will be accessible, yielding valuable information about the dynamics of the excitation process. The tunability of the synchrotron has allowed experimentalists to probe the near edge or threshold region, as well as monitoring the extended x-ray absorption fine structure (EXAFS) oscillations above threshold. This technique has yielded important new information about the geometry and chemical nature of atoms and molecules bound to the surface. The next generation of undulator source, coupled with new monochromator designs, promises to make this type of spectroscopy even more useful. With better resolution and more intensity, the dynamics of the excitation and decay process can be monitored. Recent experiments on an undulator have demonstrated the feasibility of coincidence experiments between electronic decay and ion fragmentation. The use of fluorescent detection will also provide a depth perception into the solid. Inverse photoemission has, in the last few years, begun to yield as detailed a picture of the unoccupied surface states as photoemission has for the occupied states. The series of Rydberg-like surface states trapped in the image potential have been observed for many metals, presenting a more detailed picture of the nature of this long-range surface potential. The future will see higher-resolution spectra as well as the use of spin-polarized sources. The

OCR for page 283
Advancing Materials Research characteristics of the unoccupied states are important for a complete understanding of chemical bonding at the surface. The development of high-resolution inelastic electron scattering has led the way in the study of surface vibrational modes. Simple selection rules that have been predicted theoretically and confirmed experimentally allow the experimentalist to determine the directionality of a given adsorbate vibrational mode and consequently the bonding geometry. Angle-dependent measurements are now mapping out the dispersion of intrinsic surface phonon modes as well as adsorbate-induced extrinsic phonon bands. It will not be long before direct correlations are seen between the surface phonon dispersion and reconstruction. The development of inelastic atom scattering will give a much higher resolution picture of surface phonons than can be achieved using inelastic electron scattering. Recent experiments have shown that time-resolved inelastic electron scattering can be used to monitor the time evolution of surface species. One of the most exciting prospects for the future is the marriage of laser technology to surface science. The development of tunable, intense, and ultrafast lasers will have a significant impact on surface science. Experiments that have already been completed demonstrate that tunable continuous-wave lasers can be used to measure vibrational modes on the surface with high resolution, monitor the vibrational or rotational states of desorbed molecules, measure the neutral desorption products following high-energy electron or photon excitation, or produce second harmonic generation from clean or adsorbate-covered surfaces. It is already apparent that there will be many more pump- and probe-type experiments in which, for example, a tunable infrared laser is used to preferentially excite an adsorbate molecule and a time-delayed visible laser pulse is used to monitor the second harmonic generation as a function of time after the initial excitation. These types of experiments will measure desorption times and identify intermediates on a picosecond time scale. The combination of a laser to excite and the synchrotron to probe matter holds great potential. Finally, the development of surface science has relied heavily on advances in theory. Unfortunately, most experimental probes must be strongly interactive to achieve surface sensitivity. This makes analysis of the data more complicated, requiring considerable theoretical support, and has produced a close collaboration between theorists and experimentalists. For example, advances in theory have been instrumental in the development of both inelastic electron scattering and photoemission. The new developments in atom scattering and nonlinear optical phenomena will also require considerable theoretical effort. The surface theorist must wear two hats, one for doing calculations related to a measurement technique and the other for conceptually oriented theory. Armed with an increasing data base and the progress that has been made in numerical calculations of

OCR for page 283
Advancing Materials Research the surface, the modern surface theorist is capable of actually predicting surface properties such as structure and phonon dispersion. The future will surely see a reversal of roles between the surface theorist and the experimentalist as many more theoretical predictions lead to experiments. This chapter shows the current state of techniques for “seeing” a surface. Many important and interesting areas of science will use these techniques to advance basic understanding and improve technology. The following sections present in more detail the current situation and future developments in the specific areas of surface theory, scanning tunneling microscopy, scattering experiments, electronic spectroscopy, vibrational spectroscopy, and laser-surface interactions. SURFACE THEORY Total Energies This section touches upon some of the most active areas of theoretical surface science. Certain active areas, such as the statistical mechanics of phase transitions on surfaces, have been omitted because the questions addressed are generally outside the mainstream of surface science. Much of the research described is heavily computational and aimed at specific systems. This situation should be regarded as a phase in the development of a young field and as the laying of foundations for a future period of generalization and conceptual advance. The most rapidly growing area in the theory of surfaces is the calculation of total energies. Given a geometric arrangement of atoms at a surface, the electronic energy of the system can now be calculated with an accuracy sufficient to determine preferred equilibrium structures, chemisorption bonding sites, and stiffness with respect to small displacements. The major breakthrough here came from the development of new methods to apply theory to extended systems. Parallel contributions have come from the introduction of better empirical models for extended surfaces, and from the incremental refinement of traditional quantum chemical methods for cluster models of surfaces. Local-density functional theory is an approximation to a rigorous theory of many-electron interactions, and is particularly applicable to extended systems. Early work had led to the belief that it was not sufficiently accurate for useful determinations of total energy. The development of norm-conserving pseudopotentials, a new and more accurate way of representing electron-ion interactions, led to the first mathematically converged solutions of the local-density equations. These results, for the structural energies of bulk semiconductors, gave remarkably accurate values for lattice constants, cohesive energies, high-pressure phase transitions, and phonon frequencies.

