Oxide Surfaces (The Chemical Physics of Solid Surfaces Vol. 9)

114

Table 3 Computed properties of F centers at the MgO surface property center model method EF/'"

eV

Fsc

F4c F3C

IP/^'eV

Fsc F5/ F3c F3c^

Te/'^'eV

Fsc F5c^ F4c F4c F3C

F3c^

aisoC^Mg)/"' G

K F4C

F3c^

periodic supercell embedded cluster embedded cluster periodic supercell embedded cluster periodic supercell embedded cluster periodic supercell embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster embedded cluster

LDA DFT MP2 DFT HF LDA HF LDA DFT HF DFT DFT DFT DFT CI CI CI CI CI CI HF HF HF

value 9.8 [15] 9.1 [50] 9.5 [55] 9.0 [140] 8.1 [139] 8.9 [15] 6.9 [139] 8.1 [15] 7.3 [50] 5.6 [139] 3.4 [50] 5.6 [50] 3.4 [50] 6.6 [50] 2.9 [29] 3.1 [29] 2.5 [29] 2.2 [29] 2.1 [29] 2.1 [29] 6 [139] 12 [139] 12 [139]

(a) Formation energy (removal of a neutral 0~^ atom). (b) lonizaton potential (Fpc —> Fnc"^ + e~). (c) Vertical singlet-singlet or doublet-doublet transition energy; values scaled by 15% [29] (d) Isotropic hyperfine coupling constant. Only the highest aiso values are reported for each site. Given the anisotropy of the defect, other Mg ions with smaller aiso are present.

Very recently, based on a combined use of chemical evidence, IR and EPR spectroscopies, as well as of ab initio calculations it has been suggested that oxygen vacancies form preferentially at low coordinated sites (steps, edges, comers) [133]; it has also been observed that the nature of the point defects at the surface changes dramatically depending on the pretreatment conditions [135]. Several theoretical studies have been dealing with the hyperfine coupling constants of F"^ centers in MgO. One of the earliest studies on this subject was reported by Sharma and Stoneham [136], who performed HF calculations on simple models of the surface defect and obtained values in reasonable agreement with the Tench model [122,123]. The problem has been reconsidered more recently at various levels of theory [125,39,55,137]. In general, excellent agreement is obtained for the bulk where a firm assignment is possible. On the

115

surface, aiso values of -10 G have been measured in the case of Fs"^(H) centers only while a clear experimental measure of the same value for pure F^^ centers is still lacking [125]. HF cluster calculations on the Tench model of a Fsc^ center give aiso values of about 5-6 G, Table 3, very close to those computed for the bulk [39]. The explanation for the unexpected high value recorded for Fs'^(H) centers, =10 G, has been given in terms of polarization of the electron density by the OH group [125]. This effect has been confirmed later by IR studies [138] Indeed, calculations performed at the HF level on a Fsc'^CH) center, with an OH group close to the vacancy, give aiso values of 13 G, in close agreement with the experiment [125]. On the other hand, low-coordinated paramagnetic F centers, F4c'^ or Fsc"^, do also exhibit aiso values of 12 G [139], Table 3, similar to the measured ones, consistent with the idea that these centers can form preferentially on these sites. 3.4.3. Stability and formation energies Clearly, as we mentioned above a complete characterization of Fs centers is still missing because of the high heterogeneity of the MgO surface and the abundance of possible sites. One important question is of course that of the thermodynamic stability of these defects. There is no simple way to measure the energy required to create an oxygen vacancy in a solid. Even more difficult is to establish the differences in formation energies of defects at various sites, e.g. bulk and surface. However, formation energies of anion vacancies can be computed quite accurately from first principle studies, in particular for the neutral defects. In fact, the determination of the vacancy formation for charged defects like the F^ and F^"^ centers is much more delicate for the reasons mentioned above (see § 2). Recently, two ab initio studies have addressed in a systematic way the question of formation energy, hence of the relative stability, of F centers at various sites of the MgO surface. Plane waves density functional calculations in Local Density Approximation (LDA) [15] estimated that the energy required to remove a neutral oxygen atom from various sites of MgO is 10.5 eV for a bulk six-coordinated oxygen, 9.8 eV for a surface five-coordinated oxygen, and 8.9 and 8.1 eV, respectively, for low-coordinated ions at a step (four-coordinated) or at a comer (three-coordinated) sites, Table 3. For the fivecoordinated oxygen, other accurate estimates reported recently are 9.35 eV [55], 9.07 eV [50] and 9.02 eV [140] using methods which go from embedded clusters to periodic supercell. The same trend has been found with HF cluster calculations [139] where it was also shown that the order of stability of the F centers is F > F^ > F^^. Neutral F centers are more stable because two electrons interact with the Madelung field of the crystal. F^"^ centers, on the other hand, are much less stable and are very electron deficient. The formation energies computed at the HF level for neutral F centers at bulk, surface, step and comer sites, 9.7, 8.1, 6.9, and 5.6 eV, respectively [139], are lower than those obtained

116

using the DFT approach [15] because of the absence of correlation effects. The trend in stability is directly connected to the energy required to remove an oxygen atom from the material. Thus, the formation energy of neutral vacancies follows the trend bulk (or sub-surface) > surface > step > comer; it follows that low-coordinated F centers are easier to form, hence thermodynamically more stable, than the bulk F centers. This is true also for charged vacancies but, in particular for the F^"*" centers, the difference in stability between low- and highcoordinated sites is not as high as for neutral F centers [139]. 3.4.4. Energy levels and ionization potentials An important question when one addresses the stabiUty of charged defects in insulators is that of the relative position of the neutral and charged states of a given center with respect to the vacuum level. This problem has been considered recently for the case of the MgO(lOO) surface [50]. Using cluster models embedded in polarizable ions and PC's, the ionization energies and electron affinities of oxygen vacancies located at the terraces and at the comers of the MgO surface have been determined. Fig. 7. A neutral Fsc center gives rise to an impurity level in the gap which is about 3 eV above the top of the valence band; a neutral Fsc center at a comer site lies more or less at the same energy. F"^ centers give rise to states closer to the valence band; for a terrace F"^5c the level is about 1 eV above the O 2p states, while for a F'^'sc defect the impurity level is nearly at the top of the valence band. The vertical ionization energy of a neutral Fsc defect is 3.4 eV; that of a F"^5c center 5.6 eV. Surface E, eV IP ^ I

~T

Corner

Exciton A (calc.)

IP

vacuum - level

(experimT

Exciton (experimicalc.)

-2-j -3

I

F centre

F centre

-4 -5 -6

F+ centre

Corner

Fig. 7. Energy levels of various defects at the MgO surface from embedded cluster DFT calculations. Reproduced from ref. [50]. Copyright 2000 Elsevier.

117

It is interesting that similar IP's are found for F centers located at a comer site: the IP's of Fsc and F^3c centers are 3.4 and 6.6 eV, respectively [50,139], Table 3. Thus, not only electrons trapped at low-coordinated sites may exist, but they are even more strongly bound that at the corresponding terrace sites. This result is not obvious. In fact, one would expect the energy to ionize an F center to be proportional to the MP at that site. Assuming no geometrical relaxation, the MP is smaller at low-coordinated sites and increases for more coordinated sites, opposite to what found for the vacancy ionization energy. This behavior is due to the Pauli repulsion of the trapped electrons with the surrounding ions of the vacancy [139]. This is larger for high-coordinated F centers and decreases for low-coordinated defects where the trapped electrons can extend toward the vacuum region thus reducing both the Coulomb and the exchange repulsion. The Pauli repulsion destabilizes the F center impurity level and decreases the ionization energy for the high-coordinated sites. It follows that electrons associated to low-coordinated oxygen vacancies are more strongly bound than in the corresponding high-coordinated vacancy sites because of their reduced repulsion with the surrounding [139]. We will see below that this is also the physical origin for the different excitation energies of electrons trapped at F centers located at high- and low-coordinated sites. 3.4.5. Barriers to vacancies dijfusion The barrier for oxygen migration at MgO surfaces is another topic of interest for the understanding of the defect distribution. The F^"^ centers are expected to be involved in ionic migration; neutral or singly charged vacancies in fact should give rise to higher barriers because of the repulsion of the migrating O ions with the trapped electrons. For the migration of a sub-surface F^"^ vacancy to the surface a barrier of 2.9 eV has been computed with HF cluster models [139], slightly higher than experimental measurements or band structure calculations for the migration in the bulk [141-143]. Much lower barriers have been computed for vacancy migration from high-coordinated to low-coordinated surface sites, 1.6 eV for the migration of an O ion from a terrace to another terrace site, and 0.7 eV from a step site to a terrace vacancy [139]. The barrier of 1.6 eV for the O migration on the surface of MgO is close to that measured at the grain boundary of NiO, 1.8 eV [144]. On the surface, the diffusing oxygen atom from a filled to a vacant site on the (100) terrace follows a curved trajectory and at the midpoint (transition state) is 0.6 A above the top layer [139]. The same path has been found for the diffusion of a neutral oxygen vacancy on the MgO surface from band structure DFT-LDA calculations [20]. In this case a barrier of 2.6 eV has been estimated for the diffusion process, a value which drops to about 2 eV when gradient corrections are taken into account [20]. Since in this case one is dealing with a filled vacancy, it is not surprising that the barrier is higher than for migration involving Fs^"*" vacancies.

118

All these results suggest that at high temperatures the O ions can migrate from low-coordinated sites to high-coordinated vacancies leaving an higher concentration of vacancies at the surface, step or comer sites [139]. 3.4.6. Optical spectra Both thermodynamic arguments as well as spectroscopic evidences suggest that oxygen vacancies at the MgO surface localize preferentially at lowcoordinated sites and that F centers at the terrace sites may not represent the dominant species [133]. Another possible indication of the location of F centers in MgO comes from the analysis of optical spectra. Bulk F and F"*" centers give rise to an intense absorption band around 5 eV with two components, one due to F"^ centers at 4.96 eV and one to neutral F centers at 5.03 eV [145]. Much less is known on the optical properties of surface F centers. Optical spectra of MgO surface F centers have been measured on polycrystalline materials [133,146], single crystals [126] and ultrathin MgO films [127,128,147]. Optical measurements on fine powder samples of MgO using diffuse reflectance technique showed a feature at 2.05 eV attributed to Fsc"^ centers [146]. A peak at 2.3 eV has been observed by Henrich et al. [126] on MgO single crystals using electron energy loss spectroscopy, EELS.

Fig. 8. Electron energy loss spectra of 15 ML thick MgO layers, (a) as grown; (b) after Ar"^ sputtering; (c) after additional deposition of 4 ML of Mg; (d) after deposition of 4 ML of Mg and post oxidation with O2 and consequent annealing. Reproduced from ref. [128]. Copyright 1999 Elsevier.

119

The feature was tentatively assigned to surface F centers although according to Underhill and Gallon [148] the loss is due to a surface V center. High resolution EELS studies of the defects generated in an ultrathin film of MgO grown on Mo(lOO) show three features at 1.15, 3.58 and 5.33 eV [127]. The 5.33 eV band was assigned to bulk F centers, the band at 1.15 eV to surface F centers, and the band at 3.58 eV to F aggregates [127]. By reactive deposition of Mg in oxygen atmosphere Pfntir et al. [128,147] have studied the optical spectra of MgO films with large oxygen deficiency. They observed characteristic loss peaks at 2.1 and 3.3 eV which can be attributed to color centers due to surface oxygen vacancies, Fig. 8 [128]. It is interesting to note that optical absortion bands in the same regions, 380 and 520 nm, have been observed in the UV-vis spectra of polycrystalline MgO, Fig. 9 [133]. The complexity of the spectra and the difficult interpretation due to the low sensitivity to surface defects, makes a theoretical support highly desirable in order to provide a firm attribution of the observed bands. The calculation of electronic excitations in solids is very challenging, and only recently accurate configuration interaction (CI) calculations have been reported on this problem [28,29]. To this end, finite clusters embedded in ECP and point charges have been used. A calibration of the accuracy of the results, possible for bulk F and F^ transitions where assignments are unambiguous, have shown that CI calculations tend to overestimate the optical transitions of F centers in MgO bulk by about 15% because of limitations in the size of the basis set [28]. 1001

R%

800 1000 1200 1400 1600 1800 2000 wavelength (nm)

Fig. 9. UV-vis spectra of polycrystalline MgO. (a) As prepared; (b) UV-irradiated for 1 h at RT; (c) UV-irradiated for 1 h at RT in presence of D2; (d) UV-irradiated for 1 h at 77 K in presence of D2. Reproduced from ref. [133]. Copyright 1999 Elsevier.

Scaling the computed excitation energies for surface F centers in various sites by 15% the transitions energies and the corresponding assignments are [29]: bulk Fee and F6c"^ ~5 eV, terrace Fsc and Fsc^ ~3 eV, step F4C and F4c^ -2.2-2.5 eV,

120

comer Fsc and F3/ -2.1-2.2 eV, Table 3. No evidence of transitions below 2 eV due to F centers has been found. Thus, the surface transitions require photon energies which are ~2-3 eV smaller than for the bulk [29]. The reason is that on the surface the excited state, more diffuse than the ground state, can extend toward the vacuum with less Pauli repulsion with the Mg^"^ ions than in the bulk, where the excited state is confined within the octahedral shell of Mg cations. The computed optical transitions imply that in the EELS experiment [128] not only five-coordinated F centers but also low-coordinated oxygen vacancies are present, in agreement with the results of other spectroscopic investigations [133]. In this discussion the possible role of F centers aggregates has not been considered, despite the fact that it has been suggested time ago that they are responsible for some of the optical bands observed in bulk MgO, a possibility which has been mentioned also for the surface transitions [127]. However, recent first principle band structure calculations have clearly shown that the interaction between neutral vacancies is repulsive [20]. Thus, the mechanisms which may lead to the formation of aggregates of oxygen vacancies are not clear. 3.4.7. Chemical reactivity It is only recently that the reactivity of F centers towards chemisorbed species has been investigated theoretically. Some of these studies address the mechanism and the barrier for molecular dissociation; among the systems investigated CH4 [17], H2 [149], and HCOOH [150]. The general conclusion emerging from these studies is that reactions involving bond breaking which do not occur on the regular surface, quite unreactive, may take place at defects and in particular at oxygen vacancies. Recently, an accurate study of the interaction of H2 with a Fs center located at the (100) terrace has been performed with embedded cluster HF calculations [149]. It has been found that the reaction leads to the formation of Fs(H) centers in two steps, a dissociative interaction followed by UV irradiation which removes a neutral H atom and leaves an electron trapped at the vacancy with an hydroxyl group nearby. However, the high energy barrier estimated for the process, about 1 eV, suggests that the isolated surface anion vacancy is not the most likely site for formation of Fs(H) centers and that other sites, like low-coordinated F centers or divacancies (§ 3.6), are probably more efficient in determining the reaction [149]. An interesting case of reactivity of color centers is that of the interaction with CO, O2 and N2 molecules. EPR studies have clearly shown that exposure of a MgO surface with an high concentration of F centers to these molecules results in the formation of stable or metastable radical anions [151-153]. The general reaction can be written as:

121

Fs^ + X2 -^ Fs'^ + X2

(8)

The formation of the X2' species (X2 = O2, N2, or CO) is unambiguously shown by the analysis of the EPR spectrum and by the hyperfine interactions. This process implies a net charge transfer from the vacancy to the adsorbed molecule, with formation of surface species like the CO" or N2' radical anions which do not exist as free molecules in gas-phase. The process has been investigated in detail in a series of theoretical studies [153-154]. The reaction proceeds via the transfer of one electron trapped in the surface cavity to the empty levels of the adsorbed molecule. The resulting surface complex, X2' / Fs^"^, is bound by electrostatic forces. The electron is completely transferred from the cavity to the molecule which becomes negatively charged and is attracted by the strong electrostatic potential towards the surface. Although the mechanism of the interaction is the same for the three molecules, the details of the energetics are different. O2 removes spontaneously the electrons trapped in the MgO oxygen vacancies to form the stable O2' superoxide anion [94,154]. On the contrary, CO" and N2" form only at finite temperatures and are metastable species [153, 154]. The different behavior can be rationalized in terms of the electron affinities of the three molecules, positive for O2 and negative for CO and N2. In recent years, a particular attention has been devoted to anion vacancies as possible nucleation centers for the growth of metal particles on oxides. MgO (100) is the best studied oxide surface from this point of view. It has been found that several metal atoms interact weakly with the MgO surface [155] giving rise to rapid diffusion and desorption after deposition. The nucleation and growth of metal particles is assumed to occur at defect sites [156], and the oxygen vacancies are among the best candidates. Some ab initio studies have shown indeed that the metal-oxide interaction is much stronger at the F centers than at the regular sites [157-160]. It is still debated, however, if these centers promote nucleation. For the case of Pt it has been proposed that better nucleation centers are the surface OH groups (§ 3.3) and the divacancies (§ 3.6) [112]. F centers have also been shown to activate the supported metal atoms or clusters increasing their catalytic activity in CO oxidation or acetylene trimerization reactions [161-163]. 3.5 Cation vacancies Among intrinsic defects the cation vacancy or V center is of particular interest because it favors the formation of trapped holes at neighboring O ions; at the surface this may result in particularly active O" species which can react and exhibit catalytic activity [164] (see also § 3.8). There are several studies dedicated to the V center in the bulk, both experimental [165,166] and theoretical [14,15,141,167,168], as reviewed by Chen and Abraham [169]. This center can exist in three charged states, V°, V", V^". The ground state of V°

122

corresponds to two holes located on oxygen ions at opposite sides of the vacancy, and the ground state is a triplet. The singly charged V defect is very stable and is the one which is better characterized, while mostly theoretical information exists on the diamagnetic V^' center. Compared to the F centers, much less work has been devoted to the Vs center. One reason is that these centers are supposed to be present in lower concentrations than the O vacancies [170,171] due to their high formation energy. According to various theoretical estimates the energy required to remove a Mg^^ cation from bulk MgO with formation of a V^" center is of the order of a 20-30 eV [14,21,141]. An ab initio HF cluster study of the electronic structure of V centers at the MgO (100) surface [39] has shown that the electrons associated to the V and V^' centers are occupying levels near the top O 2p valence band and that the associated holes are completely localized on the oxygen atoms around the vacancy [39], in analogy with the bulk electronic structure. Very recently, a combined periodic supercell and embedded cluster study has been performed at the HF level on the electronic structure of the cation vacancy at the MgO surface [172]. It has been found that the formation energy of the neutral V^ defect in the bulk, 15 eV, is higher than at the surface, 13.5 eV, which means that migration to the surface should be thermodynamically favored. The holes connected to the formal removal of a neutral Mg atom are localized on a p orbital pointing toward the cavity [172]. Except for a study of the interaction with isolated metal atoms [157], to the best of our knowledge there are no theoretical investigations of the reactivity of surface V centers in MgO. Clearly, this is a center which should be considered in more detail in future investigations of MgO surface defects. 3.6 Divacancies Early experiments on the EPR spectrum of paramagnetic centers at the surface of MgO by Lunsford and Jayne [130] have been interpreted in terms of an electron trapped at a surface divacancy because of the analogy with the spectrum of a Fc"^ center, attributed to a divacancy with a trapped electron in bulk MgO [173]. One of the characteristic features of the spectrum which lead to the conclusion that the unpaired electron occupies a Mg and O divacancy and not a single vacancy site is the anisotropy in the values of the hyperfine coupling constants with the ^^Mg ions. Recently, this asymmetry has been explained as due to the Fs'^(H) centers, an oxygen vacancy in proximity of an adsorbed proton [125], as described above. Nevertheless, one cannot rule out the possibility that divacancy centers exists at the surface of MgO. There are several arguments in favor of the divacancy [174]: (a) it is energetically much easier to form than a pair of isolated cation and anion vacancies; (b) due to the mobility of cation vacancies, recombination of isolated vacancies should occur at high temperatures, > 1000 K, the typical temperatures reached in the thermal

123

treatment of polycrystalline MgO; (c) the divacancy is intrinsically asymmetric, a fact that would easily explain the anisotropy of the EPR signal; (d) the divacancy is a neutral center and can be present in high concentrations without the need for compensating charges. These reasons have stimulated in the last few years a growing interest around this center [174] and its reactivity [112,175]. The divacancy is sometimes referred to as the mini-step because the removal of a MgO unit from the surface results in four-coordinated ions, Fig. 10. Two different types of divacancies can be formed, the pit, where a surface Mg and the O ion underneath are removed, and the tub, where both ions are removed from the first layer, Fig. 10. It has been found [174] that the divacancies are relatively stable defects which can form through the recombination of isolated vacancies with an energy gain of up to 12 eV. The order of stability is similar as for isolated vacancies, buUc < surface, so that migration to the surface is expected at high temperatures. Neutral divacancies have an appreciable electron affinity of almost 1 eV; this means that electrons can be trapped at these defects forming a paramagnetic center. The affinity for a second electron however is negative. This is different from single O vacancies; both Fs^"^ and Fs"^ vacancies have positive electron affinities and transform into the diamagnetic Fs^ center by electron trapping. Recently, however, it has been suggested based on spectroscopic measurements [164] that the paramagnetic divacancy, with a trapped electron, cannot be transformed into a diamagnetic center by further electron trapping, indicating that a divacancy with two trapped electrons is unstable in full agreement with the theoretical prediction.

Fig. 10. Schematic representation of a divacancy on the MgO surface. Top: pit conformation; bottom: tub conformation; x and x' indicate the positions of the removed atoms. Reproduced from ref. [174]. Copyright 1998 American Institute of Physics.

124

Of course, relaxation of neighboring ions makes an important contribution to the relative stability of neutral and charged divacancy centers. Thus, from a thermodynamic point of view the divacancy has been confirmed as an important defect center on the surface of MgO [174] and recent experimental data on the formation of surface O" ions by bleaching color centers with N2O point to the important role of these defects [164]. The study of the hyperfine coupling constants, however, resulted in a too small anisotropy of the signal, compared to experiment. In this respect the assignment of the observed EPR spectra [125] to surface divacancies with a trapped electron is not conclusive [174]. A detailed theoretical study of the reaction of a surface divacancy with H2, aimed at the identification of the Fs"^(H) color center observed in EPR, also come to the conclusion that the calculated magnetic parameters show discrepancies with the experiment [175]. Despite these open questions, there is no doubt that this defect center should be considered with great attention in future analysis of the reactivity of the MgO surface. 3.7 Impurity atoms Impurity atoms in MgO can be either cations replacing Mg^"^ or anions substituting O^'. In either case the surface reactivity is significantly modified by their presence. In the case of cation impurities, these can be of at least two kinds. If Mg^"^ is replaced by another cation with the same formal charge, e.g. Ca^"^, Ni^^ Co^\ etc., the main changes in the local electronic structure are connected to the different size of the ion, to small modifications in the chemical bonding with local increase or decrease of the covalent character, and to the presence of filled d orbitals. In fact, it has been shown by ab initio HF cluster calculations that, in bulk MgO, a cation like Ni^^ is stable against disproportionation into the +1 and +3 states [176]. If, on the other hand, a cation with a different formal charge e.g. Na"^ or Li"*^, substitutes Mg^"*^, then a deep change in the electronic structure occurs to compensate the missing charge. For monovalent cations this is the change in the anion formal charge from -2 to -1; the O' species, however, is a radical and its chemistry is fundamentally different from that of the corresponding fully reduced species, as it will be discussed in more detail in § 3.8. Let us consider first isovalent cations. MgO can dilute ions of similar size, as for instance Ni^"^ or Co^"^ forming NiO-MgO and CoO-MgO soUd solutions with an infinite range of composition. The effect of progressively replacing Mg^"^ by Ni^^ or similar cations (Co^"^, Cu^"^) on the surface properites has been investigated both experimentally [83,177,178] and theoretically [179,180]. The presence of Ni^^ cations diluted in the MgO matrix results in an efficient catalyst for nitrous oxide, N2O, decomposition; this has been attributed to the different bond strength of the Ni-0 and Mg-0 bonds at the surface [177]. Plane wave calculations on Ni-doped MgO have shown that the presence of Ni atoms on the

125

surface enhances the bonding of S-containing molecules like H2S through electronic states associated with the Ni 3d levels [181]. To check the role of the 3d orbitals of the TM cation in the chemistry of the surface the vibrational properties of adsorbed CO have been considered [83,178]. In fact, assuming that chemical forces do not play any role so that Ni^"*" and Co^"^ ions diluted in a MgO matrix experience an ionicity very similar to that of Mg^^, the frequency of CO adsorbed on Mg^"*" or Ni^"*" and Co^"^ should be practically indistinguishable. The experiments show a different situation. The peak of CO adsorbed on Ni^"*" falls at a distinctly lower frequency than that of CO on Mg^^ (2130 cm"^ versus

00

Fig. 11. Difference density maps for CO adsorbed on Mg "^ ions on a MgO (100) surface, (a) contribution from a orbitals; (b) contributions from n orbitals. Continuous lines correspond to charge accumulation, dashed lines to charge depletion with respect to the superposition of the non interaction CO and MgO densities. Reproduced from ref. [180]. Copyright 1993 Elsevier. b

( _,/' --W^^^^-

\

i

% ^--^—-' '''--'' '^—-' v,^ % %

y

Fig. 12. Difference density maps for CO adsorbed on Ni ^ ions on a MgO (100) surface, (a) contribution from a orbitals; (b) contributions from 7C orbitals. Continuous lines correspond to charge accumulation, dashed lines to charge depletion with respect to the superposition of the non interaction CO and MgO densities. Reproduced from ref [180]. Copyright 1993 Elsevier.

