Quantitative Structural Studies Of Corundum and Rocksalt Oxide

Review pubs.acs.org/CR

Quantitative Structural Studies Of Corundum and Rocksalt Oxide Surfaces D. Phillip Woodruff* Physics Department, University of Warwick, Coventry CV4 7AL, U.K. in particular, such as ion sputtering and annealing, are often ineffective for the surfaces of bulk oxides, potentially leaving heavily defected or nonstoichiometric surfaces. Nevertheless, there have been an increasing number of spectroscopic and chemical studies of oxide surfaces in the last 20 years or so, although the number of quantitative structural studies of such surfaces remains small. By far the most studied surfaces are those of rutile-phase TiO2, not only because of the practical importance of titania in heterogeneous catalysis and photocatalysis, but also because good-quality single crystals are readily available and modest heat treatment in vacuum leads to the material becoming conducting. The point defects in the bulk that lead to these conducting properties are generally believed not to influence CONTENTS the surface properties. Structural studies of surfaces of the rutile phase of titania are described in the article by Thornton, 1. Introduction A Lindsay, and Pang,1 while those of the anatase phase are 2. (0001) Surfaces of Corundum-Phase Solids C described by Diebold.2 This review therefore excludes studies 2.1. α-Al2O3(0001) D of these materials and focuses on the surfaces of oxides having 2.2. α-Cr2O3(0001) F the corundum and rocksalt bulk structures that comprise the 2.3. α-Fe2O3(0001) H two largest groups of other MxOy systems that have been 2.4. V2O3(0001) I studied. 2.5. Adsorbate Structures K A number of distinctly different experimental techniques 3. Rocksalt Structure Oxide Surfaces L have been developed to determine surface structures in a 3.1. (100) Surfaces L quantitative fashion, and as the information gained is specific to 3.2. Adsorbate Structures on (100) Surfaces N the method, some understanding of these methods and their 3.2.1. Molecular Adsorption on MgO(100) N complementary aspects is essential to evaluate the data that 3.2.2. Atomic adsorbates and the early stages emerges. Fuller details may be found elsewhere. Among the of epitaxy on MgO(100) P most established techniques are those based on conventional 3.2.3. Adsorption on NiO(100) Q diffraction of either electrons or X-rays. In both cases, the 3.3. (111) Surfaces Q information to emerge is a complete structure determination of 4. General Discussion T all atomic positions in the outermost few surface atomic layers Author Information T for those areas of the surface that have good long-range order. In Corresponding Author T surface science low energy electron diffraction (LEED)3−5 has Notes T generally been regarded as the benchmark method for surface Biography U structure determination, but the use of low energy (typically References U ∼30−300 eV) electrons can lead to charging problems for insulating surfaces. The problem is particularly acute at the lowest energies but can sometimes be overcome by the use of a 1. INTRODUCTION secondary electron “flood” gun at grazing incidence to Despite the widely recognized importance of the chemistry of discharge the surface (e.g., ref 6). An alternative way to oxide surfaces, particularly in heterogeneous catalysis, most overcome this problem for all the surface techniques involving early applications of modern ultrahigh vacuum (UHV) charged particles is to use samples that are not bulk crystals, but methods of surface science focused on metal and semiepitaxial (ultra-) thin films grown on conducting (typically conductor surfaces. There are two important reasons for this. metal) substrates.7 This approach also allows the surface to be First, most oxides are insulators, so the many electron-based regenerated in the UHV analysis chamber, by simply removing UHV surface science techniques (including electron diffraction and photoemission) are more difficult to apply due to problems Special Issue: 2013 Surface Chemistry of Oxides of surface charging. In addition, however, the preparation of clean and well-ordered stoichiometric oxide surfaces can be Received: July 25, 2012 challenging, as the standard methods applied to metal surfaces, © XXXX American Chemical Society

A

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

the film by ion sputtering and heating, and then growing a new film. Most of the available evidence suggests the surfaces of such films have the same chemical properties as those of the bulk solid oxide; although for the very thinnest films (in some cases only one or two atomic layers), this is not always the case.8 X-ray diffraction, of course, is equally applicable to conducting and insulating surfaces because there is no problem associated with charging. The general problem with X-ray diffraction as a means of surface structure determination is that X-rays are only weakly scattered by atoms, a property that ensures that they penetrate deep into a solid and are therefore well-suited to structural studies of bulk crystals. In surface X-ray diffraction (SXRD),9,10 surface specificity can be enhanced by using very grazing incidence angles below the Brewster angle for total reflection, but more generally can also be obtained by detecting the scattered intensity in regions of momentum transfer space that correspond to diffraction from the surface but not the underlying bulk. The scattered signals are nevertheless weak, even using intense and highly collimated synchrotron radiation, and as X-ray scattering cross sections vary as the square of the atomic number, detecting scattering from O atoms (and thus identifying the location of O atoms) can be particularly challenging. It is also important to note that so-called “in plane” SXRD measurements, that do not include “rod scans” of the scattered intensity as a function of momentum transfer perpendicular to the surface, provide only information on the relative lateral positions of atoms and not on the interlayer spacings, nor on the surface/substrate registry. Ion scattering methods also generally rely on long-range order to obtain surface structural information, although to a lesser degree than conventional diffraction techniques. These methods involve mainly incident H+ and He+ ions, although higher-mass inert gas ions and alkali metal ions are also used in some studies, the energy and directions of the scattering primary ions being detected. The key structural information arises from the fact that ion scattering from an atom leads to a “shadow cone” behind the atom, a region into which no ions can penetrate. If other atoms are located within this shadow cone they make no contribution to the scattering, so by varying the incident direction of the ions, these shadow cones sweep through subsurface atoms leading to variations in the scattering yield that can be related to the relative positions of the atoms. Similar effects in outgoing trajectories can shadow the detector, leading to “blocking cones”. The highest structural precision is typically achieved in so-called medium energy ion scattering (MEIS)11 using incident ion energies in the range ∼50−200 keV, for which the shadow cones are narrow (a few tenths of an Ångström unit). In low energy ion scattering (LEIS)12 using energies of ∼1−10 keV the shadow cones are much wider (∼ 1−2 Å) leading to strong surface specificity but sometimes lower precision. There are also a number of experimental techniques, primarily of interest in determining adsorbate sites and geometries, which rely only on short-range or local order, albeit on a long-range-ordered underlying solid. Two of these exploit the coherent interference of directly emitted and elastically scattered components of an outgoing photoelectron wavefield in photoemission from a core level of a surface atom. In SEXAFS or surface EXAFS (extended X-ray absorption fine structure)13,14 this scattering interference occurs at the emitter atom itself and leads to modulations in the total photoionisation cross-section with photon energy (and thus photo-

electron wavelength) which can be used to extract emitterscatterer atom nearest-neighbor distances. The dependence of the EXAFS amplitude on the direction of the polarization vector of the incident radiation also provides some information on the directions of the nearest-neighbor scatterers. In photoelectron diffraction, particularly in the energy-scan (PhD) mode,15,16 this interference occurs at the detector, and the (much larger) modulations of intensity with photon energy are also direction-dependent, providing a method to determine the complete local geometry. A variant of photoelectron diffraction that exploits forward scattering at higher energies, usually referred to as XPD (X-ray photoelectron diffraction) is particularly effective in identifying interatomic directions in thin epitaxial films. A rather different technique is X-ray standing waves (XSW)17−19 that exploits the interference of the incident and scattered X-rays at a Bragg reflection from a crystal; this creates an X-ray standing wave with a periodicity equal to that of the scattering atomic planes, which extends out beyond the surface. By detecting the photoionisation in a surface atom immersed in this standing wave, and sweeping the standing wave through this atom, its position relative to the underlying solid can be determined. One important feature of all three of these local structural probes is that, because they involve measurements of core electron binding energies that are characteristic of the photoabsorbing atom, they are element specific. Indeed, in the case of PhD, XPD, and XSW it is also possible to exploit the socalled chemical shifts in photoelectron binding energies to achieve chemical-state specificity in addition to elemental specificity. This aspect can be particularly valuable in studying oxygen-containing species on oxide surfaces, because the two distinct states of the oxygen atoms and their associated geometries on the surface can be distinguished. One important issue in all experimental studies of oxide surfaces is the problem of damage induced by the surface probe. In particular, both incident electrons and incident protons can lead to dissociation with ejection of oxygen ions and atoms through electronic excitation, the so-called DIET (desorption induced by electronic transitions) processes of electron, and photon-induced desorption. Incident ions can also lead to similar processes, in addition to damage induced by the effect of the direct ion impact and transfer of kinetic energy to surface atoms. Damage is generally believed to be more serious in incident electron techniques, such as LEED, than in incident Xray methods. This is because the damage is largely produced by low energy electrons (∼tens of eV) which are typically produced as secondary electrons more effectively using incident electrons in the LEED energy range (∼30−300 eV) than by incident X-ray photons, with energies of a few keV. It does seem, though, that all the above techniques can be used to study oxide surfaces effectively, but caution is required and the problem of radiation damage needs to be monitored carefully. Many aspects of radiation damage can be reduced, particularly for molecular adsorbates, but reducing the sample temperature during such investigations. One technique not included in this summary, which has nevertheless been extensively used in structural studies of surfaces including those of oxides, is scanning tunnelling microscopy (STM). STM has proved to be invaluable in many studies of oxide surfaces, its real-space atomic-scale images being especially valuable in studies of local defects at surfaces and their reactivity. In principle a related technique more generally applicable to the study of oxide surfaces is atomic B

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

2. (0001) SURFACES OF CORUNDUM-PHASE SOLIDS While the (110) surface of rutile-phase TiO2 is almost certainly the most studied of all oxide surface structures, the next-most studied such surface is probably α-Al2O3(0001), together with the same orientation of other corundum-phase materials, notably Cr2O3, V2O3, and α-Fe2O3 (hematite). All of these structures are rhombohedral but are usually represented by a hexagonal unit cell containing six M2O3 units. Relative to the (0001) basal plane, the structure comprises hexagonal closepacked planar layers containing 3 oxygen atoms per unit mesh alternating with buckled layers containing 2 metal atoms per unit mesh which occupy two out of the three available octahedral interstitial sites between the oxygen planes; this layer structure is denoted here as ...O3MM′O3MM′O3MM′... (see Figures 1 and 2).

force microscopy (AFM), because this technique is equally applicable to insulators. However, the resolution achieved in STM in many (but not all) studies is generally substantially superior, and so far it is this technique that has been most widely exploited to provide atomic-scale structural information. STM is not, however, a quantitative structural probe. It is intrinsically a probe of the spatial variation of the electronic properties of a surface, so while protrusions seen in images recorded in the usual mode of constant tunnelling current are commonly centered above surface atoms, this is not always the case. Moreover, the relative height of these protrusions cannot be equated with the relative heights of the underlying atoms on the surface. The technique also does not intrinsically contain any elemental or chemical information, so it is not possible, based on inspection of the images alone, to know the elemental or chemical identity of species on the surface that give rise to protrusions in the images. Despite these many limitations, however, much qualitative structural information can be gained from STM of oxide surfaces, particularly if combined with theoretical modeling. Theoretical modeling, or total-energy calculations, primarily using density functional theory (DFT), have become an increasingly important technique in the study of surface structures and have been used extensively to study oxide surfaces. Indeed, for many surfaces such calculations are the only source of structural information and as such are sometimes regarded as structure determinations in their own right. However, there is ample evidence that such methods are not universally reliable in their structural solutions, while many papers based only on these methods fail to explore a sufficient range of possible structures. In view of this only structures that have also been studied by quantitative experimental methods will be discussed in this review. In this regard it also important to note that applications of DFT, like those of most experimental methods of structure determination, are constrained by the imagination of the researcher; the optimized form of a particular structural model can be established, but if the correct model is not tested, the correct structure is not found. While structure determinations based only on DFT calculations need to be treated with particular caution, combined DFT/STM studies do allow one to have a higher degree of confidence in the results. Nevertheless, STM simulations based on DFT calculations and the Tersoff− Hamman20 approach can lead to very different images at different assumed tip spacings, while such images may not be uniquely related to a particular structural model. True quantitative experimental methods are therefore to be greatly preferred, and results obtained in this way will be emphasized in this review. In the remainder of this article the results of such structural studies are organized according to the bulk structural phases. Most structural studies to date have been on clean surfaces, seeking to establish how the bulk structure is terminated and to quantify any associated relaxations of the atomic positions, although in some cases more complex reconstructions, potentially involving stoichiometry changes, occur. In a small number of cases distinct ultrathin oxide phases have been investigated in detail as model systems for studies of surface chemistry, and these are also described briefly, although the emphasis is on studies of the surfaces of bulk or thick-film crystals. Finally, structural information on adsorbate species is given for those cases for which such data exist.

Figure 1. Side view of the corundum-phase M2O3(0001) structure showing four different possible surface terminations.

Figure 2. Plan view of the corundum-phase M2O3(0001) structure. The full lines show the surface unit mesh.

The dimensions of the associated bulk hexagonal unit cells are as follows: α-Al2O3(0001), a = 4.763 Å, c = 13.003 Å; αCr2O3(0001), a = 4.954 Å, c = 13.584 Å; α-Fe2O3(0001), a = 5.035 Å, c = 13.429 Å; V2O3(0001), a = 5.105 Å, c = 14.449 Å. A key question is at which layer the solid is terminated to create a (0001) surface. Figure 1 shows the three possible truncations of the bulk structure corresponding to an outermost surface with a complete (O3) oxygen layer, a complete buckled metal layer (full-metal), just one of the metal sublayer components of this buckled metal layer (half-metal). Notice that the first two of these terminations are both polar, and so would normally (without reconstruction) be regarded as unstable because of the divergent electrostatic surface energy.21 One further possible termination, shown in Figure 1, is that a half-metal layer termination that is capped with O atoms atop the metal atoms to produce a layer of MO species; this solution, with surface vanadyl (VO), chromyl (CrO), or C

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

the outermost four layer spacings (z(Mtop−Otop), z(Otop−M′2), z(M′2−M2), and z(M2−O2) using the labeling convention of Figure 3; see also Table 1) are significantly relaxed (by −51%,

even ferryl (FeO) species has been favored in some studies of V2O3(0001), Cr2O3(0001), and Fe2O3(0001), as described below. Notice that one feature of the corundum-phase (0001) surfaces is that for any given termination, removing one complete M2O3 layer yields a surface that is structural identical but is the mirror image. Clearly these two structures are energetically equivalent and so, on a real surface with atomic steps, must coexist with equal probability. It is, therefore, important that experimental studies of these surfaces take this domain averaging into account in their data evaluation and modeling. 2.1. α-Al2O3(0001)

Early investigations of the (001) surface of bulk α-Al2O3 surfaces using qualitative LEED6,22−24 provided evidence that the method of surface preparation, and particularly the temperature of any heat treatment, had a significant effect on the surface structure. In particular, while it was possible to produce a simple (1 × 1) surface phase, heating to increasingly higher temperatures in UHV led to the formation of (√3 × √3)R30°, (3√3 × 3√3)R30°, and (√31 × √31)R9° surface phases, believed to result from the increasing loss of oxygen from the surface, a view supported by the fact that the (1 × 1) phase can be recovered by oxygen treatment. A much more recent quantitative structure determination of the (√31 × √31)R9° phase using SXRD confirmed this interpretation.25 Specifically, this structural phase was found to comprise two Al layers on the oxide substrate with a structure similar to that of Al(111), but rotated to produce commensurability. Atomic resolution AFM images of this surface26 are also consistent with this interpretation. Nevertheless, a complete understanding of this complex structure, which appears to show significant disorder within the unit mesh, does not appear to have been achieved.27−29 The experimental SXRD data, based on in-plane measurements alone, provide a detailed picture of the arrangement of the Al atoms within the metallic overlayer but do not provide information on the interlayer spacings or the nature of the truncation of the underlying oxide substrate, although one theoretical treatment modeling this structure assumes this to be a half-metal termination.29 More recently, however, a new investigation of this surface using AFM combined with DFT calculations (that also consider the influence of different temperatures at fixed oxygen partial pressure and thus varying oxygen chemical potential) has concluded that the surface comprises a single Al layer on top of a half-metal terminated substrate.30 A SXRD investigation31 of the (3√3 × 3√3)R30° phase, aided by modeling calculations using semiempirical potentials, revealed a somewhat similar two-layer metallic layer, also involving hexagonal-close-packed Al atoms, although the degree of disorder was found to be much less than in the (√31 × √31)R9° structure. The simpler (1 × 1) surface phase has been the subject of most quantitative structure studies of the corundum surface. Detailed investigations by both SXRD32,33 and quantitative LEED,34−36 but also studies using incident ion techniques, notably time-of-flight scattering and recoiling spectrometry (TOF-SARS)37 and a form of LEIS (coaxial impact collision ion scattering spectroscopy or CAICISS),38 have provided information on the nature of the surface termination. The first of these investigations, using SXRD, concluded that the Al2O3(0001)(1 × 1) surface is half-metal terminated, but that

Figure 3. Side view of a half-metal terminated M2O3(0001) surface showing the labeling convention used here for the near-surface layers.