OCR for page 283
Advancing Materials Research The generalization of these total-energy calculations to slab representations of extended semiconductor surfaces is straightforward, and a number of groups now have these capabilities. The pseudopotential method is an efficient way of solving for the quantum-mechanical states of semiconductors and simple metals, but not of transition metals. The linear augmented planewave method, which is well suited for these materials, has also now been adapted to provide mathematically converged total energies within local-density functional theory. Although these are the most developed methods, variants using localized basis functions are being explored and should be capable of comparable accuracy. A parallel development has been the introduction of new empirical methods for total-energy calculations. So far limited to semiconductors, this approach entails a highly simplified quantum-mechanical contribution to the energy and a near-neighbor classical pair potential. Parameters are fitted to experimental data or results. These methods can be applied to far more complex surface structures than can other methods. Caution must be exercised, however, in situations where the bonding arrangements differ greatly from the fitted reference configurations. The above methods are all limited to the study of periodic surface structures. Isolated chemisorbed atoms or molecules as well as surface point defects have so far been treated by the application of quantum chemical molecular theory to clusters of atoms. The major difficulty with this approach is that the computational complexity grows rapidly with the size of the cluster, and it is seldom possible to establish convergence to the desired limit of an isolated entity on an extended surface. Recently, the first successful application of a direct approach to this class of problems was reported. It is based on local-density functional theory and uses a Green’s function method similar to one recently developed to treat point defects in bulk solids. Only a cluster consisting of the adsorbate and substrate atoms within a few screening lengths of the defect need be treated explicitly in this approach. Nevertheless, the effects of the rest of the substrate atoms are properly included. Total-energy calculations have already made substantial contributions to the solution of surface problems. One example concerns the reconstruction of the cleaved (111) surface of silicon. Calculations have shown that an old and widely accepted model for the atomic structure of this surface was unfavorable with respect to energy. A radically different arrangement, with its bonding topology altered from that of the bulk crystal, was found to result in a significant reduction in total energy. A further calculation has answered some strenuous objections to the new model by demonstrating that the energy barrier to be overcome in reaching this structure is in fact very small. Another recent success of these methods is the low-temperature reconstruction of tungsten, which involves small energies since it is destroyed by thermal motion at room temperature. The calculation showed that there was

OCR for page 283
Advancing Materials Research an arrangement of surface atoms with a slightly lower energy than the ideal arrangement of the bulk solid, and that it involved displacements of the atoms consistent with experimental results. A third example concerns the first stages of the epitaxial growth of nickel silicide on silicon. Calculations indicated that a proposed intermediate site for the nickel atoms was much less favorable than a site corresponding to the final configuration at the fully reacted interface. It can be anticipated that there will be substantial further growth in the number of groups engaged in total-energy calculations and in the variety of systems to which their results are applied. Although improvement in the basic physical approximations inherent in local-density theory is a long-term goal, the benefits of this method are far from being exhausted. Incremental improvement in computational efficiency can be expected, but trade-offs between accuracy and speed are inescapable. Substantial progress should be made in the new area of isolated adsorbates. The required numerical calculations nearly always consume large amounts of supercomputer time, and it is extremely important that such resources become more widely available. The variety and complexity of the mathematics involved make it improbable that special-purpose computers will be useful in these computations. These methods are expected to have a growing impact on all areas of surface science and related areas of technology, ranging from semiconductor device processing to catalyst design. There are many more materials for which studies of reconstruction and interface energies will be of value. These calculations can be expected to contribute to the solution of a large number of fundamental problems, such as adsorbate reactions, surface modification, and surface vibrations. Well-designed investigations should lead to a qualitative understanding of mechanisms and chemical trends and not just to the case at hand. Experimental Probes It is an unfortunate truism of surface science that surface-sensitive probes must interact strongly with matter and therefore require nontrivial theoretical effort to extract the desired information. One such area that has experienced rapid recent growth is the diffraction of monoenergetic atomic helium beams from surfaces. Though the utility of this method as a structural probe had been limited by the inability to relate the helium-surface interaction potential to surface atom positions, an approximately linear relationship between the repulsive potential acting on the helium atom and the surface charge density has been demonstrated. This quantity can be calculated from surface atom positions by employing simple approximation techniques or by methods related to total-energy calculations. Progress has also been made in the development of better ways to calculate diffraction intensities, which is itself

OCR for page 283
Advancing Materials Research a computationally intensive problem. Although advances have been impressive, better models of the weak attractive part of the potential and of the effects of the thermal motion of surface atoms on diffraction are needed. These same theoretical advances will also support a new experimental field— high-resolution inelastic helium scattering. Although surface phonon spectra can be extracted by kinematic analysis of such data, more complex calculations are required to extract the information contained in intensities and line shapes. Scanning tunneling microscopy is an exciting new development that yields real-space topographic images of surfaces at near-atomic resolution. A qualitative interpretation of these images can be made independent of theory, but theory is required to understand the interplay of atomic positions and electronic effects in determining the detailed shape of the topograph. A recently developed basic theory of low-voltage topographs also describes the effective resolution. There are many unexplored possibilities for these instruments, such as localized tunneling spectroscopy of electronic states and vibrational excitations. At high voltages the microscope tip strongly perturbs the electronic structure of the surface, and theoretical advances will be needed to unscramble such spectra. As tunneling tip preparation becomes reproducible and the atomic structure of high-resolution tips becomes known, a new level of the theory of the resolution will be possible. With the increased availability of synchrotron radiation sources, angle-resolved photoemission spectra measuring bulk- and surface-energy bands proliferate. Theoretical methods for calculating such bands have been developing since the early 1970s. All are essentially based on the local-density functional approximation and form a part of total-energy calculations. Local-density theory is not as accurate for excitations as for total energies, but reasonable agreement with measured bands has been found in many cases. Recently, the first fully systematic study of the energy bands of a real solid, based on a more sophisticated many-body approximation, was reported for silicon. The improvement in agreement with experiment was striking. The task of bringing this level of theory to bear on surface problems will certainly be addressed in the near future. Limitations of even the largest supercomputers appear to preclude a straightforward extension at present. Another class of related problems is that of localized electronic excitations, such as atomic core levels or internal valence states of adsorbed molecules. The spectra of electrons emitted following such excitations contain complex structure, which is strongly dependent upon the energy of the exciting photon. Many-body effects are often dominant in such processes. Desorption of atoms by incident photons or electrons is believed to involve related complexities. There is considerable theoretical activity in these areas, and future progress can be expected. It is obvious that considerable theoretical effort will be required to fully