2144 cm'^ [83]); furthermore, even the intensity of the peak is not the same, and for Ni^'*"-CO is larger than expected on the basis of the stoichiometry. Similar results have been obtained for Co. A rationalization in terms of a small, but non

126

negligible metal dJi-CO 271* interaction has been proposed based on DFT cluster model calculations [180], see Fig. 11 and Fig. 12. If the role of the d-7i overlap is quite modest in the Ni^^—CO surface complex, this is no longer true when a stronger 7C-acceptor molecule like NO is used as a probe. In fact, on MgO at 77 K NO is so weakly bound that that only lateral interaction products form like the (N0)2 dimers; when Ni^"^ ions are diluted in the material, Ni^"^-NO complexes form and remain stable even at room temperature [178]. While the effect of cation impurities on the surface chemistry of MgO has been investigated in detail, very little is known about anion substitution. Defect formation and excitation energies for S^' and Se^-doped bulk MgO have been calculated [182,183] but there are no data for the surface. In the bulk it has been estimated that the presence of S^" or Se^" impurities result in a outward relaxation of the Mg^"*" neighbors of 6% and 8%, respectively [182]. A recent report of the 0^"-S^" exchange reaction on MgO has been reported [184]. The reaction involves adsorption of CS2 on MgO powders and the subsequent exchange reaction with formation of COS and of S^" ions probably located at the low coordinated sites. It has been found that the basicity of the MgO surface doped with sulfur ions is drastically modified with respect to that of pure MgO [184]. The inclusion of impurity atoms in MgO is much more interesting from a chemical point of view when alkali metals are used to replace Mg ions. In fact, this results in trapped-hole centers. The M'^/O' pairs have been extensively studied in the bulk of alkaline-earth oxides by optical studies, EPR and ENDOR measurements [185,186] as well as by embedded cluster calculations [187]. The LiVO" ions create an effective dipole which polarizes the surrounding lattice, with the two ions moving toward each other. The presence of an O' radical, however, is most interesting when one is dealing with surface properties. This center in fact is very reactive and is the subject of the next paragraph. 3.8 C radical anions Paramagnetic O' anions located at the surface of ionic crystals are capable of radical cleavage processes, such as hydrogen abstraction from methane and hydrocarbon molecules adsorbed on the surface. They can also facilitate UV induced homolytic splitting of hydrogen. This particular reactivity of the O" anions has been studied extensively by means of EPR and TPD [188]. In the previous discussion we have already encountered the O' ion in connection with other defects. In fact, a neutral cation vacancy in MgO, Vs, corresponds to two trapped holes on two surface O anions; in the charged Vs" center only one O' radical is present. In the previous section we have seen that doping an alkalineearth oxide with Li or Na results in M V O ' pairs. Indeed, some of the most efficient catalysts to promote the conversion of methane into higher hydrocarbon derivatives are based on alkali-doped metal oxides, including Na/CaO and Li/MgO. At high partial pressures of methane, gas-phase oxidation of methane

127

with O2 to give coupled products such as C2H4 or C2H6 proceeds with high conversion rate at temperatures above 500 °C [189]. At low methane partial pressures the catalyst is required to give acceptable conversion rate and selectivity. The reaction proceeds via C-H bond cleavage and the centers that facilitate C-H bond breaking are unlikely to be present on a perfect MgO surface. The surface activity has been attributed to the presence of O' radical ions formed upon doping with Li. It has been proposed that methane is adsorbed at the surface and hydrogen abstracted by the O" radicals, with formation of methyl radicals [190], although there are reports that the F centers do also play a role [191]. However, the mechanisms proposed to explain all the processes occurring in real catalytic conditions comprise more than 100 steps, many of which not yet completely clarified. The elementary steps of the reaction have attracted considerable theoretical interest in recent years [192-195]. The studies have been performed at various computational levels, HF, MP2, DFT [193-195], and explicitly correlated wavefunctions (MCSCF and CI) [192] and the surface has been represented by embedded clusters [192,194,195] or by periodic slabs [193]. All the calculations agree with the fact that the LiVO' complex is the active center although most recent studies suggest that other surface species may also play a role [195]. The activation energy for the C-H bond breaking is very dependent on the method used but also on the proper inclusion of surface relaxation along the reaction profile. The values range from 4-6 kcal/mol [192] to 18 kcal/mole [193] although the best estimates are probably consistent with a low barrier [194]. The O" ions can also be produced on the MgO surface by other methods, e.g. by excitation of low-coordinated surface anions, Oic, by UV light [164,196198] or by filling anion vacancies with paramagnetic O' species [164,199]. In the first case the process involves an ionization or a charge transfer process: Oic^- + hv -> Oic" + e-

(9)

where the electron is released to an electron acceptor; under O2 atmosphere the electron can be captured by the oxygen molecule with formation of the superoxide anion, O2'. In UHV conditions, short lived 0-Mg"^ pairs form in the excitation step. Recently, the energy required to ionize O^' anions from the MgO surface have been determined theoretically from DFT cluster calculations [50]. The computed IP of a five-coordinated anion is 6.7 eV, in excellent agreement with an experimental estimate of 6.7±0.4 eV [200]. On a O3C anion at a comer site the IP decreases to 5.6 eV [50]. Thus, it is conceivable that UV irradiation will preferentially produce O' radicals at low-coordinated sites. The second mechanism to generate O" radicals on the MgO surface implies first the creation of color centers and then their bleaching with N2O:

128

(10)

Fs"" + N2O - ^ Os" + N2

In this case the decomposition of the reactive N2O molecule results in the filling of a paramagnetic vacancy site with release of the unreactive N2 molecule. Clearly, the presence of O" species at the surface of MgO is not disconnected from the existence of other defects, low-coordinated anions or O vacancies. In fact, the complex interconversion of one center into another one is one of the reasons for the difficult identification of defect centers on oxide surfaces. 3.9 (111) microfacets In contrast to nonpolar (100) surfaces of rock-salt structured ionic materials, which are thermodynamically stable, (111) polar surfaces are unstable if they remain in the bulk terminated structure due to the diverging surface potential [2,201] and undergo substantial surface reconstruction. It has been predicted that the (111) surfaces of NaCl, NiO, MgO, etc. reconstruct from the p(lxl) bulk terminated into a p(2x2) structure. Fig. 13 [201]. This kind of reconstruction has been indeed observed on NiO(lll) grown on Au(lll) [202]. Polar surfaces can also be stabilized by the presence of oxygen vacancies, impurities, hydroxyl groups [203] or in general molecular adsorbates [204,205]; on NiO, by simple heat treatment, the OH groups are desorbed and the surface reconstructs exhibiting a diffuse p(2x2) structure [203].

PlirM;?!! -.-m^m^^m. -^^KT^B- ^^f^m^^m

4^B'^V

dHi^V

t^l^V'''

Fig. 13. Structure of a p(2x2) reconstructed MgO(l 11) surface.

Recently, the surface morphology and faceting of MgO(lll) surfaces has been studied with atomic force, scanning and electron microscopies [206,207]. It has been suggested that oxygen trimers reminiscent of cyclic ozone appear on the reconstructed surface [207]. A p(2x2) reconstruction leads to the formation of a series of micropyramids, and it is possible that local arrangements of ions at the surface of MgO assume a similar morphology. A complete account of the

129

reconstruction of the polar surfaces of ionic crystals is outside the scope of this review, but here we simply want to mention the possible existence of microensembles of ions with (111) microface morphology. It has been suggested that a fundamental defect in MgO consists in a missing cation from a cube comer, leaving a (111) microface with three oxygen ions and one hole. The electronic structure of this center was then investigated theoretically in a pioneering study of Kunz and Guse [208]. The involvement of (111) microfacets was invoked also in a thermal desorption and isotope exchange reaction study of CO and CO2 on MgO where the condition needed for double exchange was identified by means of HF cluster calculations in small ensembles of three-coordinated O ions [209,210]. 4. CONCLUSIONS In this brief account we have listed a number of defects occurring on the surface of ionic oxides, in particular MgO. The experimental and theoretical activity aimed at the characterization of these defects has become quite intense in recent years, also thanks to the great advances in the preparation of oxide films in UHV under controlled conditions. Still, a detailed understanding of the relations existing between preparation methods and nature and concentration of surface defects is missing. Complex mechanisms control the interconversion of one defect into another and even the nature of a simple defect like the oxygen vacancy is still under discussion. A real control on the microscopic aspects of the chemistry of oxide surfaces will be possible only when the chemistry of defects, vacancies, irregularities will be completely identified. In this respect, ab initio calculations provide an important tool to interpret and complement the experimental information. In particular, local cluster models are particularly useful because of their computational simplicity and of the possibility to compute a great variety of observable properties, like IR, optical, photoemission NMR and EPR spectra. The design of "realistic" models of the surface defects and their validation through the calculation of observable quantities represents probably the best way to solve the problem. Acknowledgments I am indebted to A. M. Ferrari (Torino), T. Bredow (Hannover), C. Sousa and N. Lopez (Barcelona), L. Giordano, R. Soave and D. Ricci (Milano) for their substantial help in the study of defects in oxides. The cooperation with the groups of F. lUas (Barcelona), E. Giamello and C. Pisani (Turin), N. Rosch (Munich), and U. Heiz (Losanne) has been invaluable to elucidate many open questions and to link experimental with theoretical results. Financial support through the INFM Projects PAIS and PRA-ISADORA is gratefully acknowledged.

130

REFERENCES [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39]

V.E. Henrich and P.A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, Cambridge 1994. C. Noguera, Physics and Chemistry at Oxide Surfaces, Cambridge University Press, Cambridge 1996. E.A. Colboum, Surf. Sci. Rep. 15 (1992) 281. C.R. Henry, Surf. Sci. Rep. 31 (1998) 231. G. Renaud, Surf. Sci. Rep. 32 (1998) 1. D.A. Bonnel, Prog. Surf. Sci. 57 (1998) 187. S. A. Chambers, Surf. Sci. Rep. in press (2000). H. J. Freund, Faraday Discuss. 114 (1999) 1. H. J. Freund, H. Kuhlenbeck, and V. Staemmler, Rep. Prog. Phys. 59 (1996) 283. M. Baumer and H. J. Freund, Prog. Surf. Sci. 61 (1999) 127. G. Pacchioni, P.S. Bagus, and F. Parmigiani (Eds.), Cluster Models for Surface and Bulk Phenomena, NATO ASI Series B, vol. 283, Plenum, New York 1992. P.A. Cox, The Electronic Structure and Chemistry of Solids, Oxford Science PubHcations, Oxford, 1987. A. Gibson, R. Haydock, and J.P. LaFemina, Appl. Surf. Sci. 72 (1993) 285. A. Gibson, R. Haydock, and J.P. LaFemina, Phys. Rev. 50 (1994) 2582. L. N. Kantororovich, J.M. Holender, and M.J. Gillan, Surf. Sci. 343 (1995) 221. E. Castanier and C. Noguera, Surf. Sci. 364 (1996) 1. R. Orlando, R. Millini, G. Perego, and R. Dovesi, J. Molec. Catal. A 119 (1997) 253. R. J. D. Tilley, Principles and Applications of Chemical Defects, Stanley Thomes, Cheltenham, 1998. A. M. Stoneham, Theory of Defects in SoHds, Oxford University Press, Oxford 1975. F. Finocchi, J. Goniakowski, and C. Noguera, Phys. Rev. B 59 (1999) 5178. C. Pisani, F. Cora, R. Dovesi, and R. Orlando, J. Elec. Spectr. 96 (1994) 1. M. Leslie and M.J. Gillan, J. Phys. C 18 (1985) 973. A. Gorling, H.H. Heinze, S. P. Ruzankin, M. Staufer, and N. Rosch, J. Chem. Phys. 110 (1999)2785. G. Pacchioni and G. lerano, Phys. Rev. Lett. 79 (1997) 753. G. Pacchioni, G. lerano and A.M. Marquez, Phys. Rev. Lett. 81 (1998) 377. G. Pacchioni and R. Ferrario, Phys. Rev. B 58 (1998) 6090. B. B. Stepanov and K. Raghavachari, Appl. Phys. Lett. 71 (1997) 770. F. Illas and G. Pacchioni, J. Chem. Phys. 108 (1998) 7835. C. Sousa, G. Pacchioni, and F. Illas, Surf. Sci. 429 (1999) 217. A. Continenza and A. Di Pomponio, Phys. Rev. B 54 (1966) 13687. M. Magagnini, P. Giannozzi, and A. Del Corso, Phys. Rev. B 61 (2000) 2621. J. Laegsgaard and K. Stokbro, Phys. Rev. B 61 (2000) 12590. R. H. D. Nuttall and J. A. Weil, Can. J. Phys. 59 (1981) 1696. M. J. Mombourquette, J. A. Weil, and P. G. Mezey, Can. J. Phys. 62 (1984) 21. F. Sim, C. R. A. Catlow, M. Dupuis, and J. D. Watts, J. Chem. Phys. 95 (1991) 4215. G. Pacchioni, F. Frigoli and J. A. Weil, Phys. Rev. B submitted. J. Sauer, P. Ugliengo, E. Garrone, and V.R. Saunders, Chem. Rev. 94 (1994) 2095. G. Pacchioni, A.M. Ferrari, A.M. Marquez, and F. Illas, J. Comp. Chem. 18 (1997) 617. A.M. Ferrari and G. Pacchioni, J. Phys. Chem. 99 (1995) 17010.

131 [40] A.M. Ferrari and G. Pacchioni, Int. J. Quant. Chem. 58 (1996) 241. [41] M.A. Nygren, L.G.M. Pettersson, Z. Barandiaran, and L. Seijo, J. Chem. Phys. 100 (1994) 2010. [42] W.J. Stevens, H. Basch, and M. Krauss, J. Chem. Phys. 81 (1984) 6026. [43] Z. Barandiaran and L. Seijo, J. Chem. Phys. 89 (1988) 5739. [44] V. Luana and L. Pueyo, Phys. Rev. B 39 (1989) 11093. [45] M. Bom, Z. Physik 1 (1920) 45. [46] R. Pandey and J.M. Vail, J. Phys.: Condens. Matter 1 (1989) 2801. [47] J. H. Harding, A.H. Marker, P.B. Keegstra, R. Pandey, J.M. Vail, and C. Woodward, PhysicaB&C 131 (1985) 151. [48] B. G. Dick and A. W. Overhauser, Phys. Rev. 112 (1958) 90. [49] C.R.A. Catlow, M. Dixon, and W.C. Mackrodt, in Computer Simulation of Solids, C.R.A. Catlow, (Ed.), Springer, Berlin 1982, p. 130. [50] P. V. Susko, A. L. Shluger, and C. R. A. Catlow, Surf. Sci. 450 (2000) 153. [51] C. Pisani, J. Mol. Catal. 82 (1993) 229. [52] C. Pisani, F. Cora, R. Nada, and R. Orlando, Comput. Phys, Commun. 82 (1994) 139. [53] C. Pisani and U. Birkenheuer, Comput. Phys. Commun. 96 (152) 1996. [54] C. Pisani, U. Birkenheuer, F. Cora, R. Nada, and S. Casassa, EMBED96 User's Manual Universita di Torino, Torino 1996. [55] E. Scorza, U. Birkenheuer, and C. Pisani, J. Chem. Phys. 107 (1997) 9645. [56] C. Pisani, R. Dovesi and C. Roetti, in: Hartree-Fock Ab-initio Treatment of Crystalline Systems, Lecture Notes in Chemistry, Vol. 48, Springer, Heidelberg, 1988. [57] V. R. Saunders, R. Dovesi, C. Roetti, M. Causa, N. M. Harrison, R. Orlando, and C. M. Zicovich-Wilson, CRYSTAL98 User's Manual, Universita di Torino, Torino 1998. [58] I . N. Levine, Quantum Chemistry, Prentice Hall, Englewood Cliffs, 1991. [59] F. Jensen, Introduction to Computational Chemistry, Wiley-VCH, Chichester 2000. [60] R. G. Parr and W. Yang, Density Functional Theory of Atoms and Molecules, Oxford University Press, New York, 1988. [61] C. Lee, W. Yang, and R.G. Parr, Phys. Rev. B, 37 (1988) 785. [62] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [63] N. Lopez and F. Illas, J. Phys. Chem. B 102 (1998) 1430. [64] N. Lopez, F. Illas, N. Rosch, G. Pacchioni, J. Chem. Phys., 110 (1999) 4873. [65] P. Hoenberg and W. Kohn, Phys. Rev. B 136 (1964) 864. [66] O. Robach, G. Renaud, and A. Barbeir, Surf. Sci. 401 (1998) 227. [67] G. Pacchioni, Surf. Rev. & Letters, in press (2000). [68] E. Escalona-Platero, D. Scarano, G. Spoto, and A. Zecchina, Farady Discuss. Chem. Soc. 80(1985) 183. [69] S. Furuyama, H. Fuijii, M. Kawamura, and T. Morimoto, J. Phys. Chem. 82 (1978) 1028. [70] E. A. Pauskshtis, R.I. Soltanov, and E. Yurchenko, React. Kinet. Catal. Lett. 16 (1981) 93. [71] J.W. He, C.A. Estrada, J.S. Comeille, M.C. Wu, and D.W. Goodman, Surf. Sci. 261 (1992) 164. [72] F. Illas, G. Pacchioni, A. Pelmenschikov, L.G.M. Pettersson, R. Dovesi, C. Pisani, K.M. Neyman, and N. Rosch, Chem. Phys. Lett. 306 (1999) 202. [73] R. Dovesi, R. Orlando, F. Ricca, and C. Roetti, Surf. Sci. 186 (1987) 267. [74] G. Pacchioni, G. Cogliandro, and P.S. Bagus, Surf. Sci. 255 (1991) 344. [75] M. A. Nygren and L. G. M. Pettersson, J. Chem. Phys. 105 (1996) 9339.

132 [76] J. A. Mejias, A. M. Marquez, J. Fernandez Sanz, M. Femandez-Garcia, J. M. Ricart, C. Sousa, and F. Illas, Surf. Sci. 327 (1995) 59. [77] K. M. Neyman and N. Rosch, Chem. Phys. 168 (1992) 267. [78] K. M. Neyman, S. P. Ruzankin, and N. Rosch, Chem. Phys. Lett. 246 (1995) 546. [79] A. G. Pelmenschikov, G. Morosi, A. Gamba, S. Coluccia, J. Phys. Chem. 99 (1995) 975. [80] G. Pacchioni, G. Cogliandro, and P.S. Bagus, Int. J. Quant. Chem. 42 (1992) 1115. [81] L. Marchese, S. Coluccia, G. Martra, and A. Zecchina, Surf. Sci. 269/270 (1992) 135. [82] G. Pacchioni, T. Minerva, and P.S. Bagus, Surf. Sci. 275 (1992) 450. [83] D. Scarano, G. Spoto, S. Bordiga, S. Coluccia, and A. Zecchina, J. Chem. Soc. Faraday Trans. 88(1992)291. [84] L.G.M. Pettersson, M. Nyberg, J.L. Pascual, and M.A. Nygren, in: Chemisorption and Reactivity on Supported Clusters and Thin Films, R.M. Lambert and G. Pacchioni (Eds.), NATO ASI Series E, Vol. 331, Kluwer Dordrecth 1997, p. 425. [85] R. Wichtendahl, M. Rodriguez-Rodrigo, U. Hartel, H. Kuhlenbeck, and H.J. Freund, Surf. Sci. 423 (1999) 90. [86] R. Wichtendahl, M. Rodriguez-Rodrigo, U. Hartel, H. Kuhlenbeck, and H.J. Freund, Phys. Stat. Sol. (a) 173 (1999) 93. [87] P.A. Redhead, Vacuum 12 (1962) 23. [88] A. G. Pelmenschikov, G. Morosi, A. Gamba, S. Coluccia, G. Martra, and E. A. Paukshtis, J. Phys. Chem. 100 (1996) 5011. [89] G. Busca, Catalysis Today, 27 (1996) 321. [90] R. Soave and G. Pacchioni, Chem. Phys. Letters 320 (2000) 345. [91] S. Coluccia, M. Baricco, L. Marchese, G. Martra and A. Zecchina, Spectrochim. Acta 49A(1993) 1289. [92] J. H. Lunsford and J. P. Jayne, J. Chem. Phys. 44 (1966) 1487. [93] E. Giamello, P. Ugliengo, and E. Garrone, J. Chem. Soc. Faraday Trans. 1 85 (1989) 1373. [94] G. Pacchioni, A.M. Ferrari and E. Giamello, Chem. Phys. Lett. 255 (1996) 58. [95] G. Pacchioni, J.M. Ricart, and F. Illas, J. Am. Chem. Soc. 116 (1994) 10152. [96] G. Pacchioni, A. Clotet, and J.M. Ricart, Surf. Sci. 315 (1994) 337. [97] C. Sousa, J.A. Mejias, G. Pacchioni, F. Illas, Chem. Phys. Lett. 249 (1996) 123. [98] M. A. Nygren and L.G.M. Pettersson, Chem. Phys. Lett. 230 (1994) 456. [99] L. N. Kantorovich and M.J. Gillan, Surf. Sci. 374 (1997) 373. [100] D. Ochs, M. Brause, B. Braun, W. Maus-Friedrichs, and V. Kempter, Surf. Sci. 397 (1998) 101. [101] D. Ochs, B. Braun, W. Maus-Friedrichs, and V. Kempter, Surf. Sci. 417 (1998) 406. [102] X. Carrier, C. S. Doyle, T. Kendelewicz, and G. E. Brown, Surf. Rev. & Letters 6 (1999) 1237. [103] C. S. Doyle, T. Kendelewicz, X. Carrier, and G. E. Brown, Surf. Rev. & Letters 6 (1999) 1247. [104] S. Coluccia, F. Boccuzzi, G. Ghiotti, C. Mirra, Z. Phys. Chem. 121 (1980) 141. [105] E. Garrone and F.S. Stone, J. Chem. Soc. Faraday Trans. I 83 (1987) 1237. [106] H. Kobayashi, M. Yagamuchi, and T. Ito, J. Phys. Chem. 94 (1990) 7206. [107] K. Sawabe, N. Koga, and K. Morokuma, J. Chem. Phys. 97 (1992) 6871. [108] H. Kobayashi, D.R. Salahub, and T. Ito, J. Phys. Chem. 98 (1994) 5487. [109] R. Nada, A.C. Hess, and C. Pisani, Surf. Sci. 336 (1995) 353. [110] J. Goniakovski and C. Noguera, Surf. Sci. 340 (1995) 191.

133 [ I l l A. L. Shluger, P.V. Sushko, and L.N. Kantorovich, Phys. Rev. B 59 (1999) 2417. [112 A. Bogicevic and D.R. Jennison, Surf. Sci. 437 (1999) L741. [113 K. Refson, R. A. Wogelius, D. G. Eraser, M. C. Payne, M. H. Lee, and V. Milman, Phys. Rev. B 52 (1995) 10823. [114; G. E. Brown, V.E. Henrich, W.H. Casey, D.L. Clark, C. Eggleston, A. Felmy, D.W. Goodman, M. Gratzel, G. Maciel, M.I. McCarthy, K.H. Nealson, D.A. Sverjensky, M.F. Toney, and J.M. Zachara, Chem. Rev. 99 (1999) 77. [115 J. L. Anchell and A.C. Hess, J. Phys. Chem. 100 (1996) 18317. [116 C. A. Scamehom, N.M. Harrison, and M.I. McCarthy, J. Chem. Phys. 101 (1994) 1547. [117 W. Langel and M. Parrinello, Phys. Rev. Lett. 73 (1994) 504. [118 W. Langel and M. Parrinello, J. Chem. Phys. 103 (1995) 3240. [119 L. Giordano, J. Goniakowski, and J. Suzanne, Phys. Rev. Lett. 81 (1998) 1271. [120: S. Coluccia, F. Boccuzzi, G. Ghiotti, C. Morterra, J. Chem. Soc. Faraday Trans. 1 78 (1982)2111. [121 S. Coluccia, L. Marchese, S. Lavagnino, and M. Apno, Spectrochimica Acta, 43A (1987) 1573. [122 A. J. Tench and R.L. Nelson, J. Colloid Interface Sci. 26 (1968) 364. [123 A. J. Tench, Surf. Sci. 25 (1971) 625. [124: E. Giamello, D. Murphy, L. Ravera, S. Coluccia, and A. Zecchina, J. Chem. Soc. Faraday Trans. 90 (1994) 3167. [125 E. Giamello, M.C. Paganini, D. Murphy, A.M. Ferrari, and G. Pacchioni, J. Phys. Chem. 101 (1997)971. [126: V. E. Henrich, G. Dresselhaus, and H.J. Zeiger, Phys. Rev. B 22 (1980) 4764. [127 M. C. Wu, C M . Truong, and D.W. Goodman, Phys. Rev. B 46 (1992) 12688. [128 D. Peterka, C. Tegenkamp, K.M. Schroder, W. Ernst, H. Pfnur, Surf. Sci. 431 (1999) 146. [129: J. A. Weil, J. R. Bolton, and J. E. Wertz, Electron Paramagnetic Resonance, John Wiley & Sons, New York, 1994. [130: J. H. Lunsford and J.P. Jayne, J. Phys. Chem. 70 (1966) 3464. [131 W. P. Unruh and J.M. Culvahouse, Phys. Rev. 154 (1967) 861. [132: B. Henderson and J.E. Wertz, Adv. Phys. 17 (1968) 749. [133 M. C. Paganini, M. Chiesa, E. Giamello, S. Coluccia, G. Martra, D.M. Murphy, and G. Pacchioni, Surf. Sci. 421 (1999) 246. [134 M. Chiesa, M. C. Paganini, E. Giamello, and D. M. Murphy, Langmuir 13 (1997) 5306, [135 D. M. Murphy, R. D. Farley, I. J. Pumell, C. C. Rowlands, A. R. Yacob, M. C. Paganini, and E. Giamello, J. Phys. Chem. B 103 (1999) 1944. [136: R. R. Sharma and A.M. Stoneham, J. Chem. Soc. Faraday trans. 2 (1976) 913. [137 G. Pacchioni, A.M. Ferrari, and G. lerano, Faraday Discuss., 106 (1997) 155. [138 P. Hofmann and E. Knozinger, Phys. Chem. Chem. Phys. 1 (1999) 713. [139 G. Pacchioni and P. Pescarmona, Surf. Sci. 412/413 (1998) 657. [140: A. L. Shluger, L. N. Kantorovich, A. I. Livshits, and M. J. Gillan, Phys. Rev. B 56 (1977) 15332. [141 A. De Vita, M.J. Gillan, J.S. Lin, M.C. Payne, I. Stich, and L.J. Clarke, Phys. Rev. B 46 (1992) 12964. [142: W. C. Mackrodt and R.F. Stewart, J. Phys. C 12 (1979) 431. [143 S. Shirasaki and M. Harma, Chem. Phys. Ixtt. 20 (1973) 361. [144 A. Atkinson and R.I. Taylor, Philos. Mag. A 43 (1981) 979. [145 Y. Chen, R.T. Williams and W.A. Sibley, Phys. Rev. 182 (1969) 960.