+16%, −29%, and +20%) relative to an ideal bulk termination. This conclusion is fully consistent with the results of the TOFSARS study which found the same termination but was only able to establish the outermost surface layer relaxation, for which the value reported was 63 ± 12%. A rather less direct approach to the surface structure, in which the lowest-energy structure and associated electronic properties were calculated theoretically and the predicted electronic structure was then confirmed experimentally,39 also favored a half-metal termination with an outermost layer relaxation of 68%. By contrast the earliest LEED study34 concluded that the surface showed a mixed termination, with the O3 and half-metal terminations being present in the ratio 2:1, and with a much smaller relaxation of the outermost layers in the half-metal termination. However, the results of the CAICISS were found to be inconsistent with this coexistence of the O3 termination. The later LEED investigation35,36 also supported the halfmetal termination model. This most recent study involved a particularly thorough approach, exploring the effect of three different methods of preparing the surface (using atomic deuterium in situ cleaning, with different annealing conditions in vacuum, and in a partial pressure of oxygen), and also considering a range of single and mixed terminations. All three preparation methods yielded data that were best-fitted by the half-metal termination. Models tested also included mixedtermination surfaces and a half-metal termination with O atoms atop the surface Al atoms; the rationale for testing this last model was not the possibility of surface AlO species, but rather that the surface may be hydroxylated with OH species bonded to the surface metal atoms. Notice, of course, that H atoms are such weak scatterers of low-energy electrons that their presence cannot generally be detected in LEED. While this LEED analysis clearly favored a pure half-metal termination, good agreement with experimental intensityenergy spectra for the diffracted beams could only be obtained by using the so-called “split-position” technique to simulate very large surface vibrational amplitudes. In this method, two different outermost-layer spacings are assumed to have equal probability and are optimized independently in the theoryexperiment comparison. This method40−42 has been found to be effective in describing LEED from surface atoms whose vibrational amplitudes are too large to be dealt with correctly by D

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Table 1. Values of the near-Surface Interlayer Spacing Relaxations, Expressed As a Fraction of the Values in the Bulk Crystal, for the Preferred Half-Metal-Terminated Structure of α-Al2O3(0001), Obtained in Various Experimental and Theoretical Studies, as Described More Fully in the Texta method SXRD TOF-SARS theory/el. structure LEED* LEED Hartree−Fock DFT DFT DFT DFT DFT Hartree−Fock DFT shell model DFT DFT shell model shell model Hartree−Fock

year (ref) 32,33

1996 199737 199739 199834 200035,36 198754,56 199343 199444 199945 199946 200047 200048 200048 200349 200349 201050 200251 200152 200653

Δz(Altop−Otop) (%)

Δz(Otop−Al′2) (%)

−51 −63 −68 −29 −(50−53) −(46−53) −86 −87 −85 −77 −86 −(68−79) −(70−87) −(57−74) −82 −86 −(56−75) −58 −80

Δz(Al′2−Al2) (%)

+16

−29

+5 +(2−6)

−57

+3 0 +3 +11 +6 (−1)−(+6) 2−14 7−12 4 4 5−12 4 5

−54 +20 −45 −34 −49 −(27−57) −(32−54) −(42−44) −44 −45 −(41−44) −42 −49

Δz(Al2−O2) (%) +20

+25 0 +20 +18 +22 20−22 21−26 6−11 18 21 6−12 24 22

a In the LEED* study, the best-fit structure included partial coverage of an O3 termination, but only the relaxations for the half-metal terminated regions of the surface are reported here.

with values in the range 2.5−3.5 times larger than in the bulk, but another calculation51,49 found smaller enhancements and specifically concluded that the calculated anharmonicity cannot account for the discrepancy between theory and experiment in the size of the surface relaxation. While many of these theoretical studies have not addressed the issue of thermodynamic equilibrium with the surroundings, some investigations46,47,50 that specifically considered a wide range of oxygen partial pressures and associated chemical potentials concluded that the half-metal termination remains stable in essentially any accessible oxygen pressure. In addition, however, two studies45,47 considered the effect of partial pressures of hydrogen, and find that at sufficiently high H chemical potentials a hydrogenated O3 termination actually becomes more favorable. Interestingly, in one of these investigations a lower coverage of H on the half-metal terminated surface was found to reduce the calculated outermost Al−O layer spacing from 86% to ∼69% and this was suggested as a possible reason for the disagreement between experimental and theory for this surface. Of course, hydroxylation of surfaces in general, notably through interaction with water, is an important issue for all oxides, although the range of different surface preparation treatments of Al2O3(0001) explored in the more recent LEED study would suggest that not all of the surfaces studied were hydroxylated in this way. While all of the experimental studies described above were performed on bulk single crystals of α-alumina (also referred to in some papers as sapphire), there have also been extensive studies of ultrathin films of alumina grown on metallic substrates, although for most of the systems studied there has been little or no serious exploration of the surface structure. Much the most-investigated system is the ultrathin film grown on NiAl(110) by dosing in a partial pressure of oxygen with elevated temperature annealing, first characterized by Jaeger et al.56 Apparently, the Al in the NiAl alloy is preferentially

the usual Debye−Waller factor. Using this approach the outermost Al (half-layer) spacing was found to be reduced by a value in the range 50−53% relative to the bulk value, in excellent agreement with the SXRD result.32 The increase in the second layer spacing found in this study, however, was in the range 2−6%, significantly smaller than the 16% found in SXRD. The vibrational amplitudes of the surface atoms at room temperature implied by this LEED analysis are approximately 0.24 Å, roughly a factor of 2 larger than those calculated for atoms in the bulk from the known Debye temperature. Table 1 summarizes the main findings of these different studies. The problem of the termination and structure of αAl2O3(0001) has also attracted a lot of theoretical effort, and these studies all support the view that the half-metal termination is energetically preferred under all normal UHV conditions, and that the outermost Al layer, in particular, is strongly relaxed inward to the bulk although this relaxation involves little change in the Al−O bond lengths. The magnitude of the calculated surface relaxation, however, does appear to depend significantly on the exact theoretical and computational approach. Most calculations based on DFT methods43−50 (with a variety of different associated approximations) yield values of this relaxation in excess of 80% while several applications of Hartree−Fock and semiempirical shell methods39,48,49,51−53 have led to somewhat smaller values of ∼60−70%, with only the earliest such study finding a much smaller value of 46%.54,55 Table 1 summarizes the main finding of these different studies. The authors of the detailed LEED study have suggested that the unusually large surface vibrational amplitudes and associated anharmonicity seen at room temperature might account, at least in part, for the fact that these theoretically calculated surface relaxations are mostly significantly larger than the experimentally obtained values, and some molecular dynamics calculations have addressed this question. One such study52 did find very large surface vibrational amplitudes in the outermost Al layer, in particular, E

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

2.2. α-Cr2O3(0001)

oxidized and the resulting film appears to contain no Ni atoms. As this film has an estimated thickness of only 5 Å, corresponding to little more than twice the O−O layer spacing in α-Al2O3(0001) (2.17 Å), it is roughly consistent with two complete Al2O3 layers. This is thin enough for one to anticipate (as remarked above8), that the electronic (and probably the structural) properties will differ from those of bulk α-Al2O3; indeed results of STM studies of low coverages of Au on this surface indicate interaction with the underlying NiAl(110) substrate.57 Nevertheless, this film has been the subject of extensive surface science studies including the investigations of the properties of small metal particles on this oxide surface. The initial characterization led to the suggestion that the film has a similar structure to either α-Al2O3(0001) or γ-Al2O3(111) (in γ-Al2O3(111), unlike α-Al2O3(0001), Al atoms occupy both tetrahedral and octahedral sites between the close-packed O layers58), most likely with an oxygen surface termination. The γ-Al2O3(111), interpretation is favored by the results of an STM study59 as well as by vibrational spectroscopy.60 The first serious attempt to achieve a full structure determination for this film, using SXRD, concluded that it consists of “a double layer of strongly distorted hexagonal ions that hosts aluminum ions on both octahedral and tetrahedral sites with equal probability” within a domain structure related to the growth of the hexagonally ordered Al2O3 on the bcc(110) substrate.61 However, Kresse et al.62 argued that this structural solution involves some unphysically short Al−Al and Al−O bondlengths (2.08 Å and 1.51 Å, respectively) and proposed a new structure based on the results of DFT total energy calculations and STM images. Specifically, the STM images under certain tunnelling conditions indicate the presence of groups of surface atoms arranged in squares, inconsistent with the hexagonal arrangement to be expected from α-Al2O3(0001) or γ-Al2O3(111). The proposed structure is actually found to have a stoichiometry of Al10O13, with a unit mesh approximately described as commensurate with the substrate and having sides of lengths 10.93 Å and 17.90 Å with an included angle of 88.16°, although it actually appears to be incommensurate in one direction. The outermost layer in this unit mesh contains 28 O atoms, with an underlying, but almost coplanar, layer containing 24 Al atoms. This model appears to have been generally accepted, although the results of more recent experiments have provided further information. For example, atomically resolved AFM images63 provide direct support for the fact that the outermost surface layer comprises 28 O atoms within the unit mesh, while a MEIS study, supported by high-resolution core-level photoemission, yielded values of the interlayer spacings of the four (O−Al−O− Al) layers.64 Interestingly, a recent SXRD investigation of ultrathin alumina films grown on Ni(111)65 has concluded that the structure appears to be very similar to that of the films grown on NiAl(110). Notice that ultrathin alumina films can also be grown on Ni3Al(111) surfaces by simple oxidation and annealing under UHV conditions and have also been studied extensively, but there appears to have been no quantitative investigations of their surface structure, although AFM studies provide some qualitative information (e.g., ref 66). Much thicker alumina films can also be grown by more aggressive oxidation conditions and have been studied by XRD and transmission electron microscopy (TEM) on NiAl(110). These thicker films (hundreds to thousands of Å in thickness) are typically found to have the γ-Al2O3 structure,67 although transformation to α-Al2O3 can be achieved.68

Experimental surface structure determinations have been performed on the (0001) surface of α-Cr2O3 both on bulk crystals and on thin (but not ultrathin, as in the case of alumina on NiAl) epitaxial films. As for α-Al2O3(0001), the key questions being addressed in these studies have been the surface termination and the associated interlayer relaxations of the outermost atomic layers. The first LEED study, performed on epitaxial films with an estimated coverage of 40 Å grown on Cr(110), concluded that the surface has a half-metal termination with relaxations of the outermost four interlayer spacings of −60%, −3%, −21%, and +6%.69,70 In this investigation, the final stage of the sample preparation was a brief heating to 1000 K in UHV, a treatment that is relevant in the context of the results of subsequent investigations. More recently, a number of structural studies have been performed on the (0001) surface of bulk α-Cr2O3 crystals. The first such investigation,71 using SXRD, concluded that the surface structure is based on a half-metal termination, but approximately 1/3rd of the surface Cr atoms are displaced from the outermost layer sites to interstitial sites below the outermost O3 layer (Figure 4). This redistribution of the surface Cr atoms

Figure 4. Side views of the different variants of the half-metal termination of the α-Cr2O3 surface found in experimental investigations. On the left is shown the standard half-metal termination69,70 while in the center and on the right are models involving partial occupation of interstitial subsurface sites (Crint)71 and of adsorption sites (Crad)75,77 corresponding to the locations of Cr atoms in a full metal termination. The fractional occupations of the different sites in the latter two models are shown as percentages. Interlayer spacings are reported in Table 2.

(nominally Cr3+ ions) was found to lead to a much smaller (−6%) relaxation of the outermost Cr−O layer spacing; fits to the experimental data assuming a pure half-metal termination with no interstitial Cr led to large surface layer relaxations similar to those found in the earlier LEED study, but the quality of the agreement between theory and experiment for this model was significantly worse. The authors of this paper also remark that the redistribution of one-third of the surface Cr atoms provides a possible explanation for a (√3 × √3)R30° LEED pattern, previously reported to occur in the temperature range 90−150 K on films grown on Cr(110)72 (but also seen at room temperature on α-Cr2O3(0001) grown on Ag(111) to a thickness greater than ∼12 Å73); they attribute the transformation of the (√3 × √3)R30° phase to (1 × 1) to be due to an order−disorder transition. This explanation is on contrast to that of the original report of this transition72 that also F

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Table 2. Values of the near-Surface Interlayer Spacing Relaxations, Expressed As a Fraction of the Values in the Bulk Crystal, for the Preferred Surface Terminations of α-Cr2O3(0001) under UHV Conditions, Obtained in Various Experimental and Theoretical Studies, As Described More Fully in the Texta method LEED SXRD LEED SXRD Hartree−Fock Hartree−Fock DFT DFT

year (ref) 69,70

1997 199771 200975 201077 199678 200653 200079 200481

Δz(Crtop−Otop) (%)

Δz(Otop−Cr′2) (%)

Δz(Cr′2−Cr2) (%)

Δz(Cr2−O2) (%)

−60 −6 11 [Δz(Crad−Crtop) = −29%] 38 [Δz(Crad−Crtop) = −42%] −50 −49 −33 −(60−62)

−3 0 (Crint and Cr′2 coplanar) +2 −28 3 3 3 10−12

−21 −26 0 −13 0 −47 11 −(41−44)

+6 7 −1 −32 0 11 −2 6−9

a

Interlayer spacings are reported using the labelling conventions for the half-metal termination, with added information for structures involving partial occupation of Crint and Crad sites.

do not find the large inward relaxation of the outermost Cr layers that were found in the original LEED study. The SXRD measurements of the surface with 1 × 10−2 mbar O2 in the gas phase showed the structure to be significantly different, not only having terminal O atoms to produce surface chromyl species, but also with the surface Cr layer becoming more like a half-metal termination, with outermost sublayer Cr atoms in the UHV structure moving to vacant sites in the lower sublayer, although the total occupation of this layer was found to be only 38%. Under both conditions there is also some depletion of the occupancy of the outermost Cr sublayer below the O3 layer (30% under UHV conditions, 15% with oxygen present), as had been found in the earlier UHV SXRD investigation. The CrO bond length of the chromyl species was found to be 1.57 Å; the relaxation of the underlying Crtop− Otop, Otop−Cr′2, and Cr′2−Cr2 layer spacings in the presence of the chromyl O atoms were −3%, −26%, and −11%. Theoretical studies of the α-Cr2O3(0001) surface, using slab calculations that assume a (1 × 1) periodicity, have not addressed the possibility of these fractional layer occupations, but some have considered the effect of different oxygen chemical potentials in the gas phase. Two studies53,78 using ab initio Hartee-Fock theory and considering only (implicitly) the UHV surface, reached conclusions very similar to those of theoretical studies of the α-Al2O3(0001) surface. Specifically, they found that the bulk is half-metal terminated, and there is a large inward relaxation of the outermost half-metal layer relative to the bulk spacing (Table 2). The magnitude of this effect was found to be 49−50%, somewhat smaller than that obtained in theoretical treatments of Al2O3(0001) and in rather good agreement with the original LEED study. The first application of DFT to this system considered only the possibility of O3 termination,79 but two later such studies80,81 considered a range of surfaces terminations and determined the relative energetic favorability of these different terminations as a function of oxygen chemical potential (although only one of these investigations reported the optimized interlayer spacings81). Figure 5 is a simplified diagram that shows the stable phases as a function of oxygen chemical potential for α-Cr2O3(001) based on one of these calculations,81 which also includes similar diagrams for the other corundum-phase (0001) surfaces described here. In the case of Al2O3, as described above, the half-metal (HM) termination was found to be stable for all accessible values of the oxygen chemical potential. In the case of Cr2O3, the half-metal termination is also stable over a wide range of values of this potential, but at smaller (negative) values the MO (in this case CrO or chromyl) termination is

attributed it to an order−disorder transition, but with the ordering within a pure half-metal termination involving antiferomagnetic coupling of the Cr atoms in the outermost layers. A much more recent qualitative LEED study, on the other hand, suggests that the (√3 × √3)R30° phase may be due to CO adsorption from the residual gases in the vacuum.74 A more recent LEED study75 of the surface of the same crystal as that used in the SXRD investigation led to rather different conclusions. In this case the best-fit termination was found to be a full-metal termination, but with fractional occupation of the outermost two Cr sublayers (or equivalently, a half-metal termination with fractional coverage of the halfmetal sites and superimposed adsorption sites). In this case, too, no strong inward relaxation of the outermost Cr−O layer spacing was seen; indeed, this spacing was found to increase by 11%. The authors, however, reported a significant degree of disorder in this surface. In this investigation the possibility of a partial coverage of O atoms in local atop sites relative to the surface Cr atoms, corresponding to surface chromyl, CrO, species was also explored, but the analysis proved to be insensitive to the presence or absence of these species. Note that evidence for the existence of surface chromyl species on Cr2O3 thin films grown on Cr(110), under certain preparation conditions, has been found in vibrational spectroscopy.76 This question of the possible presence of surface chromyl species on the α-Cr2O3(0001) surface has recently been addressed in a very detailed SXRD investigation of surfaces of a bulk crystal prepared under the usual UHV conditions, including annealing to 1200 K and making subsequent measurements in UHV, but also subsequently studied in partial pressures of oxygen up to 1 × 10−2 mbar.77 For the surface prepared and measured under UHV conditions, the results were qualitatively consistent with the more recent LEED study, in that the best fit to experiment was achieved with a full-metal termination model with only fractional occupation of the Cr sites in the two components of the buckled terminating fullmetal layer, although the expansion of the outermost Cr−O layer was found to be 38%. Moreover, while the LEED analysis had found the fractional occupations of the outermost and underlying sublayers to be 31% and 61% (Figure 4), in the new SXRD study the corresponding values were 22% and 31%. As the method of calculating the diffracted intensities for partial site occupations in LEED (but not in SXRD) requires some significant approximations involving neglect of multiple scattering effects, and the LEED study reported significant disorder, it seems likely that the SXRD values are more reliable. Notice that both this SXRD study and the related LEED study G

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

the outermost surface and subsurface layer relaxations were found to be −60%, +12%, and −44%, whereas in the presence of the additional O atoms to produce a chromyl termination, the corresponding values were −1%, +12%, and −43%. The absence of the large relaxation between the outermost halfmetal Cr layer and the underlying O3 layer in the presence of the chromyl O atoms is consistent with the experimental results. With the exception of the third (Cr−Cr) interlayer spacing the relaxations found in these calculations for the half-metal termination are in quite good agreement with those found in the original LEED study,69,70 and as this interlayer spacing is itself very small in the bulk (0.38 Å) the large fractional spacing change actually corresponds to quite a small absolute change. Of course, both this theoretical treatment and the original LEED study only considered structures with ideal (1 × 1) periodicity. The possibility of fractional occupation of different Cr layers was not considered. In this regard, it is interesting to note that in the first SXRD study,71 applying the same (1 × 1) constraint led to a best-fit structure very similar to that of the original LEED analysis, with half-metal termination and a surface layer relaxation of −52%; releasing this constraint led to the preferred structure having ∼1/3 ML of surface Cr atoms being transferred to interstitial sites, and an associated Cr−O surface relaxation of only −6%. This highlights the existence of strong parameter coupling in these comparisons of experimental and simulated SXRD intensities, although introducing the additional parameters associated with fractional layer occupation does lead to a significant improvement in the agreement between experimental and simulated intensities.