OCR for page 283
Advancing Materials Research understand the new class of nonlinear optical experiments being conducted on single-crystal surfaces. The local fields at the surface must be calculated as a function of the laser frequency and electronic properties of the surface. The mechanism for second-harmonic generation from a clean or adsorbate-covered surface is not fully understood at the present time. The experimental tools for determining surface atomic geometry have impact on all areas, as discussed in connection with theoretical determination of surface structure. Spectroscopic tools can go further in that the electronic states they probe provide the mechanism of bonding at the surface. More information about surface chemical activity is potentially available from the interpretation of such data. The technique of “fingerprinting,” using a spectroscopic signature to identify a particular surface species such as a molecular fragment, is of value in both applied and basic surface research. The reliability of this technique depends on a thorough understanding of many-body effects on spectra. Kinetics Considerable recent progress has been made in understanding the dynamics of gas-surface interactions. The cases for which the theory is well in hand involve nonreactive situations in which the gas consists of either inert atoms or diatomic molecules that do not dissociate on the surface considered. These problems have been treated principally by molecular dynamics computer simulations using empirical classical interatomic potentials. The results have provided a fairly complete picture of energy exchange among translational, vibrational, and rotational energies of the gas and vibrational energy of the solid. They are in substantial agreement with the results of molecular-beam laser-probe experiments that measure such energy exchanges. Other theoretical approaches have recently shown that coupling to electronic excitations in metal substrates is a dominant energy-exchange mechanism only for high-frequency molecular vibrations. The long-standing puzzle of the so-called precursor state, a weakly bound state believed to precede the chemisorbed state of some molecules, may in the future be within the grasp of theorists in this field. An understanding of the kinetics of chemical reactions at surfaces is a considerably longer-term goal. The chief limitation here is the lack of knowledge of the so-called potential function, the total energy as it depends continuously on the coordinates of all the active atoms. Although pair-potential models are not realistic for dissociative reactions, it is hoped that total-energy calculations of the sort described in the first section can provide needed insight toward describing this quantity. Progress is expected to come through an interplay between preliminary calculations and progressively refined model potentials.

OCR for page 283
Advancing Materials Research Recent progress has been made in understanding the kinetics of charge transfer at surfaces. The results obtained explain ion yields in sputtering experiments and their dependence on such surface properties as work function. The new methods may provide a key to understanding ion yield in other experiments such as induced desorption and, in the longer term, the dynamics of charge transfer as it occurs in many surface chemical reactions. Static characterization of the surface provides important clues to dynamic behavior but cannot provide rate constants for surface processes. Fundamental research on surface kinetics has the potential to provide valuable directions in applied areas ranging from catalyst design to the growth of exotic materials. Competing rates determine what product will dominate or what structure will grow, and whether the yields will be of practical value. Interfaces A surface may be considered a solid-vacuum or solid-gas interface. Solid-solid interfaces are an active area of study that involves many of the same theoretical issues. Theory is stimulated by the large experimental effort in interfaces, which involves many of the same techniques and even some of the same experimentalists as surface research. Semiconductor electronic devices are essentially collections of solid-solid interfaces, and basic advances in this area are directly relevant to applied research for the electronics industry. The class of systems that has received the greatest amount of theoretical study is the semiconductor-metal interface, or Schottky barrier. Detailed local-density functional calculations have been carried out for semiconductors in contact with a simple model of a metal in which the metal ions are smeared into a uniform positive background charge. Although these theories produce correct trends for the barrier heights, they cannot be directly compared with experiment, and alternative models in which interface defects determine the barrier height have a large following. The existence of new systems in which metallic silicon compounds are grown in perfect epitaxial registry on silicon may be sufficiently ideal that they can provide a meeting ground for theory and experiment leading to the resolution of key issues. Many theoretical questions such as electron transport through metal-semiconductor interfaces remain to be addressed. The ability of the local-density functional method to give accurate band positions may be a more important factor in the theory of interfaces than in the theory of surfaces. Semiconductor energy gaps are often in error by as much as 50 percent, yet interface behavior is determined by differences in energy level that are small in relation to the gaps. Advances in generalizing more sophisticated many-body formulations to surfaces may have their biggest impact on the understanding of interfaces. Semiconductor-semiconductor interfaces, known as heterojunctions, have

OCR for page 283
Advancing Materials Research received attention since it became possible to grow them as ideal abrupt structures by molecular beam epitaxy. Such structures offer a tremendous range of possibilities for novel semiconductor device designs. Multilayered collections of these interfaces form artificial materials that can be tailored to have exotic properties. Of key interest is the alignment of the energy bands of the two materials across such interfaces. Several theoretical models for calculating this alignment exist, but they are based on conflicting mechanisms. Detailed calculations are hampered by the gap problems already discussed. Recent progress has been made in formulating improved theories of electron confinement and transport in heterojunction structures, but many important questions in this area remain to be pursued. Existing theoretical methods can do little with incommensurate interfaces or crystal interfaces with disordered materials. Passivating layers on semiconductor devices fall in this category and are of extreme technological importance. Solid-liquid interfaces are an even broader example, encompassing all of electrochemistry. It is premature to predict bold steps forward in these areas, but it is clear that they will receive increasing attention. SCANNING TUNNELING MICROSCOPE The possibility of seeing atoms has always had a romantic attraction for both scientists and nonscientists.* The scanning tunneling microscope, pushed to its ultimate capability, appears to give a real-space picture of the atomic structure at the surface. This section describes briefly how this new instrument works, presents a few research highlights, and looks to its prospects for the future. The principal tools employed in the study of surface atomic structure have made use of elastic scattering experiments, which use incident beams of photons, electrons, atoms, or ions (see the following section). While these experiments give a picture of the atomic positions at a surface, almost everyone would agree that the experiments do not “see” surface atoms. The experiments require many atoms arranged in a given structure. Imaging techniques allow one to see the surface directly. Historically the first images of atoms were obtained with the field-ion microscope. A second means by which images of atoms can be obtained is with the electron microscope. Recent experiments have obtained the images of atomic planes near a surface. The newest development in this area is the tunneling microscope. As the name suggests, this microscope operates by the mechanism of *   This section draws on an article by J.A.Golovchenko, Science 232, 48 (1986).