134 146] R. L. Nelson and J.W. Hale, Trans. Faraday Soc. 52 (1971) 77. 147] C. Tegenkamp, H. Pfnur, W. Ernst, U. Malaske, J. Wollschlager, D. Peterka, K. M. Schroder, V. Zielasek, and M. Henzler, J. Phys. Condens. Matter 11 (1999) 9943. 148] P. R. Underbill and T.E. Gallon, Solid State Commun. 43 (1982) 9. 149] A. D'Ercole, E. Giamello, C. Pisani and L. Ojamae, J. Phys. Chem. B 103 (1999) 3872. 150] M. Lintuluoto, H. Nakatsuji, M. Hada, and H. Kanai, Surf. Sci. 429 (1999) 133. 151] E. Giamello, D. Murphy, E. Garrone and A. Zecchina, Spectrochim. Acta, 49A (1993) 1323. 152] E. Giamello, D. Murphy, L. Marchese, G. Martra, and A. Zecchina, J. Chem. Soc. Faraday Trans. 1989 (1993) 3715. 153] E. Giamello, M.C. Paganini, M. Chiesa, D.M. Murphy, G. Pacchioni, R. Soave, and A. Rockenbauer, J. Phys. Chem. B, 104 (2000) 1887. 154] A. M. Ferrari and G. Pacchioni, J. Chem. Phys. 107 (1997) 2066. 155] I. Yudanov, G. Pacchioni, K. Neyman and N. Rosch, J. Phys. Chem. B 101 (1997) 2786. 156] C. T. Campbell, Surf. Sci. Reports 27 (1997) 1. 157] A. M. Ferrari and G. Pacchioni, J. Phys. Chem. 100 (1996) 9032. 158] A. V. Matveev, K. M. Neyman, I. Yudanov, and N. Rosch, Surf. Sci. 426 (1999) 123. 159] G. Haas, A. Menck, H. Brune, J. V. Barth, J. A. Venables, and K. Kern, Phys. Rev. B 61(2000)11105. 160] Y. F. Zhukovskii, E. A. Kotomin, P. W. Jacobs, A. M. Stoneham, and J. H. Harding, J. Phys.: Condens. Matter 12 (2000) 55. 161] A. Sanchez, S. Abbet, U. Heiz, W. D. Schneider, H. Hakkinen, R. N. Bennet, and U. Landman, J. Phys. Chem. A 103 (1999) 9573. 162] S. Abbet, A. Sanchez, U. Heiz, W. D. Schneider, A. M. Ferrari, G. Pacchioni and N. Rosch, J. Phys. Chem. A 103 (1999) 9573. 163] A. M. Ferrari, L. Giordano, N. Rosch, U. Heiz, S. Abbet, A. Sanchez, and G. Pacchioni, J. Phys. Chem. B in press (2000). 164] M. Sterrer, O. Diwald, and E. Knozinger, J. Phys. Chem. B 104 (2000) 3601. 165] Y. Chen, M. M. Abraham, L. C. Templeton, and W. P. Unruh, Phys. Rev. B 11 (1975) 881. 166] L. E. Halliburton, L. A. Kappers, D. L. Cowan, F. Dravnieks, and J. E. Wertz, Phys. Rev. B 8 (1973) 1610. 167] J. D. Foot, E. A. Colbourn, and C. R. A. Catlow, J. Phys. Chem. Solids, 49 (1988) 1225. 168] R. I. Eglitis, E. A. Kotomin, and G. Borstel, Phys. Stat. Sol. (b) 208 (1998) 15. 169] Y. Chen and M. M. Abraham, J. Phys. Chem. Solids, 51 (1990) 747. 170] A. Tench, and M.J. Duck, J. Phys. C: Solid State Phys. 6 (1973) 1134. 171] M. M. Abraham, Y. Chen, and W.P. Unruh, Phys. rev. B 9 (1974) 1842. 172] P. Baranek, G. Pinarello, C. Pisani, and R. Dovesi, Phys. Chem. Chem. Phys. (2000) in press. 173] J.E. Wertz, J.W. Orton, and P. Auzins, Discuss. Faraday Soc. 31 (1961) 140. 174] L. Ojamae and C. Pisani, J. Chem. Phys. 109 (1998) 10984. 175] A. D'Ercole and C. Pisani, J. Chem. Phys. 111 (1999) 9743. 176] J. Meng, J. M. Vail, A. M. Stoneham, and P. Jena, Phys. Rev. B 42 (1990) 1156. 177] Y. Izumi, T. Shimiuzu, T. Kobahashi and K. Aika, Chem. Commun. (2000) 1053. 178] A. Zecchina, D. Scarano, S. Bordiga, G. Ricchiardi, G. Spoto, and F. Geobaldo, Catal. Today 27 (1996) 403. [179] E. A. Colbourn and W. C. Mackrodt, Surf. Sci. 117 (1982) 571.

135 [180] K.M. Neyman and N. Rosch, Chem. Phys. 177 (1993) 561. [181] J. A. Rodriguez and A. Maiti, J. Phys. Chem. B 104 (2000) 3630. [182] R. Pandey, J. Zuo, and A.B. Kunz, Phys. Rev. B 39 (1989) 12565. [183] R. Pandey, J. Zuo, and A.B. Kunz, J. Mater. Res. 5 (1990) 623. [184] D. Scarano, A. Zecchina, S. Bordiga, C. Lamberti, and G. Spoto, International Workshop "Oxide surfaces", Elmau, Germany, January 1999. [185] M. M. Abraham, W. P. Unruh, and Y. Chen, Phys. Rev. B 10 (1974) 3540. [186] G. Rius and A. Herve, Solid State Commun. 15 (1974) 399. [187] J. Zuo, R. Pandey, and A. B. Kunz, Phys. Rev. B 44 (1991) 7187. [188] T. Ito, A. Kawanami, K. Toi, T. Shirakawa, and T.Tokuda, J. Phys. Chem. 92 (1988) 3910. [189] O. T. Onsager, R. Lodeng, P. Soraker, A. Anudskaas, and B. Helleborg, Catal. Today 4 (1989) 355. [190] D. J. Driscoll, W. Martir, J.X. Wang, and J.H. Lunsford, J. Am. Chem. Soc. 107 (1985) 58. [191] M. C. Wu, C M . Truong, K. Coulter, and D.W. Goodman, J. Am. Chem. Soc. 114 (1992)7565. [192] K. J. Borve and L.G.M. Pettersson, J. Phys. Chem. 95 (1991) 7401. [193] R. Orlando, F. Cora, R. Millini, G. Perego, and R. Dovesi, J. Chem. Phys. 105 (1996) 8937. [194] L. Ackermann, J.D. Gale, and C.R.A. Catlow, J. Phys. Chem. B 101 (1997) 10028. [195] M. A. Johnson, E. Stefanovich, and T.N. Truong, J. Phys. Chem. B 101 (1997) 3196. [196] E. Garrone, A. Zecchina and F. Stone, Philos. Mag. B 42 (1980) 683. [197] M. Iwamoto and J. H. Lunsford, Chem. Phys. Lett. 66 (1979) 48. [198] M. Che and A. J. Tench, Adv. Catal. 23 (1983) 1. [199] N. B. Williamson, J. H. Lunsford, and C. Naccache, Chem. Phys. Lett. 9 (1971) 33. [200] L. N. Kantorovich, A. L. Shluger, P. V. Sushko, J. Gunster, D. W. Goodman, P. Stracke, and V. Kempter, Faraday Discussion 114 (1999) 173. [201] D. Wolf, Phys. Rev. Lett. 68 (1992) 3315. [202] C. A. Ventrice, T. Bertrams, H. Hannemann, A. Brodde, and H. Neddermayer, Phys. Rev. B 49 (1994) 5773. [203] F. Rohr, K. Wirth, J. Libuda, D. Cappus, M. Baumer, and H. J. Freund, Surf. Sci. Lett. 315 (1994) L977. [204] M. Schonnenbeck, D. Cappus, J. Klinkmann, H. J. Freund, L. G. M. Petterson, and P. S. Bagus, Surf. Sci. 347 (1996) 337. [205] T. Matsumoto, J. Kubota, J. N. Kondo, C. Hirose, and K. Domen, Langmuir 15 (1999) 2158. [206] R. Plass, J. Feller, M. Gajdardziska-Josifovska, Surf. Sci. 414 (1998) 26. [207] R. Plass, K. Egan, C. Collazo-Davila, D. Grozea, E. Landree, L. D. Marks, and M. Gajdardziska-Josifovska, Phys. Rev. Lett. 81 (1998) 4891. [208] A. B. Kunz and M. P. Guse, Chem. Phys. Lett. 45 (1977) 18. [209] R. Huzimura, Y. Yanagisawa, K. Matsumura, and S. Yamabe, Phys. Rev. B 41 (1990) 3786. [210] Y. Yanagisawa, K. Takaoka, and S. Yamabe, J, Chem. Soc. Faraday Trans. 90 (1994) 2561.

136

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

Chapter 4

Theory of physical and chemical behavior of transition metal oxides: vanadium and molybdenum oxides K. Hermann" and M. W^itko*' ^Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg 4-6, 14 195 Berlin, Germany. ^Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences, ul. Niezapominajek, 30239 Cracow, Poland.

1. INTRODUCTION Transition metal oxides are well known for their large variety of physical and chemical properties [1-3]. Many of these materials undergo phase transitions with interesting structural, electronic and magnetic behavior [3]. Some exhibit high temperature superconductivity and exciting optical properties or high catalytic activity [2]. Among these, vanadium and molybdenum oxides represent an important class of materials due to their large variety in crystal structures and physical/chemical properties. They cover a wide range of electronic properties from metals to semiconductors and insulators and are, therefore, used in many technological applications. Examples are vanadium oxide based electrical and optical switching devices, write-erase media, light detectors, critical temperature sensors, infrared spatial light modulators [4-9], and even vanadia constituents of surfaces of medical Ti-Al-V implants [10]. Further, molybdenum trioxide forms n- and p-type semiconductor phases [11] which are of technological interest. The rich and diverse chemistry of these oxides results mainly from two interrelated factors. First, the metal ions can adopt different formal oxidation states in the oxides, ranging from +2 to +5 in the case of vanadium and from +4 to +6 in the case of molybdenum. Second, the ions can exhibit quite different

137

coordination geometries described by octahedra, pentagonal bipyramids, square pyramids, or tetrahedra, where the different local units share comers and/or edges and/or faces. Many of these oxides are described by a defined composition (single or mixed valency oxides) while others exhibit a wide range of compositions. Vanadium and molybdenum oxides are also extremely interesting from a chemical point of view because of their wide exploitation in many catalytic processes of industrial relevance. Among several factors responsible for the catalytic behavior we mention the mobility of surface/lattice oxygen atoms, the existence of Lewis acid-base sites as well as different reactivity of different crystal faces [2, 12-15]. The versatility of the properties of both oxides, when combined with other elements such as bismuth, cobalt, aluminum, or alkali metals, leads to their use as active and selective catalysts in many reactions belonging to redox or acid-base processes. Redox processes include reactions where oxygen is involved as well as those where only hydrogen participates. Examples of the former group are selective oxidation, oxidation of different functional groups, oxidative condensation or dehydrogenation or dehydrocondensation, oxyhydration, or epoxidation. Examples of the latter are hydrogenation, hydrogenolysis and reduction. Among acid-base reactions the most important are isomerization (both structural isomerization and ring contraction), disproportionation (metathesis), decomposition (dehydration, dehydrogenation, dehydrocyclization, dehydrocondensation) and etherification. Vanadia-based compounds are used as components of various catalysts in mild oxidation, ammoxidation, and dehydrogenation of hydrocarbons and other organic compounds [16]. They are also efficient catalysts in oxidation of S02to SO3, naphthalene or oxylene to phthalic anhydride and more recently n-butane to maleic anhydride [16]. Vanadia-based catalysts seem also promising for the oxidation of toluene to benzaldehyde, methanol to formaldehyde and to methyl formate, as well as for the removal of NOx by selective reduction with NH3 [4, 17]. Molybdenum oxides are used, either as pure materials or in combination with other elements (Bi, V, Co, Al) as highly active and selective catalysts for many reactions of very different type. Examples of commercial processes are isomerization, polymerization, and production of formaldehyde and acronitrile [2, 12-13]. In particular, molybdenum trioxide, M0O3, serves as a selective catalyst for the partial oxidation of hydrocarbons and alcohols [18-32]. Its catalytic behavior results from several interrelated factors, where the valence of the Mo ions together with their local environment and the type of exposed crystal plane are the most important. Each of the differently oriented M0O3 surfaces contains different active oxygen sites that participate in various elementary steps of catalytic reactions. The role of these surface sites has become the subject of lively discussion in recent years, see Ref. [12] and references therein. For

138

example, in selective oxidation of propene to acroleine [19-27] two subsequent conversion steps proceed on different crystal planes of M0O3 [22]. Hydrogen abstraction takes place at the basal (001) or (100) crystal planes whereas nucleophilic addition of oxygen occurs at the basal (010) plane [25, 26]. In both steps, surface oxygen sites form Lewis centers with strong basic character [2, 33] leading to surface hydroxyl or water species and to the oxygenated precursor of aldehyde / ketone [33]. In typical catalytic reactions involving vanadium or molybdenum oxides (as well as oxides of other transition metal elements of groups V, VI, and VII) the oxide surfaces undergo reduction and oxygen vacancies are formed [1-3, 12, 14, 34]. When the concentration of these point defects surpasses a certain critical value at the surface they propagate into the substrate bulk. This may give rise to crystallographic shear (CS) planes [1, 14] where the linkage between polyhedra is changed towards higher connectivity, e.g. from comer- to edgesharing, which results in effect in annihilation of vacancies. The CS planes constitute extended two-dimensional defects in the bulk. As a consequence, the crystal can be considered as being composed of slabs of comer-sharing polyhedra of more or less undistorted parent stmcture, separated by CS planes. If the spacing between the CS planes is regular, the parent stmcture is transformed into a mixed valency oxide. Depending on the actual spacing different intermediate oxides may be formed which are usually described as homologous series. As an example we mention the series McnOsn-p appearing in ReOs like stmctures. Here, n is the width of the slab of parent stmcture as given by the number of coordination octahedra in unbroken rows between CS planes, and p measures the number of anionic sites eliminated by the crystallographic shear. Despite the enormous importance of vanadium and molybdenum oxides as catalysts many details of their microscopic behavior are still under debate [4, 12, 16]. This is due to the large structural complexity of these materials as well as to their diverse electronic behavior that makes quantitative studies, both experimental and theoretical, rather challenging. In the present chapter, which deals with theoretical concepts applied to vanadium and molybdenum oxide surfaces, we will restrict the discussion to binary oxide systems. So far, mixed metal oxide systems have not been studied by quantitative theory. Theoretical methods that have been used to study oxide surfaces can be classified according to the approximations made in the system geometry where two different concepts are applied at present, local cluster and repeated slab models. Local cluster models are based on the assumption that the physical/chemical behavior at selected surface sites can be described by finite sections cut out from the oxide surface. These sections (surface clusters) are treated as fictitious molecules with or without additional boundary conditions to take the effect of environmental coupling into account. Therefore, their electro-

139

nic structure can be calculated by modem quantum chemical methods. Here methods using various ab initio density functional theory approaches [35] have proven to be particularly successful in recent work. Repeated slab models [36] rely on an exact two-dimensionally periodic arrangement of all atoms and molecules at the oxide surface. Within this approach, a slab of full translational periodicity and finite thickness (surface slab) describes the surface system. For computational convenience surface slabs are repeated perpendicular to their surface with vacuum separating adjacent slabs. This yields an altogether threedimensionally periodic system with a large supercell which can be studied by modem bulk methods of solid state theory. Here ab initio density functional theory methods, such as the full potential linear augmented plane wave (FPLAPW) method [37-39] have been applied exclusively. In the following, we discuss recent concepts and theoretical results concerning microscopic properties of vanadium and molybdenum oxide surfaces. While the two elements form different classes of oxides their surfaces exhibit numerous stmctural and electronic similarities, such as microscopic surface binding, adsorption, or oxygen vacancies, which we will point out accordingly.

2. VANADIUM OXIDES Binary oxides of vanadium, VxOy, form a large family of materials with different stoichiometry where vanadium ions differ by their formal valence charge as well as by their local coordination. Single valency oxides, where all vanadium ions occur with the same valence charge, include four oxides, VO, V2O3, VO2, and V2O5. The monoxide VO appears as a cubic (rocksalt) lattice [40, 41] and vanadium appears with a formal charge +2. The sesquioxide V2O3 forms an antiferromagnetic monoclinic phase [42] below 150 - 170K and a paramagnetic trigonal comndum phase [43-45] yielding in both phases a formal charge +3 for vanadium. The dioxide VO2 crystallizes in a monoclinic (distorted rutile) phase [46-50] below 337K and in a tetragonal mtile phase [47, 51-55] resulting always in a formal charge +4 for vanadium. In the pentoxide V2O5, which forms an orthorhombic layer-type lattice [56-58], vanadium assumes its highest formal valence charge, +5. In addition to single valency oxides there are many phases of intermediate composition where vanadium ions of different and even fractional valence charge occur. These mixed valency oxides may have well defined composition and bulk crystal structure. Examples are Magneli phases, which form homologous series Vn02n-i or Vn02n+i and have been associated with shear plane formation [59, 60]. Single crystals of Vn02n-i have been observed for 2 < n < 9. In these systems the average valence charge of vanadium is +(4-2/n), ranging

140

between +3 and +4, which is described by ion ratios V'^/V^"^ = (n-2)/2. Apart from the single valency oxide V2O3 (n = 2) examples are monoclinicVsOs [61, 62], triclinic V4O7 [63-65], V5O9 [66, 67], V6O11 [68], V7O13 [68], and V9O17 [69]. Single crystals of Vn02n+i oxides have been observed for 2 < n < 6. Here the average valence charge of vanadium is +(4+2/n), ranging between +4 and +5, which is described by ion ratios V'^/V^'^ = (n-2)/2. Apart from the single valency oxide V2O5 (n = 2) examples are monoclinicVsOv [70], orthorhombic V4O9 [71], and YeOu where a monoclinic low temperature phase [72-74] and an orthorhombic phase at high temperatures [75] is observed. Further, single crystals of monoclinic V12O26 [76] and of monoclinic V14O6 [77] have been identified in the experiment. Point defects in vanadium oxides, such as oxygen vacancies, can order themselves and give rise to superstructures or complexes. This has been observed in several oxides (VO with about 20% oxygen vacancies being a good example [1]). Further, superstructures, as e. g. V52O64 or V244O320 [1], are formed under oxygen-rich conditions where the oxygen sublattice is almost complete. However, some vanadium ions occupy tetrahedral sites where each tetrahedrally coordinated cation has four vacant octahedral sites as nearest neighbors and the cluster is topologically similar to the rocksalt structure. In addition to bulk material, vanadium oxide cluster ions, such as V20(4.6/, V30(6V. V40(8-ii)\ VsOdi.n/, V60(i3-i5)\ and YjO^.e-m" [78, 79], can be prepared in gas phase. The chemical reactivity of these species shows a distinct dependence on cluster size. While the smaller clusters are reacting quite easily with other molecules, the reactivity of the larger systems is decreased with the exception of oxygen-rich clusters, which can release molecular oxygen upon collision with reactant gas. All of the vanadium oxides described above are interesting in their physical/chemical properties and may be used in many applications. However, the following discussion will be restricted to single valency oxides that have been dealt with in quantitative theoretical studies. 2.1. Vanadium oxide bulk systems 2.1,1. Geometric structure Single crystals of vanadium oxide can be classified in their geometric structure, apart from their lattice definition by lattice and lattice basis vectors, by the occurrence of specific elementary VOx building units. All single valency oxides are characterized by a common octahedral V06 unit. The different crystal lattices differ by the distortion of the octahedral units (as a result of actual V-0 distances and 0-V-O angles) as well as by the connectivity of the units in the

141

crystal (linking of comers, edges, or faces). This becomes evident in the ballsand-sticks models of the four oxides, VO, V2O3, VO2, V2O5, shown in Fig. 1. In the monoxide VO, which forms a cubic rocksalt lattice [40, 41] with two atoms (1 V + 1 O) in the crystal unit cell, the VOg octahedron is regular with all distances dv-o = 2.06 A and angles Z(OVO) = 90°, see Fig. la. Here all adjacent VOe units are connected by their edges where oxygen centers share six octahedra each and are six-fold coordinated with respect to their vanadium neighbors. The dioxide VO2 exits as a tetragonal rutile crystal above 337K [47, 51-55J with six atoms (2 V + 4 O) in the crystal unit cell. This structure is distorted slightly by shear, becoming monoclinic below 337K [46-50, 80] and doubling its crystal unit cell (twelve atoms, 4 V -h 8 O). In the tetragonal lattice the regular V06 octahedron is elongated sightly along its 4-fold axis with four (two) distances dv-o = 1-90 A (1.95 A) and all angles Z(OVO) = 90°. In the monochnic lattice the VOg octahedron is distorted yielding three distances dy-o each in the range 1.76 - 1.87 A and 2.01 - 2.05 A, respectively and angles Z(OVO) varying between 78° and 99°. In both crystal lattices adjacent VOg units are

Fig 1 Geometric structure of (a) cubic bulk VO ((100) netplane stacking), (b) monochmc bulk VO2 ((Oil) stacking), (c) trigonal bulk V2O3 ((0001) stacking), and (d) orthorhombic bulk V2O5 ((010) stacking). Vanadium and oxygen centers are shown as light large and dark small balls, respectively. Octahedral VOg and bipyramidal VzOg units are indicated by black lines.

142

connected by their edges or comers, see Fig. lb. The oxygen centers share three octahedra each and are three-fold coordinated with respect to their vanadium neighbors with distances varying by 3% (tetragonal) and 10% (monoclinic) respectively. The sesquioxide V2O3 crystallizes as a monoclinic phase [42] below 150 170K and as a trigonal corundum phase [43-45] above this temperature with ten atoms (4 V + 6 O) in the crystal unit cell. In the trigonal lattice the VOe octahedron is distorted with three distances dv-o each at 1.96 A and 2.06 A, respectively and angles Z(OVO) varying between 82° and 97°. Here adjacent VOe units are connected by their edges or comers, see Fig. Ic. The oxygen centers share four octahedra each and are four-fold coordinated with respect to their vanadium neighbors with distances varying by 5% in the trigonal lattice. The pentoxide V2O5 forms an orthorhombic layer-type lattice [56-58] with 14 atoms (4 V + 10 O) in the crystal unit cell. In this crystal lattice the WOe octahedra are distorted substantially with one small distance dy-o = 1.59 A and one very large value, dy-o = 2.79 A, while the other V-0 distances range beo

o

tween 1.78 A and 2.02 A. Angles Z(OVO) involving the vanadyl oxygen at short distance amount to 105° while those including the farthest oxygen lie between 73° and 77°. As in V2O3, adjacent V06 units are connected by edges or comers. However, in V2O5 three oxygen centers share three and two octahedra, respectively. This results in oxygen centers being singly, two-, and three-fold coordinated. The large distortion of the octahedra actually suggests the use of V2O8 bipyramids as building units to describe the lattice stmcture which reflects the layer-type lattice more appropriately. The oxide layers of bulk V2O5 shown in Fig. Id extend parallel to the (010) netplanes of the lattice. This experimental stmcture of bulk V2O5 [56-58] has been verified recently in theoretical total energy [81-83] as well as molecular dynamics studies [84, 85]. Note that depending on the choice of the orthorhombic crystal axes the layer net plane orientation may also be denoted by (001). The latter corresponds to an interchange of lattice constants b and c as proposed in Ref. [86]. In this chapter we have adopted an altemative nomenclature used e.g. in Ref. [87]. 2.7.2. Electronic properties The electronic stmcture of bulk vanadium oxides is determined to a major extent by the amount of d electron occupation in the vanadium ions. In the ideal three-dimensional periodic bulk, electrons are described by Bloch states with energy dispersions reflected in band stmctures and corresponding densities of states (DOS). These quantities can be calculated with high accuracy by modem band stmcture and total energy methods based on the density functional theory (DFT) method.

143

The vanadium atom is described in its ground state by an electron configuration [Ar] 3d^ 4s^. Therefore, vanadium cations V^^ for 2 < p < 5 are given by a configuration [Ar] 3d^"^ resulting in an empty (p = 5) or partially filled 3d shell (p < 5). This result of the free ions is also reflected in the electronic ground states of the bulk oxides where vanadium possesses a formal charge +p. As a consequence, the amount of occupied V 3d electron states is expected to increase in going from V2O5 (corresponding formally to V^^ ions) to VO (formally V^"^ ions). This effect becomes obvious in results from recent band structure studies on the single valency oxides of vanadium discussed in the following. The electronic structure of bulk V2O5 has been studied extensively by both semi-empirical and ab initio techniques. This includes periodic Hartree-Fock (HF) pseudopotential calculations [88], tight binding studies within a perturbative approach [89], non-empirical atomic orbital method studies [90], ab initio orthogonalized linear combination of atomic orbitals (OLCAO) work [91], and self-consistent augmented spherical wave (ASW) [82, 83] as well as ab initio full-potential linear augmented plane wave (FP-LAPW) calculations [81]. These studies yield overall good agreement with the measured photoemission and optical properties. As an example, Fig. 2 shows the band structure and total density of states (DOS) of geometry optimized bulk V2O5 obtained by ab initio FPLAPW calculations [81]. The electron states with negative energies between 0.0 and -5.5 eV describe the occupied valence band region, characterized by dominant oxygen 2p and rather little vanadium contributions. The integration of the DOS over this valence band region yields close to 60 electrons per unit cell (containing 10 oxygen ions) which is consistent with a 2p^ valence configuration of ionic oxygen O^". The conduction band region starts at about 2 eV above the highest occupied electron level (defined by 8 = 0 in Fig. 2) and is determined dominantly by (empty) vanadium 3d type electron states. This indicates a 3d^ valence configuration of ionic V^"^ and confirms altogether the ionic nature of a (V^"^)2(0^")5 compound suggested by formal valence charges. However, a detailed binding analysis based on the electronic density and wavefunctions [81] shows clearly that the actual amount of atom charging is smaller than the formal valence charges indicate. As a consequence, V-O binding in bulk V2O5 is determined not exclusively by ionic coupling but contains also sizeable covalent contributions. This will be discussed in more detail in Sec. 2.2.2. The energy region of the calculated V2O5 valence bands is found to have a width of 5.5 ± 0.5 eV, cp. Fig. 2. This matches very well with values from angle-integrated [92] as well as angle-resolved ultra-violet photoemission spec-

144 6.04.0\^

2.0-

£

.0-

z

LU

-2.0-4.0rx -6.0-

uz

r

u R

Y

r

0.0

15.0

k point path

orthorhombic BZ

Fig. 2. Band structure and total DOS of bulk V2O5. The energy bands are shown for characteristic paths connecting high symmetry points of the irreducible part of the orthorhombic Brillouin zone (BZ) which is included at the bottom. All energies e(k) are taken with respect to that of the highest occupied state. The DOS is given in states per unit volume and per eV.

-7.0

-5.0

-3.0

1.0

Energy [eV]

Fig. 3. Angle resolved He-II ultraviolet photoemission spectrum of a ViOsCOlO) surface sample taken at normal incidence from Refs. [93, 94].