Figure 5. Simplified phase diagrams showing the range of stabilities of different terminations of four α-M2O3(0001) surfaces as a function of oxygen chemical potential. Also shown are the corresponding oxygen pressures at two temperatures relevant to sample preparation conditions. For Al2O3 the diagram is derived from data from refs 44, 46, and 47. For Cr2O3 and Fe2O3, the diagrams are based on the GGA +U results of ref 81, and for V2O3, the results of ref 114 have been used.

2.3. α-Fe2O3(0001)

favored, while at the smallest (negative) values, corresponding to the highest oxygen partial pressures, the O3 termination becomes the most stable. It is important to stress that the exact value of the chemical potential at which the phase boundaries occur in these calculations is sensitive to a range of features of the theoretical approach and as the associated oxygen partial pressure is exponentially related the chemical potential, the exact conditions corresponding to these boundaries should be regarded as indicative rather than precise. In the case of Cr2O3 (and Fe2O3), a particularly dramatic shift is found depending on whether calculations are performed in a standard GGA (generalized gradient approximation) or GGA+U (including the Hubbard U correction). In Figure 5 the results obtained from the GGA+U calculations, which appear to describe several properties more satisfactorily, are shown. Bearing in mind these caveats, it is clear that the theory predicts that the half-metal termination should be the stable surface phase on Cr2O3(0001) following high-temperature annealing in UHV, while the chromyl terminated surface may be expected to occur at higher oxygen partial pressures, under conditions broadly consistent with the most recent SXRD experimental results. As noted above, these calculations did not consider the possibility of fractional occupation of the outermost Cr layers and so it is not possible to establish any total agreement (or otherwise) between theory and experiment; indeed, there are some striking differences between the findings of the different experimental studies. Nevertheless, the influence of the oxygen partial pressure is certainly consistent with the experimental finding of CrO termination and a half-metal-like termination at high and low oxygen pressures. The calculations also provide information on the surface relaxations associated with the favored structures. In particular, in the half-metal termination

Early investigations of the surfaces of both bulk α-Fe2O3(0001) (hematite) and epitaxial thin films, using qualitative LEED82,83 and STM,84−88 showed significant variations in the nature of the surface ordering depending on the method of surface preparation. A key issue seems to have been that different preparations lead to different oxides being formed. Some of these studies83,85 indicated the creation of Fe3O4(111) at the surface, or even the coexistence of Fe2O3(0001) and FeO(111) phases,86 although later studies do appear to have identified preparation conditions that lead to the expected αFe2O3(0001)(1 × 1) terminations and have provided some clarification of the existence of “biphase” surfaces.89 The first attempt to achieve an experimental quantitative surface structure determination, using quantitative LEED,90 based on epitaxial films grown on Pt(111), sought to determine the influence of the partial pressure of oxygen on the surface termination. As shown in Figure 5, a theoretical determination of the equilibrium surface phase diagram81 indicates that at low pressures (larger negative oxygen chemical potentials) a halfmetal termination is predicted, whereas at higher oxygen partial pressures (smaller negative chemical potentials) a ferryl (Fe O) termination is expected. As remarked above, the location of the phase boundaries for Fe2O3 (and Cr2O3) is sensitive to the exact method of computation; the diagram in Figure 5 is based on GGA+U calculations, but GGA calculations shift the phase boundaries significantly to the left (to larger negative chemical potentials) to produce a diagram more like that shown for Cr2O3 in Figure 5. In particular, the boundary between the halfmetal and FeO terminations shifts to an oxygen chemical potential of approximately −1 eV, as also found in an independent calculation of this phase diagram91 by a different H

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Table 3. Values of the near-Surface Interlayer Spacing Relaxations, Expressed As a Fraction of the Values in the Bulk Crystal, for the Half-Metal Termination of α-Fe2O3(0001) Obtained in Various Experimental and Theoretical Studies, as Described More Fully in the Text method LEED HM solution XPD LEED DFT DFT DFT

year (ref)

Δz(Fetop−Otop) (%)

90

2001 199994 200975 199892 200491 200481

−27 (Fetop

−79 −41 only 50% occupation) −57 −51 −53

Δz(Otop−Fe′2) (%)

Δz(Fe′2−Fe2) (%)

Δz(Fe2−O2) (%)

4 18 4 7 6 22

35 −8 −9 −33 −32 −31

−28 47 7 15 13 34

So far, therefore, only the half-metal termination predicted by theory to be the most stable phase at low oxygen partial pressures has been found in the experimental investigations that led to clear conclusions. At high oxygen pressures the strongest experimental evidence for the existence of the O3 termination is the STM images88 showing coexistence of two different domains at intermediate pressures with height differences indicative of different terminations, while higher and lower pressures led to one or other of these domains becoming dominant. None of these investigations have identified the ferryl termination predicted by theory to occur in equilibrium, albeit only over a rather narrow range of oxygen chemical potentials. However, an infrared absorption spectroscopy (IRAS) study95 has identified a vibrational mode assigned to the Fe−O stretching frequency of this species on the surface of an epitaxial film (thickness ∼50−70 Å) grown on Pt(111) in a pressure of 10−3 mbar of oxygen at ∼1050 K. In this regard we should note that while the chromyl, CrO, species is wellknown in the chemistry of Cr compounds (as is the vanadyl, VO, species in vanadium compounds), the FeO species are much rarer, believed to occur mainly in heme-containing enzymes.

group at almost exactly the same time, also using DFT. Notice that, also like the phase diagram for Cr2O3 in Figure 5, an O3 termination is found for the highest oxygen partial pressures, with the ferryl termination being predicted to occur over a rather narrow range of chemical potential values. An earlier calculation of this phase diagram92 which did not consider the possibility of a FeO termination located the half-metal/O3 phase boundary at a chemical potential values of approximately −0.8 eV, essentially consistent with these later calculations. The first quantitative LEED study of Fe2O3(0001)90 proved to be rather inconclusive in that for both of the two different surfaces studied, one involving formation in a partial pressure of oxygen of 10−5 mbar, the other in a pressure of 1 mbar, the measured diffracted beam intensity spectra were found to be fitted comparably well by simulations based on different termination models with different interlayer relaxations. For the surface prepared at the lower pressure, the best agreement was found for an O3 termination, but a fit of comparable quality was obtained for the half-metal termination; however, complementary data from other techniques led the authors to conclude that the surface was probably hydroxylated. An STM study88 of films prepared in a similar way provided rather clear evidence of the coexistence of domains of different terminations, believed to correspond to half-metal and O3 terminations, as indicated by layer height differences that were not multiples of the bulk repeat distances to be expected for a single type of termination. More detailed interpretation of the images led the authors to suggest that at the lower oxygen pressures (10−5 mbar) only the half-metal termination was seen, whereas at the highest pressures (1 mbar) an O3 termination dominated. While this interpretation is broadly consistent with the theoretical predictions of Figure 5, it is interesting to note that an investigation of Fe2O3 films grown epitaxially on α-Al2O3(0001) by oxygen-plasma-assisted molecular beam epitaxy (OPA-MBE) were found to be Feterminated, despite being cooled to near room temperature in the highly oxidizing plasma.93 Indeed, quantitative structural information has been obtained from the surface of a film prepared by this method by the technique of XPD. The conclusions of this study94 were that the surface is half-metal terminated with outermost interlayer spacing relaxations of −41%, +18%, −8%, and +47%, with no evidence of an O3 termination. These values compare rather favorably with the results of the three DFT-based calculations, summarized in Table 3. A more recent quantitative LEED study75 also found a halfmetal termination at the surface of a bulk α-Fe2O3(0001) crystal, although with evidence for only ∼50% occupation of the outermost Fe layer. The near-surface interlayer relaxations found in this study (Table 3) are significantly smaller than those predicted by theory or found in the XPD experiments.

2.4. V2O3(0001)

Although there have been extensive studies of V2O3(0001) surfaces, these seem to have been entirely based on epitaxial thin films. Most of these have been grown on metal surfaces including Pd(111),96 Rh(111),97 W(110),98 Au(111),98 and Cu3Au(100),99 although films have also been grown on αAl2O3(0001).100 Extensive studies of the spectroscopic and chemical properties of these surfaces have been published, particularly by the groups of Netzer in Graz and of Freund in Berlin, and most of this work up to 2003 has been reviewed101 by the Netzer group. Some work has been conducted on ultrathin (∼1−2 layer) films, including one quantitative LEED structural study (combined with DFT calculations)102 of (4 × 4) and (2 × 2) phases at V coverages of 0.31 and 0.5 ML corresponding, respectively to V5O14 and V2O3 stoichiometries, but most attention has been focused on somewhat thicker films that are believed to be more representative of the bulk solid. The films have been grown by vapor deposition of V in a partial pressure of oxygen gas, typically at elevated temperatures and with specific annealing recipes, which mostly lead to (1 × 1) terminations as judged by LEED or STM, although (√3 × √3)R30° phases have been reported that are believed to be associated with an oxygen-rich surface produced by exposure to oxygen at elevated temperatures.103 The conventional wisdom reflected in much of this literature is that the (1 × 1) asprepared surface in these studies is vanadyl-terminated, with O atoms atop the surface V atoms of an otherwise half-metal termination. The main evidence for this is the observation in vibrational spectroscopy, both by high-resolution electron I

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

energy loss spectroscopy (HREELS)104 and by IRAS,98 of an absorption band at ∼127−130 meV that is attributed to the V− O stretching mode of a surface vanadyl species. An analysis of NEXAFS (near-edge X-ray absorption fine structure) from this surface, combined with the results of theoretical simulations, also supports the vanadyl-termination model.105,106 In addition, however, it has been found that the as-prepared surface is chemically inactive, with no evidence of chemisorption of a range of adsorbate molecules, whereas after electron beam radiation there is a significant increase in such reactivity.98,107,108 STM studies clearly show that this electron beam treatment can lead to controlled atomic-scale changes to the surface, with the apparent removal of surface atoms that can lead, under certain conditions, to a new (1 × 1) termination;109,110 this has been attributed to the removal of the terminating O atoms of the surface vanadyl species, the “reduced” surface having exposed V atoms (in a half-metal termination) that are more reactive. The fact that reaction of the reduced surface with oxygen restores the original surface condition supports the view that the electron beam irradiation does, indeed, lead to oxygen atom removal. The first attempt to undertake an experimental quantitative determination of the surface structure of the as-prepared V2O3(0001) surface was performed using scanned-energy mode photoelectron diffraction (PhD). As remarked in the Introduction, this technique is best-suited to the study of adsorption structures, because core level photoemission allows one to distinguish photoelectrons that arise only from the adsorbate species. On a clean surface that lacks any significant surface core level shift the PhD structural information (in this case from O 1s and V 2p photoemission) arises from both surface and subsurface emitter atoms, significantly reducing the incisiveness of the technique to determine the surface structure. Nevertheless, this approach has proved to be quite effective in a study of the TiO2(110) surface.111 Its application to V2O3 led to the conclusion that the surface was either half-metal or vanadyl terminated,112 the technique proving insensitive to the presence or absence of the vanadyl O atoms. One notable feature of both of these alternative best-fit structures was a strong (−30%) relaxation of the V−O interlayer spacing between the half-metal layer and the underlying O3 layer, almost identical to that obtained in one theoretical calculation113 for the half-metal termination. We note, though, that another theoretical calculation114 predicts a much larger relaxation (−64%) for this interlayer spacing in the half-metal termination, while both calculations indicate that in the presence of the vanadyl O atoms this relaxation should be significantly reduced (to −17% and −24%, respectively; See Table 4.). While the idea that the V2O3(0001) surface should be either half-metal or vanadyl terminated is clearly consistent with the results of both experimental and theoretical studies of the alumina, chromia and hematite surfaces described above, it is not consistent with the theoretically computed equilibrium phase diagram, reported by Kresse et al.,114 and shown in a simplified form in Figure 5. In fact, the full phase diagram shows that there are no values of the oxygen chemical potential that lead to either a half-metal or a (1 × 1) vanadyl termination being in equilibrium with the surroundings. This conclusion is also supported by the results of independent calculations for ultrathin epitaxial films of V 2 O 3 (0001) grown on αAl2O3(0001).115 Calculations based on partial coverage of a vanadyl species (either one or two such species per (√3 ×

Table 4. Values of the near-Surface Interlayer Spacing Relaxations, Expressed As a Fraction of the Values in the Bulk Crystal, for the Half-Metal (HM) and Vanadyl (VO) Terminations of V2O3(0001) Obtained from the Experimental PhD Investigation and Two Theoretical Studies, As Described More Fully in the Texta method/ model PhD/ HM DFT/ HM DFT/ HM PhD/ VO DFT/ VO DFT/ VO

year (ref)

Δz(Vtop− Otop) (%)

Δz(Otop− V2′ ) (%)

Δz(V′2− FV) (%)

Δz(V2− O2) (%)

2007112

−30/0

−6/−5

8/−3

4

2003113

−30

0

−26

−3

2004114

−64

13

−45

13

2007112

−30

5/7

−11/−8

9/−3

2003113

−17

−5

−6

−3

2004114

−24

−1

−30

8

a Note that in the PhD study optimization from two different starting structures for both the half-metal and vanadyl-terminated models led to tow slightly different sets of interlayer spacing values.

√3)R30° unit mesh, corresponding to coverage of 1/3 ML and 2/3 ML) do indicate these surfaces may be stable at larger (negative) oxygen chemical potential (Figure 2) although at the smallest (negative) oxygen chemical potentials an O 3 termination is found to be most stable, albeit involving a reconstruction (Figure 6) in which V atoms from the outermost

Figure 6. Side view of the reconstructed-O3 termination model of V2O3(0001) found to be stable at low negative values of the oxygen chemical potential in independent theoretical studies.114,115 The arrow indicates the movement of second layer V3′ atoms into the first layer to become V″2 .

half-metal component of the second buckled metal layer are displaced up into the outermost buckled layer, leading to a O3V3O3 termination, with a structure very similar to that of bulk VO2. It is this reconstruction114,115 that stabilizes the otherwise highly polar O3 termination. In fact clear support for this reconstructed-O3 termination has come from a recent report of two independent ion scattering studies,116 one using MEIS, the other using a special form of LEIS. This investigation provides clear evidence that simulations of the data from both techniques based on either the half-metal or vanadyl termination models are inconsistent with experiment, whereas simulations for this reconstructed O3 model agree well with experiment. The interlayer spacings found in MEIS study117 (the higher-energy ions penetrating the subsurface layers) are also in good agreement with the theoretically predicted values as shown in Table 5. Notice that in this table the actual layer J

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

indicate that the O atoms that are hydroxylated are in the underlying O3 layer. While this data interpretation was based on calculations that assumed the presence of half-metal terminations (with or without vanadyl O atoms), the PhD technique is primarily sensitive to the location of scattering atoms that are “behind” the emitter atom relative to the detection direction, so the presence of this half-metal overlayer may be expected to have a rather weak influence on the PhD data. As such, the data may be expected to be compatible with hydroxylation of the terminating-layer O atoms in the reconstructed-O3 model. There is at least one further piece of evidence that appears, at least at first sight, to be incompatible with an O3 termination, namely the STM images that show a (1 × 1) periodicity. If the surface is half-metal or vanadyl terminated the images may be expected to show a single atomic protrusion per surface unit mesh, corresponding to the coverage of these species. An O3 termination, however, has three O atoms per surface unit mesh, so one might expect these three atoms to be resolved in STM images. Indeed, simulated images of both a simple and reconstructed O3 termination,117 using the standard Tersoff− Hamman20 method, do suggest that these surfaces should lead to multiple protrusions in STM images within a single surface unit mesh. However, important variations in the atomic-scale images from the V2O3(0001) surface have been reported both as a function of imaging conditions (notably variation of the bias voltage at fixed tunnelling current120) and in coexistent surface regions at fixed imaging conditions.121 In particular, images that show significant structure within the unit mesh have been reported (e.g., Figure 4 of ref 120), and these may be more compatible with those predicted for the reconstructed-O3 termination. In this context a very recent investigation122 of the surface of a thin film of Rh2O3(0001) grown epitaxially on Rh(111), is of interest. No quantitative surface structure determination has been conducted on this surface. Nevertheless, using a combination of STM, qualitative LEED, photoemission, and DFT calculations, the authors of this study concluded that this surface also shows a reconstructed-O3 termination essentially identical to that shown in Figure 3, with a RhO2 terminating layer. In this case simulated STM images appear to show only a single atomic-scale protrusion per surface unit mesh, similar to that seen experimentally.