OCR for page 283
Advancing Materials Research tunneling. Electrons tunnel from a small metallic tip held above the surface, across a vacuum barrier, into the surface if the tip is biased negatively with respect to the surface or vice versa if the bias is changed. The vertical resolution is achieved by the exponential dependence of the tunneling current on the tip-to-surface separation. A typical variation in current is an order of magnitude for every angstrom of separation. An image in the lateral direction is achieved by scanning the tip across the surface. The lateral resolution is determined primarily by the size and shape of the “probe” or tip. Ultimately, experimentally realizable resolution in the vertical direction is dictated by the electronic and vibrational stability of the individual instrument, typically 0.1 angstrom. The horizontal or lateral resolution (2 to 3 angstroms) depends not only on the stability but upon the size and shape of the tip. An “image” is usually obtained by moving the tip across the surface with a piezoelectric x-y scan while maintaining a constant current (fixed height) with a third piezoelectric device. The many experimental difficulties associated with vibrations, electronic stability, and tip shape will not be discussed here. A fundamental problem of this technique is one of interpretation: How does one translate tunneling images into pictures that reflect the identity and position of individual atoms of an unsolved structure? Solutions tend to be part science, part art, and the subject of continuous discussion. Both an advantage and a disadvantage of scanning tunneling microscopy is that the pictures obtained are in general a mixture of spectroscopy and microscopy. It is tempting to say that the tunneling microscope sees atoms, but, in fact, electrons tunneling from the Fermi energy of the tip see the spatial characteristics of the local density of states of the surface at an energy level equal to the bias voltage, which is higher than the Fermi energy. Therefore, the image is a function of bias voltage. This cross between microscopy and spectroscopy has tremendous advantages if we can learn to use it properly. Keeping in mind the caveats listed above, the following few paragraphs describe experimental observations, beginning with structural studies where the reconstruction of clean metal and semiconductor surfaces has been observed. These observations include images of gold, silicon, germanium, and gallium arsenide. The steps on a Si(111) surface have been imaged and show how the surface defects causing the surface reconstruction are incorporated into the step edge. Images of germanium-silicon alloys indicate that at the surface there is an ordered GeSi alloy. These structural studies are just beginning to present a picture of the role of defects such as step edges and vacancies on such surface processes as epitaxial growth and catalysis. Preliminary steps are being taken to image foreign atoms on surfaces, including large molecules such as DNA or viruses. The tunneling microscope can also provide images of the change in the electronic distribution at the surface. Two recent examples hint at future

OCR for page 283
Advancing Materials Research to adjust to the presence of the projectile. These conditions are fulfilled when using protons with an energy above, say, 25 keV. The technique becomes surface specific through appropriate alignment of the ion beam with the target. The magnitude of the backscattered flux will therefore be proportional to the number of surface atoms visible to the ion beam, assuming that the atoms are frozen in their rigid lattice positions. For nonzero vibrational amplitudes, the number of visible atoms (and hence the backscattered flux) will increase as the shadowing of near-surface atoms becomes less effective. A measurement of the backscattered flux will therefore provide information about displacements in the surface in the direction perpendicular to the beam. If, for example, the atoms in the surface have moved laterally, they will shadow the atoms in the second layer less completely. This will result in an increase of the backscattered flux, and the magnitude of the increase will depend on the relative movements of the atoms. A particularly useful variant of ion scattering is the measurement of the angular distribution of the backscattered flux: in the directions joining second-layer atoms with those in the first layer, the flux from the second layer will be blocked by that from the first layer, resulting in a reduced flux (“blocking dip”) at those particular directions. The positions of these blocking dips are therefore, to a first-order approximation, a direct measure of the relative positions of these two atoms. To quantify such data, the scattering process can be simulated on a computer using Monte Carlo techniques. To perform these types of experiments, a new class of high-resolution ion energy analyzers were constructed. The principal advantage of these new instruments is that they enable quantitative depth analysis of Rutherford backscattering with unprecedented resolution (3–5 angstroms). Another advantage of the technique is that of mass dispersion. For example, in studies of binary systems, it is useful to have a separate signal from each of the two constituents rather than a superposition of the two signals. Ion scattering is a local probe, as the interaction is kinematic. X-ray scattering, on the other hand, is a diffraction phenomenon and yields information about the long-range order. It is only recently that x-ray scattering has been applied to surface structural problems. The difficulty is that x rays are deeply penetrating and that the surface signal is therefore only a small part of the total diffracted intensity. This technique is therefore particularly suitable for studies of systems where the bulk and the surface have different periodicities. However, only with intense sources is it possible to obtain an acceptable signal-to-noise ratio. Even the strongest laboratory source available today (60-kW rotating anode) gives, from the (7×7) structure on Si(111), only 10–4 diffracted photons per second in a typical reflex, making analysis impossible. However, wigglers now under construction at various synchrotron radiation sources, will increase this flux by 5 orders of magnitude, yielding approximately 10 counts per second (cps). For heavier ma-