145

troscopy (ARUPS) [93, 94] experiments on V2O5. As an example, Fig. 3 shows recent ARUPS data for single crystal V2O5 using He-II light at normal incidence with respect to the (010) oriented surface [93, 94]. The emissionspectrum has an overall width of 6 eV and can be described by a superposition of mainly three peaks. A major central peak lies at about -3.2 eV below the onset of the spectrum (taken as the energy zero in Fig. 3) and two satellite peaks are found at -1.2 and -4.3 eV, respectively. The nature of these peaks is connected with the different coordination of the oxygen ions in V2O5 as will be discussed in detail in Sec. 2.2.2. From photodesorption experiments V2O5 is known to be a semiconductor with a visible band gap of Eg = 2.35 eV [95] at low temperatures. This is confirmed by values obtained from other measurements, for example, optical absorption (Eg = 2.30 eV) [96] and optical reflectance measurements (Eg = 2.38 eV) [97]. The FP-LAPW calculations [81] yield a direct band gap of 2.3 eV at F and an indirect gap (corresponding to a transition R to F, see Fig. 2) of 1.9 eV which seems to be in good agreement with the experimental gap values obtained for V2O5. However, this agreement, found also in other theoretical studies [82, 83, 98], is considered fortuitous. Bandstructure methods, using DFT schemes within the local density approximation (applied in the theoretical studies), are not expected to describe electronic excitations to a high accuracy [99]. There are two narrow split-off conduction bands at about 2 eV above the highest occupied electron level and about 0.5 eV below the upper conduction band range which have been discussed in the literature [83, 91, 100]. These bands are characterized by rather small dispersion widths corresponding to large effective electron masses and localized band states. The latter are described by dominant V 3dxy character (t2g symmetry) with small O 2px,y admixing. The presence of these split-off conduction bands hints at possible excitations to localized electron states and is of interest for conductivity as well as optical absorption experiments on bulk V2O5 [83, 91, 100]. The electronic structure of bulk VO2 has been examined in several theoretical studies [101-105] in connection with a structural phase transition at 337K (monoclinic to tetragonal rutile). This is accompanied by a semiconductor-tometal transition. As an example. Fig. 4 shows the total densities of states (DOS) of bulk VO2 for the monoclinic (left diagram) and tetragonal rutile (right diagram) lattice geometry obtained by ab initio FP-LAPW calculations [105]. Here the experimentally known geometries of the two bulk phases have been used, see Sec. 2.1.1. The electron states with negative energies (between -1.5 and -7.5 eV in the monoclinic and between -1.9 and -8.0 eV in the tetragonal rutile geometry) represent an occupied valence band region described by dominant oxygen 2p and almost no vanadium contributions. This is completely analogous to the valence band region found for V2O5 and reflects a 2p^ valence

146 10.0

5.0

10.0

DOS [a.u.:

15.0

5.0

DOS [a.u.]

10.0

Fig. 4. Total DOS of bulk VO2 for the monoclinic (left diagram) and tetragonal rutile (right diagram) lattice geometry. The energy zero coincides with the energy of the highest occupied state. The DOSs are given in states per unit volume and per eV.

configuration of ionic oxygen O^'. However, in contrast to the V2O5 system, in VO2 the energy region about the highest occupied electron levels (e « 0) is characterized by (occupied and empty) vanadium 3d type electron states. This is consistent with a 3d^ valence configuration of ionic V"^ and with an overall (V^^)(0^")2 compound suggested by formal valence charges. The DOS curves show subtle differences between the two different phases near (8 = 0). The rutile phase shows a continuous increase of the DOS from 8 = -0.5 eV with increasing energy that evidences a metallic state. In contrast, the monoclinic phase yields finite DOS values above 8 = -0.4 eV with a 0.1 eV wide deep minimum just above 8 = 0 indicating a small gap semiconductor. These findings can give a qualitative picture of the electronic consequences (semiconductor-to-metal transition) of the structural phase transition observed for bulk VO2. However, the actual size of the semiconductor band gap, 0.6 eV found in the experiment [92], cannot be reproduced by the band structure calculations. This has been explained by limitations of the local density approximation used in the theoretical treatment [104]. Further, the band structure calculations cannot uniquely identify the electronic origin of the first order phase transition for bulk VO2 nor the nature of the monoclinic phase (Peierls vs. Mott-Hubbard insulator) which is still under debate [104]. The electronic structure of bulk V2O3 has been studied by different theoretical methods [106-109]. The main goal was to explain the physics of the Mott transition at Tc= 150 - 170K where the material transforms from an antiferromagnetic insulating phase with monoclinic structure below Tc to a paramagnetic

147 10.0

5.0 10.0 DOS [a.u.]

15.0

Fig. 5. Total DOS of bulk V2O3 for the trigonal corundum lattice geometry. The energy zero coincides with the energy of the highest occupied state. The DOS is given in states per unit volume and per eV.

metallic phase with trigonal corundum structure. While the monoclinic low temperature phase is quite involved and not fully understood in its electronic properties, the paramagnetic metallic phase seems quite clear and can be accounted for by band structure calculations. As an example, Fig. 5 shows the total DOS curve of bulk V2O3 for the trigonal corundum lattice geometry obtained by ab initio FP-LAPW calculations [105] where the experimentally known geometry has been used, see Sec. 2.1.1. The characterization of the two main band regions shown in Fig. 5 is completely analogous to what has been found for the V2O5 and VO2 bulk systems and differs only in the relative position of the highest occupied electron levels. Electron states with negative energies between -3.5 and -8.3 eV define an occupied valence band region determined by dominant oxygen 2p and negligible vanadium contributions, thus, reflecting a 2p^ valence configuration of ionic oxygen O^". In addition, there is a valence band region between -1.6 and 0.0 eV, which originates from occupied vanadium 3d type electron states. This region is larger than that of bulk VO2 and is consistent with a 3d^ valence configuration of ionic V^^ and with an overall (V^"')2(0^-)3 compound suggested by formal valence charges. The electronic structure of bulk VO has been calculated by different band structure methods [110-114] and using correlated electron procedures [115]. This Mott-Hubbard metal, which forms a rocksalt type lattice, is the simplest of all single valence oxides of vanadium and has been treated theoretically already a long time ago. As an example, Neckel et al. [114] have published results from self-consistent APW calculations for the experimentally known lattice geometry

148

of VO, which continue the trends in the electronic parameters found for the previously discussed vanadium oxides. The VO bandstructure reveals a valence band region between 10.3 and 5.7 eV below the Fermi level described by oxygen 2p contributions. In addition, there is another valence band region between 4.1 eV below the Fermi level and the Fermi level where the band states are vanadium 3d type. The region is larger than that of bulk VO2 and V2O3 reflecting a larger V 3d occupation. This is confirmed by data from charge analyses which yield a V 3d contribution of 2.8 electrons in the oxide, somewhat larger than suggested by a (V^'^)(0^') configuration from formal valence charges.

2.2. Vanadium oxide surfaces 2.2.1. Geometric structure So far, theoretical studies on vanadium oxide surfaces have focused exclusively on single crystal surfaces which are described by low Miller indices and are believed to be energetically favorable. These surfaces are most easily accessible by theory due to their relatively simple geometry although their relevance as to catalytic activity has been doubted. Single crystal surfaces of the cubic monoxide VO, see Fig. 6, include the

Fig. 6. Geometric structure of the (100), (111), and (110) oriented single crystal surfaces of cubic VO. Vanadium and oxygen centers are shown as light large and dark small balls, respectively, with connecting sticks indicating nearest neighbor relations.

149

non-polar (100) surface described by planes containing square lattices of both V and O ions in a checkerboard arrangement. The polar (111) surface is determined by alternating planes of either V or O ions only in hexagonal lattice positions where the surface can be V or O terminated. Further, the open (110) surface can be characterized by planes with rectangular lattices containing chains of alternating V and O ions. So far, no theoretical studies involving VO surfaces have been published. The only single crystal surface of the trigonal corundum phase of the sesquioxide V2O3, which has been examined so far in preliminary cluster studies, is the (0001) surface, see Fig. 7. This surface is described by alternating rows of two hexagonal V planes (A, A') very close together and quasi-hexagonal O planes (B, whose atom density is three times as large as that of the V planes) in a stacking sequence ...BAA'BAA'B... This allows three different ideal surface terminations, ...BAA', ...AA'B, and ...A'BA where the latter reflects the ion arrangement with the smallest average Madelung potential at the surface. This corresponds to a surface arrangement of one hexagonal V plane on top of a quasi-hexagonal O plane (containing 3-fold coordinated oxygen 0(3)) as shown in Fig. 7. This surface geometry is very similar to that found in experimental LEED studies on the (0001) surface of trigonal Cr203 where in addition major relaxations of the top layers along the surface normal have been observed [116].

0(3)

Fig. 7. Geometric structure of the (0001) oriented single crystal surface of trigonal V2O3. for a surface termination A'BA. Vanadium and oxygen centers are shown as light large and dark small balls, respectively, with connecting sticks indicating nearest neighbor relations.

150

0(3)

Fig. 8. Geometric structure of the (Oil) oriented single crystal surface of the monoclinic (distorted rutile) phase of the VO2 for a surface termination A'BA. Vanadium and oxygen centers are shown as light large and dark small balls, respectively, with connecting sticks indicating nearest neighbor relations.

Theoretical studies on single crystal surfaces of the monoclinic (distorted rutile) phase of the dioxide VO2 have been restricted to the (Oil) surface (in the crystal definition according to Andersson [80]), see Fig. 8. This surface is described approximately by alternating rows of two quasi-rectangular O planes (A, A') and one VO plane (B, whose atom density is twice as large as that of the O planes) in a stacking sequence ...BAA'BAA'B... This allows three different ideal surface terminations, ...BAA', ...AA'B, and ...A'BA where the latter, see Fig. 8, is expected to be the energetically lowest in analogy with the surface structure of rutile type Ti02(l 10) [117]. This leads to 5- and 6-fold coordinated vanadium centers, V(5), V(6), and two- as well as three-fold coordinated oxygen, 0(2) and 0(3), respectively, at the ideal bulk terminated VO2(011) surface as indicated in Fig. 8. The (Oil) surface of monoclinic VO2 is, in its structure, very close to the (110) surface of tetragonal (rutile) VO2 with identical local surface elements such as rows of bridging oxygen and 5- as well as 6-fold coordinated vanadium ions. However, no theoretical studies on the high temperature rutile phase have been published so far. In the pentoxide V2O5, forming a rather open orthorhombic layer-type lattice, the only single crystal surface which has been examined in detail is the (010) surface, see Fig. 9. (It should be noted again that in this chapter the definition of Miller indices is based on orthorhombic crystal axes as introduced in Ref. [87], see also Sec. 2.1.1.). This surface, which represents the preferred cleavage plane of the crystal, is determined in its structure by local V-0 binding

151

0(1)

0(2)

0(2')

Fig. 9. Geometric structure of the (010) oriented single crystal surface of orthorhombic V2O5. Vanadium and oxygen centers are shown as light large and dark small balls, respectively, with connecting sticks indicating nearest neighbor relations. Differently coordinated surface oxygen centers, 0 ( l - 3 , 2',3'), are labeled accordingly.

rather than by stacking of the crystal planes (discussed above for the compact oxides) resulting in an altogether rough surface shape. Here six planar atom layers (4 oxygen, 2 vanadium layers) form a crystal slab which is repeated along to the (010) netplane direction. These crystal slabs can also be described by periodic arrangements of edge and comer sharing VO5 pyramids sticking out at both sides of the slab as discussed in Sec. 2.1.1. In each V2O5 slab there are three structurally different oxygen centers, terminal (vanadyl) oxygen, 0(1), coordinated to one vanadium atom through a short bond and bridging oxygen, 0(2) / 0(3), coordinated to two or three vanadium atoms. This leads to five different oxygen centers at the ideal bulk terminated ¥265(010) surface as indicated in Fig. 9. Terminal oxygen centers 0(1) are located above vanadium centers. Further, oxygen centers 0(2), 0(2') bridge two vanadyl groups pointing into the bulk and sticking out of the surface respectively, while oxygen centers 0(3), 0(3') are connected to three vanadyl groups pointing into different directions. In addition, the surface exposes bare vanadium centers connected to 0(2) andO(3, 3') centers.

152

2.2.2, Physical and chemical properties Theoretical studies on various physical and chemical parameters of vanadium oxide surfaces have been performed using both repeated slab as well as local cluster models. However, the vast majority of studies has focused on the (010) surface of the pentoxide, V2O5. This is a result of rather little experimental information on details of VxOy surface systems other than ¥205(010) that has attracted great interest due to its possible importance in catalytic applications. Further, most of the theoretical work has been based on cluster type studies where local surface behavior, in particular near surface oxygen sites, is discussed in detail. This will be reflected in the following discussion.

(a) Properties of clean surfaces The electronic structure and binding at vanadium oxide surfaces can be quantified within the DFT framework by analyses of the Kohn-Sham orbitals and corresponding electron densities resulting from periodic slab or local cluster calculations. In particular, cluster studies offer all traditional quantum mechanical tools, such as population analyses [118] or bond orders [119, 120], to examine inter-atomic binding at the surface. As an example, Table 1, lists atom charges from Mulliken populations and Mayer bond orders of substrate clusters V10O31H12 and V20O62H24, modeling one and two layers of the ¥265(010) surface, see Fig. 10. The data are obtained from DFT calculations [121-123] using the revised Perdue-Burke-Emzerhof (RPBE) functional [124, 125]. A compariTable 1 Atom charges q and bond orders p of substrate clusters V10O31H12 and V20O62H24, modeling the ¥205(010) surface, see text. All data refer to V, O surface species near the cluster center. The differently coordinated oxygen species 0(1), 0(2), 0(3) are indicated in Fig. 9. The two entries of q(V) for V20O62H24 correspond to vanadium of the first and second layer respectively. All values are given in atomic units. V10O31H12

V20O62H24

q(V)

1.59

1.66,1.78

q(0(l))

-0.34

-0.34

q(0(2))

-0.70

-0.70

q(0(3))

-0.87

-0.88

P(0(1)-V)

2.04

2.05

P(0(2)-V)

0.82

0.82

P(0(3)-V)

0.49

0.44

153

Fig. 10. Geometric structure of the clusters V10O31H12 (a, one layer) and V20O62H24 (b, two layers) representing a section at the ¥205(010) surface. The V (O) atoms are shown as large (small) shaded balls while very small balls refer to hydrogen atoms used to saturate oxygen atoms at the cluster periphery. For a definition of the differently coordinated surface oxygen centers, 0(1) to 0(3), see Fig. 9.

son of the calculated atom charges and bond orders shows very close similarity between the one- and two-layer clusters, which indicates that electronic interlayer binding is rather weak confirming the experimental result of a layer-type substrate material. All vanadium centers are positively charged and oxygen centers are negative. Vanadium ions are described as V^'^"^ to V^'^"^ where the variation reflects the location inside the cluster. Further, the negative oxygen charges scale with coordination, O^"^' for singly coordinated oxygen 0(1), O^"^' for doubly coordinated oxygen 0(2), O^'^" for triply coordinated oxygen 0(3). This indicates for the ¥205(0x0) surface that bridging oxygen sites are more nucleophilic than terminal vanadyl sites, which becomes important in view of the reactivity of the different sites with respect to surface chemical reactions. Altogether, the cluster data yield charging of the different cluster ions which is much smaller than formal valence charges, V^"^ and O^", would suggest. This reflects the well-known fact that MuUiken charges and formal valence charges can yield the same qualitative picture but may not be compared on a quantitative basis due to the absence of a unique definition of atom charges in compound material. In addition, the small MuUiken charges indicate sizeable covalent contributions to interatomic binding inside the surface layers of V2O5(010). This is confirmed by the bond order results of Table 1, which reveal the general picture based on atom coordination. Bonds between terminal oxygen, 0(1), and vanadium yield bond order values close to 2 which suggests double bonds. Bonds between bridging oxygen, 0(2), and each of their two vanadium neighbors result in bond order values close to 1, corresponding to two single bonds per oxygen. Further, V - 0 bond orders involving bridging atoms, 0(3), give values of 0.5 - 0.6 per bond revealing weaker than single bonds.

154

The electronic structure of the surface clusters in the valence region is determined by occupied Kohn-Sham valence orbitals being dominantly O 2sp type with very little V 3d admixture. Their energy range corresponds to valence energy widths A of about 5.5 eV. For comparison, very recent FP-LAPW band structure calculations [81] yield A = 5.35 eV for the V2O5 bulk and A = 5.05 eV for ¥205(010) single layer slabs, which is rather close to the cluster results. This indicates that the two different model concepts, periodic slabs and local clusters, yield the same electronic structure of the ¥205(010) surface. The energetic distribution of the Kohn-Sham valence levels in the clusters can be described by a total density of states ntot(e) (DOS) while its atom decomposition is accounted for by corresponding partial (atom projected) densities of states nA(e) (PDOS) [122, 126]. Fig. 11 shows DOS and PDOS curves for the V10O31H12 cluster, see Fig. 10, where the vanadium contributions as well as those from the differently coordinated oxygen atoms, terminal 0(1) and combined bridging 0(2) + 0(3), are included [126]. The curves have been smoothed by gaussian level broadening (FWHM of 0.4 eV) and the energy region of occupied levels is separated from that of the empty levels (gray background) by a thin vertical line. The total DOS in the energy region between -13 and -7 eV shows a multi-peak structure reflecting the above described dominantly O 2sp derived valence band region, in very good agreement with DOS results from larger clusters as well as from bulk V2O5 and ¥205(010) single layer slabs [81-83, 127]. (Note that the additional peaks below -13 eV correspond to split-off cluster levels arising from 12.0

3

CO

O Q

-16.0

-13.0

-10.0 -7.0 Energy [eV]

-4.0

Fig. 11. Total DOS and atom projected PDOS curves of the V10O31H12 cluster. For further details see text.

155

bond saturation of peripheral oxygen atoms by hydrogen terminator atoms and have to be considered cluster artifacts.) The PDOS curve referring to terminal oxygen 0(1) (dotted line in Fig. 11) is described by an overall confined (~ 3 eV wide) distribution that is concentrated near the center of the valence region with smaller contributions above the center. In contrast, the PDOS's of bridging oxygen 0(2,3) yield a broad distribution covering the full energy range of the valence region. Thus, the 0(1) derived cluster contribution shows a dispersion width smaller than that of bridging 0(2,3) species. This may be explained by the spatial distribution of the different oxygen centers in the crystal and their effective inter-atomic binding [94]. Recent angle resolved UPS experiments on a freshly cleaved ¥265(010) surface sample [93, 94] yield spectra at normal emission which exhibit three major peaks in the O 2sp valence region, see Fig. 3. This is consistent with the overall shape of the calculated total DOS curve of the V10O31H12 cluster shown in Fig. 11 and suggests that the most prominent central peak in the experimental spectrum may be assigned to emission from mainly terminal oxygen, 0(1), while the two peripheral peaks at the top and bottom of the valence band region are characterized as mixtures of vanadium with 0(2) and 0(3) induced emission intensity. This interpretation can help to identify the type of surface oxygen that participates in specific reaction steps in combined reaction and photoemission experiments. Table 2 Atom charges q and bond orders p of substrate cluster V17O42H16 modeling the monoclinic VO2(011) surface, see text. All data refer to V, O surface species near the cluster center. The differently coordinated oxygen species 0(2), 0(3) as well as the 5- and 6-fold coordinated vanadium species V(5), V(6) are indicated in Fig. 12a. The two and three entries in the bond orders reflect the variation between slightly different V-O distances. All values are given in atomic units. V17O42H16

q(V(5))

1.49

q(V(6))

1.57

q(0(2))

-0.46

q(0(3))

-0.71

p(0(2)-V)

0.95, 0.99

p(0(3)-V)

0.55,0.65,0.70

156

Fig. 12. Geometric structure of (a) the V17O42H16 cluster representing a section at the VO2(011) surface and (b) the V11O33H33 cluster representing a section at the V2O3(0001) surface. The V (O) atoms are shown as large (small) shaded balls while very small while balls refer to hydrogen atoms used to saturate peripheral oxygen atoms. The differently coordinated surface oxygen centers, 0(2), 0(3), and the 5- and 6-fold coordinated vanadium centers V(5), V(6) are labeled accordingly.

The electronic structure of the monoclinic VOiCOl 1) surface has been examined in preliminary studies [128] using clusters of different size and shape. As an example, Table 2, lists atom charges from MuUiken populations and Mayer bond orders for the substrate cluster V17O42H16, see Fig. 12a. The data are obtained from DFT calculations [128] using the RPBE functional [124, 125]. These results reveal local charging and V-O binding at the VO2(011) surface which is quite similar to that found for ¥205(010). Positive vanadium ions are described as V^'^"^ (5-fold coordinated) and V^"^"^ (6-fold coordinated) and the negative oxygen charges scale with coordination, O^'^" for bridging oxygen 0(2), O^*^" for bridging oxygen 0(3). As for the ¥205(010) surface clusters, charging of the different cluster ions in the V17O42H16 cluster is found to be much smaller than formal valence charges of the VO2 substrate, V"^ and O^", suggest. This has implications identical to those discussed before. In particular, the data evidence that the ion charges of vanadium differ only little (by 0.1 electrons) between the 5- and 6-fold coordinated surface species. Therefore, discrimination between the two must be based on covalent rather than ionic binding behavior. Furthermore, the vanadium charges determined from poputions in the VO2(011) surface clusters are decreased only slightly compared to those of the ¥205(010) surface clusters. This is in contrast to the standard chemical characterization in terms of formal valence charges +4 and +5 in the two substrates and confirms the importance of covalent binding in the substrates. The bond order results of Table 2 reflect again the general picture based on atom coordination. Bonds between doubly coordinated oxygen, 0(2), and each of their two vanadium neighbors result in bond order values close to 1, sugges-

157

Table 3 Atom charges q and bond orders p of substrate cluster V11O33H33 modeling the trigonal ¥203(0001) surface, see text. All data refer to V, O surface species near the cluster center. The 3-fold coordinated oxygen species 0(2) is indicated in Fig. 12b. The three entries in the bond orders reflect the variation between different V-0 bonds. All values are given in atomic units. V11O33H33

q(V)

0.82

q(0(3))

-0.67

P(0(3)-V)

0.50, 0.58, 0.95

ting two single bonds per oxygen. Further, V-O bond orders involving triply coordinated atoms, 0(3), give values of 0.5 - 0.7 per bond revealing weaker than single bonds. However, the V-O bond orders in the VO2(011) surface clusters are, altogether, somewhat larger than corresponding values found for the ¥205(010) surface clusters, see Table 1. The suggests that covalent binding becomes more important at the expense of ionicity at VOiCOll) that is consistent with V and O charges of the dioxide being smaller than those of the pentoxide. Preliminary cluster studies [129] on the electronic structure of the trigonal ¥203(0001) surface substantiate the above concepts. As an example. Table 3, lists atom charges from MuUiken populations and Mayer bond orders for the substrate cluster V11O33H33, see Fig. 12b. The data are obtained from DFT calculations [129] using the RPBE functional [124, 125]. Here positive vanadium ions closest to the surface are described as V^"^^, and the negative oxygen as O^'^". The vanadium charges determined from populations are the smallest of the three oxide surfaces, which goes along with the formal valence charge +3 of vanadium in V2O3 and emphasizes the increased importance of covalent binding in this substrate. In addition, the V-O bond order results of Table 3 reflect the coordination of the oxygen ions as discussed before.

(b) Surface oxygen vacancies Catalytic properties of vanadium oxide based catalysts depend strongly on their ability to provide surface oxygen as a reactant. Therefore, theoretical studies of physical and chemical (catalytic) properties of the structurally different surface oxygen centers can help to elucidate details of the reaction steps. Moreover, calculations of electronic, structural, and energetic properties of oxygen vacancies created by oxidation reactions are of vital importance as will be discussed in the following.

158

It should be noted that oxygen vacancies at oxide surfaces are of major importance for the physical and chemical behavior of these systems, apart from consequences for catalytic reactivity. As examples we mention point defects at silicon dioxide [130] or the existence of color centers at MgO surfaces which are due to localized electron states at oxygen vacancy sites and have been discussed extensively in the literature [131]. Surface oxygen vacancies have been studied extensively for the ¥205(010) surface by cluster calculations [122-123, 126]. Due to the existence of five structurally non-equivalent oxygen sites, 0(1-3, 2',3'), at ¥205(010), see Fig. 9, this surface may exhibit five different types of surface oxygen vacancies. These can be calculated by removing corresponding oxygen species from the surface clusters and re-evaluating the electronic structure of the resulting vacancy clusters. Here frozen vacancy states are obtained from calculations where all atom positions are kept frozen at the initial cluster geometry. In addition, relaxed vacancy states are determined by geometry optimizations allowing all cluster atoms to rearrange according to lowest total energy. As an illustration, Table 4 lists results from calculations based on a V10O31H12 substrate cluster [126], see Fig. 10, where oxygen has been removed from sites 0(l-3). Oxygen vacancy energies, ED^(0) (frozen geometry) and ED^''^(0) (relaxed geometry), are determined by appropriate cluster total energy differences (based on DFT calculations using the RPBE functional) using the ground state of atomic oxygen as a reference. All vacancy energies are found to be rather large. Frozen values range between 7 and 8 eV and vacancy induced relaxation decreases these energies by only 0.5 to 0.8 eV. This suggests that oxygen is bound very strongly to its substrate environment and may not be removed in one single step during an oxidation reaction. However, pre-adsorbed atoms or molecules may weaken surTable 4 Oxygen vacancy energies, EB^'^\0), for the surface sites 0(1-3) at the ViOsCOlO) surface with and without surface relaxation. All data are obtained for the V10O31H12-O (= V10O30H12) model cluster [126]. In addition, the table contains values of charges q^^'^^^V) and relaxation displacements Ar(V) of the central V atom(s) closest to the vacancy. Energies are given in eV and lengths are in A. Vacancy site

ED'(0)

ED'(O)

q'(V)

q^(V)

Ar(V)

0(1)

7.16

6.47

1.38

1.51

0.50

0(2)

7.95

7.19

1.26

1.36

0.70

0(3)

7.01

6.49

1.28

1.33

0.08

159

face binding of the oxygen and hence may make it available for oxidation processes as will be discussed later on. Figs. 13a-d visualize geometric consequences of substrate relaxation as well as charge transfer due to vacancy formation at the ¥205(010) surface where V10O31H12 based vacancy clusters are used as models of different vacancy sites 0(l-3). All cluster ions are displayed by shaded balls where ball radii represent atom charges as obtained by populations. As a main result, the relaxation effect is always found to be locally confined. If oxygen is removed from a (singly coordinated) terminal 0(1) site, see Fig. 13a, b, relaxation causes the vanadium center below the vacancy site (hatched ball) to move perpendicular to the surface towards the substrate by about 0.5 A. Thus, the vanadium penetrates further into the substrate and can interact with the nearest terminal oxygen of the

Fig. 13. Relaxed geometry of the V10O31H12-O ( = V10O30H12) cluster with (a, b) an 0(1) vacancy, (c) an 0(2) vacancy, and (d) an 0(3) vacancy. The views are perpendicular (plots a, c, d) and parallel (plot b) to the ¥205(010) surface, respectively. Cluster atoms are shown by shaded balls where ball radii represent atom charges. Dark (light) shading refers to negative (positive) charge while the radii give the amount of charge. The central V species next to the 0(1) vacancy is emphasized by hatching in plots a, b. The white balls behind the relaxed cluster describe the system without vacancy.