Table 5. Comparison of Interlayer Spacings Found Experimentally and Theoretically for the Reconstructed O3Terminated Mode of V2O3(0001) method/ model

year (ref)

z(Otop− V′2) (Å)

z(V′2− V″2 ) (Å)

z(V″2 − V2) (Å)

z(V2− O2) (Å)

z(O2− V3) (Å)

MEIS DFT

2012117 2003114

0.91 0.86

0.18 0.19

0.18 0.19

0.98 0.99

1.27 1.44

spacings are given, as it is less easy to define the changes (relaxations) in these spaces relative to the bulk, becaues of the way the surface structure differs from the bulk. We note, though, that the V−O interlayer spacings in the bulk are 0.985 Å, so the outermost such layer spacing (between the Otop and V2′ metal sublayer) is relaxed by −8% according to the experimental data and by −13% according to the DFT calculations. This same model is also supported by an investigation using a technique of ‘ion beam triangulation’ based on variations in the secondary electron emission as a function of azimuthal incidence angle of grazing incidence 25 keV H atoms118 in which the same interlayer spacings were used but the results indicated small (0.13 ± 0.03 Å) lateral displacements of the surface O atoms. While these new results favoring the reconstructed-O3 termination model are broadly consistent with the theoretically predicted surface phase diagram, typical sample preparation conditions falling in the appropriate range of oxygen chemical potentials, they are clearly at variance with the established wisdom of the many earlier spectroscopic, imaging, and chemical studies. We note, though, that typical surface preparation conditions do lie quite close to the phase boundary with the low-coverage vanadyl termination (and the lowest coverage tested theoretically was 0.33 ML), so at least partial vanadyl coverage could result from some surface preparations. This partial coverage would also be consistent with the results of vibrational spectroscopic studies, which provide no information of the surface coverage of the contributing vanadyl species. Moreover, an O3 termination would be expected to be rather unreactive, while electron beam irradiation could also lead to removal of outer-layer oxygen atoms and enhanced reactivity; this situation is qualitatively similar to the previously proposed removal of vanadyl O atoms from a (1 × 1) vanadylterminated surface. In fact the only previous quantitative experimental structural investigation providing partial support for the (1 × 1)-vanadyl termination model is the PhD study. The data from these experiments have been reanalysed to establish whether or not they may also be compatible with the reconstructed-O3 termination.117 This reanalysis indicates that the PhD data are not compatible with this model if all the layer spacings found in the MEIS study are correct, but some modification of these layer spacing may be compatible with the measurements from both techniques. However, the results of another PhD study,119 namely of a hydroxylated V2O3(0001) surface, may be seen, with hindsight, as providing further evidence for an O3 termination. Specifically, in this investigation PhD data were collected from the chemically shifted O 1s photoemission associated with the surface OH species, providing a means of identifying the location of the hydroxylated O atoms on the surface. For a vanadyl-terminated surface, one would expect the vanadyl O atoms to be preferentially hydroxylated, leaving the OH species atop the outermost (half-metal) V atoms. The PhD data are clearly not compatible with this model, but instead

2.5. Adsorbate Structures

While there have been extensive studies of adsorption on these corundum-phase (0001) surface both experimentally and theoretically, there are very few experimental quantitative structure determinations of these systems. Most notable among these are some rather disparate investigations of hydroxylated surfaces. On α-Al2O3(0001), a SXRD investigation123 of the surface under ambient conditions after an aqueous wash, believed to be more than sufficient to achieve full hydroxylation, appears to be the only quantitative experimental study of a water-exposed surface. The main conclusion of this work is that the alumina surface under these conditions is no longer half-metal terminated, as in its clean state, but is O3 terminated, while the fit to the experimental data also indicated that a layer of O atoms was present approximately 2.3 Å above the surface. The authors infer that this corresponds to a layer of molecular water adsorbed on the hydroxylated surface. Notice, of course, that SXRD is insensitive to the presence or location of the H atoms K

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

hydroxylated O3 layer is actually that of the reconstructed O3 termination. Notice that, in marked contrast to α-Al2O3(0001) and α-Fe2O3(0001), the as-prepared V2O3(0001) surface appears to be completely unreactive to water vapor. This distinction provides further evidence of their different terminations, the implication being that there are no exposed metal atoms on V2O3(0001). If the surface is reduced by electron beam irradiation, water reaction does occur leading to a partially hydroxylated surface,130 and indeed water dissociation can occur at the interface between the as-prepared surface and a condensed multilayer of water in the presence of soft Xray (synchrotron) radiation.119 Beyond these few studies of hydroxylated surfaces, there is a dearth of quantitative structural information on other adsorbed species on these surfaces. There are a number of pure DFT studies, but in the absence of experimental data it is difficult to assess the likely reliability of their conclusions. An EXAFS investigation of low coverages of Ni on α-Al2O3(0001)131 led to the conclusion that the Ni atoms bond to 3-fold coordinated O sites (with a Ni−O bond length of 1.98 ± 0.03 Å) and not to the surface Al atoms. Other data are far more fragmentary; for example, an investigation of CO adsorption on α-Cr2O3(0001) using angle-resolved photoemission132 led to the conclusion that the molecule has its axis strongly tilted relative to the surface normal, quite unlike CO adsorption on most metal surfaces, but the data do not even provide qualitative information on the adsorption site. On the V2O3(0001) surface extensive attempts to identify the adsorption site of several molecular species, including CO, using the PhD technique, proved fruitless.133 In this case, though, these molecules are only found to react with the reduced surface following electron beam irradiation, and the almost complete absence of PhD modulations in the data may be due to the local disorder surrounding the defect sites at which adsorption occurs.

within these layers. Associated with this structure was a significant (53%) reduction in the amplitude of buckling of the outermost full-metal layer below the O3 termination. An AFM study of the α-Al2O3(0001)/liquid water interface124 showed that the (1 × 1) periodicity is retained, but provided no quantitative information on atomic positions or terminating species. Despite there being several DFT studies of the adsorption and dissociation of water on αAl2O3(0001),45,125−127 these investigations are mostly concerned with interaction with the clean half-metal terminated surface and do not appear to address the possibility of substantial reconstruction, although in a molecular dynamics study,128 in particular, it is suggested that the water interaction may create highly mobile species, such as Al(OH)3, that may then diffuse away from the surface, leaving the underlying oxygen termination. This study128 also explicitly considers the situation in which each surface Al atom in the clean surface halfmetal termination is replaced by three H atoms, creating the (hydroxylated) oxygen-terminated surface found in the SXRD study. A SXRD study of α-Fe2O3(0001) in equilibrium with a water-saturated He atmosphere at ambient pressure, combined with DFT calculations, has also been performed.129 In this case, the authors found evidence for the coexistence of domains of two different terminations, one similar to that found in the αAl2O3(0001) study, with an oxygen-terminated surface, but the other in which the half-metal Fe atoms of the clean surface remain in place, but the underlying O atoms were believed to be hydroxylated. The associated DFT calculations actually indicated that the most stable termination in equilibrium with water is the half-metal termination, with each surface Fe atom bound to three OH species, while all O atoms in the underlying O3 layer are also fully hydroxylated, leading them to suggest that the mixed-domain termination found in the experiments may be due to dissolution during sample preparation. On V2O3(0001), the PhD investigation of the hydroxylated surface119 produced by exposure to atomic H has already been mentioned, the results indicating that it is the O atoms in the outermost O3 layer that are hydroxylated (Figure 7). The analysis for this structure assumed that the half-metal Vtop atoms were present on the surface, but in the light of the most recent investigations of the clean surface it seems likely that the

3. ROCKSALT STRUCTURE OXIDE SURFACES 3.1. (100) Surfaces

Another group of oxide surfaces that have been studied extensively comprises the (100) surfaces of oxides having the rocksalt structure, especially MgO but also NiO, CaO, CoO, and MnO. The rocksalt structure has a F-cubic lattice with an atomic basis of one metal atom and one oxygen atom, one at (0,0,0), the other at (1/2, 1/2, 1/2). The single lattice parameter, a, of each of the oxides considered here are as follows: MgO, 4.211 Å; NiO, 4.168 Å; CaO, 4.810 Å; CoO, 4.267 Å; MnO, 4.445 Å; SrO, 5.160 Å; BaO, 5.523 Å. These crystals cleave along the low-energy (100) face, which presents a nonpolar surface containing equal numbers of oxygen anions and metal cations. Despite the problems associated with surface charging, the first quantitative structural studies of these surfaces, using LEED intensity analysis, were some of the very earliest applications of this method of surface structure determination in the 1970s.134−140 While some more recent structural studies of these surfaces have been performed on epitaxial thin films, most work has been on the surfaces of bulk single crystals. Weak half-order LEED beams have been reported from NiO(100) and attributed to antiferromagnetic ordering,141 but in general, all these surfaces show a (1 × 1) surface unit mesh, indicating that no substantial reconstruction occurs. The key structural question therefore concerns the surface layers relaxation, including differential relaxation of the surface anions and cations that leads to a “rumpling” of the

Figure 7. Schematic perspective view of the V2O3(0001)−OH structure as determined by PhD.119 The PhD measurements determined the location of the hydroxylated O atoms, but the positions of the H atoms are identified only in the associated DFT calculations. L

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

terms of a rather large relaxation of (−15 ± 3)% but little or no rumpling,147 but a later ion scattering study, using the higher energies of MEIS (97 keV H+) capable of providing more precise structural data, reported small relaxation and rumpling values consistent with the LEED results.148 Rather precise values and generally consistent values for these parameters have also been reported using SXRD.149 Finally, and most recently, an investigation using a novel technique of fast atom diffraction, a method that is sensitive to the surface rumpling but not the relaxation relative to the underlying substrate, reported a rumpling amplitude of (0.03 ± 0.03)%.150 There have, of course, been many theoretical studies of these relaxation and rumpling phenomena, beginning with simple shell model calculations151 which indicated that the relaxation was probably smaller than −1% but the rumpling may be ∼5%, depending on the potentials used. These early ideas were based on the electrostatic model of Verwey152 suited to the treatment of alkali halides, but later analytical models that seek to interpret the results of self-consistent electronic structure calculations have been designed to account for both electrostatic and covalent effects.153 An extensive listing of the many theoretical treatments of MgO(100) has recently been provided by Schüller et al.;150 in Table 6 the results of several of the most recent calculations, all based on DFT,154−157 are reported. Although there is significant scatter in the experimental values, much of this variation lies within the estimated precision of the measurements, and the general level of agreement between the trends of experimental and theoretical values to show very small values of both the relaxation and the rumpling is rather good. The fact that this is true for a range of different methods of surface preparation suggests that there is no significant influence of adsorption (such as hydroxylation). One interesting feature of several of the DFT calculations is that they investigated not only the MgO(100) surface, but also the (100) face of CaO, SrO and BaO and found that the sign of the rumpling on MgO and these other three surfaces is predicted to be opposite155,157 (Table 7); relatively, the surface O species are displaced outward on MgO(100) and inward on the other three faces. One further study158 found the same effect when comparing only MgO and CaO. Unfortunately, there appear to be no experimental data for SrO(100) and BaO(100), and the only such investigation of CaO(100), based on a quantitative LEED study,139 reported that the comparison between experimental and simulated intensity data were

surface. This effect is shown schematically in Figure 8, in a slightly simplified form, in which it is assumed that only the

Figure 8. Side view of a rocksalt structure (100) oxide metal surface showing the surface relaxation and rumpling.

outermost layer relaxes and rumples. Here dB is the bulk interlayer spacing, while dM and dO are the outermost layer spacings to the metal and oxygen atoms. The relaxation is then given by δrel = 0.5(dM + dO) − dB, while the rumpling parameter is δrump = (dO − dM). Notice that with these definitions an inward relaxation of the average outermost layer to the bulk is defined as negative (as in the discussion of the corundum-phase surfaces), while positive rumpling corresponds to the case in which the outermost O atoms are higher on the surface than the outermost metal atoms. The main results of experimental determinations of the surface relaxation and rumpling of MgO(100) are summarized in Table 6. Most of the early studies used the LEED technique but somewhat different methods of surface preparation. The first such study was based on data from a surface created by in situ cleavage, and explored only the surface relaxation and not the rumpling.134−136 The essential absence of any significant relaxation was also found in a slightly later LEED study, based on three different methods of surface preparation.142 However, a further LEED study, of an air-cleaved surface, subsequently subjected to further cleaning in situ, did find evidence for a small rumpling amplitude of 2% in the surface, but with no detectable relaxation.143 Later LEED studies seemed to favor very little relaxation but a small rumpling amplitude.144,145 The last of these investigations145 also found evidence for a possible rumpling in the second layer of (0.2 ± 2)%. An investigation of the MgO(100) surface using reflection high-energy electron diffraction suggested that following heating to 300 °C a rumpling amplitude of up to 6% occurred,146 but a LEED study of this same heat treatment failed to detect any change.142 In contrast to the evidence from these experiments that there is little or no surface relaxation at this surface, the results of a CAICISS experiment using 1 keV He+ ions was interpreted in

Table 6. Comparison of Experimental and Theoretical Relaxation and Rumpling Parameter Values for MgO(100) method

year (ref)

δrel(1) (%)

LEED LEED LEED LEED LEED CAICISS MEIS SXRD fast atom diffraction DFT DFT DFT DFT

1974−6134−136 1983142 1982143 1991144 1998145 1988147 1994148 1998149 2012150 2003154 2004155 2005156 2005157

−0.3 ± 1.6 0−2.5 0 ± 0.75 1±2 −0.2 ± 0.7 −15 ± 3 −1.0 ± 1.0 −0.56 ± 0.35 −2.9 −0.5 −1.4 0.003 M

δrump(1) (%) 0 2±2 5 ± 2.5 3.3 ± 1.5 small 0.5 ± 1.0 1.07 ± 0.5 0.03 ± 0.03 3.8 1.8 0.95 2.27

δrel(2) (%)

0.2 ± 2

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

the relaxations are a consequence of the in-plane compressive stress associated with the pseudomorphic growth. There has also been one quantitative LEED study of the surface of a thicker (22 ML) FeO(100) film grown epitaxially on Ag(100),166 which indicates a marginally significant rumpling amplitude. While the issue of the sign and magnitude of the surface layer rumpling of these rocksalt oxide surfaces has clearly generated a lot of theoretical activity and a significant number of quantitative experimental investigations, the overall conclusion is clearly that the effects are small and in many cases lie at the limits of detection of most experimental methods. Detailed comparisons of the results using different methods appear to indicate conflicting results, but few of the associated differences can be regarded as really significant. Lopes et al.166 have suggested that there is a simple linear relationship between the magnitude of the rumpling and the oxide lattice parameter; a curious feature of their plot of experimental values according to this relationship is that the scatter of data points is far smaller than the estimated errors of the individual results, perhaps implying that the precision of many of the measurements is better than has been estimated. Nevertheless, there are certainly some significant discrepancies between experimental values, most notably in the case of MEIS and LEED studies of NiO(100) and MnO(100), with the two techniques yielding rumpling amplitudes of opposite sign. One possible factor that may be relevant to this disagreement is that the MEIS studies were of the surface of bulk crystals, while those conducted by LEED were on epitaxial films. These two different approaches lead to different challenges in preparing well-ordered and clean stoichiometric surfaces, and it is far from clear which method is yielding the more reliable results.

Table 7. Predicted Values of the Surface Rumpling Amplitude, δrump(1), for the (100) Surfaces of MgO, CaO, SrO, and BaO in Three Different DFT Calculationsa year/reference

MgO (%)

CaO (%)

SrO (%)

BaO (%)

2.0 1.8/2.2 2.27

−0.5 −0.6/−0.2 −0.68

−1.5/−1.3 −2.26

−1.8/−1.6 −4.89

158

2000 2004155 2005157 a

In the case of the 2004 paper of Broqvist et al. two values are given arising from the use of two different functional.

insensitive to positive values of the rumpling, but that no negative values were tested. LEED structural studies have also been performed on NiO(100), MnO(100), CoO(100), and FeO(100) surfaces, and the main results regarding their surface relaxation are summarized in Table 8. All four oxides are antiferromagnetic in the bulk below the relevant Néel temperatures (525, 116, 291, and 198 K, respectively). Antiferromagnetic ordering is thought to account for the observation of weak half-order diffracted beams in LEED from some NiO(100) surfaces,141 but although some early theoretical work appeared to suggest such measurements could be interpreted in these terms,159 no detailed measurements have been made. Analysis of the integral order beam intensities from NiO(100) indicates that there is no structural modification of the surface other than possible relaxation and rumpling within the (1 × 1) unit mesh; a slight inward relaxation of is favored, with no detectable rumpling.137,138,160 DFT calculations predict small first-layer relaxation and rumpling amplitudes.161 On MnO(100) surfaces, two experimental studies find rumpling amplitudes of opposite sign. In particular, a MEIS study of the surface of a bulk crystal162 yielded a rumpling amplitude of −3.6% with no significant relaxation of this outermost layer, whereas a LEED study163 of thin films (of thickness of 24 ML and 48 M) grown epitaxially on Ag(100)) found a rumpling amplitude of +(4.8 ± 2.0)% with a interlayer relaxation of (1.0 ± 0.6)%. A DFT calculation for this surface161 predicted very little rumpling but a small relaxation. An early LEED study of the (100) surface of a bulk CoO crystal164 concluded that, within the limited precision of ±3%, the surface showed no relaxation or rumpling and was ideally bulk-terminated. Unfortunately, the only more recent studies of CoO are of epitaxial films that were so thin that this was judged to influence the surface structure. In particular, a LEED study165 of the surface of a CoO(100) film with an average thickness of approximately 4 ML, grown on Ag(100), was found to display significant outward relaxation of the outermost two layers by 6% and 2%, respectively with associated rumpling amplitudes of 3% and less than 1%, but the authors argue that

3.2. Adsorbate Structures on (100) Surfaces

Despite the interest in the adsorption properties of these surfaces there are very few quantitative experimental structure determinations; the few results that are available relate to MgO(100) or NiO(100). 3.2.1. Molecular Adsorption on MgO(100). On MgO(100) the adsorption of water under UHV conditions has been shown by LEED167,168 and helium-atom168,169 scattering (HAS) to lead to two ordered phases, namely c(4 × 2) in the temperature range 100−180 K and p(3 × 2) at temperatures of 185−221 K. This latter phase was found to contain a glide symmetry plane.168,170 While vibrational (infrared) spectroscopy was used to infer some information regarding the molecular orientations, no detailed picture of the adsorption structure was obtained. However, a much more complete structural model of the p(3 × 2) phase was obtained in a quantitative LEED study, aided by semiempirical potential calculations.171 The resulting model is

Table 8. Comparison of Reported Relaxation and Rumpling Parameter Values for the (100) Surfaces of NiO, MnO, CoO, and FeO surface NiO(100) MnO(100)

CoO(100) FeO(100)

method LEED DFT MEIS LEED DFT LEED LEED

year (ref) 138

1977 2003161 2004162 2005163 2003161 1979164 2007166 N

δrel(1) (%)

δrump(1) (%)

(−3−0) ± 5 −1.4 0.1 ± 0.7 1.0 ± 0.6 −1.5 0±3 −2.4 ± 4.6

0±5 −2.5 −3.6 ± 0.7 4.8 ± 2.0 −0.5 0±3 3.9 ± 3.2

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

of 1 × 10−8 mbar, while above 51 K only (1 × 1) periodicity is seen, attributed to a CO lattice gas. An intermediate ordered phase reported in the LEED study was not confirmed by HAS. Vibrational spectroscopy179 also shows a distinct change around 51 K, suggestive of at least two distinct local CO geometries in the ordered phase; the combination of these results led to the suggestion that in the c(4 × 2) phase there are six partially tilted CO molecules per unit mesh. The many theoretical studies, as reviewed and addressed previously180,181 have concluded that the CO bonds atop surface Mg atoms through the C atom in a perpendicular orientation. Direct experimental evidence for this adsorption site comes from an STM study,182 but in this case the epitaxial MgO(100) films studied, grown on Ag(100), were only 2 ML thick; as such, it is quite possible that the underlying metallic substrate has a significant influence on the bonding, making this result not directly comparable with studies on thicker crystals. A similarly incomplete picture of the adsorption of CO2 on MgO(100) is provided by the limited amount of experimental information. LEED observations183,184 show that in the temperature range ∼60−80 K there is an ordered (2√2 × √2)R45° phase that appears to have two perpendicular glide lines, implying that the space group symmetry is pgg. This has led to the suggestion that the molecules are arranged in a herringbone pattern on the surface, with each O atom bound to nearly surface Mg ions. Polarization-dependent infrared spectroscopy led to the suggestion that the molecules are tilted with respect to the surface by 27 ± 10°.184 Above 93 K the ordered phase is lost. However, there is no specific experimental evidence to identify the local adsorption site, and there are only theoretical studies that address this issue (e.g., refs 185 and 186). A very different surface treatment, exposing the MgO(100) surface to up to 260 Torr of CO2 at room temperature for 15 min led to the identification by C K-edge NEXAFS of a surface carbonate species.187 The polarization-angle dependence of these spectra indicated either no orientational order of a tilt angle of ∼55°. One adsorption system for which there is a complete experimental structure determination on MgO(100) by quantitative LEED is that of acetylene (C2H2) adsorption.188 The structure determination of the ordered (2 × 2) phase seen at 88 K was aided by semiempirical potential calculations which gave rather good agreement with the optimized LEED structure shown in Figure 10. Interestingly, the results were also found to

of a water monolayer containing six molecules per unit mesh with their planes parallel to the surface and the O atoms close to atop Mg surface atoms. It is important to note, of course, that LEED (like almost all surface structural techniques) is insensitive to the location of the H atoms, and is essentially a probe of the location of the O atoms relative to the underlying MgO surface. As such, these experiments provide no direct information on the state of dissociation of the adsorbate layer. The layer spacing of the overlayer O atoms, relative to the outermost MgO surface, was found to be 2.05−2.22 ± 0.05 Å. Subsequent DFT modeling of this p(3 × 2) phase has led to broadly similar models172 of an intact water layer, but later studies173,174 have proposed that the water layer is partially dissociated. Most recently such calculations have been performed for both the p(3 × 2) and c(4 × 2) phases,175 with the resulting models being compared with a range of experimental spectroscopic data, particularly including vibrational spectroscopy which does provide information on the presence of vibrational modes associated with H in the layer. These models of the two phases contain six and ten water molecules per unit mesh, respectively, with two of these molecules in each case being dissociated to produce four hydroxyl species. Figure 9 shows the resulting model of the p(3 × 2) phase.