OCR for page 283
Advancing Materials Research terials like the (1×2) reconstruction of Au(110), the intensity will increase to 106 cps. There are few fields of science where such enormous intensity enhancements have taken place in such a short time. Though few mainstream surface systems have yet been studied with x-ray diffraction, these early experiments have provided valuable symmetry information. Additionally, unique information has been obtained about step densities on surfaces. Structural determinations, however, must rely rather heavily on measurements of the diffracted intensity as a function of momentum transfer. Considering these problems, it is not surprising that the actual atomic positions derived from x-ray data are being refined with other structural techniques, a situation that will change dramatically within a few years. Results Surface structural investigations are at a stage where different investigators, in different laboratories, using different techniques, studying the same system, can obtain the same results. The multilayer relaxations on Cu(110) and Ni(110) discussed below have been determined with LEED and with ion scattering, and the two experiments agree with respect to all structural parameters to better than 0.01 angstrom. This length scale is useful for discussing structural changes in chemical reactions in the gas phase. It therefore appears likely that within the next several years we will be investigating the geometrical adsorbate-substrate response to a chemical reaction with the accuracy necessary to have an impact on problems of this kind. Another class of systems where ion scattering and LEED agree extremely well includes the (110) faces of GaSb, InAs, and GaAs. These surfaces consist of equal numbers of anions and cations located in an ideal crystal in the plane of the surface. It has been known for several years that, at the surface, the anion moves out and the cation moves in. It is now known that the axis joining the two forms an angle of 29 degrees with the surface plane and the agreement between the two techniques is 1 degree or better. The two techniques have different strengths, nevertheless: ion scattering has excellent sensitivity to lateral movements, and LEED is sensitive to normal displacements. The surface structure of metals was long considered a relatively uninteresting field. Metallic screening at the surface, it was argued, is so short ranged that only minor rearrangements of the atoms would be expected at the surface. This view has changed tremendously over the last few years with the observation of phase transitions on clean metal surfaces as well as those induced by adsorbates. Surface melting has been observed at a temperature far below the bulk melting temperature. Even the structure of ideal surfaces has shown some unexpected features. It has been found, for example, that on clean metal surfaces the separation

OCR for page 283
Advancing Materials Research of the outermost layer is smaller than the average bulk value, while the separation between layers two and three is expanded, and so on in an oscillatory fashion. These results, which (as mentioned above) have been obtained on several surfaces by different groups using both LEED and ion scattering, are in qualitative agreement with theoretical predictions and are caused by the rearrangement of the electronic charge density at the surface. More dramatically, on high-index surfaces such as certain iron faces, the entire first layer of atoms is displaced horizontally while the bulk periodicity is maintained, again as predicted theoretically. Several metal surfaces— W(100), Mo(100), Pd(110), Au(110)—change their two-dimensional superstructure as a function of temperature. Intense theoretical and experimental work is being performed to characterize these new structures. Adsorbates have also been found to induce reconstructions. A particularly interesting case is the (110) face of copper, where oxygen adsorption doubles the size of the unit cell in one azimuth and alkali metals or hydrogen double it in another. This therefore suggests that surface structure can to a certain extent be “custom made” by a suitable choice of adsorbates. The advance of the next decade will be due largely to the more widespread use of ion scattering and the coming revolution in x-ray diffraction. This coming of age is very timely: electron and vibrational spectroscopy are at a stage where all three techniques can be used to attack problems in surface phase transitions on an unprecedented scale. These techniques will make possible new approaches to the study of structural phenomena in surface chemistry and much greater detail in the investigation of surface structure of multicomponent systems, including silicides and alloys. ELECTRONIC SPECTROSCOPIES Spectroscopic tools for measuring the energies of electronic transitions fall into two classes depending upon the excitation source—monochromatic photons or electrons. Since any discussion of surface experiments involving incident photons must focus on synchrotron radiation sources, the following discussion is divided into two parts: synchrotron radiation experiments and all others. With the exceptions of inverse photoemission and laser-induced absorption or multiphoton ionization, the sources of excitation in these two classes of surface experiments are, respectively, photons and electrons. Synchrotron Radiation Experiments For convenience, synchrotron radiation experiments are divided into two categories—valence-band and core-level spectroscopy. Valence-band angle-resolved photoemission has become to electronic structure what x-ray scattering is to crystallography and neutron scattering

OCR for page 283
Advancing Materials Research is to phonon structure. In general, given a single crystal, the three-dimensional bulk or the two-dimensional surface band structure can be determined. The unique capabilities of a polarized, tunable, and intense radiation source enable the experimenter to tune to any portion of the two- or three-dimensional band structure and to determine the orbital symmetry of each state. Intrinsic surface states or resonance of almost every conceivable type have been identified: dangling-bond or back-bond surface states on semiconductors, free-electron or d-band surface states on metals, exchange split surface states on ferromagnetic metals. Obviously the surface electronic properties are intimately related to the surface structure. Photoemission data first led to the proposal of the chain model of the Si(111) surface. Recent high-resolution angle-resolved experiments showed that a metal-insulator transition accompanied the order-disorder transition on Ge(001). In general, angle-resolved photoemission has become an established tool for probing the occupied electronic states of the surface. It has recently been used also in the study of many materials phenomena, including intermetallic surfaces, rare-earth compounds, intercalated compounds, thin metallic layers on metals or semiconductors, coadsorption systems, two-dimensional phases of physisorbed molecules, segregated bulk impurities, and alkali-promoted adsorption. Yet several important improvements can and will be made. First, the energy and momentum resolution should be improved through the use of better monochromators and analyzers. The most important improvement in angle-resolved photoemission will come when more efficient spin-polarized detectors are incorporated into the electron analyzers. The new undulator insertion devices on electron storage rings, which will produce a sufficient number of photons to offset the lower efficiency of the spin detector, should make spin-polarized photoemission as routine in a couple of years as photoemission is now. The next-generation insertion device source coupled with developments in monochromator design and soft x-ray optics will have a dramatic effect on core-level spectroscopy. It is reasonable to expect enhancements in signal strength for x-ray photoelectron spectroscopy, usually called electron spectroscopy for chemical analysis (ESCA), by 102 to 103 as well as significant improvements in energy resolution. The resolution of ESCA systems is limited to approximately 0.2 eV by the inherent properties of the crystal monochromator. Design studies demonstrate that new soft x-ray monochromators should operate in the energy range of 100 to 1000 eV with better than 0.1 eV resolution. This increase in intensity and resolution will make it easier to observe low surface concentrations or to study samples that are rapidly contaminated. But the newfound resolution will also open up the possibility of studying the dynamics of core-hole ionization by measuring the inherent line shape. Recent data indicate that the width of the peak in the core-level photoelectron spectra from an adsorbed molecule is not due to lifetime broad-