160

second substrate layer to form a V - O - V bond bridge between the two crystal layers. This effect has been confirmed by calculations using a vacancy cluster V20O62H24-O ( = V20O62H24) modeling two adjacent crystal layers [5, 126]. Here the V - O - V inter-layer bond bridge is found to be very similar to bridge bonding at the 0(2) surface site as confirmed by bond order and charge analyses [132]. Thus, vacancy formation at the first crystal layer of the ¥205(010) surface can increase the electronic coupling with the second layer. This may be the starting point of major surface relaxation and may result eventually in a reconstruction of the whole surface region. If oxygen is removed from a (doubly coordinated) bridging 0(2) site, see Fig. 13c, the strongest relaxation shifts occur for the two vanadium centers adjacent to the vacancy which move laterally by 0.6 A such that the vacancy opening is enlarged. However, the lattice topology of the ¥205(010) surface is conserved suggesting that a single 0(2) vacancy will not introduce major surface restructuring. Finally, the relaxation geometry of the 0(3) vacancy, given in Fig. 13d, can be understood by the same mechanisms discussed for the other vacancies. A closer inspection of the ball radii in Figs. 13a-d reveals that the positive charges of the vanadium ions closest to each oxygen vacancy are smaller than corresponding values of the cluster without the vacancy. This is also clear from a comparison of the charge values q(V) given in Tables 1 and 4 and evidences chemical reduction of the metal sites. A more detailed description of the effect becomes possible by (P)DOS and orbital analyses of the vacancy clusters. As an example. Fig. 14 shows (P)DOS curves obtained for the V10O31H12 based vacan12.0

3 C/)

O Q

-16.0

-13.0

-10.0 -7.0 Energy [eV]

-4.0

Fig. 14. Total DOS and atom projected PDOS curves of the valence band region for the V10O31H12 based vacancy cluster modeling an 0(2) surface vacancy, see text.

161

cy cluster modeling an 0(2) surface vacancy. A comparison of these curves with those of the initial substrate cluster, see Fig. 11, shows clear differences in the energy region of the highest occupied cluster levels. While in the initial cluster the highest levels are at the top of the O 2sp valence region and refer to oxygen type orbitals, those of the vacancy cluster are located at about 2 eV above the O 2sp valence region and are characterized by vanadium 3d orbitals. This shows that the oxygen vacancy leads to increased local vanadium 3d occupation, substantiating the microscopic picture of chemical reduction of the metal sites observed also in photoemission experiments [133]. Extended studies of local electronic states of 0(2) and 0(3) vacancies at the ViOsCOlO) surface [134] based on the V10O31H12 cluster have been carried out in search of possible color center states. These calculations do not yield any low-lying electron states localized at the vacancies and reminiscent of color centers as found at MgO surfaces [131] in accordance with experimental evidence [135]. This can be rationalized by simple geometric arguments [126] based on the size of the surface vacancies at ¥205(010) and by the strength of the electrostatic Madelung potential near the vacancies. Details of possible diffusion of surface oxygen vacancies are of great importance for a microscopic understanding of structural transformations at the V2O5 surface which happen as a consequence of oxidation reactions. Diffusion processes, where oxygen, residing at a surface site, diffuses to a nearby oxygen vacancy site, have been studied in some detail by cluster models [126]. As an illustration, we mention results from V10O31H12 based vacancy clusters with a model path 0(1) -^ 0(2')vac at the ¥205(010) surface. This corresponds either to diffusion of oxygen species from an 0(1) site filling a nearby vacancy at 0(2') or to diffusion of an oxygen vacancy at 0(2') to a nearby 0(1) site. From calculations of total cluster energies for oxygen moving along the path its transition state geometry T and the corresponding diffusion energy barrier EB can be evaluated. Fig. 15 visualizes the 0(1) -^ 0(2')vac diffusion path by small dark balls with the transition state geometry T indicated by a large dark ball. This path is found to yield the smallest barrier, only 0.5 eV, which can be easily released by typical surface reactions. Thus, 0(2^'^) vacancies, created by oxidation reactions at the ¥205(010) surface may be annihilated quite easily by oxygen transfer from terminal 0(1) sites. This result is rather interesting in connection with the identification of reactive surface sites. Depending on the experiment, one may not be able to distinguish between an oxidation reaction step occurring at an 0(1) site and one occurring at an 0(2) site followed by diffusion of oxygen from 0(1) to the vacancy at 0(2). However, more detailed calculations are needed to substantiate these findings.

162

Fig. 15. Diffusion of surface oxygen filling a neighboring vacancy at the ¥205(010) surface. The model path 0(1) -^ 0(2')vac is visualized for a V10O31H12 substrate cluster and connects sites 0(1) (geometry A) with 0(2') (geometry B). Oxygen positions along the path are shown by small black dots with the transition state geometry (T) emphasized by a larger dark ball.

(c) Adsorption Atomic and molecular adsorption at vanadium oxide surfaces have been studied theoretically using both periodic slab and cluster models where so far studies are restricted to the pentoxide, V2O5, as a substrate due to its possible importance in catalytic applications as mentioned before. Further, adsorbate species include in all cases atoms (H [122-123, 126, 136-142], O (see below)) or rather small molecules (O2 (see below), H2O [143-144], NH3 [145-147], NO [146, 148], C2H4 [149], propene (CsHg) [140], toluene (C6H5CH3) [140]) that are of catalytic interest but also small enough to make meaningful calculations feasible. Hydrogen adsorption at the ¥265(010) surface has been studied extensively by both cluster [122-123, 126, 136-140] and slab models [141] due to its catalytic relevance. Under reaction conditions hydrogen will always be present at vanadium oxide surfaces giving rise to surface hydroxyl groups. These groups may desorb from the surface or recombine forming water species that desorbs. Here it is interesting to find out which of the oxygen sites are involved in the adsorption and formation of surface OH or H2O species and which of these species will desorb most readily. These questions have been addressed in previous theoretical DFT studies [122-123, 126, 140] based on a V10O31H12 substrate cluster. Here one or two hydrogen atoms are added to one of the five different surface sites, 0(l-3, 2',3'), and local geometries of the resulting surface OH and H2O groups are optimized in total energy calculations. Fig. 16 visualizes resulting equilibrium geometries for both species and for different

163

OH, H20atO(1)

OH, HgO at 0(3) OH, H2O at 0(2) Fig. 16. Sketch of computed equilibrium geometries of hydrogen adsorption at different oxygen sites of the ViOsCOlO) surface. Single hydrogen adsorption (surface OH) as well as adsorption of two hydrogen atoms (surface H2O) are included. The results are obtained from optimizations of V10O31H12+H and V10O31H12+2H cluster models, respectively. The surface species, OH or H2O, are shown by haded balls while the surface lattice is sketched by light balls.

surface sites. In addition, Table 5 lists energetic results from the calculations. Here EB(H) denotes the binding energy of a single hydrogen adsorbed at a given oxygen site 0(l-3, 2', 3') while EB(H') gives the binding energy of a hydrogen adsorbing at an existing surface OH group to yield surface H2O. Further, EB(H) refers to the combined adsorption energy of two hydrogen species to form surface H2O. All energies are determined from total energy differences of the appropriate clusters where one or two isolated hydrogen atoms in their ground states are taken as reference. The EB(H) values of Table 5 show that hydrogen Table 5 Hydrogen adsorption energies, EB(H), EB(H'), and EB(2H), for the V10O31H12 substrate cluster with H, 2H adsorbed at different oxygen sites 0 ( l - 3 , 2', 3')- For definitions of the adsorption energies see text. All energies are given in eV. Site

EB(H)

EB(H')

EB(2H)

0(1)

2.34

1.56

3.90

0(2)

2.21

0.66

2.87

0(3)

1.88

1.08

2.96

0(2')

0.54

--

--

0(3')

0.76

--

--

164

can adsorb at all oxygen sites. Binding is found to be rather strong at the 0(1), 0(2), and 0(3) sites, with EB(H) values between 1.88 eV and 2.34 eV, which suggests rather stable surface hydroxyl groups at these sites. In contrast, the small EB(H) values found for the 0(2') and 0(3') sites reveal fairly weak adsorptive binding. The latter sites are located close to two vanadyl groups that obviously makes surface OH formation less likely to happen. The equilibrium geometries of surface OH resulting from the calculations, see Fig. 16, do not reflect the symmetry of the corresponding oxygen sites before adsorption. The 0-H distances of all surface groups are quite close to those of typical OH containing systems [150] while the distances between oxygen and its nearest vanadium neighbors at the V2O5 surface are enlarged. Thus, hydrogen adsorption, which results in strong O-H binding, weakens the V-0 bonds near the adsorption site. This is confirmed by bond order analyses [126], which show that for example the V=0 double bond at the 0(1) site is reduced to a single bond and the weaker V-0 bonds at 0(2, 3) are reduced even further. At all oxygen sites, hydrogen adsorption leads to the accumulation of additional negative charge at the corresponding surface oxygen and the surface OH species becomes slightly negative, OH^^" to OH^ '^' from population analyses. In addition, the positive charges at vanadium neighbors are decreased slightly which can be understood by hydrogen induced chemical reduction of the metal species. This is in complete analogy with the reduction effect caused by oxygen vacancies discussed above. Densities of states of the adsorbate clusters reveal additional occupied states, at about 2 eV above the oxygen 2sp valence region, which are characterized as V 3d type. The hydrogen induced reduction effect has been confirmed by recent photoemission experiments [151]. The EB(H') values of Table 5 show that a second hydrogen can adsorb at the three surface OH groups involving the 0(l-3) sites which results in surface H2O. However, corresponding O-H binding energies of the second hydrogen are always found to be smaller than those of the first. The largest EB(H') value, 1.56 eV, is obtained for the terminal 0(1) site yielding a combined adsorption energy EB(2H) = 3.9 eV. The EB(2H) values of the other sites are considerably smaller which is explained by the fact the surface oxygen in two- and three-fold coordinated sites is less flexible in its ability to rehybridize compared to singly coordinated terminal oxygen 0(1). It should be noted that the present definitions of hydrogen adsorption energies refer to adsorption starting from atomic species. If molecular hydrogen gas is exposed to the ¥205(010) surface the energetics of corresponding adsorption processes has to include H2 -^ H + H dissociation. This reduces the computed adsorption energies EB(H) for single H adsorbates by half of the dissociation energy of free H2 (V2 D(H2) = 2.25 eV [150]). As a consequence, positive adsorption energies (stable adsorption states) are obtained only for the 0(1) and

165

0(2) sites while 0(2'), 0(3), and 0(3') site adsorption states are metastable. Further, energies EB(2H) for adsorption of two H species at a given oxygen site would have to be reduced by D(H2) = 4.5 eV which, based on the present results, yields only metastable adsorption states. Surface oxygen which, as a result of catalytic oxidation at the V2O5 surface, is incorporated into a hydrocarbon will be removed from the surface leaving a vacancy behind. Here a possible removal scenario could consist of a combined process where hydrogen is adsorbed at the surface and the resulting surface OH or H2O group desorbs. The energetics of such processes can be studied within the cluster approach by combining the results obtained for the vacancy clusters with those describing hydrogen adsorption. Within the surface cluster approach, desorption processes can be described in their energetic consequen(V205)surf ^

( V 2 0 5 ) s u r f / 0 ^ " + Ogas

(A)

or (V205)surfHads ^

(V205)surf/0^^^ + (OH)gas

(B)

or (V205)surf2Hads " ^ (V205)surf/0^^^ + (H20)gas

( Q

ces by desorption energies E D ( 0 ) , ED(OH), and ED(H20), respectively, which are given by corresponding cluster total energy differences. Based on the V10O31H12 substrate cluster [122, 123, 126] desorption energies of the different processes are listed in Table 6. Note that process (A) describes oxygen vacancy formation and, therefore, desorption energies ED(0) of Table 6 are identical to oxygen vacancy formation energies ED^O) of Table 4. As discussed before, desorption energies ED(0) are rather large (6.5 eV to 7.2 eV) for all surface oxygen sites, which shows that it is quite difficult to remove oxygen by itself from Table 6 Desorption energies E D ( 0 ) , E D ( O H ) , and E D ( H 2 0 ) for different oxygen sites at the V2O5(010)

surface. All data are obtained from calculations based on the V10O31H12 substrate cluster [122, 123, 126]. For definitions of the different quantities see text. All energies are given in eV. site

ED(0)

ED(OH)

ED(H20)

o(i)

6.47

3.98

0.48

0(2)

7.19

4.57

0.16

0(3)

6.49

3.54

-0.44

0(2')

7.19

2.90

—

0(3')

6.49

2.43

—

166

the ideal ¥205(010) surface. However, corresponding energies ED(OH), for desorption of OH and H2O, leaving an oxygen vacancy behind, are always found to be smaller. For OH removal, ED(OH) values vary between 2.4 eV and 4.6 eV and for H2O removal, ED(H20) values range even between -0.5 eV (metastability at the 0(3) site) and 0.5 eV. These decreased desorption energies originate obviously form weakening of V-0 bonds at the V2O5(010) surface caused by hydrogen adsorption. Altogether, the cluster results suggest that the presence of pre-adsorbed hydrogen at the vanadium oxide surface facilitates oxygen removal substantially and can, thus, contribute to an enhanced yield of oxygenated products near vanadia based surfaces. Atomic oxygen adsorption at the ¥205(010) surface has been studied recently by cluster models [134] in an attempt to describe details of healing of surface oxygen vacancies. These surface vacancies can be annihilated by either diffusion in the bulk or by binding of adsorbing oxygen, where in the latter case both atomic and molecular oxygen are possible candidates. The healing process ED(H20)

(V205)surf/0^^^ + Ogas " ^ (V205)surf

(A)

is the inverse of vacancy formation and, therefore, its energetics are determined by quantities ED(0) discussed above. Alternatively, one could envision a twostep process (V205)surf/0^^^ + (02)gas " ^ ( V 2 0 5 ) s u r f / 0 ^ " + (02)ads ^ (V205)surf + Oads " ^ (V205)surf + Ogas

(B)

where molecular oxygen adsorbs at the surface followed by desorption of atomic oxygen leaving a healed surface behind. Here the intermediate state corresponds to either O2 filling an oxygen vacancy or to O being adsorbed at an oxygen surface site. The latter healing process (B) has been considered in recent theoretical DFT studies [134] based on a V10O31H12 substrate cluster. The intermediate state describing atomic oxygen adsorption deserves special attention. In calculations, atomic oxygen is added at different oxygen sites of the V10O31H12 cluster and the geometry of the resulting local O2 surface group is optimized in total energy calculations. The resulting equilibrium geometries of the O2 surface species, visualized in Fig. 17, yield O2 lying flat on the surface at the 0(1, 2, 2') sites and protruding out of the surface at the 0(3, 3') sites. Its inter-nuclear distance is increased at all surface sites by about 0.2 A compared to the distance in gas phase O2, thus suggesting a per-oxo type species. Table 7 lists energetic results from the calculations. Here EB(0) denotes the binding energy of a single oxygen adsorbed at a given oxygen site 0(1-3, 2', 3') while EB(02/vac) gives

167

02 at 0(3') 02atO(3) \ r . r ^

O2 at 0(2')

02atO(1)

/

O2 at 0(2) Fig. 17. Sketch of computed equilibrium geometries of atomic oxygen adsorption at different oxygen sites of the ViOsCOlO) surface reflecting also molecular oxygen adsorption at respective oxygen vacancies. The results are obtained from optimizations of the local O2 surface species on a V10O31H12 substrate cluster. The O2 surface species is shown by shaded balls while the surface lattice is sketched by light balls.

binding energies of an O2 molecule at corresponding oxygen vacancy sites. The data are determined by total energy differences of the appropriate clusters. Obviously, atomic oxygen can adsorb at all oxygen sites where strong binding is found for the 0(1, 2, 2') sites. However, at all sites EB(0) values are smaller than corresponding EB(02/vac) values. This reflects the fact that oxygen vacancy energies at the V2O5(010) surface are, for all sites, larger than the dissociation energy of gas phase O2. It suggests further that the healing scenario (B) is an exothermic process. Table 7 Binding energies E B ( 0 ) , EeCOi/vac) for different oxygen sites at the ViOsCOlO) surface. All data are obtained from calculations based on the V10O31H12 substrate cluster [134]. For definitions of the different quantities see text. All energies are given in eV. site

EB(0)

EeCOi/vac)

0(1)

1.82

2.59

0(2)

1.78

3.33

0(3)

0.35

0.98

0(2')

1.78

3.34

0(3')

0.36

0.98

168

Water adsorption at the ViOsCOlO) surface has been studied by semi-empirical cluster models [143] as well as by periodic DFT slab models [144]. The DFT studies show that molecular adsorption of H2O can happen at all surface oxygen sites as well as at the 5-fold coordinated vanadium site. At the vanadium site hydrogen bonding is found to be most important for the H2O adsorption while at the oxygen sites donation from oxygen to the water molecule seems to dominate the adsorbate binding. Dissociation of H2O is found to be rather unfavorable at the ¥205(010) surface which is explained by pronounced Coulombic effects. For both molecular and dissociative water adsorption the vanadyl oxygen 0(1) is identified as the most favorable site. This interpreted as the 0(1) site being of highest catalytic activity in contrast to other theoretical work [121, 136-140, 152-153]. Ammonia adsorption at the ¥205(010) surface has been examined in DFT studies using very small cluster models [146, 147] as well as using periodic slabs [145]. In the slap studies it is assumed that surface oxygen sites are covered with hydrogen and adsorption of NH3 at the resulting surface OH groups (Bronsted acid sites) is considered. The calculations show clearly that NH3 is bound to the Bronsted acid sites fairly strongly forming an N H / ion where binding is the strongest at the terminal 0(1) site. On the other hand, NH3 can stabilize above the 5-fold coordinated vanadium site where a small binding energy is combined with rather large adsorbate - substrate equilibrium distances. This suggests polarization and hydrogen binding while dative binding involving the NH3 lone pair orbital seems unlikely in contrast to the claim made by the authors [145]. 3. MOLYBDENUM OXIDES Molybdenum oxides form a large group of materials resulting from the ability of molybdenum to possess different formal oxidation states as well as different local coordination [2-3, 12, 34]. Of particular importance is the coexistence of different oxidation states of Mo that leads to mixed valency oxides. Amongst the binary oxides MOxOy there are only two single valency oxides, M0O3 (Mo^"^) (low-pressure orthorhombic a-phase [154] as well as high-pressure monoclinic P phase [155, 156]) and monoclinic 6-M0O2 (Mo"^) [157]. Successive reduction of M0O3 to M0O2 gives rise to intermediate mixed valency oxides which are based on the formation of crystallographic shear (CS) planes [12, 34, 154, 158160]. These oxide have been suggested to play an important role in catalysis [161, 162]. During high-temperature reduction, removal of oxygen from M0O3 creates oxygen vacancies that can interact to order into equally spaced parallel

169

CS planes across which the M0O6 octahedra share edges rather than comers. These extended defects exist even at modest temperature due to the large mobility of oxygen vacancies, Further, the defects are stabilized by relaxation as a result of Mo cation displacement from their centers of symmetry to interstices. Homologous series of molybdenum oxides [1, 3, 12, 34, 154, 158, 160, 163, 164], which are based on CS plane formation starting from a ReOs-type structure, include MOnOsn-i and M0n03n-2 systems. Single crystals of oxides MOnOsn-i have been observed for 4 < n < 9 where n determines the thickness of Re-Os-type slabs separated by CS planes. Here the CS planes are perpendicular to lattice direction a with inter-planar distances corresponding to the height of n/2 MoOe octahedra [158]. In these systems the average valence charge of molybdenum is +(6-2/n), ranging between +4 and +6, which is described by ion ratios MO^VMO"^ = (n-l)/l. Apart from the single valency oxide M0O2 (n = 1) examples are M09O26 [158, 165] (with low-temperature triclinic ^ phase and high-temperature monoclinic p' phase), monoclinic M08O23 [154, 158, 164, 166-171], M07O20, MoeOn (not prepared as single crystal), tetragonal M05O14 [154], and M04O11 [154, 158, 167, 171-175] (with low-temperature monoclinic r| phase and high-temperature orthorhombic 7 phase). Both M04O11 structures are ReOa-based but adjacent slabs of octahedra are connected by M0O4 tetrahedra and the orientation varies between slabs. Electronic properties of MonOsn-i bulk systems are studied experimentally in [176-186]. Single crystals of oxides Mon03n-2 contain CS planes where M0O6 octahedra share faces rather than edges [154, 159, 187]. In these systems the average valence charge of molybdenum is +(6-4/n), which is described by ion ratios MO^VMO"^ = (n-2)/2. The only example observed so far is triclinic Moi8052[154, 167, 188-189]. In addition to these intermediate oxides, reduced phases built of M0O6 octahedra and pentagonal columns have been characterized. Examples include the low-temperature phase K-M017O47 [154] and M010O28 [154, 158]. Further, metallic molybdenum oxide M03O with cubic structure has been described [190]. Apart from bulk material, evaporation of M0O3 can produce cyclic species like M03O9, M04O12 or M05O15 with the structures based on comer-shared M0O4 tetrahedra [191-194]. Finally, molybdenum oxide clusters have been prepared as neutral and ionic species in gas phase and studied by quantum chemical methods at various levels of theory. This includes MoO [195-198], MoOn\ n=l-3 [199], M0O3, Mo04^", M0O4H2, Mo205^"^, and M02O6, [200] where the calculations provide a satisfactory description of stmctural, energetic, and spectroscopic properties. The following discussion will be restricted to the two single valency oxides of molybdenum which have been accounted for by quantitative theoretical work.

170

3.1. Molybdenum oxide bulk systems 3.1.1. Geometric structure Single crystals of the two single valency molybdenum oxides, M0O2 and M0O3, can be characterized both by arrangements of distorted M0O6 octahedra. The corresponding crystal lattices differ basically by the connectivity of these octahedra and by their Mo-0 distances as well as 0-Mo-O angles. This is obvious from the balls-and-sticks models of the two oxides shown in Figs. 18,19. The trioxide M0O3 forms an orthorhombic crystal lattice where the elementary cell contains four chemical M0O3 units [154, 201-202]. It has a layered structure with weakly binding bilayers parallel to the (010) crystal plane. Each bilayer consists of two interleaved planes of comer-linked distorted MoOe octahedra (see Fig.l, one bilayer is marked by shaded atom balls) where octahedra from adjacent planes share edges. The octahedra are described by Mo-0 distances dMo-o = 1-67, 1.73, 2*1.95, 2.25, and 2.33 A where the largest distance refers to oxygen and molybdenum of different planes. Alternatively, each bilayer can be discussed in terms of comer-sharing sheets are composed of ribbons of octahedra sharing edges with two adjacent M0O4 tetrahedra along the (001) direction. The tetrahedra form infinite strings that are combined to increase

Fig. 18. Geometric structure of orthorhombic M0O3 with netplane stacking along the (010) direction. Molybdenum (oxygen) centers are shown by large (small) balls where those of the topmost bi-layer are shaded in light/dark gray. Non-equivalent oxygen centers, 0 ( 1 , 2, 3), are labeled accordingly. An elementary unit cell of the crystal is included to the right.

171

the coordination of Mo to 6. The resulting octahedra and comers with octahedra of parallel ribbons of both sides. In this description the layer structure of M0O3 results from the existence of free unshared comers (one per octahedron) in one layer that are pointing between those of neighboring layers. Bulk molybdenum trioxide contains three structurally different oxygen sites, terminal oxygen 0(1) coordinated to only one Mo center, and two bridging oxygen sites, doubly 0(2) and triply 0(3) coordinated to molybdenum centers, see Fig. 18. These sites can be identified by infrared and Raman spectroscopy which allow a quantitative assignment by corresponding group frequencies for stretching vibrations and can correlate bond lengths with force constants [203, 204]. The dioxide M0O2, crystallizes in a monoclinic lattice that deviates only slightly from the tetragonal rutile structure and is characteristic for several early transition metal dioxides like VO2. Its elementary cell contains four chemical M0O2 units [157] and almost equal Mo-O bond lengths describe the octahedral Mo environments. The M0O6 units form rows in which they are connected by edges and the octahedra of adjacent rows are linked by comers. Further, every second row is rotated by 90° about the (Oil) direction, see Fig. 19, lb. Molybdenum dioxide contains only one type of oxygen sites, 0(3) linking three metal centers.

Mo

0(2)

0(3)

^-.JL^/A

unit cell

Fig. 19. Geometric structure of monoclinic M0O2 with netplane stacking along the (Oil) direction. Molybdenum (oxygen) centers are shown by large (small) balls where those of the topmost layer are shaded in light/dark gray. Non-equivalent oxygen centers, 0(2, 3), are labeled accordingly. An elementary unit cell of the crystal is included to the right.

172

3.1.2. Electronic properties The electronic structure of bulk molybdenum oxides is determined by the amount of 4d electron occupation of the Mo ions analogous to vanadium oxides discussed above. The molybdenum atom is described in its ground state by an electron configuration [Kr] 4d^ 5s\ As a consequence, molybdenum cations Mo^"*^ for 4 < p < 6 are characterized by configurations [Kr] 4d^"P with a partially filled (p < 6) or an empty (p = 6) 4d shell. The 4d occupation is reflected in the electronic ground states of the bulk oxides where Mo possesses a formal charge +p. As a result, the amount of occupied Mo 4d electron states is expected to increase in going from M0O3 (corresponding formally to V^"^ ions) to M0O2 (formally V"^ ions). Although modem total energy methods allow us to obtain band structures and corresponding densities of states (DOS) for molybdenum oxides, the electronic structure of these materials has been examined in only very few theoretical studies. Electronic ground state as well as structural properties of bulk M0O3 have been studied by ab initio periodic Hartree-Fock (HF) methods [205, 206] with different electron corrections. Here geometry optimizations yield the experimental bulk structure and confirm the weak attractive interaction between adjacent bilayers. Further, interatomic Mo-O binding was found to depend on bond distance, ranging from strongly covalent binding for the shortest distances to predominantly ionic for the longest. In addition to the electronic structure also the formation of shear structures in M0O3 was studied theoretically by the SCF-SW-Xa method using very small model clusters [196, 207]. Here a mechanism of oxygen removal and change in linkage of M0O6 octahedra from comer to edge linked was assumed and an interpretation of the corresponding core level photoemission spectra of M0O3 was given. Bulk M0O2 displays a striking metal-insulator stmctural transition that was the subject of extensive theoretical as well as experimental work. The electronic bulk stmcture was studied by periodic [208] as well as cluster DFT methods [209, 210]. The electronic properties of molybdenum dioxide were found to be dominated by strong hybridization of O 2p and crystal-field split Mo 4d states with bands near the Fermi energy originating almost exclusively from Mo 4d t2g-type orbitals.