Figure 9. Plan view of a single unit mesh of the model of the MgO(100)p(3 × 2)-water phase recently proposed.175 For clarity the O atoms in the outermost substrate layer (Osub) are distinguished from those arising from the adsorbed water (Oad). The unit mesh contains four intact water molecules adsorbed on surface Mg atoms, two OH species adsorbed on surface Mg atoms, and two OH species resulting from H adsorption on surface O atoms.

A second quantitative structural study of water on MgO(100)176 was performed on a surface following exposure to liquid water or a high vapor pressure of water vapor at room temperature. The investigation of this very differently prepared surface exploited the technique of XSW, although in the absence of chemical-state sensitivity or intrinsic surface specificity in the measurements, a novel method was used to attempt to separate the surface and bulk components of the oxygen-related absorption profiles and their associated structural information. The O atoms in the adsorbed surface layer were found to lie with a spacing above the outermost substrate layer equal to that of the bulk. The results were interpreted in terms of a fully hydroxylated surface with OH species atop the surface Mg atoms with a Mg−O bond length identical to that of the bulk. Despite the interest in CO as a probe molecule of surfaces, most studies of CO adsorption on MgO(100) have focused on the strength and nature of the weak surface bonding, with little hard experimental information on the structure. Both LEED177 and HAS178 studies established that an ordered c(4 × 2) phase is formed at temperatures below 40 K in a CO partial pressure

Figure 10. Plan view of the MgO(100)(2 × 2)-C2H2 adsorption structure as determined by LEED.188 Note that only the C atoms of the molecules are shown, as the LEED technique is insensitive to the location of the H atoms. O

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

be in rather good agreement with earlier neutron diffraction experiments189 in the relative positions of the molecules within the adsorbed layer (neutron diffraction being unable to detect the position of this layer relative to the substrate). The LEED results yielded an adsorbate−substrate layer spacing of 2.50 ± 0.05 Å. A close parallel with a prior neutron diffraction study190 was also found in a HAS study191 of adsorption phases of molecular hydrogen adsorbed on MgO(100) in the temperature range 7− 12 K, identifying ordered c(2 × 2), c(4 × 2), and c(6 × 2) phases; however, no quantitative information on the substrate registry was obtained by these methods. 3.2.2. Atomic adsorbates and the early stages of epitaxy on MgO(100). In addition to these studies of molecular adsorbates on MgO there have also been quite a number of investigations of metal atoms on the surface. In many cases, though, deposited metal atoms on oxide surfaces produce local clusters, and while such systems are potentially of interest in terms of the chemical reactivity of supported nanoparticles, obtaining quantitative information on the atomic-scale structure of the interface is beyond the scope of most surface probes, although grazing incidence small-angle Xray scattering (GISAXS), a technique related to SXRD, has been shown to be capable of characterizing the particle shapes and sizes and does contain information on the interfacial structure.192 However, quantitative structural studies are mostly restricted to identifying the local geometry of isolated adatoms, or the structure of well-defined long-range ordered single atomic layers. Several structural studies of Ca on MgO(100) have been conducted on surfaces in which the Ca has segregated as an impurity from the substrate under suitable thermal treatments. This approach seems to have avoided the clustering that has been found to occur in Ca deposition on the surface.193 The first study of this segregated surface194 exploited a variant of the LEIS technique, using a 1 keV beam of incident neutral He atoms. The azimuthal-angle dependence of the scattering from the Ca atoms indicated that they occupied Mg substitutional sites in the surface at a height of 0.4 ± 0.1 Å above the surrounding surface, while the surface Mg atoms appeared to be essentially coplanar with their O atomic neighbors. While no ordered surface superstructure associated with the Ca segregation was seen in LEED in this study, a later SXRD investigation149 found the segregation led to an ordered (√2 × √2)R45° structure. The most likely reason for this apparent inconsistency is a difference in Ca surface coverage. In the LEIS study Ca segregation was seen to increase with increasing temperature, most probably limited by the kinetics of segregation, but the highest temperature used was 1200 °C. By contrast, in the SXRD study the surface had been annealed in air at 1500 °C for many hours prior to lower-temperature cleaning in UHV. Indeed, a more recent LEIS study195 which also found only a (1 × 1) LEED pattern from the segregated surface, established by a combination of LEIS and conventional Rutherford backscattering measurements that under these conditions the Ca surface coverage was approximately 0.2 ML, the Ca atoms being located entirely in the outermost layer. By contrast, an ordered (√2 × √2)R45° structure would imply a coverage of 0.5 ML. The analysis of the diffracted beam intensities in the SXRD study, assuming that Ca occupies half the Mg substitutional sites in the outermost layer alone (Figure 11) in an ordered (√2 × √2)R45° phase, led to the conclusion that the Ca atoms lie above the surrounding

Figure 11. Schematic view of the MgO(100)(√2 × √2)R45°-Ca structure showing the distortion parameter δz.

outermost-layer O atoms (δz(Ca) = 0.63 ± 0.03 Å), while the surface Mg atoms lie below these surface O atoms (δz(Mg) = −0.066 ± 0.14 Å). Early theoretical calculations, considering the energetics of Ca segregation in different ordered surface phases, favor the creation of the ordered (√2 × √2)R45° segregation phase such that the “calcium impurity moves substantially out of the surface to relieve the lattice strain”196 caused by the fact that the Ca2+ ion is significantly larger than Mg2+. More recent DFT calculations197 have provided detailed quantification of this statement, with significant rumpling found to occur in several of the outermost layers as a result of the Ca segregation, and the specific distortions found for the (√2 × √2)R45° (δz(Ca) = 0.55 Å, δz(Ca) = −0.16 Å), agree rather closely with those of the SXRD experiments, although some significant subsurface distortions reported in the SXRD study are found to be much smaller in the DFT calculations. A quantitative LEED study of the early stages of growth of Fe on MgO(100)198 at room temperature led to the conclusion that the growth is pseudomorphic, with the first layer Fe atoms lying atop surface O ions at a spacing of 2.0 Å (Figure 12),

Figure 12. Schematic view of the MgO(100) surface showing the local site, atop a surface O atom, favored in several studies of Fe, Ag, and Pd adsorption.

somewhat shorter than the distance in bulk FeO (2.14 Å) or in Fe3O4 (2.10 Å). This adsorption site is consistent with the results of ab initio calculations199 that find, however, a significantly longer Fe−O distance of 2.30 Å. The LEED study indicated that successive layers grow initially in a bct structure, transforming to bcc for thickness great than ∼10 Å. A later SXRD study200 found, by contrast, that the Fe film grows initially with the lateral lattice parameter 3% less than that of the MgO surface and much closer to that of bulk Fe, while the thicker films also showed significant islanding. The methods of surface preparation of deposition in the two experimental studies were rather different however. Most notably, in the LEED study the MgO surface was first cleaned by elevated temperature (800 °C) annealing in a partial pressure (1 × 10−3 Pa) of oxygen until the initial carbon contamination seen in P

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Auger electron spectroscopy had been removed. By contrast, the surface studied by SXRD was cleaned in UHV only be heating to 360 °C, a procedure less likely to ensure removal of possible adsorbates or contaminants. There have also been a number of investigations of Ag deposition on MgO(100); a lattice mismatch of only 3% appears to favor epitaxy. An early investigation of the interface between MgO(100) and an Ag epitaxial film with a thickness of 1000 Å using transmission electron microscopy indicated the presence of misfit dislocations to accommodate the lattice strain and showed that the Ag atoms occupy local sites atop either the surface O or Mg atoms.201 Early ab initio calculations clearly favored the O-atop site (Figure 12) with the Ag−O distance in the range 2.34−2.70 Å,202−204 in good agreement with a SEXAFS study at low coverage which found a value for this bond length of 2.53 ± 0.05 Å.205 More recently a series of SXRD206,207 and GISAXS208 studies have provided detailed characterization of the structure and growth of Ag films of increasing thickness on Mg(100). SXRD studies of Pd growth on MgO(100) have also identified the local Pd site at the interface of epitaxial films as atop the surface O atoms with a layer spacing of 2.22 ± 0.03 Å.209,210 A similar investigation of Ni growth on MgO(100)211 identified this same site for the Ni atoms at the interface with a layer spacing of 1.82 ± 0.02 Å, close to the value found in ab initio calculations of 1.87 Å.212 GISAXS has provided detailed characterization of the growth and morphology of Pt on MgO(100).213 3.2.3. Adsorption on NiO(100). Studies of adsorption on NiO(100) have focused on molecular adsorbates, and experimental results emerging from the adsorption of CO and NO have proved to be particularly challenging for theory. The initial structural information for both adsorbates came from NEXAFS studies, which indicated that while CO adsorbs with its axis perpendicular to the surface,214 the axis of NO is strongly tilted215 from the surface normal by approximately 45°. Both studies were performed on NiO(100) films grown on Ni(100), but despite the fact that these films, the faces of which are tilted relative to the underlying metal, were believed to be heavily defected, the behavior of NO adsorption in particular seemed to be identical to that on a surface of a bulk oxide.215,216 Subsequently, the PhD technique was applied to similarly prepared films to determine the adsorption geometry more completely. Specifically, CO217,218 was found to adsorb atop surface Ni atoms with a tilt of the molecular axis of 12 ± 12° and a C−Ni bond length of 2.07 ± 0.02 Å, while NO218−220 was found to bond to the same site but with a molecular tilt of 59(+31/−17)° and a N−Ni bond length of 1.88 ± 0.02 Å (Figure 13). Clearly the tilt angles obtained in these studies are entirely consistent with the earlier NEXAFS studies, and the adsorption sites are consistent with previous theoretical studies based on Harteree-Fock cluster calculations,215,221,222 but the adsorption bondlengths differ considerably from the theoretically predicted values. Specifically, the theoretical value for the N−Ni bond length is 0.22 Å longer215 than the experimental value, while for the C−Ni bond length the theoretical values are 0.42 Å221 and 0.79 Å222 longer than experiment. Achieving a good description of NiO and its adsorption properties using more modern DFT methods is a significant challenge and a number of more recent papers223−226 have addressed the underlying problems and methods that can provide much better agreement with the experimental bondlengths. A similar PhD study of the adsorption of NH3

Figure 13. Model of the NiO(100)/NO local adsorption geometry as determined by the PhD technique.218

on NiO(100) was also conducted;217,218 in this case too, the adsorption site was found to be atop a surface Ni atoms with a N−Ni bond length of 2.06 ± 0.02 Å; the results for this adsorbate also indicated a significant relaxation perpendicular to the surface of the bonding Ni atom of 0.11 ± 0.02 Å. An early investigation227 of the interaction of H2S with NiO(100) using qualitative LEED and S coverage determination, led to the proposal that, following a high exposure (∼2 × 104 mbar.s) at 570 K, the oxide surface was reduced and the surface layer was effectively a single-layer Ni(100)c(2 × 2)-S phase in which S atoms occupy 4-fold coordinated hollow sites, essentially identical to the equivalent ordered phase on the metallic Ni(100) surface. In particular, LEED indicates that the overlayer is incommensurate due to the shorter Ni−Ni separation in the metal than in the oxide. This conclusion was subsequently supported by a SEXAFS investigation228 that found a Ni−S bond length of 2.21 ± 0.02 Å while the Ni−Ni in-plane distance was determined to be 2.77 ± 0.09 Å, 6 ± 4% smaller than in the underlying NiO(100). 3.3. (111) Surfaces

While the (100) faces of rocksalt structures comprise layers with equal numbers of anions and cations, leading to a nonpolar surface, this is not true for the (111) faces which comprise alternating layers of pure anions and pure cations. An ideal bulk termination of such a polar surface has a divergent electrostatic surface energy and thus is unstable and must undergo some form of electronic or structural reconstruction (e.g.229). The generally favored solution to this instability is the so-called octopolar (2 × 2) reconstruction, first proposed for NaCl(111),230 and shown schematically in Figure 14.

Figure 14. Plan view of the octopolar (2 × 2) reconstruction proposed for the (111) faces of rocksalt structure oxides. The full lines show the (2 × 2) unit mesh. The three (100) nanofacets formed in each unit mesh are shown in white. Q

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Experimental structural studies of a MgO(111)(1 × 1) surface using a combination of high energy electron diffraction and photoelectron diffraction,245 and quantitative LEED,234 have both concluded that the surface structure is consistent with a hydroxylated bulk termination. Of course, none of these techniques are able to detect the presence of the surface H atoms explicitly, but the outermost layer spacings found in the experiments are consistent with the relaxations predicted in ab initio calculations for the hydroxylated surface. Note that in neither study was the surface specifically exposed to water vapor, although the sample preparations did involve annealing in air. Surface hydroxylation also appears to stabilize a (1 × 1) surface phase on NiO(111). In particular, and early qualitative LEED study246 of the growth of NiO(111) on Ni(111) found the surface showed only weak and diffuse half-order diffraction peaks, while electron energy loss spectroscopy showed the presence of surface OH species; annealing largely removed the OH but enhanced the (2 × 2) diffracted beams which the authors suggested may correspond to the predicted230 octopolar reconstruction. A combined LEED and STM study of similar films, aided by XPS to identify surface hydroxyl species, also concluded that a (1 × 1)−OH surface phase exists on NiO(111).247 Similar behavior, of a mainly (1 × 1) phase after initial preparation, and increasing (2 × 2) coverage after vacuum annealing, was found in a SXRD study of the surface of a bulk NiO(111) crystal,248 the surface initially being prepared by annealing in air at 1000 °C. Subsequently, the same group conducted a full SXRD structure determination of the NiO(111)(2 × 2) phase,249 using both a bulk NiO crystal and a thin (∼5 ML) film grown epitaxially on Au(111). Although the diffraction experiments were performed under UHV conditions, the surface of the bulk crystal was studied after ex situ preparation (annealing in air at 1300 K) and showed some surface contamination; subsequent in situ cleaning was found to degrade the surface order unacceptably for the SXRD study. The NiO epitaxial film was prepared in situ. Analysis of the data from the surface of the bulk crystal showed good agreement for the Ni-terminated octopolar reconstruction. Interestingly, the diffracted beam intensities of the integral order beams were found to be essentially identical to those measured earlier from the nominal (1 × 1) surface,248 indicating that this surface actually also shows the octopolar reconstruction but that the poor lateral order meant the (2 × 2) half-order diffracted beams were not detectable. It appears, therefore, that the nominal (1 × 1) diffraction pattern was not consistent with an unreconstructed (1 × 1) hydroxylated bulk termination, as implied by the similarity with the earlier LEED study. On the surface of the thin film the results for the (2 × 2) phase were somewhat different; specifically, on this surface, the analysis indicated that the surface comprised both Niterminated and O-terminated octopolar reconstructions, separated by single (rather than double) step heights, with coherent interference of the X-ray scattering between the different regions. An alternative mode of analysis of these data from the thin-film surface conducted by another group250 led to a somewhat different conclusion for the structure of this surface, the outermost layer being judged to have the same structure as the Ni-terminated octopolar reconstruction, but with different occupations of the second and third layers. Recently, a combined TED and DFT study of the (111) surface of bulk NiO crystals236 has also questioned the