OCR for page 283
Advancing Materials Research ening. Therefore, a line-shape analysis will make visible phonon and electron excitation sidebands. The tunability of the synchrotron source has enabled experimentalists to use the core-level absorption spectra of adsorbed atoms and molecules to study the bonding configuration on the surface. Core-to-bound excitations have a specific symmetry, which can be probed by a specific polarization direction. The direction of the polarization vector with respect to the surface when the core-to-bound excitation is allowed (or forbidden) transparently fixes the orientation of the molecule. Changes in the energy position of the core-to-bound excitation have been used to measure changes in the intermolecular bond length. The small oscillations in the absorption cross section far above the threshold—EXAFS—is used to determine distances to the nearest and the next-nearest neighbor. Polarization dependence of the surface EXAFS signal can define the directions of the bonds. More intense sources will improve the signal-to-noise ratio for these experiments, thus allowing the study of moieties of lower surface concentrations. The use of fluorescence detection and of higher-energy monochromators will give these types of experiments a depth perception into the bulk. It is now possible to study the dynamics of core-hole decay and ion fragmentation (photostimulated desorption) following the absorption of a soft x-ray photon. Consider the situation in which the photon energy is tuned to a core-to-bound excitation of either an adsorbed molecule or of the solid (exiton). If the lifetime of the core hole is shorter than the time required for the excited electron to move away from the hole, then the excited electron can participate in the decay process. When this happens the final state is a single hole on the atom or molecule originally excited. This single-hole final state is exactly the same as created by the direct photoionization of a valence electron. However, the direct photoemission process samples the global nature of the wave function, and the deexcitation process views the valence states through the window of the localized core hole. Because there have been only a few experiments of this type, it is too early to evaluate the true potential of this technique. One may conclude, however, that this deexcitation spectroscopy may be used to probe the local electronic configuration. The details of the deexcitation spectra are already yielding new information about the dynamics of core-hole decay. The second new development is the ability to measure the ion fragmentation products in coincidence with the energy-resolved electron deexcitation spectra. These measurements can make a unique identification of the hole configuration with the fragmentation products. Laboratory Sources Inverse photoemission occurs when an incident electron loses energy as it interacts with the surface, a process resulting in an emitted photon. This

OCR for page 283
Advancing Materials Research spectroscopy probes the unoccupied states in a region between the Fermi energy and the vacuum level that is not accessible by other techniques. Just as in the case of photoemission, angle-dependent inverse photoemission has been used to map the two- and three-dimensional unoccupied bands of the surface and bulk, respectively. The greatest potential in surface science for inverse photoemission appears to be in the area of surface chemistry, where the position and dispersion of unoccupied energy levels are as important as the occupied levels. The coupling of these types of studies with the ability to use spin-polarized sources to investigate magnetic changes induced by chemisorption yields a powerful and complementary surface tool. Inelastic electron scattering (electronic excitations) was one of the earliest probes used in studies of surfaces. Yet, it has not developed into a commonly used surface technique. Several recent developments indicate that this technique may become more important in the future. Measurements of the valence-band losses for an adsorbed molecule as a function of the incident electron energy and scattering angle indicate that surface losses may be separable from bulk losses and that dipole and nondipole excitation can be identified. New theoretical interest in the inelastic-scattering cross sections from oriented molecules may help unravel the data. It is now obvious that core-level loss spectra can be used in conjunction with photoabsorption data, with the advantage that the electron-loss experiment can detect optically forbidden transitions. It is also clear that 3e-e or 3e-ion coincident experiments will soon be conducted on a surface. The first coincidence experiment between the energy-resolved Auger decay and the photostimulated desorption products was just reported. As described in the foregoing discussion of synchrotron radiation, these coincidence experiments will determine the hole configuration responsible for desorption. New technology will make it possible to do gas-phase e-2e experiments on the surface such that the energy-resolved, scattered, and emitted electrons are detected in coincidence. This e-2e experiment is capable of measuring the momentum distribution of a specific energy level or, used in a different mode, capable of measuring the surface EXAFS signal. The improvements in source intensity and detector efficiency will bring on line more coincidence or two-probe experiments. Unoccupied surface states could be probed with a two-photon process in which molecules can be vibrationally excited and then ionized. VIBRATIONAL SPECTROSCOPY OF THE CRYSTAL SURFACES In molecular physics, as well as condensed-matter physics, access to data on the frequencies and nature of the vibrational modes of molecules or crystals has led to major new insights into their geometry, the nature of their chemical bonds, and the dynamic properties of, or energy transfer between, the basic constituents. Various forms of vibrational spectroscopy have been applied