3.2. Molybdenum oxide surfaces 3.2.1. Geometric structure Theoretical studies on molybdenum oxide surfaces have focused exclusively on single crystal surfaces of M0O3 and M0O2 that are described by low Miller indices. This is due to their relatively simple geometry considering the

173

M0-15O56 cluster

Fig. 20. Geometric M0O3. Molybdenum oxygen sites, 0 ( 1 , 2, model calculations is

Mo

0(1)

0(2)

0(3)

structure of the (010) oriented single crystal surface of orthorhombic (oxygen) centers are shown by large (small) balls and non-equivalent 3), are labeled accordingly. The M015O56 surface section used in cluster shown by shaded balls.

very complex bulk structures of molybdenum oxides. Single crystal surfaces of orthorhombic M0O3, considered by theory, are of (010) and (100) orientation. The (010) surface, which represents the preferred cleavage plane of the crystal and is chemically quite inert, is determined in its structure by three differently coordinated oxygen ions while molybdenum ions as such are not exposed, see Fig. 20. Oxygen ions appear as terminal (molybdenyl) oxygen species, 0(1), coordinated to one molybdenum center at a distance dMoo = 1-67 A. Bridging oxygen, 0(2), is coordinated asymmetrically to two Mo centers at distances dMo-o = 1-74, 2.25 A and bridging oxygen, 0(3), is coordinated to three Mo centers with two equal (dMo-o = 1-95 A) and one longer distance (dMo-o = 2.33 A). The (100) surface is rather different in its structure from the (010) surface, see Fig. 21. It is much rougher and less compact and there are many structurally different metal as well as oxygen sites. Molybdenyl oxygen ions 0(1) are similar to those of the (010) surface. However, they do not cover all Mo ions leaving bare metal sites at the (100) surface. There are also symmetrically bridging oxygen centers that are coordinated to two Mo centers of the surface and couple weakly with a third Mo surface center. These centers correspond structurally to triply coordinated oxygen 0(3) at the (010) surface. Finally, there are terminal oxygen centers which correspond to molybdenyl 0(1) of the (010)

174

(010)

M0H5O56 cluster

^(100) -^— M05O24 cluster

Fig. 21. Perspective view of the (010) and (100) surfaces of orthorhombic M0O3. Molybdenum (oxygen) centers are shown by large (small) balls. The M015O56 and M06O24 surface section used in cluster model calculations are shown by shaded balls.

MO11O39 Cluster

Mo(6) Mo(5)

0(3)

0(2)

Fig. 22. Geometric structure of the (Oil) oriented single crystal surface of monoclinic M0O2. Molybdenum (oxygen) centers are shown by large (small) balls and non-equivalent sites, V(5, 6), 0(2, 3), are labeled accordingly. The M011O59 surface section used in cluster model calculations is shown by shaded balls.

175

surface but point parallel to the surface. Theoretical studies on single crystal surfaces of monoclinic M0O2 have been restricted to the (Oil) surface, see Fig. 22. This surface contains 5- and 6fold coordinated molybdenum centers, Mo(5), Mo(6), and two sets of two- as well as three-fold coordinated oxygen, 0(2) and 0(3), as indicated in Fig. 22. The surface does not contain singly coordinated oxygen.

3.2.2. Physical and chemical properties Up to now, theoretical work on molybdenum oxide surfaces has been restricted to only few periodic slab as well as local cluster studies on the single valence oxide surfaces MoO3(010), (100) and MoO2(011). As for vanadium oxide, there is very little experimental information on quantitative microscopic details of these complex surfaces despite their importance concerning catalytic applications. The following discussion of theoretical results will be based exclusively on cluster studies that have yielded rather detailed microscopic insight.

(a) Properties of clean surfaces Electronic properties of the MoO3(010) surface have been examined by cluster model studies [211] using a M015O56H22 cluster, see Fig. 23a, which describes a bilayer section of the bulk, cp. Fig. 18. This cluster has been proven to yield electronic properties that are size converged. In fact, previous studies have

Mo(6)

0(3)

Fig. 23. (a) Geometric structure of the M015O56H22 cluster modeling a bi-layer section at the MOOBCOIO) surface, (b) Geometric structure of the M011O56H34 cluster modeling a one-layer section at the MoO2(011) surface. The central molybdenum atom and its neighboring nonequivalent surface oxygen centers, 0(l-3), are labeled accordingly.

176

shown [212] that clusters as large M07O30H18 are already sufficient to account for local electronic properties of the surface. The M015O56H22 cluster contains all structurally different oxygen centers in their complete surface environment. Further, hydrogen terminators are placed at the cluster periphery to saturate dangling oxygen bonds, thus accounting for electronic embedding of the surface cluster. The representation of the local MoOsCOlO) surface geometry by the M015O56H22 cluster can be tested by geometry optimizations in which all cluster atoms are allowed to rearrange except the peripheral hydrogen terminators. These optimizations yield equilibrium positions for the surface atoms that are very close to those of the bulk termination with inter-atomic distances varying by less than 0.05 A. The electronic structure of the clusters is determined by ab initio DFT methods [35, 213-214] using both local [215] and gradient corrected (RPBE [124, 125]) functional with model core potentials for Mo [216]. In addition to total energies and equilibrium geometries, detailed analyses of the electronic structure in the clusters are performed using Mullilcen populations [118] and Mayer bond order indices [119, 120]. Further, total density of states (DOS) and atom projected partial densities of states (PDOS) are evaluated for the interpretation of experimental photoemission spectra of M0O3 surface systems as discussed below. Table 8 lists geometric and electronic parameters of the M015O56H22 cluster with its geometry taken from the experimental bulk geometry [154] as well as Table 8 Selected atom charges q, bond orders p and distances do-Mo, (A) of the M015O56H22 cluster representing MoOsCOlO) surface. For a definition of the two sets of data, "bulk termination" and "optimized cluster", see text. Results refer to DFT calculations using the RPBE approximation. bulk termination

optimized cluster

q(Mo)

2.23

2.16

q(0(l))

-0.48

-0.48

q(0(2))

-0.74

-0.72

q(0(3))

-0.99

-0.97

1 p(0(l)-Mo) 1 do(i)-Mo

1.66 1 1.67

1.7411.69

p(0(2)-Mo) 1 do(2)-Mo

1.05 1 1.73 0.33 1 2.25

1.13 11.75 0.3112.33

p(0(3)-Mo) 1 do(3)-Mo

0.26 1 1.94 0.44 1 1.94 0.31 1 2.33

0.28 11.97 0.45 1 2.03 0.32 1 2.28

177

for the optimized geometry (all atom positions except for the hydrogen terminators are optimized according to the lowest energy). Results of atom charges from MuUiken populations and Mayer bond orders are given for selected atoms closest to the cluster center. All data are found to differ only little between the bulk terminated and optimized cluster calculations. The populations yield positively charged molybdenum ions, described as Mo^'^"^ with minor charge variation (Aq = 0.2) inside the cluster, and negative oxygen ions. The negative oxygen charges scale with coordination, O^'^" for singly coordinated oxygen 0(1), O^^" for doubly coordinated (asymmetrically) bridging oxygen 0(2), and O^^' for triply coordinated (symmetrically) bridging oxygen 0(3). This result indicates that the largest ionic contributions to binding at the MoO3(010) surface involve symmetrically bridging 0(3) sites, which is of importance in view of the reactivity of the different sites with respect to chemical reactions. Altogether, local charging of the different cluster atoms is found to be much smaller than formal valence charges suggest. This indicates that inter-atomic binding in M0O3 is described by both ionic and sizeable covalent contributions. Covalent contributions to Mo-O binding can be estimated roughly from corresponding bond order indices. The data of Table 8 confirm the general picture based on simple valence concepts revealing three types of Mo-O bonds. Bonds between terminal oxygen, 0(1), and Mo yield bond order values close to 2, which suggests double bonds and is consistent with the single coordination of 0(1). The asynunetrically bridging oxygen 0(2) binds to two neighboring Mo centers with bond order values of 1.0 for the center at closer and 0.3 for that at further distance, corresponding to one single and one much weaker bond that is reasonable in view of the environmental geometry of the 0(2). The symmetrically bridging oxygen 0(3), which is coordinated to three neighboring Mo centers (two of the upper and one of the lower part of the bi-layer), yields bond order values of 0.3 - 0.4 indicating three bonds which are similar in strength and considerably weaker than single bonds. If the total amount of covalent binding of each oxygen center with its Mo neighbors is defined by the sum of bond orders, the highest value is obtained for 0(1) (sum = 1.7), followed by 0(2) (sum = 1.4) and 0(3) (sum = 1.0). This correlates nicely with the increase in negative charge from 0(1) to 0(2) to 0(3). Altogether, the results suggest for the different oxygen centers at the MoO3(010) surface that an increase in coordination with respect to Mo neighbors leads to an increased ionic and decreased covalent binding character. The calculated energy of the highest occupied orbitals (HOMO), 8HOMO, of the M015O56H22 cluster amounts to 6.1 eV. Based on general DFT theory [35] the quantity -8HOMO represents the first cluster ionization potential which converges with increasing cluster size towards the workfunction of the MoO3(010) surface. The theoretical value of -8HOMO is reasonable in view of typical work-

178

function values of transition metal oxide surfaces (e. g. 6.4 ± 0.7 eV for ¥205(010) [94, 217]), while experimental workfunction values for MoOsCOlO) do not seem to exist in the literature. The character of the HOMO and lowest unoccupied orbital (LUMO) of the M015O56H22 cluster becomes obvious from population results. The analysis shows that the HOMO is described by dominant 2sp contributions of oxygen at the cluster periphery while the LUMO can be characterized by 4d contributions of molybdenum. This indicates that optical absorption leading to electronic HOMO-LUMO excitations is combined with an oxygen to metal charge transfer at the MoO3(010) surface. Further, adding electronic charge to the surface will result in an increased 4d occupation of the metal ions at the surface. This can be connected with chemical reduction and is important later on in connection with oxygen vacancy formation. The electronic structure of the M015O56H22 cluster can also be characterized by energy level distributions of the occupied Kohn-Sham valence orbitals. Fig. 24 shows the total DOS curve for the valence band region of the cluster together with its PDOS decomposition into Mo and O contributions. The DOS of the valence band region extends from -12.5 to -6.1 eV and is described by a multi-peak structure (regions A-D). The decomposition shows clearly that dominant contributions are due to oxygen while molybdenum contributions are smaller and confined to the lower part of the valence band region. Population analyses identify the oxygen contributions as O 2p and those of molybdenum as Mo 4d that is to be expected based on the valence configuration of the atoms

(a) total Mo 0

4.0

B . 0

CO

02.0

Q

A

1

0.0 -18.0

^

E]

/"' '

-14.0 -10.0 Energy [eV]

-6.0

-18.0

-14.0 -10.0 Energy [eV]

Fig. 24. Total DOS and atom projected PDOS curves of the M015O56H22 cluster, (a) Total DOS (thick solid) with decomposition into Mo (thin solid) and O (dashed) contributions, (b) PDOSs of the differently coordinated oxygen centers 0(1) (solid), 0(2) (dotted), and 0(3) (dashed). A gaussian level broadening of 0.4 eV is applied and the energetic position of the highest occupied cluster orbital EHOMO at -6.1 eV is marked by a thin vertical line.

179

and their formal valence in M0O3 bulk. There is an additional set of levels between -15.0 and -12.8 eV (region D, below the valence band region) which is due to split-off orbitals as a consequence of the bond saturation of peripheral oxygen by hydrogen in the clusters and therefore has to be considered a cluster artifact. Altogether, the valence band region assumes a width of A = 6.4 eV in the present cluster. Width A is expected to converge with increasing cluster size towards the total valence band width of the extended MoOsCOlO) surface. This is consistent with recent MoOsCOlO) slab and bulk [206, 218] calculations as well as experimental Hell photoemission studies [206]. The valence band region can be characterized in more detail using PDOS curves shown in Fig. 24b. Here the oxygen PDOS is decomposed into contributions of the differently coordinated surface oxygen sites 0(l-3), allowing a distinction of the three main peak regions A, B, C in the total DOS curve. The energetically highest region A (about 2 eV below the upper valence band edge, £HOMO) is described by contributions from oxygen of all three coordination types with that of 0(3) being most pronounced. The central region B (~ 4 eV below £HOMO) is characterized by oxygen with dominating 0(1) contributions in addition to small molybdenum ingredients. The lowest region C (about 5.5 eV below 8HOMO) has dominant 0(3) as well as molybdenum character. Region E, denoting unoccupied levels and simulating the conduction band region of the substrate within the cluster approach, starts at 1.8 eV above the upper valence band edge. This is consistent with the well known experimental result of bulk M0O3 forming a small band gap insulator. Fig. 24a shows in addition that the lower part of region E is dominated by molybdenum contributions that are described as Mo 4d type based on orbital analyses. This confirms the qualitative picture of the electronic structure of M0O3 being described roughly by a Mo 4d type conduction band region in addition to the O 2sp type valence band region. The present conclusions about electronic properties of the MoO3(010) surface are confirmed by other cluster model studies using ab initio HartreeFock methods [219] together with full geometry optimization. The results from the M02O11H10, M03O16H14, and M07O32H22 surface clusters confirm mixed ionic and covalent Mo-0 binding and suggest symmetric bridging oxygen as a probable active site for insertion into propylene. Further, the electronic structure of the MoO3(010) surface has been treated by periodic DFT calculations using plane-wave pseudopotential methods in a repeated slab model [218]. Here the binding analysis based on crystal orbital overlap populations indicates also that Mo-O binding is a combination of ionic and covalent contributions for all oxygen species. However, the analysis suggests that covalent Mo-O binding is stronger for the terminal oxygen that for either of the two bridging oxygens [218].

180

Relationships between chemical composition, surface geometry, and electronic properties of molybdenum oxide surfaces can be discussed by comparing surfaces of different orientation. As an example, electronic properties of the MoOsClOO) surface have been calculated for appropriate surface clusters, such as M06O24H12 shown in Fig. 21, using DFT methods [220]. A comparison of these results with those of MoOsCOlO) derived surface clusters [220] shows that electronic differences between both surfaces are mainly due to different atom arrangement while local charging and binding properties seem surface independent. In particular, geometrically equivalent oxygen centers at the MoOsCOlO) and (100) surfaces are similar in their electronic behavior. Therefore, the electronic structure at the two surfaces is determined mainly by their detailed atom arrangement and is not influenced by major charge redistributions due to substrate surface binding. As mentioned above, the MoOsCOlO) surface has been found to be chemically inert at room temperature. Valence band photoemission spectra for this surface [201] show negligible emission in the bulk band gap region. However, when the surface is exposed to electron or ion bombardment, even of relatively low energies, or to ultraviolet photons [221] the surface becomes reduced. Many reduced oxides of molybdenum have been observed and found to exhibit significant emission in the bulk band gap region. The band gap emission must involve electron states of predominant Mo 4d character. Angle-resolved UPS studies on band gap defect states conclude [202] that these states are associated with extended defects rather than with point defects, and are presumably due to the shear-plane structure of the reduced phases. One example of these phases is M0O2 resulting from deep reduction of molybdenum trioxide and being always present as domains at M0O3 surfaces [222, 223]. The reduction effect can be seen in model studies using surface clusters which simulate local sections of the MoO2(011) surface. As an illustration. Table 9 lists atomic charges from Mulliken populations and Mayer bond orders for the Moii039H34cluster cut out of the MoO2(011) surface (see Figs. 22, 23b). The data are compared with those of a M015O56H22 cluster modeling the MoO3(010) surface. The comparison suggests local charging and Mo-O binding which is quite similar between the MoO2(011) and MoO3(010) surface. In the MoO2(011) surface cluster positive molybdenum ions are described by Mo^'^"^ (6-fold coordinated) and Mo''^^ (5-fold coordinated) while negative oxygen charges scale with coordination, O^^' for doubly and O^'^" for triply coordinated oxygen. (Note that at the monoclinic MoO2(011) surface there are two very slightly differing 0(2) and 0(3) species and the results in Table 9 refer to averages.) As for the MoO3(010) surface clusters, charging of the different cluster ions in the M011O39H34 cluster is much smaller than formal valence charges of the M0O2 substrate, Mo"^"^ and O^", suggest. In addition, charging of the 5- and 6-fold coordinated Mo ions differs only by little. There-

181

Table 9 Comparison of selected atom charges q, bond orders p, and distances do-Mo, (A) of two surface clusters representing MoOiCOll) and MoOsCOlO) surfaces, respectively. The results refer to DFT calculations using the LDA approximation. MOii039H34for

M015O56H22 for

MoO2(011)

MoOsCOlO)

q(Mo(6))

1.28

1.89

q(Mo(5))

1.54

1.89 -0.43

q(0(l)) q(0(2))

-0.52

-0.62

q(0(3))

-0.79

-0.87 1.73 1 1.67

p ( 0 ( l ) - M 0 ) 1 do(2)-Mo p(0(2)-M0) 1 do(2)-Mo

0.87 1 2.06 0.92 1 2.07

1.21 1 1.73 0.42 1 2.25

p(0(3)-M0) 1 do(3)-Mo

0.64 1 2.06 0.54 1 2.07 0.59 1 1.98

0.38 1 1.94 0.50 1 1.94 0.38 1 2.33

fore, discrimination between the two should be based on covalent rather than ionic binding behavior. A comparison between the Mo ion charges (determined from populations) of the MoOaCOl 1) and MoOsCOlO) surface clusters shows that those of the reduced oxide (M0O2) are decreased by only 0.3 electrons. This is in contrast to the standard chemical characterization in terms of formal valence charges and confirms further the importance of covalent binding in the substrates. The general picture based on atom coordination at the MoO2(011) surface is reflected in the bond order results listed in Table 9. Bonds between doubly coordinated oxygen and each of its two molybdenum neighbors result in bond orders close to 1 indicating single bonds. Further, Mo-O bonds involving triply coordinated oxygen yield bond orders of 0.5-0.6 revealing weaker than single bonds. A comparison of bond orders in the M0O2 and M0O3 surface clusters suggests that in the case of M0O2 covalent binding becomes more important at the expense of ionicity which it consistent with molybdenum and oxygen charges of the dioxide being smaller than those of the trioxide. Additional studies have been published to explain factors that influence solid acidity in pure molybdenum oxides and those supported on silica-alumina [224]. These studies are based on cluster models of respective acid sites where electronic properties are evaluated by self-consistent HF, electron-correlated

182

M0ller-Plesset (MP2) and local density functional (LDA) methods. Here it was possible to identify relationships between structural transformations in the molybdenum trioxide tetrahedra and changes in Bronsted-Lewis acidity. In addition, the experimentally observed influence of the support composition on the acid strength of the supported molybdenum oxide catalyst was interpreted by the calculated charge redistribution and by molecular orbital energies. The proposed mechanism of acidity changing with molybdenum loading was supported by the agreement between calculated and measured infrared vibration frequencies.

(b) Surface oxygen vacancies The oxidation of organic molecules at the surface of a molybdenum oxide catalyst, which involves surface oxygen, creates surface oxygen vacancies. Here the initial state of the catalyst may be restored by molecular oxygen from gas phase (reoxidation) or the density of oxygen vacancies may grow and give rise to extended (rather than point) defects [14, 154, 158]. These defects may lead to the annihilation of oxygen vacancies by formation of CS planes. For M0O3 surfaces this process is postulated to be essential for the ability of the oxide to in(a)

0(1) vacancy

0(2) vacancy

0(3) vacancy

(b) 0(1) vacancy

0(2) vacancy

0(3) vacancy

Fig. 25. (a) Sketch of vacancies at different surface oxygen sites 0(l-3) in the M015O56H22 cluster. Cluster atoms of the top (bottom) part of the bi-layer are shown as shaded (white) balls where the ball size decreases from Mo to O to H. The vacancies are sketched as black dots, (b) Relaxed geometry of the M015O56H22-O (=Moi5056H2i ) clusters with different vacancy sites.

183

sert oxygen into the adsorbed species in a selective oxidation process [12, 33, 225]. This is supported by quantum chemical calculations on M02O10 clusters [196] where changes in topology by transforming M0O6 octahedra from comersharing (with a vacancy) to edge-sharing favor the edge-linked geometry. However, formation of CS planes cannot exclude the existence of oxygen vacancies at M0O3 surfaces [226]. Therefore, studies on electronic properties in the presence of surface oxygen vacancies are of a great importance. In recent cluster studies [212] the electronic and geometric structure of the three different types of oxygen vacancies, that can exist at the ideal MoOsCOlO) surface, have been examined by DFT calculations using the RPBE approach. These studies are based on a M015O55H22 substrate cluster where oxygen is removed from one of the three sites 0(l-3), see Fig. 25a. In a first set of calculations the electronic structure of the indented cluster is evaluated with its atom centers kept fixed at their positions of the bulk termination (frozen geometry). In subsequent calculations all atoms of the indented cluster (except the peripheral hydrogen terminators) are allowed to relax in response to the oxygen removal (relaxed geometry). As a result. Table 10 lists oxygen vacancy energies, ED^(O), EDXO) corresponding to frozen and relaxed geometries. The vacancy energies are determined by appropriate cluster total energy differences using the ground state of atomic oxygen as a reference. Obviously, oxygen vacancy energies are rather large for all surface oxygen sites. This indicates that it is quite difficult to remove oxygen by itself from the ideal MoO3(010) surface. The frozen geometry value ED^(0) is largest, 7.6 eV, for the singly coordinated terminal 0(1) site, while for two bridging sites 0(2), 0(3) smaller values, 7.1 and 6.8 eV, are obtained. Table 11 lists selected atom charges from MuUiken populations and Mayer bond orders of the M015O56H22 based vacancy clusters representing the different oxygen vacancies 0(1-3) at the MoO3(010) surface (frozen geometries). For comparison, corresponding results for the vacancy free M015O56H22 cluster with bulk termination (denoted "substrate") are included. Table 10 Results of vacancy energies, E/'^(0), of the different oxygen surface sites 0(l-3) with and without surface relaxation, for definitions see text. The data are obtained from DFT calculations on the M015O56H22 cluster representing a local section of the MoO3(010) surface. All energies are given in eV. Site 0(1) 0(2) 0(3)

ED'(0)

EDXO)

7.64 7.09 6.81

4.84 5.10 6.35

184

Table 11 Selected atom charges q, bond orders p, and distances do-Mo (in A) of the M015O56H22-O (= M015O55H22) clusters representing oxygen vacancies 0(1-3) at the MoO3(010) surface. The results refer to DFT calculations using the RPBE approach and frozen cluster geometries. The fourth column, denoted "substrate", gives data of the M015O56H22 cluster for comparison 0(1) vac.

0(2) vac.

0(3) vac.

substrate

q(Mo)

1.57

1.32

1.70

2.23

q(0(l))

--

-0.46

-0.48

-0.48

q(0(2))

-0.66

~

-0.73

-0.74

q(0(3))

-0.96

-0.96

--

-0.99

p(Mo-0(l))ldMo-o(i)

--

1.9711.67

1.75 1 1.67

1.6611.67

p(M0-0(2)) 1 dMo-0(2)

1.45 11.73

~

1.11 11.73

1.05 11.73

p(Mo-0(3)) 1 dMo-o(3)

0.4611.94

0.3611.94

-

0.2611.94

The atom charges evidence that the existence of a surface oxygen vacancy always reduces the central Mo atom(s) closest to the vacancy (i. e. positive Mo charges are decreased) where the effect amounts to 0.5 - 0.9 electrons. In contrast to the sizeable charge reduction at the metal ions, the oxygen ions near the vacancy experience only small changes in their charge state. In all cases, oxygen ions become less negative with charge differences below 0.1 electrons. A comparison of the bond order results for the different oxygen vacancy clusters with those of the vacancy-free cluster shows that the largest effects occur when the 0(1) or 0(2) vacancies are created. For the 0(1) vacancy the bond order between the central Mo center of the asymmetric bridging oxygen 0(2) increases considerably, suggesting a Mo-0(2) bond strength near the 0(1) vacancy similar to that of the original Mo-O(l) bond. Thus, in terms of Mo-0 binding the asymmetric bridging oxygen 0(2) takes the place of the original terminal 0(1) atom, which has been removed in the vacancy creation process. On the other hand, the existence of an 0(2) vacancy results in strengthening of a nearby Mo-O(l) bond. These findings become important for the vacancy properties when surface relaxation is included in the vacancy calculations [211-212, 227]. Geometric consequences of substrate relaxation due to vacancy formation at the 0(1-3) sites are sketched in Fig. 25b for the M015O56H22 based vacancy cluster. A comparison with the initial cluster geometry, shown in Figs. 23a, 25a, evidences that in all cases relaxation is locally confined. The response of the substrate to the creation of a triply coordinated 0(3) vacancy is rather small. In contrast, removing an oxygen from the terminal 0(1) or doubly coordinated 0(2) sites leads to significant changes, where for both sites the same relaxed

185

substrate geometry is achieved. Here a terminal oxygen species, which originates from either the 0(1) or 0(2) site (complementary to the site where the oxygen has been removed), stabilizes in the interstitial region between the two sites. This suggests that a clear distinction between the two types of oxygen vacancies may not be possible in the experiment. An alternative view of the chemical reduction effect at the metal sites near oxygen vacancies becomes possible by orbital analyses and atom projected partial densities of states (PDOS) of the vacancy clusters. As an example. Fig. 26 shows DOS and PDOS curves of the valence band region obtained for the M015O56H22 based vacancy cluster with a triply coordinated 0(3) vacancy. A detailed comparison of Fig. 26 with Fig. 24 (giving the (P)DOS results of the vacancy-free cluster) shows only very small differences over the whole range of the oxygen 2sp dominated valence band region except for its upper edge. In the region near the HOMO at -6.1 eV the occupied orbitals of the vacancy-free cluster are determined by O 2sp contributions whereas in the vacancy cluster the highest occupied orbitals (below 5.3 eV) are characterized as Mo 4d type. An analysis of these metal derived valence orbitals (unoccupied in the vacancy-free cluster) reveals further that their charge distribution is located mainly at the three Mo sites adjacent to the vacancy. Thus, the calculations show that the oxygen vacancy leads to increased local molybdenum 4d occupation. This result is also found in the (P)DOS and orbital analyses of the vacancy clusters for 0(1) and 0(2) sites. It confirms the picture of chemical reduction of adjacent metal

-14.0 -10.0 Energy [eV]

14.0 -10.0 Energy [eV]

Fig. 26. Total and atom projected (P)DOS curves of the valence band region fo the M015O56H22 based vacancy cluster with an 0(3) surface vacancy, (a) Total DOS (thick solid) with decomposition into Mo (thin solid) and O (dashed) contributions, (b) PDOSs of the differently coordinated oxygen centers 0(1) (solid), 0(2) (dotted), and 0(3) (dashed). A gaussian level broadening of 0.4 eV is applied and the energetic position of the highest occupied cluster orbital £HOMO at -5.3 eV is marked by a thin vertical line.