By removing 3/4 of the outermost layer atoms and 1/4 of the second layer, the resulting (2 × 2) structure presents (100) nanofacets that are, themselves, nonpolar, and this structure is found theoretically to be energetically stable. Notice that there are two possible versions of this octopolar reconstruction, one metal-terminated (shown in Figure 14), the other oxygenterminated. DFT calculations indicate that the relative stability of these structures depends on the oxygen chemical potential for both MgO(111)231 and NiO(111),232 although on NiO(111) these two terminations are found to be almost degenerate in energy over a wide range of oxygen chemical potential. However, while a (2 × 2) reconstruction has been observed in many experimental studies of these surfaces, there is certainly no universal acceptance that these octopolar reconstructions are consistent with all the data. One complication is certainly that structural differences have been seen in experiments performed under different conditions of temperature and oxygen partial pressure,233 but also there is considerable debate regarding the influence of hydroxylation of the surface (intentional or otherwise).234−237 Early attempts to prepare MgO(111) surfaces for investigation by LEED238 led to surfaces interpreted as presenting (100) facets, although the orientation of these facets was later called into question.239 The first attempt to undertake a quantitative structure determination of this surface240 appears to have been using transmission electron diffraction (TED), following surface preparation by annealing (at 1450−1650 °C) under a vacuum of ∼5 × 10−7 mbar. Three different surface phases were identified with increasing annealing temperature, namely (√3 × √3)R30°, (2 × 2) and (2√3 × 2√3)R30°. On the basis of the interpretation of the TED data they concluded that all three phases involved differing arrangements of equilateral oxygen trimers on the surface and that previously proposed models based on nanofacets for these three phases241,242 (including the (2 × 2) octopolar reconstruction) gave poor fits to the experimental data. A later investigation of the MgO(111)(2 × 2) phase233 by SXRD, combined with theoretical calculations, found a distinct difference in the surface structure at high and low temperatures (∼500 and 100 K) under partial pressures of oxygen. At low temperature the best fit was found for a 1:1 mixture of the oxygen-terminated octopolar reconstruction and a structure comprising a (0001) Mg epitaxial layer on the O-terminated octopolar surface. This Mg overlayer structure was found to cover increasing amounts of the surface as the temperature was raised. A newer application of the TED technique, but in this case combined with surface characterization by XPS and the use of complementary DFT calculations, was used to investigate both the (√3 × √3)R30° and (2 × 2) phases of MgO(111).235 A key finding of the XPS was evidence for OH species on the surface; no deliberate exposure to water vapor was involved in the surface preparation, but one step of the preparation involved annealing in air. They conclude that the data from the (2 × 2) phase, in particular, is consistent with the Mg overlayer model found in the SXRD study, and not with the oxygen trimer model proposed in the earlier TED study. However, they also conclude that the presence of water and hydroxylation of the surface is a major factor in determining the observed surface structures. In fact the influence of hydroxylation on the structure of MgO(111) had previously been investigated in the context of stabilization of an unreconstructed (1 × 1) surface.243,244 R

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

importance of hydroxylation in the (2 × 2) phase and in a (√3x√3)R30° phase detected in this study. The authors argue that the results of the direct inversion of the TED data are not consistent with the Ni-terminated octopolar structure favored in the SXRD study but may be consistent with the epitaxial overlayer model proposed for MgO(111). On the basis of their DFT calculations, they argue that such a structure could only be stable on a hydroxylated surface. Further DFT calculations237 of the equilibrium surface structure as a function of oxygen and water chemical potential also point to the stability of hydroxylated surfaces (Figure 15) under a wide range of oxygen and hydrogen chemical potential, although the octopolar reconstructions are found to be stable at high temperatures and H-poor conditions.

be more representative of the bulk have also been investigated on this same substrate.255−257 Interestingly, the initial growth in an oxygen-rich atmosphere was found to lead to a Co3O4(111) film, but subsequent annealing in UHV led to a transformation to CoO(111). A reversible phase transition from (1 × 1) to (√3 × √3)R30° is found to occur on cooling below 50 °C. However, a detailed structure determination of the (1 × 1) surface on films up ∼130 Å led to the conclusion that there is a lateral registry shift just below the surface such that the outermost surface layers actually have the structure of the wurtzite CoO phase, rather than the rock-salt structure of the underlying material. While the use of thin films reduces the electrostatic energy associated with a polar surface, these films are sufficiently thick that the role of the underlying interface is unlikely to play a major role in the surface structure. Unlike MgO(111) and NiO(111), it appears that the more modest surface structural modification, with surface rumpling, the interfacial registry shift, and the (√3 × √3)R30° ordering, are sufficient to stabilize the surface, and there is no evidence of the (2 × 2) octopolar reconstruction. Studies of FeO(111) surfaces seem to have all used epitaxial films grown on a range of substrates including Mo(111),258 Cu(100),259 Fe(110),260,261 and Pt(100)262 although by far the most studies have been on Pt(111).263−266 In general the FeO films are only stable to thicknesses ∼2 ML, transforming to Fe3O4 at larger thicknesses. Most attention has been focused on FeO(111) films ∼2 ML on Pt(111) with the main structural information coming from STM, XPD, and LEED. The large (10%) interfacial lattice mismatch leads to the formation of coincidence lattice structures. Both STM modeling,267 as well as the XPD results,265 indicate that the films are oxygenterminated with no reconstruction; it is suggested that for such ultrathin films the electrostatic energy is insufficient to render the polar termination unstable. The XPD results also indicate a strong contraction (from 1.25 Å to 0.68 Å) of the outermost O−Fe layer spacing relative to the bulk oxide that contributes to this stability. A quantitative LEED study of the (2 × 9) coincidence lattice FeO(111) monolayer structure on Pt(100) also favors oxygen termination of the surface. Particular attention in the study of these films has been to their ferromagnetic or antiferomagnetic properties and the relationship of the magnetic and structural properties.261,268,269 A recent study of these ultrathin films on Pt(111)270 indicates that further oxidation can lead to a FeO2 island superstructure that may be a precursor to transformation to an α-Fe2O3(0001) film, but the investigation provided no quantitative experimental structural information. In contrast to this work on ultrathin FeO films, there is also one very recent report of the growth on FeO films with a thickness of up to 16 ML grown epitaxially on MgO(111).271 The films were grown under UHV conditions, using a 30 Å homoepitaxial MgO film grown on the bulk MgO substrate prior to FeO growth, yet both this MgO(111) layer and the FeO(111) film showed a clear (1 × 1) LEED pattern, indicative of unreconstructed surfaces. There have been very few structural studies of adsorbates on these rocksalt MO(111) surfaces other than the effect of hydroxylation mentioned above. NEXAFS measurements from CO and NO on NiO(111)252 have been performed, leading to inferred angles between the molecular axes and the surface normal of ∼45° and 54°, respectively. In view of the fact that CO is known to adsorb on NiO(100) with its axis perpendicular to the surface, this was taken to infer that the

Figure 15. Schematic plan view of the hydroxylated NiO(111)(1 × 1)H structure favored by DFT calculations.

Overall, the true structures of the (2 × 2) phases of both MgO(111) and NiO(111) remain unclear. Most studies have been performed under conditions that may well lead to hydroxylated surfaces, so while several experimental investigations do favor the octopolar reconstructions or minor variations of these structures, it is difficult to know the extent to which they may be influenced by surface hydrogen. Of course this question of the role of hydroxylation is common in all studies of oxide surfaces, but the rocksalt oxide (111) surfaces appear to be particularly reactive in contrast, for example, to the (100) surfaces of the same materials.243,244 Moreover, in view of the expected instability of an unreconstructed termination of the clean (111) surface, it is tempting to attribute all apparently unreconstructed surfaces to the influence of adsorption, with hydroxylation being regarded as the most likely cause. The behavior of the CoO(111) appears to be rather different. Attempts to study the surface of a bulk crystal using SXRD251 led to the conclusion that the surface was covered by a film of ∼50 Å thickness of Co3O4; removing this by UHV annealing and ion bombardment led to the formation of metallic Co clusters, while annealing this surface in a partial pressure of oxygen only restored the Co3O4 film. However, rather more work has been done on epitaxial films. An early study of CoO(111) grown on Co(0001)252 was concerned mainly with the adsorption on CO and NO on the surface, and while the authors assumed that the surface (which was hydroxylated) possessed the octopolar reconstruction, the only real evidence for this was the orientation of the adsorbed CO molecules (determined by NEXAFS) that was consistent with CO being oriented perpendicular to the (100) nanofacets of this structure. The structure of novel ultrathin films on an unreconstructed (1 × 1) Ir(100) surface, comprising only one of two atomic layers, have been investigated in depth by quantitative LEED, supported by STM and other techniques.253,254 However, somewhat thicker films that are likely to S

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

theory does predict that a surface with 1/3 ML of VO should be stable on the surface at closely similar conditions, and it is certainly possible that calculations for a lower coverage of this species may prove to be even more stable. There may therefore be no conflict with the fact that vibrational spectroscopy shows evidence of some vanadyl coverage while the recent structure studies fail to find evidence for these vanadyl species. In this regard, we may also note that the evidence for fractional occupation of some surface sites found experimentally for Cr2O3 and Fe2O3 have not been explored theoretically. Of course, such calculations require the use of much larger unitmesh structures, but should now be readily accessible to modern computational resources. On the rocksalt oxide (100) surfaces much attention has been devoted to measuring and calculating the surface layer relaxations and rumpling. Comparisons of theory and experiment, and indeed of theory and theory, and experiment and experiment, indicate that the effects may be too small to be determined with sufficient reliability. In fact many of the experimental studies are from the early years of development of the methodology of the quantitative LEED technique, but the small number of more recent experimental studies do not appear to be sufficiently more precise to resolve the outstanding, but rather subtle, questions. On the rocksalt oxide (111) surfaces, there are a significant number of experimental studies of MgO(111) and NiO(111) that favor some or all of the surface having the octopolar (2 × 2) reconstruction first proposed on theoretical grounds, but at least some of these studies may well be of surfaces that are at least partially hydroxylated; the fact that there is evidence of a stable (1 × 1)-OH phase on these surfaces further confuses the experimental situation. The results for the epitaxial film of CoO(111) indicating a modified wurtzite-like stacking of the outermost layer are particularly intriguing. Quantitative structural information on adsorbate structures on these surfaces is sparse, although a few reasonable trends emerge. On the nonpolar rocksalt oxide (100) surfaces simple molecules, such as CO and NO bond atop the surface metal atoms, whereas as metal atoms bond atop the surface O atoms. Ca on MgO(100) occupies Mg substitutional sites in the surface, but again bonds to surface O atoms. Nevertheless, in this area in particular, many more experimental studies are required.

underlying surface possessed the octopolar reconstruction, leading to (100) nanofacets at approximately the correct angle to provide the same local geometry. However, the authors noted that the surface was hydroxylated prior to the exposure to these two adsorbate species, so there must be significant ambiguities in this interpretation. There have also been quite a number of investigations of metal deposition on these surfaces, but in many cases, it seems likely that the initial surfaces were hydroxylated, and none of these studies involve true quantitative experimental structure determinations. A growing interest in the properties of metallic nanoparticles, including single metal atoms, on oxide surfaces has led to STM studies that do provide some structural information. In particular, STM studies of single Au adatoms on the 2 ML FeO(111) film on Pt(111)272 have identified preferred adsorption sites that relate to the relative registry of the oxide/substrate registry within the coincidence lattice, and a recent theoretical study273 has sought to generalize the understanding of such adsorption to other metals and to MgO(111) monolayers.

4. GENERAL DISCUSSION While the number of structural studies of oxide surfaces has greatly increased in the last 20 years, it is clear that there remain many open questions, even concerning the structure of the clean surfaces. One factor that seems to be particularly relevant to the structure of the rocksalt oxide (111) surfaces is the issue of hydroxylation. The possible influence of surface OH species on the structure (and chemistry) of oxide surfaces is a general one, but the rocksalt oxide (111) surfaces appear to especially reactive to molecular water that is present in abundance in air at ambient pressures, but is also a residual component in UHV. It appears that rocksalt oxide (100) surfaces, and at least some of the corundum-phase (0001) surfaces, are likely to be less influenced by this problem. In the case of the corundum-phase (0001) surfaces, there seems to be broad consistency of the surface structures found experimentally with those predicted theoretically as a function of oxygen chemical potential, although there are certainly detailed discrepancies. Of course, these calculations predict the conditions for thermodynamic equilibrium, and achieving true equilibrium in the experiments may be challenging, particularly if there are substantial kinetic barriers to a change of structure or if substantial changes in surface stoichiometry are required. Much of the work on these surfaces has been performed on thin epitaxial films, prepared in situ, and growth is intrinsically a nonequilibrium process, yet it may be easier to a achieve an equilibrium surface with a supply of metal and oxygen atoms to the surface during growth, than by modifying the surface of a bulk crystal by applying changes in the oxygen partial pressure and temperature. In fact the general trend of the experimental structural findings is consistent with the calculated phase diagrams. For Al2O3 a half-metal termination is predicted for essentially all oxygen chemical potentials and is the preferred solution to most experiments. For Cr2O3 and Fe2O3 either halfmetal or chromyl and ferryl terminations are predicted, dependent on the oxygen chemical potential, broadly compatible with the experimental results. For V2O3, the results are more controversial, in that while theory indicates that the reconstructed O3 termination should be stable under conditions similar to those of the surface preparation, and recent experiments support this view, the conventional wisdom until recently was that the surface is vanadyl terminated. In fact

AUTHOR INFORMATION Corresponding Author

*E-mail: d.p.woodruff@warwick.ac.uk. Notes

The author declares no competing financial interest. T

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

Biography

(26) Barth, C.; Reichling, M. Nature 2001, 414, 45. (27) Chame, A.; Lançon, F.; Paliti, P.; Renaud, G.; Vilfan, I.; Viallain, J. Int. J. Mod. Phys. B 1997, 11, 3657. (28) Vilfan, I.; Lançon, F.; Villain, J. Surf. Sci. 1997, 392, 62. (29) Jarvis, E. A. A.; Carter, E. A. J. Phys. Chem. B 2001, 105, 4045. (30) Lauritsen, J. V.; Jensen, M. C. R.; Venkataramani, K.; Hinnemann, B.; Helveg, S.; Clausen, B. S.; Besenbacher, F. Phys. Rev. Lett. 2009, 103, No. 076103. (31) Vilfan, I.; Deutsch, T.; Lançon, F.; Renaud, G. Surf. Sci. 2002, 505, L215. (32) Guenard, P.; Renaud, G.; Barbier, A.; Gautier-Soyer, M. Surf. Rev. Lett. 1998, 5, 321. (33) Guenard, P.; Renaud, G.; Barbier, A.; Gautier-Soyer, M. MRS Proc. 1996, 437, 15. (34) Toofan, J.; Watson, P. R. Surf. Sci. 1998, 401, 162. (35) Walters, C. F.; McCarty, K. F.; Soares, E. A.; Van Hove, M. A. Surf. Sci. 2000, 464, L732. (36) Soares, E. A.; Van Hove, M. A.; Walters, C. F.; McCarty, K. F. Phys. Rev. B 2002, 65, 195405. (37) Ahn, J.; Rabalais, J. W. Surf. Sci. 1997, 388, 121. (38) Suzuki, T.; Hishita, S.; Oyoshi, K.; Souda, R. Surf. Sci. 1999, 437, 289. (39) Puchin, V. E.; Gale, J. D.; Shluger, A. L.; Kotomin, E. A.; Günster, J.; Brause, M.; Kempter, V. Surf. Sci. 1997, 370, 190. (40) Over, H.; Motitz, W.; Ertl, G. Phys. Rev. Lett. 1993, 70, 315. (41) Hertel, T.; Over, H.; Bludau, H.; Gierer, M.; Ertl., G. Phys. Rev. B 1994, 50, 8126. (42) Stampfl, C.; Scheffler, M.; Over, H.; Burchhardt., J.; Nielsen, M.; Adams, D.; Moritz, W. Phys. Rev. B 1994, 49, 4959. (43) Manassidis, I.; De Vita, A.; Gillan, M. J. Surf. Sci. 1993, 285, L517. (44) Godin, T. J.; LaFemina, J. P. Phys. Rev. B 1994, 49, 7691. (45) Di Felice, R.; Northrup, J. E. Phys. Rev. B 1999, 60, R16287. (46) Batyrev, I.; Alavi, A.; Finnis, M. W. Faraday Discuss. 1999, 114, 33. (47) Wang, X.-G.; Chaka, A.; Scheffler, M. Phys. Rev. Lett. 2000, 84, 3650. (48) Wander, A.; Searle, B.; Harrison, N. M. Surf. Sci. 2000, 458, 25. (49) Marmier, A.; Lozovoi, A.; Finnis, M. W. J. Eur. Ceram. Soc. 2003, 23, 2729. (50) Kurita, T.; Uchida, K.; Oshiyama, A. Phys. Rev. B 2010, 82, 155319. (51) Marmier, A.; Finnis, M. W. J. Phys.: Condens. Matter 2002, 14, 7797. (52) Baudin, M.; Hermansson, K. Surf. Sci. 2001, 474, 107. (53) Sun, J.; Stirner, T.; Matthews, A. Surf. Coat. Technol. 2006, 201, 4205. (54) Pisani, C.; Causà, M.; Dovesi, R.; Roetti, C. Prog. Surf. Sci. 1987, 25, 119. (55) Causà, M.; Dovesi, R.; Pisani, C.; Roetti, C. Surf. Sci. 1989, 215, 259. (56) Jaeger, R. M.; Kuhlenbeck, H.; Freund, H.-J.; Wuttig, M.; Hoffmann, W.; Franchy, R.; Ibach, H. Surf. Sci. 1991, 259, 235. (57) Kulawik, M.; Nilius, N.; Freund, H.-J. Phys. Rev. Lett. 2006, 96, 036103. (58) Levin., I.; Brandon, D. J. Am. Ceram. Soc. 1998, 81, 1995. (59) Ceballos, G.; Song, Z.; Pascual, J. I.; Rust, H.-P.; Conrad, H.; Bäumer, M.; Freund, H.-J. Chem. Phys. Lett. 2002, 359, 41. (60) Lee, M. B.; Lee, J. H.; Frederick, B. G.; Richardson, N,V. Surf. Sci. 2000, 448, L207. (61) Stierle, A.; Renner, F.; Streitel, R.; Dosch, H.; Drube, W.; Cowie, B. C. Science 2004, 303, 1652. (62) Kresse, G.; Schmid, M.; Napetschnig, E.; Shishkin, M.; Köhler, L.; Varga, P. Science 2005, 308, 1440. (63) Simon, G. H.; König, T.; Nilius, M.; Rust, H.-P.; Heyde, M.; Freund, H.-J. Phys. Rev. B 2008, 78, 113401. (64) Nishimura, T.; Hoshino, Y.; Okazawa, T.; Kido, Y. Phys. Rev. B. 2008, 77, 073405.