OCR for page 283
Advancing Materials Research to the study of adsorbates on surfaces (and, in some instances, atoms within the surface itself) for nearly two decades. However, the information that could be extracted from the data has been limited, and compromised by the limited resolution of probes used for this purpose. In the past five years qualitative advances in experimental method have led to an explosive expansion in the data available, and we are in a position to explore many new questions that simply could not be addressed in the recent past. The near future looks exciting indeed. For the first time, it has proved possible to perform detailed studies of the dispersion curves of surface phonons on clean and adsorbate-covered surfaces. This is done through the use of particle probes in energy-loss experiments that are a surface analogue of the neutron-scattering experiments that have proved so powerful in studying crystals. Two complementary methods may be used to study surface phonons. One method makes use of highly monoenergetic, well-collimated beams of neutral helium atoms directed at the surface. Their kinetic energy is in the thermal range, and energy resolutions of a few tenths of a millivolt (three or four wave numbers) can be realized. Surface phonons have already been studied for a range of materials, from insulators to metals. Shortly after helium beams were used in the first studies of surface phonons, a second method was demonstrated. Electron energy-loss spectroscopy as a function of scattering angle was used to study clean and adsorbate-covered nickel surfaces. The key experimental development was the use of much higher energies (50 eV-300 eV) than are used in the more traditional spectrometer. One must use such high energies to probe well out into the Brillouin zone, and the experimental challenge is to produce high-energy beams, sufficiently monoenergetic to allow resolution of the low-energy losses associated with scattering from surface phonons. The two methods are complementary in that the helium beams may be used to study rather low-frequency surface phonons whose excitation energy is far too small to be resolved by the electron energy-loss method. One such example is the study of vibrational modes of rare-gas monolayers, bilayers, and trilayers adsorbed on Ag(111) by using helium beams. Here dispersion curves are studied in the frequency range from 10 cm–1 (1.2 meV) to 30 cm–1 (3.7 meV). The electron energy-loss method can detect modes as “soft” as 30 cm–1 (3.7 meV), but is best suited to the study of the spectral range above 100 cm–1 (12 meV). There is no upper bound to the frequency range that may be explored with electrons, whereas multiphonon scattering obscures single phonon features in spectra taken with helium beams above roughly 250 cm–1 (30 meV). Electrons can also resolve modes in which the atomic motion is predominantly parallel to the surface, and the extraction of useful information from the spectra will be assisted by the appearance of a quantitative theory of the excitation process, which can be used to predict energy regimes within which excitation cross sections are enhanced.

OCR for page 283
Advancing Materials Research In the near future we may expect these two methods to be applied to the study of areas of surface dynamics where little information is available at present. The dynamics of the surface may be studied as its temperature is raised to the point where melting begins. A modest improvement in the resolution of the helium-beam method will allow the magnitude and temperature variation of the surface-phonon line width to be studied at various points in the two-dimensional Brillouin zone. This will provide fundamental insights into the nature of vibrational energy transfer between adsorbed layers and the underlying substrate. Many new questions such as these will be addressed in the coming years. Infrared spectroscopy offers, in principle, much higher resolution than either of the particle probes discussed above. The line width of narrow high-frequency adsorbate modes may thus be studied, along with subtle splittings that contain important information about the symmetry of the adsorption site. This technique has also developed impressively in recent years; there have been high-resolution studies of hydrogen adsorbed on silicon surfaces, and subtle line-shape anomalies have been resolved in recent work. Thermal emission studies of adsorbate-covered surfaces have led to the quantitative study of the low-frequency vibration of carbon monoxide on the Ni(100) surface. The development of tunable continuous-wave infrared lasers will have a large impact on the infrared spectroscopy of surfaces. The appearance of particle probes that can explore the dispersion of surface phonons (and surface resonances) throughout the Brillouin zone and the dramatic improvement in the infrared technique have created a new era of surface-vibrational spectroscopy. The field is evolving rapidly at this point, and the next five years will see numerous important results emerge from the laboratory. The time scale for collecting an inelastic loss spectrum from a surface can be reduced considerably by using multichannel detection and dispersion compensation allowing real-time data collection. This technique has already been used with temperature pulsing to study the kinetics of decomposition versus desorption of an adsorbed molecule. Since electron-loss spectroscopy (ELS) is site and molecular specific, the time evolution of the molecular orientation, binding site, and intermediates on the surface can be followed. The single-shot time resolution at present is approximately 5 ms, whereas a reversible process can be resolved at the 1-ms level. Future experiments will use pulsed gas beams, higher pressure, and laser excitations and will probably improve the time resolution by at least one order of magnitude. Light scattering (Raman and Brillouin) has been a mainstay of vibrational spectroscopy in solid-state physics and can also be applied to surface and interface studies. Under conditions outlined in the literature, the signals can be enhanced enormously, though appreciable roughness of a surface or interface is a key requirement. Thus, at least in this mode, light-scattering spectroscopy is not used to study the clean, carefully prepared surface struc-