186

centers induced by oxygen vacancies and observed already in the population results. The existence of additional occupied states of Mo character, located above the O 2sp derived valence region, is relevant for the interpretation of experimental photoemission spectra of molybdenum oxide surfaces. According to the results of the cluster studies additional photoemission intensity above the valence band region may be indicative of chemical reduction of the metal centers, leading to lower oxidation states, where the effect can be introduced by oxygen vacancies or by different chemical composition of the oxide. This has been verified in UPS experiments on differently prepared MoOsCOlO) surfaces in comparison with measurements of other single and mixed valency molybdenum oxide samples [212].

(c) Adsorption Atomic and molecular adsorption at molybdenum oxide surfaces have been studied theoretically using both periodic slab and cluster models where so far studies are restricted to the trioxide, M0O3, as a substrate. Further, adsorbate species include in all cases atoms (H [138, 218, 227-229]) or small molecules (methanol [230, 231], CO [206], H2O[206], CH2 [232], CH3 [232], CH4 [233], OCH3 [234-236], C3H5 [228, 229]) which are of catalytic interest. Hydrogen adsorption at the MoO3(010) surface has been examined extensively by both cluster [138, 227-229] and slab models [218]. Under reductive or oxidative conditions of catalytic reactions, the surface/bulk structure of M0O3 or M0O2 can transform from one oxide to the other rapidly [237]. There are several possible processes that lead to reduction of M0O3 to lower valency oxides. One process starts with adsorption of hydrogen at surface oxygen sites and results in surface hydroxyl or water species. Under reaction conditions these surface species may desorb leaving oxygen vacancies behind. Here it is interesting which of the differently coordinated oxygen sites at the M0O3 surface is dominantly involved in hydrogen adsorption and OH/H2O desorption, see e. g. [238]. This question has been addressed in cluster model studies of hydrogen adsorption at 0(1-3) sites of the MoO3(010) surface, see Fig. 18. Atomic hydrogen is added to one of the different surface oxygen sites, 0(1-3), of a M015O56H22 model cluster, see Fig. 21, and the local geometry of the resulting surface OH species is optimized in DFT total energy calculations [227] using the RPBE functional. Fig. 27 visualizes calculated equilibrium geometries of the different surface OH species. In addition, Table 12 lists electronic and energetic parameters from the calculations. Here EB(H) denotes the adsorption energy of hydrogen at a given

187

0(2)H

(010) view

0(3)H

0(1 )H

(001) view

Fig. 27. Sketch of calculated equilibrium geometries of hydrogen adsorption at different oxygen sites 0(1-3) of the MoOsCOlO) surface, yielding surface OH. The results are obtained from DFT optimizations using M015O56H22H adsorbate clusters. Darker shaded balls show the surface OH species while light balls refer to the surface lattice environment.

oxygen site 0(l-3), calculated from corresponding total energy differences, and Table 12 Selected atom charges q from populations and Mayer bond orders p of the M015O56H22H clusters modeling adsorption of hydrogen at surface sites 0(l-3) at the MoO3(010) surface. The table includes energies EB(H) for H adsorption and ED(OH) for OH desorption, based on relaxed cluster geometries and obtained by DFT calculations using the RPBE functional. Further, data for the substrate M015O56H22 cluster as well as for isolated OH are included. 0(1)H

0(2)H.

0(3)H

Surface / OH

Q(Mo)

2.17

1.84

2.11

2.23

Q(0(1))

-0.81

-0.47

-0.46

-0.48

Q(0(2))

-0.74

-1.04

-0.73

-0.74

Q(0(3))

-0.99

-0.98

-1.18

-0.99

Q(H)

0.49

0.56

0.59

0.42/0.24

P(Mo-0(l))

1.60

1.67

1.75

1.66

1 P(Mo-0(2))

1.05/0.32

0.98/0.86

0.99/0.35

1.05/0.33

0.25/0.43/0.3

0.26/0.45/0.3

0.12/0.15/0.3

0.26/0.44/0.3

0.71

0.54

0.43

0.80/0.91

-

P(Mo-0(3)) P(H-O) EB(H)

(eV)

E D ( O H ) (eV)

1.62

1.43

1.34

3.92

3.74

3.32

188

is the desorption energy of OH (leaving a surface oxygen vacancy). Table 12 shows that hydrogen can adsorb at all surface oxygen sites where binding is found to be strong (with binding energies of 1.3 - 1.6 eV) suggesting rather stable surface hydroxyl species. The equilibrium geometries of the surface OH groups, see Fig. 27, do not reflect the symmetry of the corresponding oxygen sites. The O-H distances are always quite close to those of typical OH containing molecules (similar to hydrogen adsorption at vanadium pentoxide discussed above). Further, the distances between oxygen and its nearest vanadium neighbors at the M0O3 surface are enlarged. This indicates that hydrogen adsorption weakens the Mo-0 bonds near the adsorption site. The effect is most pronounced for the triply coordinated oxygen site 0(3) where hydrogen is tilted lying almost flat at the surface. This may be an initial step for penetration of the adsorbate into the substrate leading to molybdenum oxide bronzes, see for example [239]. At all oxygen sites, hydrogen adsorption leads to accumulation of negative charge at the corresponding site and surface OH appears as a negative species, see Table 12. In addition, the positive charges at vanadium neighbors are decreased slightly, which can be described as hydrogen induced chemical reduction. The reduction effect is also obvious from (P)DOS curves of the adsorbate cluster where additional occupied Mo 4d type states occur [227]. In analogy to the previous discussion for vanadium pentoxide surface oxygen vacancies at the M0O3 surface may be formed in a two-step process where, in the first step, hydrogen is adsorbed and, in the second, OH is desorbed. Table 12 includes calculated desorption energies ED(OH) for the different oxygen sites. These energies range between 3.3 and 3.9 eV and are all considerably smaller than corresponding oxygen vacancy energies of 4.8 to 6.4 eV, see Table 10. The energy decrease is obviously due to weakening of Mo-0 bonds at the MoO3(010) surface caused by hydrogen adsorption. This suggests that the presence of pre-adsorbed hydrogen facilitates oxygen removal, completely analogous to the findings for vanadium pentoxide. The interaction of atomic hydrogen with the MoO3(010) surface was also studied using periodic DFT methods in pseudopotential plane-wave calculations 218]. Here the inclusion of full surface relaxation is found to be important even for a qualitative description of adsorbate bonding. The results confirm that hydrogen is most strongly bound at the 0(1) site followed by 0(2) and 0(3). Similar to the discussion above, hydrogen was predicted to diffuse into bulk forming molybdenum oxide bronzes, HxMo03 [1, 239]. Methanol adsorption at MoO3(010), (001), and (100) surfaces has been studied in cluster models using the semi-empirical ENHT method [230]. The calculations suggest that both molecular and dissociative adsorption may occur at the different surfaces. The stability of molecular adsorption complexes increases from the (100) surface to (001) and (010) if the latter contains oxygen ED(OH)

189

vacancies. At the (100) and (001) surface, which contain unsaturated oxygen, dissociative adsorption is energetically preferred over molecular adsorption. The oxidative dehydrogenation of methanol to formaldehyde on M0O3 surfaces was also studied by quantum chemical ab initio methods [231]. It was concluded that dual dioxo Mo=0 sites (existing at MoO3(010)) are necessary for the reaction to happen where each of these sites extracts one hydrogen in a sequence of steps. The importance of 0x0 Mo=0 groups for the olefin methathesis reaction as well as for hydrocarbon oxidation was also pointed out by quantum chemical studies on small molecular models, such as CI2M0O2 [240, 241]. Theoretical studies on adsorption and oxidation reactions of CH3 and CH2 fragments at the MoO3(100) surface [232] have also been connected with methanol adsorption. The adsorption energy and binding of these fragments was computed using a methodology based on the semi-empirical atomic superposition and electron delocalization molecular orbital theory (ASED-MO). Here both homolytic and heterolytic mechanism of C-H bond breaking are discussed. On the MoO3(010) surface, exposing mainly oxygen ions, the interactions between the CH3/CH2 fragments and the surface is found to be weak. In contrast, on other M0O3 surfaces, containing bare molybdenum ions, lone pair orbitals of the fragments can hybridize with the metal ions to form stronger C-Mo bonds. The adsorption of H2O and CO has been studied using periodic ab initio Hartree-Fock and correlated methods [206] where the five-fold coordinated molybdenum centers at the MoO3(100) surface were chosen as adsorption sites. The calculations show that adsorption of both molecules is rather weak and determined by electrostatic binding. Numerous theoretical studies are devoted to details of C-H activation at M0O3 surfaces. In particular, C-H activation and bond breaking in adsorbed CH4 at the MoO3(010) surface was considered in periodic DFT studies using pseudopotential plane-wave methods [233]. The calculations indicate that methyl is adsorbed most strongly above terminal oxygen sites 0(1) at MoO3(010) resulting in a surface methoxy species. On the other hand, methyl adsorbing above the asymmetric bridging 0(2) site yields formaldehyde with hydroxyl binding near the 0(1) site. Finally, methyl above the symmetric bridging 0(3) site reacts to form a formyl species with water near the 0(1) site. The understanding of C-H bond activation was also the goal of theoretical studies on methoxy adsorption at MoO3(100) surfaces based on a M03O11 cluster model using the serm-empirical ASED-MO method [234, 235]. In particular, the mechanism of C-H bond breaking during the reaction of methane near a surface oxygen vacancy is considered where methane is found to adsorb dissociatively. Further, photon assisted C-H activation in methoxy, adsorbed near bare molybdenum centers at M0O3 and M08O24 surfaces, has been treated by ASEDMO calculations on a Mo30i3^" model cluster [236].

190

Adsorption of allyl, C3H5, species at the MoOsClOO) and (100) surfaces has been examined in DFT cluster studies [228, 229]. Altogether, the adsorptive interaction is found to be rather weak at both surfaces. At the (010) surface, the adsorbate binds most strongly above asymmetric bridging oxygen sites 0(2) where the allyl plane extends perpendicular to the surface. This yields a weakening of both C-H and adjacent C-C bonds while the C-C bond further away becomes stronger. The asymmetric C3 skeleton is qualitatively similar to that in acroleine and can be interpreted as an initial step in the oxidation process. At the (100) surface, the adsorbate binds most strongly above bare molybdenum sites with the allyl plane parallel to the surface. As a result, there is no trend towards asymmetric C-C bonds which makes the allyl oxidation less likely to happen. The different activity of the MoO3(100) and (100) surfaces with respect to allyl adsorption was confirmed further by charge sensitivity analyses [242].

SYNOPSIS Vanadium and molybdenum oxides, which represent commercially important materials, form a fascinating class of compounds concerning their crystal structure as well as their electronic properties. The knowledge of these properties is a necessary prerequisite to understand their physical and chemical behavior. In single valency oxides of both vanadium and molybdenum the building units are distorted Me06 octahedra linked by comers or/and edges. The richness of arrangements of these octahedra is a source for the existence of differently coordinated oxygen ions and has an important impact for surface properties of the oxide systems. Despite the importance of vanadium and molybdenum oxides as technologically relevant materials many details of their microscopic behavior are still under debate. According to theoretical studies, single valency oxides are compounds described by mixed ionic and covalent character of metal - oxygen bonds. Charging of metal and oxygen ions is smaller than expected from their formal valence charges and oxygen ionicity is found to scale with the coordination numbers of the corresponding sites. The character and energetic distribution of the highest valence orbitals of these oxides, characterized also by densities of states (DOS), are correlated with the formal oxidation state of the metal ions. The analysis reveals differences in electronic properties as well as transformations from metal to semiconducting and insulating behavior for different oxidation states. Theoretical studies on surface oxygen vacancies show that the oxygen species is bound quite strongly to the surface. Thus, it is rather difficult to remove surface oxygen by itself. However, pre-adsorbed atoms or molecules, atomic hydrogen serves as an example, can facilitate oxygen removal resulting in surface vacancies. Further, surface oxygen vacancies are able to diffuse at the

191

oxide surface or move into the bulk and they can be easily refilled by adsorbing oxygen from gas phase. In conclusion, the present theory on vanadium and molybdenum oxide surfaces, while it is not yet fully developed, can already give important input towards a microscopic understanding of their physical and chemical behavior.

ACKNOWLEDGEMENT This work has been supported by the Deutsche Forschungsgemeinschaft through its Sonderforschungsbereich 546, and by grant No. 3T09A 14615 of the State Committee for Scientific Research in Poland.

REFERENCES [I] [2]

[3] [4]

[5] [6]

[7] [8] [9] [10] [II] [12]

[13] [14] [15]

C.N.R. Rao and B. Raven, "Transition Metal Oxides", VCH Publishers, New YorkWeinheim-Cambridge, 1995. H.K. Kung, in: B. Delmon, J.T. Yates (Eds.), "Transition Metal Oxides: Surface Chemistry and Catalysis", Studies in Surface Science and Catalysis, Elsevier, Amsterdam, 1989, Vol. 45. V.E. Henrich and P.A. Cox, "The Surface Science of Metal Oxides", University Press, Cambridge, 1994. B. Grzybowska-Swierkosz, F. Trifiro, J.C. Vedrine (Eds.), "Vanadia Catalysts for Selective Oxidation of Hydrocarbons and Their Derivatives", J. Appl. Catal. 157 (1997) 1-420 and references therein. E.E. Chain, Appl. Opt., 30 (1991) 2782. L.A. Gea, L.A. Boatner, J.D. Budai and R.A. Zuhr, in: de. D.B. Poker, D. Ha, Y.T. Cheng, L.R. Harriott, T.W. Sigmon (Eds.), "Ion-Solid Interactions for Materials Modification and Processing Materials", Research Society, Boston, 1995, pp.215. D. Yin, N. Xu, J. Zhang and X. Zheng, J. Phys., D29 (1996) 1051. V.L. Galperin, LA. Khakhaev, F.A. Chudnovskii and E.B. Shadrin, Sov. Tech. Phys. Lett., 18 (1992) 329. C.E. Lee, R.A. Atkins, W.N. Gibler and H.F. Taylor, Appl. Opt., 28 (1991) 4511. G.C. Granqvist, Sohd State Ionics, 70 (1994) 678. A.K. Pandit, M. Prasad, T.H.Ansari, R.A. Singh and B.M. Wanklyn, Sol. State Comm., 80 (1991) 125. E.R. Braithwaite, J. Haber (Eds.), "Molybdenum: An Outline of its Chemistry and Uses", Studies in Inorganic Chemistry Vol. 19, Elsevier, Amsterdam-LausanneNew York-Oxford-Shannon-Tokyo, 1994, and references therein. R. Pearce, W. R. Patterson (Eds.), "Catalysis and Chemical Processes", WileyHalsted, New York, 1981. A. Bielanski and J. Haber, "Oxygen in Catalysis", Marcel Dekker, New York, 1991. J. Haber in: P. Ruiz, B. Delmon (Eds.,), "New Developments in Selective Oxidation

192

[16] [17]

[18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34]

[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48]

by Heterogeneous Catalysis", Studies in Surface Science and Catalysis, Vol. 72, Elsevier Science Publishers (1992). see e. g. G. Ertl, H. Knozinger and J. Weitkamp, "Handbook of Heterogeneous Catalysis", VCH/Wiley Publishing, New York 1997. B. Grzybowska-Swierkosz, J. Haber (Eds.), "Vanadia Catalysts for Processes of Oxidation of Aromatic Hydrocarbons", PWN - Polish Scientific Publishers, Warsaw, 1984. J. Haber, E. Lalik, Catal. Today, 33 (1997) 119. J. Haber, B. Grzybowska, J. Catal., 28 (1973) 489. B. Grzybowska, J. Haber and J. Janas, J. Catal., 49 (1977) 150. J. Haber, Pure Appl. Catal., 50 (1978) 923. K. Bruckman, R. Grabowski, J. Haber, A. Mazurkiewicz, J. Sloczynski and T. Wiltowski, J. Catal., 104 (1987) 71. F. Weiss, J. Marion, J. Metzger and J. M. Cognion, Kinet. Katal., 14 (1973) 45. R.K. Grasseli, J.D. Burrington, Adv. Catal., 30 (1981) 133. J.C. Volta, J.L. Portefaix, Appl. Catal., 18 (1985) 1. J.C. Volta, J.M. Tatibouet, J. Catal., 93 (1985) 467. M. Abon, J. Massardier, B. Mingot, J.C. Volta, N. Roquet and O. Bertrand, J. Catal., 134(1992)542. M.A. Banares, J.L.G. Fierro and J.B. Moffat, J. Catal., 142 (1993) 406. A. Parmaliana, F. Arena, J. Catal., 167 (1997) 75. A. Bielanski, M. Najbar, Appl. Catal., 157 (1997) 223. J.M. Tatibouet, J.E. Germain, J. Catal., 72 (1981) 375. A. Baiker, D. Gasser, Z. Phys. Chem., 149 (1986) 119. B. Grzybowska-Swierkosz, Appl. Catal. A:, General 157 (1997) 409. A.M. Chippindale, A.K. Cheetham in: E.R. Braithwaite. J. Haber (Eds.), "The Oxide Chemistry of Molybdenum", Studies in Inorganic Chemistry 19, Elsevier, Amsterdam-Lausanne-New York-Oxford-Shannon-Tokyo, 1994. J.K. Labanowski, J.W. Andzelm (Eds.), "Density Functional Methods in Chemistry", Springer, New York, 1991. J.R. Chelikowski. M. Schluter, S.G. Louie and M.L. Cohen, Solid. State Comm., 17 (1975) 1103. P. Blaha, K.H. Schwarz, P. Sorantin and S.B. Trickey, Comput. Phys. Commun., 59 (1990)399. B. Kohler, S. Wilke, M. Scheffler, R. Kouba and C. Ambrosch-Draxl, Comput. Phys. Commun., 94 (1996) 31. M. Petersen, F. Wagner, L. Hufnagel, M. Scheffler, P. Blaha and K. Schwarz, Comp. Phys. Commun., 126 (2000) 294, H. Hartmann, W. Maessing, Zeitsch. Anorg. Allg. Chemie, 266 (1951) 98. R.E. Lohman, C.N.R. Rao and J.M. Honig, J. Phys. Chem., 73 (1969) 1781. P.D. Dernier, M. Marezio, Phys. Rev., B2 (1970) 3771. M.G. Vincent, K. Yvon, Acta CrystalL, A36 (1980) 808. R.E. Newnham, M.Y. de Haan, Zeitsch. Kristallographie, Kristallgeometrie, Kristallphysik, Kristallchemie, 117 (1962) 235. L.W. Finger, R.M. Hazen, J. Appl. Phys., 51 (1080) 5362. G. Anderson, Acta Chem, Scand., 8 (1954) 1599. S. Westman, Acta Chem. Scand., 15 (1961) 217. J.M Longo, P. Kierkegaard, Acta Chim. Scand., 24 (1970) 420.

193 [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82]

F. Theobald, R. Cabala and J. Bernard, J. Solid State Chem., 17 (1976) 431 Y. Oka, T. Yao, N. Yamamoto, Y. Neday and A. Hayashi, J. Solid State Chem., 105 (1993)271. R. Heckingbottom. J.. Linnett, Nature (London), 194 (1962) 678. Y. Oka, T. Yao, and N. Yamamoto, J. Solid State Chem., 86 (1990) 116. K.D. Rogers, Powder Diffraction, 8 (1993) 240. T. Yao, Y. Oka and N. Yamamoto, J. SoHd State Chem., 112 (1994) 196. D.B. McWhan, M. Marezio, J.P. Remeika and P.D. Dernier, Phys. Rev., BIO (1974) 490. W. Bystroem, K.A. Wilhelmi and O. Brotzen, Acta Chem. Scand., 4 (1950) 1119. R. Enjalbert, J. Galy, Acta Cryst., C42 (1986) 1467. H.G. Bachmann, F.R. Ahmed, and W.H. Barnes, Zeitsch. Kristallographie, Kristallgeometrie, Kristallphysik, Kristallchemie, 96 (1936) 9. G. Anderson, A. MagneH, Acta Chem. Scand., 7 (1953) 154. S. Anderson, B. CoUen, U.Kuylenstiema and A. Magneli, Acta Chem. Scand., 11 (1957) 1641. S.H. Hong, S. Asbrink, Acta Cryst, B38 (1982) 713. S. Asbrink, Acta Cryst, B36 (1980) 1332. M. Marezio, D.B. McWhan, P.D. Dernier and J.P. Remeika, J. Solid State Chem,, 6 (1973) 419. M. Marezio, J.L. Hodeau, J. Solid State Chem., 23 (1978) 253. H. Horiuchi, M. Tokonami, N. Morimoto and K. Nagasawa, Acta Cryst., B28 (1972) 1404. M. Marezio, P.D. Dernier, D.B. McWhan and S. Kachi, J. Solid State Chem., 11 (1974)301. Y. Le Page, P. Bordet and M. Marezio, J. Solid State Chem., 92 (1991) 380. H. Horiuchi, N. Morimoto and M. Tokonami, J. SoUd State Chem., 17 (1976) 407. H. Kuwamoto, N. Otsukan and H. Sato, J. Solid State Chem., 36 (1981) 132. K. Waltersson, B. Forslund, K.A. Wilhelmi, S. Andersson and J. Galy, Acta Cryst., B30 (1974) 2644. K.A. Wilhelmi, K. Waltersson, Acta Chem. Scand., 24 (1970) 3409. I. Kawafa, M.Ishii, M.Saeki, N. Kimizuka, M. Najano-Onoda and K. Kato, Acta Cryst, B34 (1978) 1037. K.A. Wilhelmi, K. Waltersson and L. Kihlborg, Acta Chem. Scand., 25 (1971) 2675. P.D. Dernier, Material Research Bulletin, 9 (1974) 955. T. Ohno, Y. Nakamura and S. Nagakura, J. Solid State Chem., 56 (1985) 318. F. Uebi, Helv. Chim. Acta, 31 (1948) 8. L.N. Galkin, V.V. Vavilova and L.E. Fykin, Izvestiya Akademii Nauk SSSR, Neorganicheskie Materialy, 13 (1977) 1839. R.C. Bell, K.A. Zemski and A.W. Castlemam, J. Phys. Chem., A103 (1999) 2992, J. Cluster Sci., 10 (1999) 509. R.C. Bell, K.A. Zemski, K.P Kerns, H.T. Deng and A.W. Castleman, J. Phys. Chem., A102 (1998) 1733. G. Anderson, Acta Chem. Scand., 10 (1956) 623. A. Chakrabarti, K. Hermann, R. Druzinic, M. Witko, F. Wagner and M. Petersen, Phys. Rev., B59 (1999) 10583. V. Eyert, in: M. Springborg (Ed.), "Density Functional Methods : Applications in

194

[83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94]

[95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [Ill] [112] [113] [114]

Chemistry and Materials Science", Wiley, Chichester, 1997 and references therein. V. Eyert, K.H. Hock, Phys. Rev., B57 (1998) 12727. X. Yin, A. Endou, R. Miura, A. Fahmi, I. Gunji, R. Yamauchi, M. Kubo, K. Teraishi and A. Miyamoto, Comp. Mat. Sci., 14 (1999) 114. H. Chiba, K. Nishidate, M. Baba, N. Kumagai, T. Sato and K. Nishikawa, SoHd State Commun., 110 (1999) 497. R.W.G. Wyckoff, "Crystal Structures", Interscience Publishers, John Wiley & Sons, Inc., New York-London-Sydney, 1965. R. Ramirez, B. Casal, L. Utrera and E. Ruiz-Hitzky, J. Phys. Chem., 94 (1990) 8960. J. Y. Kempf, B. Silvi, A. Dietrich, C. R. A. Cadow and B. Maigret, Chem. Mater., 5 (1993)641. W. Lambrecht, B. Djafari-Rouhani, M. Lannoo and J. Vennik, J. Phys. C.: SoUd State Phys., 13(1980)2485. D. W. Bullett, J. Phys. C : Solid State Phys., 13 (1980) L595. J.C. Parker, D.J. Lam, Y.N. Xu and W.Y. Ching, Phys. Rev., B42 (1990) 5289. S. Shin, S. Suga, M. Taniguchi, M. Fujisawa, H. Kanzaki, A. Fujimori, H. Daimon, Y. Ueda, K. Kosuge and S. Kachi, Phys. Rev., B41 (1990) 4993. B. Tepper, Ph.D. Thesis, Free University, Berlin, 2000. K. Hermann, M. Witko, R. Druzinic, A. Chakrabarti, B. Tepper, M. Eisner, A. Gorschluter, H. Kuhlenbeck and H.J. Freund, J. Electr. Spectr. Rel. Phen., 98/99 (1999) 245. N. Van Hieu, D. Lichtman, J. Vac. Sci. Technol., 18 (1981) 49. S.F. Cogan, N.M. Nyugen, S.J. Perrotti and R.D. Rauh, J. Appl. Phys., 66 (1989) 1333. A.Z. Moshfegh, A. Ignatiev, Thin Sohd Films, 198 (1991) 251. S. Atzkem, S.V. Borisenko, M. Knupfer, M.S. Golden, J. Fink, A.N. Yaresko, V.N. Antonov, M. Klemm and S. Horn, Phys. Rev., B to be published. M. Hybertsen, S.G. Louie, Comments Cond. Matter Phys., 13 (1987) 5. W. Lambrecht, B. Djafari-Rouhani and J. Vennik, J. Phys. C : Solid State Phys., 14 (1981)4785. E. Caruthers, L. Keinman, Phys. Rev., B7 (1973) 3760. M. Gupta, A.J. Freeman and D.E. EUis, Phys. Rev., B16 (1977) 3338. A.V. Nikolaev, Y.N. Kostrubov and B.V. Andreev, Fiz. Tverd. Tela, 34 (1992) 3011 [Sov. Phys. Solid State, 34 (1992) 1614]. R.M. Wentzcovitch, W.W. Schulz and P.B. Allen, Phys. Rev. Lett., 72 (1994) 3389. A. Chakrabarti and K. Hermann, to be published. J.M. Honig, Acta Phys. Pol., A 97 (2000) 141. K. Itoh, J. Phys. Soc. Japan, 68 (1999) 322. S.Y. Ezhov, V.I. Anisimov, D.I. Khomskii and G.A. Sawatzky, Phys. Rev. Lett., 83 (1999)4136. M. Catty, G. Sandrone, Faraday Discuss., 106 (1997) 189. V. Heine, L.F. Mattheiss, J. Phys., C4 (1971) L191. L.F. Mattheiss, Phys. Rev., B5 (1972) 290. S. Tewari, Solid State Commun., 11 (1972) 1139. D.J. Kraan, Solid State Commun., 15 (1974) 991. A. Neckel, P. Rastl, R. Eibler, P. Weinberger and K. Schwarz, J. Phys., 9 (1976) 579.