Phil Woodruff completed his first degree at the University of Bristol before moving to the (then new) University of Warwick for his PhD, working on the solid/melt interface. He then switched his interest to the solid/vacuum interface and established a research group working on the structural, electronic and chemical properties of wellcharacterised surfaces, using a wide range of techniques. For some years, his primary focus has been on quantitative surface structure determination, mainly of molecular adsorbates on metal and oxide surfaces. Much of this work has involved the use of established and new synchrotron-radiation based methods. In addition to his role as Professor of Physics at the University of Warwick, he held visiting positions at the Fritz Haber Institute in Berlin during the period 1999−2011.

REFERENCES (1) Thornton, G.; Lindsay, R.; Pang, C. Chem. Rev. 2013, this issue. (2) Diebold, U. Chem. Rev. 2013, this issue. (3) Pendry, J. B. Low Energy Electron Diffraction; Academic Press: New York, 1974. (4) Van Hove, M. A.; Weinberg, W. H.; Chan, C.-M. Low-Energy Electron Diffraction: Experiment, Theory, and Surface Structure; Springer, Berlin, 1986. (5) Heinz, K. Rep. Prog. Phys. 1995, 58, 637. (6) French, T. M.; Somorgai, G. A. J. Phys. Chem. 1970, 74, 2489. (7) Chambers, S. A. Surf. Sci. Rep. 2000, 39, 105. (8) Freysoldt., C.; Rinke., P.; Scheffler, M. Phys. Rev. Lett. 2007, 99, 086101. (9) Feidenhans’l, R. Surf. Sci. Reports 1989, 10, 105. (10) Robinson, I. K.; Tweet, D. J. Rep. Prog. Phys. 1992, 55, 599. (11) van der Veen, J. F. Surf. Sci. Rep. 1985, 5, 199. (12) Niehus, H.; Heiland, W.; Taglauer, E. Surf. Sci. Rep. 1993, 17, 213. (13) Stöhr, J. In X-ray Absorption, Principles, Techniques, Applications of EXAFS, SEXAFS and XANES; Prins, R., Koeningsberger, D. C., Ed.; Wiley: New York, 1988, 443. (14) Woodruff, D. P. Rep. Prog. Phys. 1986, 49, 683. (15) Woodruff, D. P.; Bradshaw, A. M. Rep. Prog. Phys. 1994, 57, 1029. (16) Woodruff, D. P. Surf. Sci. Rep. 2007, 62, 1. (17) Zegenhagen, J. Surf. Sci. Rep. 1993, 18, 199. (18) Woodruff, D. P. Prog. Surf. Sci. 1998, 57, 1. (19) Woodruff, D. P. Rep. Prog. Phys. 2005, 68, 743. (20) Tersoff, J.; Hamman, D. R. Phys. Rev. B 1985, 31, 805. (21) Noguerra, C. J. Phys.: Condens. Matter 2000, 12, R367. (22) Charig, J. M. Appl. Phys. Lett. 1967, 20, 139. (23) Chang, C. C. J. Appl. Phys. 1968, 39, 5570. (24) Gautier, M.; Durand, J. P.; Pham Van, L.; Guittet, M. J. Surf. Sci. 1991, 250, 71. (25) Renaud, G.; Villette, B.; Vilfan, I.; Bourret, A. Phys. Rev. Lett. 1994, 73, 1825. U

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

(101) Surnev, S.; Ramsey, M. G.; Netzer, F. P. Prog. Surf. Sci. 2003, 73, 117. (102) Klein, C.; Kresse, G.; Surnev, S.; Netzer, F. P.; Schmid, M.; Varga, P. Phys. Rev. B 2003, 68, 235416. (103) Schoiswohl, J.; Sock, M.; Surnev, S.; Ramsey, M. G.; Netzer, F. P.; Kresse, G.; Andersen, J. N. Surf. Sci. 2004, 555, 101. (104) Surnev, S.; Kresse, G.; Sock, M.; Ramsey, M. G.; Netzer, F. P. Surf. Sci. 2001, 495, 91. (105) Kolczewski, C.; Hermann, K. Theor. Chem. Acc. 2005, 114, 60. (106) Kolczewski, C.; Hermann, K.; Guimond, S.; Kuhlenbeck, H.; Freund, H.-J. Surf. Sci. 2007, 601, 5394. (107) Abu Haija, M.; Guimond, S.; Romanyshyn, Y.; Uhl, A.; Kuhlenbeck, H.; Todorova, T. K.; Ganduglia-Pirovano, M. V.; Döbler, J.; Sauer, J.; Freund, H.-J. Surf. Sci. 2006, 600, 1497. (108) Bandara, A.; Abu-Haija, M.; Höbel, F.; Kuhlenbeck, H.; Rupprechter, G.; Freund, H.-J. Top. Catal. 2007, 46, 223. (109) Göbke, D.; Romanyshyn, Y.; Guimond, S.; Sturm, J. M.; Kuhlenbeck, H.; Döbler, J.; Reinhardt., U.; Gandiglia-Pirovano, M. V.; Sauer, J.; Freund, H.-J. Angew. Chem., Int. Ed. 2009, 48, 3695. (110) Yang, B.; Nilius, M.; Freund, H.-J. J. Phys. Chem. C. 2011, 115, 3404. (111) Kröger, E. A.; Sayago, D. I.; Allegretti, F.; Knight, M. J.; Polcik, M.; Unterberger, W.; Lerotholi, T. J.; Hogan, K. A.; Lamont, C. L. A.; Woodruff, D. P. Phys. Rev. B 2007, 75, 195413. (112) Kröger, E. A.; Sayago, D. I.; Allegretti, F.; Knight, M. J.; Polcik, M.; Unterberger, W.; Lerotholi, T. J.; Hogan, K. A.; Lamont, C. L. A.; Woodruff, D. P. Surf. Sci. 2007, 601, 3350. (113) Czekaj, I.; Hermann, K.; Witko, M. Surf. Sci. 2003, 545, 85. (114) Kresse, G.; Surnev, S.; Schoiswohl, J.; Netzer, F. P. Surf. Sci. 2004, 555, 118. (115) Todorova, T. K.; Ganduglia-Pirovano, M. V.; Sauer, J. J. Phys. Chem. B 2005, 109, 23523. (116) Window, A. J.; Hentz, A.; Sheppard, D. C.; Parkinson, G. S.; Niehus, H.; Ahlbehrendt, D.; Noakes, T. C. Q.; Bailey, P.; Woodruff, D. P. Phys. Rev. Lett. 2011, 107, 016105. (117) Window, A. J.; Hentz, A.; Sheppard, D. C.; Parkinson, G. S.; Woodruff, D. P.; Unterberger, W.; Noakes, T. C. Q.; Bailey, P.; Ganduglia-Pirovano, M. V.; Sauer, J. Surf. Sci. 2012, 606, 1716. (118) Seifert, J.; Meyer, E.; Winter, H.; Kuhlenbeck, H. Surf. Sci. 2012, 606, L41. (119) Kröger, E. A.; Sayago, D. I.; Allegretti, F.; Knight, M. J.; Polcik, M.; Unterberger, W.; Lerotholi, T. J.; Hogan, K. A.; Lamont, C. L. A.; Cavalleri, M.; Hermann, K.; Woodruff, D. P. Surf. Sci. 2008, 602, 1267. (120) Niehus, H.; Blum, R.-P.; Ahlbehrendt, D. Surf. Rev. Lett. 2003, 10, 353. (121) Surnev, S.; Kresse, G.; Sock, M.; Ramsey, M. G.; Netzer, F. P. Surf. Sci. 2001, 495, 91. (122) Blomberg, S.; Lundgren, E.; Westerström, R.; Erdogan, E.; Martin, N. M.; Mikkelsen, A.; Andersen, J. N.; Mittendorf, F.; Gustafson, J. Surf. Sci. 2012, 606, 1416. (123) Eng, P. J.; Trainor, T. P.; Brown, G. E., Jr.; Waychunas, G. A.; Newville, M.; Sutton, S. R.; Rivers, M. L. Science 2000, 288, 1029. (124) Gan, Y.; Franks, G. V. J. Phys. Chem. B 2005, 109, 12474. (125) Fernández, E. M.; Ehlitis, R. I.; Borstel, G.; Balbás, L. C. Comput. Mater. Sci. 2007, 39, 587. (126) Ranea, V. A.; Schneider, W. F.; Carmichael, I. Surf. Sci. 2008, 602, 268. (127) Ranea, V. A.; Carmichael, I.; Schneider, W. F. J. Phys. Chem. C 2009, 113, 2149. (128) Hass, K. C.; Scheider, W. F.; Cyrioni, A.; Andreoni, W. Science 1998, 282, 265. (129) Trainor, T. P.; Chaka, A. M.; Eng, P. J.; Newville, M.; Waychunas, G. A.; Catalano, J. G.; Brown, G. E., Jr. Surf. Sci. 2004, 573, 204. (130) Abu Haija, M.; Guimond, S.; Uhl, A.; Kuhlenbeck, H.; Freund, H.-J. Surf. Sci. 2006, 600, 1040. (131) Ijima, K.; Koike, Y.; Chun, W.-J.; Saito, Y.; Tanizawa, Y.; Shido, T.; Iwasawa, Y.; Nomura, M.; Asakura, K. Chem. Phys. Lett. 2004, 384, 134.

(65) Prévot, G.; Le Moal, S.; Bernard, R.; Croset, B.; Lazzari, R.; Schmaus, D. Phys. Rev. B. 2012, 85, 205450. (66) Hamm, G.; Barth, C.; Becker, C.; Wandelt, K.; Henry, C. R. Phys. Rev. Lett. 2006, 97, 126106. (67) Zhang, Z.; Li, L.; Yang, J. C. Acta Mater. 2011, 59, 5905. (68) Vlad, A.; Stierle, A.; Kasper, N.; Dosch, H.; Rühle, M. J. Mater. Res. 2006, 21, 3047. (69) Rohr, F.; Bäumer, M.; Freund, H.-J.; Mejias, J. A.; Staemmler, V.; Müller, S.; Hammer, L.; Heinz, K. Surf. Sci. 1997, 372, L291. (70) Rohr, F.; Bäumer, M.; Freund, H.-J.; Mejias, J. A.; Staemmler, V.; Müller, S.; Hammer, L.; Heinz, K. Surf. Sci. 1997, 389, 391. (71) Gloege, Th.; Meyerheim, H. L.; Moritz, W.; Wolf, D. Surf. Sci. 1999, 441, L917. (72) Bender, M.; Ehrlich, D.; Yakovkin, I. N.; Rohr, F.; Bäumer, M.; Kuhlenbeck, H.; Freund, H.-J.; Staemmler, V. J. Phys.: Condens. Matter 1995, 7, 5289. (73) Priyantha, W. A. A.; Wadill, G. D. Surf. Sci. 2005, 578, 149. (74) Redding, S. Honors Thesis, Texas State University, San Marcos, Texas, U.S.A., 2009; https://digital.libarary.txstate.edu/handle/10877/ 3272. (75) Lübbe, M.; Moritz, W. J. Phys.: Condens. Matter 2009, 21, 134010. (76) Dillmann, B.; Rohr, F.; Seiferth, O.; Klivenyi, G.; Bender, M.; Homann, K.; Yakovkin, I. N.; Ehrlich, D.; Bämer, M.; Kuhlenbeck, H.; Freund, H.-J. Faraday Discuss. 1996, 105, 295. (77) Bikondoa, O.; Moritz, W.; Torrelles, X.; Kim, H. J.; Thornton, G.; Lindsay, R. Phys. Rev. B 2010, 81, 205439. (78) Rehbein, C.; Harrison, N. M.; Wander, A. Phys. Rev. B 1996, 54, 14066. (79) Cline, J. A.; Rigos, A. A.; Arias, T. J. Phys. Chem. B 2000, 104, 6195. (80) Wang, X.-G.; Smith, J. R. Phys. Rev. B 2003, 68, 201402. (81) Rohrbach, A.; Hafner, J.; Kresse, G. Phys. Rev. B 2004, 70, 125426. (82) Kurtz, R.; Henrich, V. E. Surf. Sci. 1983, 129, 345. (83) Ladd, R. J.; Henrich, V. E. Surf. Sci. 1988, 193, 81. (84) Eggleston, C. M.; Hochella, M. F. Am. Mineral. 1992, 77, 911. (85) Condon, N. G.; Murray, P. W.; Leibsle, F. M.; Thornton, G.; Lennie, A. R.; Vaughan, D. J. Surf. Sci. 1994, 310, L609. (86) Condon, N. G.; Leibsle, F. M.; Lennie, A. R.; Murray, P. W.; Vaughan, D. J.; Thornton, G. Phys. Rev. Lett. 1995, 75, 1961. (87) Condon, N. G.; Leibsle, F. M.; Lennie, A. R.; Murray, P. W.; Parker, T. M.; Vaughan, D. J.; Thornton, G. Surf. Sci. 1998, 397, 278. (88) Shaikhutdinov, Sh. K.; Weiss, W. Surf. Sci. 1999, 432, L627. (89) Lanier, C. H.; Chiaramonti, A. N.; Marks, L. D.; Poeppelmeier, K. R. Surf. Sci. 2009, 603, 2574. (90) Ketteler, G.; Weiss, W.; Ranke, W. Surf. Rev. Lett. 2001, 8, 661. (91) Bergemeyer, W.; Schweiger, H.; Wimmer, E. Phys. Rev. B 2004, 69, 195409. (92) Wang, X.-G.; Weiss, W.; Shaikhutdinov, Sh. K.; Ritter, M.; Petersen, M.; Wagner, F.; Schlögl, R.; Scheffler, M. Phys. Rev. Lett. 1998, 81, 1038. (93) Chambers, S. A.; Yi, S. I. Surf. Sci. 1999, 439, L785. (94) Thevuthasan, S.; Kim, Y. J.; Yi, S. I.; Chambers, S. A.; Morais, J.; Denecke, R.; Fadley, C. S.; Liu, P.; Kendelewicz, T.; Brown, G. E., Jr. Surf. Sci. 1999, 425, 276. (95) Lemire, C.; Bertarione, S.; Zecchina, A.; Scarano, D.; Chaka, A.; Shaikhutdinov, S.; Freund, H.-J. Phys. Rev. Lett. 2005, 94, 166101. (96) Surnev, S.; Vitali, L.; Ramsey, M. G.; Netzer, F. P.; Kresse, G.; Hafner, J. Phys. Rev. B 2000, 61, 13945. (97) Schoiswohl, J.; Sock, M.; Eck, S.; Surnev, S.; Ramsey, M. G.; Netzer, F. P. Phys. Rev. B 2004, 69, 155403. (98) Dupuis, A.-C.; Abu Haiji, M.; Richter, B.; Kuhlenbeck, H.; Freund, H.-J. Surf. Sci. 2003, 539, 99. (99) Niehus, H.; Blum, R.-P.; Ahlbehrendt, D. Phys. Status Solidi A 2001, 187, 151. (100) Biener, J.; Bäumer, M.; Madix, R. J.; Liu, P.; Nelson, E. J.; Kendelewicz, T.; Brown, G. E., Jr. Surf. Sci. 1999, 441, 1. V

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

(132) Xu, C.; Dillman, B.; Kuhlenbeck, H.; Freund, H.-J. Phys. Rev. Lett. 1991, 67, 3551. (133) Unterberger, W.; Kröger, E. A.; Allegretti, F.; Lerotholi, T. J.; Woodruff, D. P. Unpublished results (134) Legg, K. O.; Prutton, M.; Kinniburgh, C. J. Phys. C: Solid State Phys. 1974, 7, 4236. (135) Kinniburgh, C. G. J. Phys. C: Solid State Phys. 1975, 8, 2382. (136) Kinniburgh, C. G. J. Phys. C: Solid State Phys. 1976, 9, 2695. (137) Netzer, F. P.; Prutton, M. J. Phys. C: Solid State Phys. 1975, 8, 2401. (138) Kinniburgh, C. G.; Walker, J. A. Surf. Sci. 1977, 63, 274. (139) Prutton, M.; Ramsey, J. A.; Walker, J. A.; Welton-Cook, M. R. J. Phys. C: Solid State Phys. 1979, 12, 5271. (140) Prutton, M.; Walker, J. A.; Welton-Cook, M. R.; Felton, R. C.; Ramsey, J. A. Surf. Sci. 1979, 89, 95. (141) Palmberg, P. W.; DeWames, R. E.; Vredevoe, L. A.; Wolfram, T. J. Appl. Phys. 1969, 40, 1158. (142) Urano, T.; Kanaji, T.; Kaburagi, M. Surf. Sci. 1983, 134, 109. (143) Welton-Cook, M. R.; Berndt, W. J. Phys. C: Solid State Phys. 1982, 15, 5691. (144) Blanchard, D. L.; Lessor, D. L.; LaFemina, J. P.; Baer, D. R.; Ford, W. K.; Guo., T. J. Vac. Sci. Technol., A 1991, 9, 1814. (145) Ferry, D.; Suzanne, J.; Panella, V.; Barieri, A.; Van Hove, M. A.; Bibérian, J.-P. J. Vac. Sci. Technol., A 1998, 16, 2261. (146) Gotoh, T.; Murakami, S.; Kinosita, K.; Murata, Y. J. Phys. Soc. Jpn. 1981, 50, 2063. (147) Nakamatsu, H.; Sudo, A.; Kawai, S. Surf. Sci. 1988, 194, 265. (148) Zhou, J. B.; Lu, H. C.; Gustafsson, T.; Häberle, P. Surf. Sci. 1994, 302, 350. (149) Robach, O.; Renaud, G.; Barbier, A. Surf. Sci. 1998, 401, 227. (150) Schüller, A.; Blauth, D.; Seifert, J.; Busch, M.; Winter, H.; Gärtner, K.; Włodarczyk, R.; Sauer, J.; SDierka, M. Surf. Sci. 2012, 606, 161. (151) Welton-Cook, M. R.; Prutton, M. Surf. Sci. 1978, 74, 276. (152) Verwey, E. J. W. Recl. Trav. Chim. Pays-Bas 1946, 65, 521. (153) Goniakowski, J.; Noguera, C. Surf. Sci. 1995, 323, 129. (154) Xu, Y.; Li, J.; Zhang, Y.; Chen, W. Surf. Sci. 2003, 525, 13. (155) Broqvist, P.; Grönbeck, H.; Panas, I. Surf. Sci. 2004, 554, 262. (156) Quintanar, C.; Caballero, R.; Castaño, V. M. Int. J. Quantum Chem. 2005, 102, 820. (157) Skorodumova, N. V.; Hermansson, K.; Johansson, B. Phys. Rev. B 2005, 72, 125414. (158) Alfonso, D. R.; Snyder, J. A.; Jaffe, J. E.; Hess, A. C.; Gutowski, M. Phys. Rev. B 2000, 62, 8318. (159) Walker, J. A.; Kinniburgh, C. G.; Matthew, J. A. D. Surf. Sci. 1977, 78, 221. (160) Welton-Cook, M. R.; Prutton, M. J. Phys. C: Solid State Phys. 1980, 13, 3993. (161) Momida, H.; Oguchi, T. J. Phys. Soc. Jpn. 2003, 72, 588. (162) Okazawa, T.; Kido, Y. Surf. Sci. 2004, 556, 101. (163) Soares, E. A.; Paniago, R.; de Carvalho, V. E.; Lopes, E. L.; Abreu, G. J. P.; Pfannes, H.-D. Phys. Rev. B 2005, 73, 035419. (164) Felton, R. C.; Prutton, M.; Tear, S. P.; Welton-Cook, M. R. Surf. Sci. 1979, 88, 474. (165) Schindler, K.-M.; Wang, J.; Chassé, A.; Neddermeyer, H.; Widdra, W. Surf. Sci. 2009, 603, 2658. (166) Lopes, E. L.; Abreu, G. J. P.; Paniago, R.; Soares, E. A.; de Carvalho, V. E.; Pfannes, H.-D. Surf. Sci. 2007, 601, 1239. (167) Heidberg, J.; Redlich, B.; Wetter, D. Ber. Bunsen. Gesell. Phys. Chem. 1995, 99, 1333. (168) Ferry, D.; Glebov, A.; Senz, V.; Suzanne, J.; Toennies, J. P.; Weiss, H. J. Chem. Phys. 1996, 105, 1697. (169) Ferry, D.; Glebov, A.; Senz, V.; Suzanne, J.; Toennies, J. P.; Weiss, H. Surf. Sci. 1997, 377, 634. (170) Xu, C.; Goodman, D. W. Chem. Phys. Lett. 1997, 265, 341. (171) Ferry, D.; Picaud, S.; Hoang, P. N. M.; Girardet, C.; Giordano, L.; Demirdjian, B.; Suzanne, J. Surf. Sci. 1998, 409, 101. (172) Giordano, L.; Goniakowski, J.; Suzanne, J. Phys. Rev. Lett. 1998, 81, 1271.