OCR for page 283
Advancing Materials Research tures that are the backbone of modern surface science. However, on clean, high-quality surfaces, Raman spectroscopy has been used to study a small number of adsorbed monolayers, and in some cases signals from a single monolayer have been reported. There is still an unresolved question concerning the possibility of tuning the laser into an electronic excitation to do resonant Raman spectroscopy. If it is possible to do resonant Raman spectroscopy on adsorbed atoms or molecules, then Raman spectroscopy will become useful for surface science. If this is not possible, then Raman spectroscopy is likely to emerge as a useful probe of systems that contain a few monolayers of adsorbates in molecular form. Brillouin spectroscopy offers promise in the study of the liquid-solid interface, as do certain sophisticated forms of modern multibeam nonlinear spectroscopy (coherent antistokes Raman spectroscopy, for example). During the next decade, light-scattering and nonlinear laser spectroscopies may be expected to play an increasingly important role in the study of the “ideal” surface or interface environment, though at present the amount of unambiguous information derived from those methods is limited. LASER-SURFACE INTERACTION The last decade saw tremendous growth of interest in the area of laser interaction with surfaces. The high-intensity, coherent, and directional nature of laser beams provides unique possibilities for studying, controlling, and modifying surfaces. The advances are generally on two fronts: development of laser techniques for probing surfaces, and development of laser techniques for processing the surfaces of materials. Numerous methods of laser surface probing have been invented in the past five years, and most of them are spectroscopic. Photoacoustic and photothermal spectroscopies rely on surface resonant absorption of laser photons and subsequent conversion of these photons into acoustic waves or heat, which are then detected. These techniques have been used to obtain vibrational spectra of adsorbed molecules and to probe the surface states of reconstructed semiconductor surfaces. Photodesorption spectroscopy is based on resonant excitation of adsorbed molecules and subsequent desorption of the molecules from the surface. If the desorption is a resonant process, then this technique can also be used to obtain vibrational spectra of adsorbed molecules. Furthermore, valuable information about the laser desorption process, which is essential to an understanding of laser material processing, can be obtained from laser-stimulated desorption. Raman scattering and fluorescence spectroscopies have also been applied to surface studies. The recent development of detectors and computer-assisted detection technology has dramatically improved the sensitivity of such mea-

OCR for page 283
Advancing Materials Research surements. With but a marginal improvement in the technique, it will become easy to obtain a Raman spectrum from a monolayer of adsorbed molecules on a smooth surface. This is very exciting, as one would then have a relatively simple optical technique that could yield the vibrational spectrum of adsorbed molecules in a single scan. These laser techniques have, at least in principle, the advantages over other surface tools in that they possess a high spectral resolution, can probe an interface between two dense media as long as one of them is transparent, and are applicable to time-resolving studies if a pulsed laser is used. The spectrum of adsorbed molecules provides direct information about the geometry and strength of the molecule-substrate interaction, which is the basis of all surface reactions, including catalysis, corrosion, and adhesion. Another effective method for studying molecule-surface interaction is to measure the redistribution of energy in various degrees of freedom resulting from collisions of molecules with a surface. The highly sensitive laser spectroscopic techniques are ideal for such measurements. The state-selective detection ability of these techniques makes it possible to measure not only the translational energy distribution of the molecules before and after collisions but also their internal energy distribution. Thus, the exchange of rotational, vibrational, and electronic energy states of the molecule resulting from surface collision can be probed directly. Such information is important for the understanding of surface reactions and is difficult to obtain by conventional means. Nonlinear optical effects can also be used to probe surfaces. Among the few possibilities that have been explored, second-harmonic generation appears to be the most sensitive and versatile. Its surface specificity is based on the fact that this process is generally forbidden in a bulk with inversion symmetry but always allowed at a surface or interface. This technique has numerous proven advantages over other surface tools. It is nondestructive, capable of in situ remote sensing with subpicosecond time resolution, and applicable to all types of interfaces. This technique can be used to measure the adsorption isotherm; the symmetry of surface structure and molecular arrangement; the spectrum and orientation of adsorbed molecules; the time development of adsorption, desorption, and epitaxial growth; and many other properties. In addition, this technique provides the prospect of probing real-time surface dynamics and reactions with ultrafast time resolution on a wide range of interfaces. Applications of this technique to science and technology are limited only by the imagination. Examples include studies of electrochemistry at a liquid-solid interface, surfactants at liquid-liquid and liquid-solid interfaces, biological membranes, oxidation of silicon surfaces, and surface corrosion and catalysis. All these problems are intimately related to the future progress of modern science and technology. Laser surface techniques provide new ways to probe surfaces and molecule-

OCR for page 283
Advancing Materials Research surface interactions. They can yield valuable information complementary to that obtained by other means. The field is still in the beginning stage: the existing techniques need further development, and many other possible techniques are yet to be explored. The practical importance of this research area is evidenced by the recent entrance of many industrial and national laboratories into the field. Lasers already play a significant role in the area of surface processing. Although most of the laser techniques are still in the developmental stage, a few have reached the production line. Laser annealing is probably the best-known process: the recrystallization of the surface after laser-induced melting can yield high-quality crystalline films. With a pulsed laser, it is even possible to obtain selectively a crystalline or amorphous film depending on the duration of the pulse. Laser heating of surfaces is also the basis of a number of other laser surface-processing techniques. These include laser-enhanced etching, laser-enhanced electroplating, and laser alloying. In the former two cases, a continuous-wave laser beam can enhance the processing speed by orders of magnitude. In the latter, pulsed laser alloying with extremely rapid quenching can yield alloy films obtainable by ordinary methods. In all these cases, laser heating has the advantage of being able to heat a local spatial region selectively to a very high temperature. The spatial resolution of such laser processing techniques would be limited only by the degree to which the laser can be focused. Aside from laser heating, laser-induced chemical reactions can also be used to modify surfaces. Photochemical surface etching or deposition is one example. Micron-size metal or nonmetal structure can be inscribed on a surface at high speed using such a process. Laser ablation can also be an effective method for surface processing. With ultraviolet lasers, the technique can carve extremely sharp surface patterns on polymers or biological materials. This is believed to be the result of laser-induced bond breaking rather than thermal evaporation. All of these laser processing techniques hold considerable promise as practical tools for surface treatment. Many industrial labs have already invested heavily in the development of these techniques. One might think that the basic physics underlying these laser processes must be well understood. This is not true, however, and universities in collaboration with industrial laboratories can expect to make significant contributions.