(173) Delle Site, L.; Alavi, A.; Lynden-Bell, R. M. J. Chem. Phys. 2000, 113, 3344. (174) Lynden-Bell, R. M.; Dell Site, L.; Alavi, A. Surf. Sci. 2002, 496, L1. (175) Włodarczyk, R.; Sierka, M.; Kwapień, K.; Sauer., J.; Carrasco, E.; Aumer, A.; Gomes, J. E.; Sterrer, M.; Freund, H.-J. J. Phys. Chem. C 2011, 115, 6764. (176) Liu, P.; Kendelewicz, T.; Nelson, E. J.; Brown, G. E., Jr. Surf. Sci. 1998, 415, 156. (177) Audibert, P.; Sidoumou, M.; Suzanne, J. Surf. Sci. 1992, 273, L467. (178) Gerlach, R.; Glebov, A.; Lange, G.; Toennies, J. P.; Weiss, H. Surf. Sci. 1995, 331, 1490. (179) Heidberg, J.; Kandel, M.; Meine, D.; Wildt, U. Surf. Sci. 1995, 331, 1467. (180) Pacchioni, G. Surf. Rev. Lett. 2000, 7, 277. (181) Qin, C. Chem. Phys. Lett. 2008, 460, 457. (182) Yang, B.; Lin, X.; Gao, H.-J.; Nilius, N.; Freund, H.-J. J. Phys. Chem. C 2010, 114, 8997. (183) Suzanne, J.; Panella, V.; Ferry, D.; Sidoumou, M. Surf. Sci. 1993, 293, L912. (184) Heidberg, J.; Meine, D.; Redlich, B. J. Electron Spectrosc. Relat. Phenom. 1993, 64, 599. (185) Jensen, M. B.; Pettersson, L. G. M.; Swang, O.; Olsbye, U. J. Phys. Chem. B 2005, 109, 16774. (186) Daub, C. D.; Patey, G. N.; Jack, D. B.; Sallabi, A. K. J. Chem. Phys. 2006, 124, 114706. (187) Carrier, X.; Doyle, C. S.; Kendelewicz, T.; Brown, G. E., Jr. Surf. Rev. Lett. 1999, 6, 1237. (188) Ferry, D.; Hoang, P. N. M.; Suzanne, J.; Birérian, J.-P.; Van Hove, M. A. Phys. Rev. Lett. 1997, 78, 4237. (189) Coulomb, J. P.; Lahrer, Y.; Trabelsi, M.; Mirebeau, I. Mol. Phys. 1994, 81, 1259. (190) Degenhardt, D.; Lauter, H. J.; Haensel, R. Jpn. J. Appl. Phys. Suppl. 1987, 26, 341. (191) Skofronick, J. G.; Toennies, J. P.; Traeger, F.; Weiss, H. Phys. Rev. B 2003, 67, 035413. (192) Renaud, G.; Lazzari, R.; Revenant, C.; Barbier, A.; Noblet, M.; Ulrich, O.; Leroy, F.; Jupille, J.; Borensztein, Y.; Henry, C. R.; Deville, J.-P.; Scheurer, F.; Mane-Mane, J.; Fruchart, O. Science 2003, 300, 1416. (193) Zhu, J.; Farmer, J. A.; Ruzycki, N.; Xu, L.; Campbell, C. T.; Henkelman, G. J. Am. Chem. Soc. 2008, 130, 2314. (194) Souda, R.; Aizawa, T.; Ishizawa, Y.; Oshima, C. J. Vac. Sci. Technol. A 1990, 8, 3218. (195) Souda, R.; Hwang, Y.; Aizawa, T.; Hayami, W.; Oyoshi, K.; Hishita, S. Surf. Sci. 1997, 387, 136. (196) Masri, P.; Tasker, P. W.; Hoare, J. P.; Harding, J. H. Surf. Sci. 1986, 173, 439. (197) Yan, Y.; Chisholm, M. F.; Pennycook, S. J.; Pantelides, S. T. Surf. Sci. 1999, 442, 251. (198) Urano, T.; Kanaji, T. J. Phys. Soc. Jpn. 1988, 57, 3403. (199) Li, C.; Freeman, A. J. Phys. Rev. B 1991, 43, 780. (200) Lairson, B. M.; Payne, A. P.; Brennan, S.; Rensing, N. M.; Daniels, B. J.; Clemens, B. M. J. Appl. Phys. 1995, 78, 4449. (201) Trampert, A.; Ernst, F.; Flynn, C. P.; Fischmeister, H. F.; Rühle, M. Acta Metall. Mater. 1992, 40, S227. (202) Li., C.; Freeman, A. J.; Fu, C. L. Phys. Rev. B 1993, 48, 8317. (203) Schönberger, U.; Andersen, O. K.; Methfessel, M. Acta. Matall. Mater. 1992, 40, S1. (204) Smith, J. R.; Hong, T.; Srolovitz, D. J. Phys. Rev. Lett. 1994, 72, 4021. (205) Flank, A. M.; Delaunay, R.; Lagarde, P.; Pompa, M.; Jupille, J. Phys. Rev. B 1996, 53, R1737. (206) Renaud, G.; Guénard, P.; Barbier, A. Phys. Rev. B 1998, 58, 7310. (207) Robach, O.; Renaud, G.; Barbier, A. Phys. Rev. B 1999, 60, 5858. W

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX

Chemical Reviews

Review

(208) Revenant, C.; Renaud, G.; Lazzari, R.; Jupille, J. Phys. Rev. B 2009, 79, 235424. (209) Renaud, G.; Barbier, A.; Robach., O. Phys. Rev. B. 1999, 60, 5872. (210) Renaud, G.; Barbier, A. Surf. Sci. 1999, 433, 142. (211) Barbier, A.; Renaud, G.; Robach, O. J. Appl. Phys. 1998, 84, 4259. (212) Pacchioni, G.; Rösch, N. J. Chem. Phys. 1996, 104, 7329. (213) Olander, J.; Lazzari, R.; Jupille, J.; Mangili, B.; Goniakowski, J.; Renaud, G. Phys. Rev. B 2007, 76, No. 075409. (214) Cappus, D.; Klinkmann, J.; Kuhlenbeck, H.; Freund, H.-J. Surf. Sci. 1995, 325, L421. (215) Kuhlenbeck, H.; Odörfer, G.; Jaeger, R.; Illing, G.; Menges, M.; Mull, Th.; Freund, H.-J.; Pölchen, M.; Staemmler, V.; Witzel, S.; Scharfschwerdt, C.; Wennemann, K.; Liedtke, T.; Neumann, M. Phys. Rev. B 1991, 43, 1969. (216) Bäumer, M.; Cappus, D.; Illing, G.; Kuhlenbeck, H.; Freund, H.-J. J. Vac. Sci. Technol. A 1992, 10, 2407. (217) Kittel, M.; Hoeft, J. T.; Bao, S.; Polcik, M.; Toomes, R. L.; Kang, J.-H.; Woodruff, D. P.; Pascal, M.; Lamont, C. L. A. Surf. Sci. 2002, 499, 1. (218) Hoeft, J. T.; Kittel, M.; Polcik, M.; Bao, S.; Toomes, R. L.; Kang, J.-H.; Woodruff, D. P.; Pascal, M.; Lamont, C. L. A. Phys. Rev. Lett. 2001, 87, No. 086101. (219) Polcik, M.; Lindsay, R.; Baumgärtel, P.; Terborg, R.; Schaff, O.; Kulkarni, S.; Bradshaw, A. M.; Toomes, R. L.; Woodruff, D. P. Faraday Discuss. 1999, 114, 141. (220) Lindsay, R.; Baumgärtel, P.; Terborg, R.; Schaff, O.; Bradshaw, A. M.; Woodruff, D. P. Surf. Sci. 1999, 425, L401. (221) Pacchioni, G.; Cogliandro, G.; Bagus, P. S. Surf. Sci. 1991, 255, 344. (222) Pöhlchen, M.; Staemmler, V. J. Chem. Phys. 1992, 97, 2583. (223) Pacchioni, G.; Di Valentin, C.; Dominguez-Ariza, D.; Illas, F.; Bredow, T.; Klüner, T.; Staemmler, V. J. Phys.: Condens. Matter 2004, 16, No. S2497. (224) Rohrbach, A.; Hafner, J.; Kresse, G. Phys. Rev. B. 2004, 69, 075413. (225) Rohrbach, A.; Hafner, J. Phys. Rev. B 2005, 71, 045405. (226) Valero, R.; Gomes, J. R. B.; Truhlar, D. G.; Illas, F. J. Chem. Phys. 2010, 132, 104701. (227) Steinbrunn, A.; Dumas, P.; Colson, J. C. Surf. Sci. 1978, 74, 201. (228) Woodhead, A. P.; Harte, S. P.; Haycock, S. A.; Muryn, C. A.; Wincott, P. L.; Dhanak, V. R.; Thornton, G. Surf. Sci. 1999, 420, L138. (229) Noguera, C. J.Phys.: Condens. Matter 2000, 12, R367. (230) Wolf, D. Phys. Rev. Lett. 1992, 68, 3315. (231) Zhang, W.-B.; Tang, B.-Y. J. Phys. Chem. C 2008, 112, 3327. (232) Zhang, W.-B.; Tang, B.-Y J. Chem. Phys. 2008, 128, 124703. (233) Finocchi, F.; Barbier, A.; Jupille, J.; Noguera, C. Phys. Rev. Lett. 2004, 92, 136101. (234) Poon, H. C.; Hu, X. F.; Chamberlain, S. E.; Saldin, D. K.; Hirschmugl, C. J. Surf. Sci. 2006, 600, 2505. (235) Ciston, J.; Subramanian, A.; Marks, L. D. Phys. Rev. B 2009, 79, 085421. (236) Ciston, J.; Subramanian, A.; Kienzle, D. M.; Marks, L. D. Surf. Sci. 2010, 604, 155. (237) Ebensperger, C.; Meyer, B. Phys. Status Solidi B 2011, 248, 2229. (238) Henrich, V. E. Surf. Sci. 1976, 57, 385. (239) Plass, R.; Feller, J.; Gajdardziska-Josifovska, M. Surf. Sci. 1998, 414, 26. (240) Plass, R.; Egan, K.; Collazo-Davila, C.; Grozea, D.; Landree, E.; Marks, L. D.; Gajdardziska-Josifovska. Phys. Rev. Lett. 1998, 81, 4891. (241) Watson, G. W.; Kelsey, E. T.; de Leeuw, N. H.; Harris, D. J.; Parker, S. C. J. Chem. Soc. Faraday Trans. 1996, 92, 433. (242) Pojani, A.; Finocchi, F.; Goniakowski, J.; Noguera, C. Surf. Sci. 1997, 387, 354. (243) Refson, K.; Wogelius, R. A.; Fraser, D. G.; Payne, M. C.; Lee, M. H.; Milman, V. Phys. Rev. B 1995, 52, 10823.

(244) Wander, A.; Bush, I. J.; Harrison, N. M. Phys. Rev. B 2003, 68, 233405. (245) Lazarov, V. K.; Plass, R.; Poon, H.-C.; Saldin, D. K.; Weinert, M.; Chambers, S. A.; Gajdardziska-Josifovska, M. Phys. Rev. B 2005, 71, 115434. (246) Rohr, F.; Wirth, K.; Libuda, J.; Cappus, D.; Bäumer, M.; Frend, H.-J. Surf. Sci. 1994, 315, L977. (247) Kitakatsu, N.; Maurice, V.; Hinnen, C.; Marcus, P. Surf. Sci. 1998, 407, 36. (248) Barbier, A.; Renaud, G. Surf. Sci. 1997, 392, L15. (249) Barbier, A.; Mocuta, C.; Kuhlenbeck, H.; Peters, K. F.; Richter, B.; Renaud, G. Phys. Rev. Lett. 2000, 84, 2897. (250) Erdman, N.; Warschkow, O.; Ellis, D. E.; Marks, L. D. Surf. Sci. 2000, 470, 1. (251) Mocuta, C.; Barbier, A.; Renaud, G. Appl. Surf. Sci. 2000, 162, 56. (252) Schönnenbeck, M.; Cappus, D.; Klinkmann, J.; Freund, H.-J.; Petterson, L. G. M.; Bagus, P. S. Surf. Sci. 1996, 347, 337. (253) Gubo, M.; Ebensperger, C.; Meyer, W.; Hammer, L.; Heinz, K. J. Phys.: Condens. Matter 2009, 21, 474211. (254) Ebensperger, C.; Gubo, M.; Meyer, W.; Hammer, L.; Heinz, K. Phys. Rev. B 2010, 81, 235405. (255) Meyer, W.; Hock, D.; Biedermann, K.; Gubo, M.; Müller, S.; Hammer, L.; Heinz, K. Phys. Rev. Lett. 2008, 101, 016103. (256) Meyer, W.; Biedermann, K.; Gubo, M.; Hammer, L.; Heinz, K. Phys. Rev. B 2009, 79, 121403. (257) Biedermann, K.; Gubo, M.; Hammer, L.; Heinz, K. J. Phys.: Condens. Matter 2009, 21, 185003. (258) Corneille, J. S.; He, J.-W.; Goodman, D. W. Surf. Sci. 1995, 338, 211. (259) Maeda, T.; Kobayashi., Y.; Kishi, K. Surf. Sci. 1999, 436, 249. (260) Cappus, D.; Hassel, M.; Neuhaus, E.; Heber, M.; Rohr, F.; Freund, H.-J. Surf. Sci. 1995, 337, 268. (261) Koike, K.; Furakawa, T. Phys. Rev. Lett. 1996, 77, 3921. (262) Shaikhutdinov, Sh.; Ritter, M.; Weiss, W. Phys. Rev. B 2000, 62, 7535. (263) Galloway, H. C.; Benítez, J. J.; Salmeron, M. Surf. Sci. 1993, 298, 127. (264) Ritter, M.; Ranke, W.; Weiss, W. Phys. Rev. B 1994, 60, 1527. (265) Fadley, C. S.; Van Hove, M. A.; Hussain, Z.; Kaduwela, A. P. J. Electron Spectrosc. Relat. Phenom. 1995, 75, 273. (266) Ritter, M.; Ranke, W.; Weiss, W. Phys. Rev. B 1998, 57, 7240. (267) Galloway, H. C.; Sautet, P.; Salmeron, M. Phys. Rev. B 1996, 54, 11145. (268) Giordano, L.; Pacchioni, G.; Goniakowski, J.; Nilius, N.; Rienks, D. L.; Freund, H.-J. Phys. Rev. B 2007, 76, 075416. (269) Zhang, W.; Li, Z.; Luo, Y.; Yang, J. J. Phys. Chem. C 2009, 113, 8302. (270) Giordano, L.; Lewandowski, M.; Groot, I. M. N.; Sun, Y.-N.; Goniakowski, J.; Noguera, C.; Shaikhutdinov, S.; Pacchioni, G.; Freund, H.-J. J. Phys. Chem. C 2010, 114, 21504. (271) Gurgul, J.; Młyńczak, E.; Spiridis, N.; Korecki, J. Surf. Sci. 2012, 606, 711. (272) Nilius, N.; Rienks, E. D. L.; Rust, H.-P.; Freund, H.-J. Phys. Rev. Lett. 2005, 95, 066101. (273) Goniakowski, J.; Noguera, C.; Giordano, L.; Pacchioni, G. Phys. Rev. B 2009, 80, 125403.

X

dx.doi.org/10.1021/cr3002998 | Chem. Rev. XXXX, XXX, XXX−XXX