195 [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139] [140] [141] [142]

[143] [144] [145] [146] [147] [148] [149] [150] [151]

R. Dagys, A. Kancerevicius, Phys. Rev., B53 (1996) 997. C. Rebhein, N.M. Harrison and A. Wander, Phys. Rev., B54 (1996) 14066. M. Li, W. Hebenstreit, U. Diebold, M.A. Henderson and D.R. Jennison, Faraday Discuss., 114(1999)245. R.S. Mulliken, J. Chem. Phys., 23 (1955) 1833, 1841, 2388, 2343. I. Mayer, Chem. Phys. Lett., 97 (1983) 270. I. Mayer, J. Mol. Struct. (Theochem), 149 (1987) 81. A. Michalak, M. Witko and K. Hermann, Surf. Sci., 375 (1997) 385. K. Hermann, M. Witko, R. Druzinic, Faraday Disc, 114 (1999) 53. K. Hermann, M. Witko, R. Druzinic and R. Tokarz, Topics in Catal., 11/12 (2000) 67. J.P. Perdew, K. Burke, M. Emzerhof, Phys. Rev. Lett., 77 (1996) 3865. B. Hammer, L.B. Hansen and J.K. Norskov, Phys. Rev., B 59 (1999) 7413. K. Hermann, M. Witko, R. Tokarz and R. Druzinic, Appl. Phys., (2000) in print. X. Yin, A. Fahmi, A. Endou, R. Miura, L Gunji, R.Yamauchi, M. Kubo, A. Chatterjee and A. Miyamoto, Appl. Surf. Sci., 130-132 (1998) 539. A. Haras, M. Witko and K. Hermann, unpublished. I. Czekaj, M. Witko and K. Hermann, unpublished. G. Pacchioni, A. Basile, J. Non-Cryst. Solids, 254 (1999) 17. see Chapter 3 of this book. R. Druzinic, Ph.D. Thesis, Free University, Berlin (2000). Z.M. Zhang, V.E. Henrich, Surf. Sci., 321 (1994) 133. R. Druzinic, K. Hermann, unpublished. H.J. Freund, private communication. M. Witko, K. Hermann, J. Mol. Catal., 81 (1993) 279. M. Witko, K. Hermann and R. Tokarz, J. Electr. Spectr. Rel. Phen., 69 (1994) 89. K. Hermann, A. Michalak and M. Witko, Catalysis Today, 32 (1996) 321. M. Witko, R. Tokarz and K. Hermann, PoUsh J. Chem., 72 (1998) 1565. M. Witko, K. Hermann and R. Tokarz, Catalysis Today, 50 (1999) 553. K. Hermann, A. Chakrabarti, R. Druzinic and M. Witko, Phys. Stat. Sol., 173 (1999) 195. M. Witko, K. Hermann, R. Tokarz, R. Druzinic und A. Chakrabarti, in: N. Russo, D. Salahub (Eds.), "Metal Ligand Literactions in Chemistry, Physics, and Biology", NATO Science Series C, Vol. 546, Kluwer Academic Pubhshers, Dordrecht, p. 417. V.A. Ranea, J.L. Vicente, E.E. Mola and R.U. Mananu, Surf. Sci., 441 (1999) 498. X. Yin, A. Fahmi, H. Han, A. Endou, S.S.C. Ammal, M. Kubo and A. Miyamoto, J. Phys. Chem., 103 (1999) 3218. X. Yin, H. Han, L Gunji, A. Endou, S.S.C. Ammal, M. Kubo and A. Miyamoto, J. Phys. Chem., 103 (1999) 4701. F. Gilardoni, J. Weber and A. Baiker, J. Phys. Chem., AlOl (1997) 6069. F. Gilardoni, J. Weber and A. Baiker, Int. J. Quantum Chem., 61 (1997) 683. X. Yin, H. Han and A. Miyamoto, Phys. Chem. Chem. Phys., 2 (2000) 4243. D.C. Sayle, D.H.Gay, A.L. Rohl, C.R.A. Catlow, J.H. Harding, M.A. Perrin and P. Nortier, J. Mat. Chem., 6 (1996) 653. K.P. Huber and G. Herzberg, "Molecular Spectra and Molecular Structure Constants of Diatomic Molecules", Van Nostrand, New York, 1979. see Chapter 8 of this book.

196 [152] [153] [154] [155] [156] [157] [158]

[159] [160] [161] [162] [163]

[164] [165] [166] [167] [168] [169] [170] [171] [172] [173] [174] [175] [ 176] [177] [178] [179]

[180]

A. Da Costa, C. Mathieu, Y. Barbaux, H. Poelman, G. Dalmai-Vennik and L. Fiermans, Surf. Sci., 370 (1997) 339. M. Witko, K. Hermann, in: V. C. Corberan, S. V. Bellon (Eds.), Studies in Surface Science and Catalysis, Vol. 82 Elsevier 1994, p. 75. L. Kihlborg, Ark.Kemi, 21 (1963) 357, 443, 461, Acta Chem Scand., 17 (1963) 1485. J.B. Parrise, E.M. McCarron m, R. Von Dreele and Goldstone, J. Sol. State Chem., 93 (1991) 193. McCarron m., J.C. Calabrese, J. Sol. State Chem., 91 (1991) 121. B.G. Brant, A.C. Skapski, Acta Chem Scand., 21 (1967) 661. A. Magneli, Acta Crystallogr., 6 (1953) 133, 495, Nature, 115 (1950) 356, Arkiv. Kemi, 1 (1949) 223, 1 (1950) 513, 21 (1963) 365, Acta Chim.Scand., 2 (1948) 501, 861. L.A. Bursill, Proc. R. Soc. (London), A311 (1969) 267. J.B. Goodenough, Prog. Solid State Chem., 5 (1971) 145. E.M. Gaigneaux, M.J. Genete, P. Ruiz and B. Delmon, J. Phys. Chem., B104 (2000) 5724. J.C. Volta, O. Bertrand and N. Floquet, J. Chem. Society - Chem. Comm., 19 (1985) 1283. J.B. Goodenough in: H.F. Barry, P.C.H. Mitchell (Eds.), Proc. Climax 4th Intern. Conf. Chemistry and Uses of Molybdenum, Climax Molybdenum Co., Ltd. London 1982, p.l A.J.H. Komdeur, L. de Boer and S. van Smaalens, J. Phys. Cond. Matter, 2 (1990) 45. M. Sato, M.Onoda and Y. Matsuda, J. Phys., C20 (1987) 4763. H. Fujishita, M. Sato, S. Sato and H. Hoshinos, J. Sol. State Chem., 66 (1987) 40. R.L. Smith, G.S. Rohrer, J. Sol. State Chem., 124 (1996) 104. H. Sowa, W. Streurer and J.L. Deboer, Phase Transition, 47 (1994) 1. E. Canadell, M.H. Whangbo, hiorg. Chem., 29 (1990) 2256. M. Sato, H. Fujishita, S. Sato and S. Hoshino, J. Phys. C: Solid State Phys., 19 (1986) 3059. M. Sato, K. Nakao and S. Hoshino, J. Phys. C: Solid State Phys., 17 (1984) L817. R. Knorr, U. Mueller, Zeitsch. Anorg. Allgem. Chemie, 621 (1995) 541. M. Ghedira, H. Vincent, M. Marezio, J. Marcus and Furcaudot, J. Sol. State Chem., 56 (1985) 66. H.K. Fun, P. Yang, M. Sasaki, M. Inoue and H. Kadomatsu, Acta Cryst. Sec.C: Crystal Struct. Comm., 55 (1999) 841. H.K. Fun, P. Yang, M. Sasaki, M. Inoue and H. Kadomatsu, Powder Diffraction, 14 (1999)284. H. Guyot, Physica, B261 (1999) 261. E. Canadelle, M.H. Whangbo, C. Schlenker and C. Escribefilippini, Inorg. Chem., 28 (1989) 1466. J.P. Sorbier, H. Tortel, Physica, B194 (1994) 1299. M. Nakatake, M. Tamura, H. Namatame, M. Taniguchi, Y. Ueda, K. Morikawa, T. Mizokawa, A. Fujimori, H. Negishi and M. Inoue, J. Electr. Spectr. and Related Phen., 78 (1996) 485. Z. Zhu, V.C. Long, J.L. Musfeldt, X. Wei, J. Sarrao, Z. Fisk, H. Negishi, M. Inoue, H.J. Koo and M.H. Whangbo, J. de Physique IV, 9 (1999) 251.

197 [181] [182] [183] [184] [185] [186] [187] [188] [189] [190] [191] [192] [193] [194] [195] [196] [197] [198] [199] [200] [201] [202] [203] [204] [205] [206] [207] [208] [209] [210] [211] [212] [213] [214]

M. Sasaki, M. Inoue, N. Miyajima, Y. Mishima and H. Negishi, Physica, B284 (2000) 1720. A. Terrasi, M. Marsi, H. Berger, F. Gautjier, L. Forro, G. Margaritondo, R.J. Kelley and M. Onellion, Zeitsch. fur Physik B: Cond. Matter, 100 (1996) 493. Z. Zhu, S. Chowdhary, V.C. Long, J.L. Musfeldt, H.J. Koo, M.H. Whangbo, X. Wei, H. Negishi, M. Inoue, J. Sarro, and Z. Fisk, Phys. Rev., B61 (2000) 10057. M. Koyano, A. Miyata and H. Hara, Physica, B284 (2000) 1663. S. Ohara, H. Negishi and M. Inoue, Phys. Stat. Sol., B72 (1992) 419. E. Canadell, M.H. Whangbo, Interm. J. Mod. Phys., B7 (1993) 4005. L.A. Bursill, Acta Cryst, A28 (1972) 187. G.S. Rohrer, W. Lu, R.L. Smith and A. Hutchinson, Surf. Sci., 292 (1993) 261. O. Bertrand, P. Dufour, N. Floquet and L.C. Dufour, Phys. Stat. Sol., A71 (1982) 511. N. Schoenberg, Acta Chem.Scand., 8 (1954) 617. J. Berkowitz, M.G. Ingraham and W.A. Chupta, J. Chem. Phys., 26 (1957) 843. E.F. Fialko, A.V. Kikhtenko and V.B. Goncharov, Organometallics, 17 (1998) 25. E.F. Fialko, A.V. Kikhtenko, V.B. Goncharov and K.L Zamaraev, J. Phys.Chem., AlOl (1997) 8607. D.L. Neikirk, J.C. Fagerh, M.L. Smith, D. Mosman and T.C. Devore, J. Mol. Struct, 244 (1991) 165. Ch.W. Bauschlicher and P.S. Bagus, J. Chem. Phys., 82 (1985) 3265. E. Broclawik, J. Haber, J. Catal., 72 (1981) 379. E. Broclawik, D. Salahub, J. Mol. Catal., 82 (1993) 117. E. Broclawik, D. Salahub, Mem. J. Quant. Chem. Symp., 26 (1992) 393. I. Kretzschmar, A. Fiedler, J.N. Harvey, D. Schroder, and H. Schwarz, J. Phys. Chem., AlOl (1997) 6252. A.C. Tsipis, Phys. Chemistry Chem. Phys., 2 (2000) 1363. L.E. Firment, A Ferretti, Surf. Sci., 129 (1983) 155. L.E. Firment, A. Ferretti, M.R. Cohen and R.P. Merrill, Langmuir 1, (1985) 166. L. Seguin, M. Figlarz, R. Cavagnat and J.C. Lassegues, Spectr. Acta Part A: Mol. and Biolomol. Spectr., 51 (1995) 51. G.A. Nazri, C. Julien, Sol. State Ionics, 53 (1992) 376. F. Cora, A. Patel, N.M. Harrison, C, Roetti and C.R.A. Catlow, J. Mat. Chem., 7 (1997) 959. A. Papakondylis, Ph. Sautet, J. Chem. Phys., 100 (1996) 10681. E. Broclawik, A.E. Fotti and V.H. Smith Jr., J. Catal., 51 (1978) 380, 67 (1981) 103. V. Eyert, R. Homy, K.H. Hock and S. Hom, J. Phys: Cond. Matter, 12 (2000) 4923. T.A. Sasaki, T.Soga and H. Adachi, Phys. Stat. Sol., Bl 13 (1992) 647. I.G. Falkov, S.G. Gagarin and O.P Yablonskii, Zhum. Neorgan. Khimi, 27 (1982) 1101. R. Tokarz-Sobieraj, K. Hermann and M. Witko, in preparation. R. Tokarz-Sobieraj, K. Hermann, M. Witko, A. Blume, U. Wild, G. Mestl, A. Knop-Gericke and R. Schlogl, in preparation. N. Godbout, D.R. Salahub, J.Andzelm and E. Wimmer, Can. J. Phys., 70 (1992) 560. The DFT-LCGTO program package DeMon was developed by A. St.-Amant and D. Salahub (University of Montreal). Here a modified version with extensions by L. G.

198

[215] [216] [217] [218] [219] [220] [221] [222] [223] [224] [225] [226] [227] [228]

[229] [230] [231] [232] [233] [234] [235] [236] [237] [238] [239] [240] [241] [242]

M. Pettersson and K. Hermann is used. S.H. Vosko, L. Wilk and M. Nusair, Can. J. Phys., 58 (1980) 1200. J. Andzelm, E. Radzio and D.R. Salahub, J. Chem. Phys., 83 (1985) 4573. G. Grymonprez, L. Fiermans, and J. Vennik, Surf. Sci., 36 (1973) 370. M. Chen, U.V. Wagmare, C M . Friend and E. Kaxiras, J. Chem. Phys., 109 (1998) 6854. S. Yuan, J. Wang, Y. Li and S. Peng, to be pubhshed. A. Michalak, K. Hermann and M. Witko, Surf. Sci., 366 (1996) 323. T.H. Fleisch, G.W. Zajac, J.O. Schreiner and G.J. Mains, Appl. Surf. Sci., 26 (1986) 488. A. Katrib, V. Logic, N. Saurel, P. Wehrer, L. Hilaire and G. Maire, Surf. Sci., 377 (1997) 754. A. Katrib, P. Leflaive, K. Hilaire and G. Maire, Catal. Lett., 38 (1996) 95. S. Ramani, J.F. Richardson and R, Miranda, Modeling and Simulations in Mat. Sci. and Eng., 7 (1999) 459. B. Delmon, Catalysts Deactivation, 111 (1997) 39. E. Serwicka, J. Solid State Chem., 51 (1984) 300, Crit. Rev. Surf. Science, 1 (1990) 27. R. Tokarz-Sobieraj, M. Witko and K. Hermann, in preparation K. Hermann, M. Witko and A. Michalak, Proc. on Symp. "Advances and Aplications of Computational Chemical Modeling to Heterogeneous Catalysis" ACS Meeting, San Francisco, Vol.42. (1997), p. 106. K. Hermann, M. Witko and A. Michalak, Catal. Today, 50 (1999) 567. A. Rahmouni, C. Barbier, J. Mol. Struct. (THEOCHEM), 33 (1995) 359. J.N. Allison, W.A. Goddard m, J. Catal., 92 (1985) 127. B. Mgoyen, A. Juan and N. Castellani, J. Catal., 190 (2000) 14. M. Chen, C M . Friend and E. Kaxiras, J. Chem. Phys., 112 (2000) 9617. S.P. Mehandru, A.B. Anderson, J.F. Brazdil and R.K. Grasselli, J. Phys. Chem., 91 (1987) 2930. A.B. Anderson, D.W. Edwin, Y. Kim, R.K. Grasselli, J.D. Burrington and J.F. Brazdil, J. Catal., 93 (1985) 222. A.B. Anderson, N.K. Ray, J. Am. Chem. Soc, 107 (1985) 253. T. Ressler, O. Timpe, T. Neisius, J. Find, G. Mestl, M. Dieterle and R. Schogl, J. Catal., 191 (2000) 75. A. Guerrero-Ruiz, I. Rodriguez-Ramos, P. Ferreira-Aparicio, M. Abon and J.C Volta, Catal. Today, 32 (1996) 223. H. Vincent, M. Marezio, in: C Schlenker (Ed.), "Low -Dimensional Electronic Properties of Molybdenum Bronzes and Oxides", Kluwer, Dordrecht, 1989, p.49. A.K. Rappe, W.A. Goddard m., J. Am. Chem. Soc, 102 (1980) 5115, 104 (1982) 448, 3287. A.K. Rappe, W.A. Goddard m, Nature, 285 (1980) 311. R.F. Nalewajski, A. Michalak, J. Phys. Chem., 100 (1996) 29976, 102 (1998) 636.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

199

Chapter 5

Geometry of adsorbates on metal oxide surfaces R. Lindsay, B.G. Daniels and G. Thornton Surface Science Research Centre and Chemistry Department, Manchester University, Manchester Ml3 9PL, United Kingdom 1. INTRODUCTION Currently, considerable research effort, both basic and applied, is being devoted to understanding surfaces of metal oxides on an atomic scale. This work is partly motivated by the huge significance of these surfaces in a variety of commercially important applications which can potentially benefit from such detailed knowledge. These include heterogeneous catalysis, corrosion inhibition, electronic component manufacture and the emerging fields of nanoand bio- technologies. Along with other chapters in this volume, this contribution is concerned with fundamental studies on such surfaces, performed on single crystal metal oxides (or highly oriented thin films). Here we focus on quantitative structural determinations of adsorbates on these surfaces, knowledge of which is a necessary prerequisite for comprehending, and eventually predicting various surface phenomena e.g. surface reaction mechanisms. There is a wide variety of experimental surface science techniques for quantitative structural determinations (see for example Refs. 1 and 2), and many of them have been applied to studying adsorbate/metal oxide systems, as will be seen below. It should be noted that we have excluded work involving scanning probe microscopy (SPM) which normally provides only a qualitative or at best a semi-quantitative picture of adsorption geometry. The extent to which adsorption geometries have been elucidated varies widely, ranging from merely extracting the angular orientation of an adsorbate to precise determination of the atomic coordinates. The amount of structural information obtained depends both upon the experimental technique employed, and the depth of analysis. Compared to the sizeable volume of quantitative structural determinations of adsorbates on metals and semiconductors (see for example Ref 3) the number

200

of similar studies on metal oxides is somewhat restricted. In our opinion this limited archive of literature is not due to a lack of interest in this area but rather the following: (i) (ii) (iii) (iv)

Insulating nature of metal oxides Difficulties with surface preparation Degradation of surface during measurement Lack of structural data for the clean surfaces

The insulating nature of metal oxides (point (i)) has restricted the number of such studies since a majority of the relevant experimental techniques involve the incidence and/or emission of charged particles, usually electrons. In an insulator this loss/gain of charged particles normally results in significant surface charging due to the low electrical conductivity of the solid, and this severely distorts the data, making structure determination usually an intractable problem. For some oxides the problem can be overcome simply by increasing electrical conductivity via reduction of the bulk through thermal treatment (e.g. Ti02) [4], whereas for other metal oxides the problem is less easily solved. One way around it is to employ photon in/photon out techniques such as surface x-ray diffraction (SXRD) or fluorescence yield surface extended x-ray absorption fine structure/near edge x-ray absorption fine structure (SEXAFS/NEXAFS). Another option is to conduct studies on highly ordered ultra thin metal oxide films grown on conducting substrates (e.g. NiO(100)/Ni(100) [5]). This approach allows one to obtain surfaces which can mimic, if prepared correctly, the geometric and electronic structure of a single crystal metal oxide surface rather well, whilst avoiding charging problems. Difficulties with surface preparation (point (ii)) mean that for many metal oxides it has been, and still is in many cases, a significant task just to prepare clean metal oxide surfaces reproducibly and reliably. In fact, currently one could state that the preparation of most metal oxide surfaces, other than those of some well studied binary metal oxides, largely remains something of a black art. Commonly, in situ surface preparation involves cycles of inert gas ion sputtering and high temperature annealing, perhaps followed by annealing in a partial pressure of O2 to restore surface stoichiometry. However, for compositionally more complex metal oxides, including high temperature superconductors, the most promising techniques are either in situ cleavage or thin film growth. An excellent illustration of the difficulties associated with the surface preparation of metal oxides is provided by the (110) surface of rutile Ti02, which has long been considered a prototypical metal oxide surface. From qualitative low energy electron diffraction (LEED) data it was believed.

201

for a number of years, that one could relatively easily prepare the unreconstructed (1x1) surface phase [4]. Recent scanning tunneling microscopy (STM) studies (e.g. Refs 6,7), however, have demonstrated that LEED alone cannot be relied upon to confirm unequivocally surface quality. STM images show that surface geometry depends strongly upon both in situ preparation conditions, and the thermal history of the sample, in particular the extent of bulk reduction of the single crystal. Degradation of metal oxide surfaces during measurement (point (iii)), is seemingly rather a common problem, in particular for molecular adsorbate covered surfaces. The damage occurs due to impinging (or outgoing) photons/electrons/ions destroying the surface through one or more of three processes: desorption, dissociation and disordering. For example the (2x1) LEED pattern formed following acetic acid adsorption on Ti02(l 10)1x1 reverts to (1x1) within 30 seconds using standard LEED optics [8]. Destruction on this timescale normally renders the collection of appropriate data almost impossible, since data acquistion times are normally significantly longer. There are, however, methods to circumvent this problem. For example for LEED studies, like the one just described, one may utilise a modified LEED optics, which operates with electron beam current densities several orders of magnitude lower than normal, thus exposing the sample to a much lower dose of electrons during the measurement. The viability of this approach has been demonstrated for weakly bound molecular adsorbates on MgO(lOO) [9,10]. Lack of structural data for clean metal oxide surfaces (point (iv)), has also hindered quantitative structural determinations of adsorbates on these surfaces. This is particularly true for those techniques providing a rather complete description of the adsorption geometry (e.g. quantitative LEED (LEED-IV)), since at least some knowledge of the clean surface geometry is usually necessary for structure elucidation. The reasons for the lack of structural data for clean metal oxide surfaces are basically identical to the previous three points (i-iii) for adsorbate covered surfaces. As the existence of this review demonstrates, the various experimental problems, described above, have not discouraged everybody from attempting to perform structural determinations of adsorbates on metal oxide surfaces. The following sections of this paper detail such work. It is structured such that there is a separate section on each metal oxide for which quantitative structural determinations of adsorbates have been carried out. We have restricted ourselves to only those studies involving monolayer/submonolayer adsorbate coverages, but have included atoms, molecules and metal adsorption in this coverage regime.

202

2. MAGNESIUM OXIDE - MgO Under normal conditions MgO exhibits the rocksalt structure. Cleavage of MgO occurs preferentially along the non-polar (100) face (see Fig. 1), revealing a relaxed surface which is not far removed from the ideal bulk termination. The other low Miller index surfaces, (111) and (110), have significantly higher surface energies, and have a tendency to reconstruct to display (100) facets, at least in vacuum [11]. Only the structure of adsorbates on the (100) surface have been investigated quantitatively so far. The primary hurdle to overcome for such studies is that MgO is highly insulating. Although thin films of MgO(lOO) have been utilised to circumvent this problem in a number of other surface studies, most notably by Goodman et al (see for example Ref. 12), somewhat surprisingly nobody has exploited this technique for quantitative structural determinations of adsorbates. The most probable reason for this concerns the degree to which the films are representative of the single crystal (100) surface [13]. 2.1. MgO(lOO) - H2O Water adsorption on MgO(lOO) has been subject of two recent quantitative structural determinations [9,14]. The first of these is a LEED-IV study of a p(3x2) phase [9] formed following exposure of a vacuum-cleaved surface at 200 K to 5x10'^ mbar of water for 30 min, and then maintaining a water partial pressure of 1x10"^ mbar. This procedure resulted in a stable water overlayer of coverage 1 monolayer (ML) (where 1 ML is equal to the number of surface

Fig. L A schematic plan view of the (100) surface of MgO (or NiO).

203

Fig. 2. Theoretically derived minimum energy configuration for a (3x2) unit cell of H2O on MgO(lOO) [9].

Mg cations) suitable for LEED-IV measurements, which were performed using low current LEED optics. The automated tensor LEED method was utilised to determine the surface geometry [15]. Both the atomic coordinates of the water molecules and the atoms in the top MgO layer were varied during the optimisation, a total of 54 structural parameters. The results of a semiempirical potential calculation, described in the same paper, were used as the initial geometry for the structural optimisation. A schematic diagram of the minimum energy configuration obtained from this calculation for the (3x2) unit cell is displayed in Fig. 2. The water molecules lie approximately flat (i.e. the H-O-H molecular plane is approximately parallel to the MgO(lOO) surface) with the oxygen atoms sitting almost directly above surface Mg cations at a perpendicular height of 2.11 A. The authors state that this overlayer geometry is adopted as it favours hydrogen bonding between molecules adsorbed in consecutive rows. The optimum, LEED-IV derived, coordinates of the oxygen atoms in the water monolayer are listed in Table 1 along with the coordinates obtained from the calculation (scattering from hydrogen atoms was ignored as is typical in LEED-IV studies of H-containing molecules, since it is a very weak electron scatterer). From this table it can be seen that the experimentally determined structure is very similar (including the estimated errors) to that obtained from the calculation: perpendicular heights of the oxygens range from 2.05 ± 0.05 A to 2.22 ± 0.05 A. The surface MgO layer, which was not optimised in the theoretical calculation, exhibits only small displacements from bulk termination: less than 0.05 ± 0.05 A along the surface normal and of the order of 0.1 ± 0.15 A in the surface plane. One final point to note is that the

204

Table 1 LEED-IV derived coordinates of oxygen atoms in water molecules (Mj) for MgO( 100)3x2H2O. The error bars are estimated to be ± 0.15 A for the in-plane x,y distances (see Fig. 2 for definition of x and y axes) and ± 0.05 A for the surface-atom distances z. The positions obtained from semi-empirical calculations are in parentheses (from Ref 9). Molecule

x(A)

Ml

0.19(0.00) 2.88 (2.89) 2.90(3.10) 0.06 (0.00) 5.81 (6.05) 5.89 (5.84)

M2 M3 M4 M5 M6

y(A) 0.20 (0.21) 0.23 (0.00) 2.85 (2.92) 3.28(3.19) 2.85 (2.98) 5.94 (5.90)

z(A) 2.22(2.11) 2.05(2.11) 2.14(2.11) 2.07(2.11) 2.11(2.11) 2.10(2.11)

authors suggest that the results of their structural optimisation should be regarded with some caution due to the limited data set and the large number of parameters. The other structural determination following water adsorption on MgO(lOO) was performed employing normal incidence x-ray standing waves (NIXSW) by Liu et al [14]. Again cleaved MgO(lOO) surfaces were employed. The water dosing conditions were, however, completely different from the study described above. Measurements were recorded following (i) exposure to a partial pressure of 1x10"-^ Torr of water for 3 min at 300 K, and (ii) immersion in water at 300 K for 10 min. Only (1x1) LEED patterns were observed after either preparation, although the intensity variation of the diffraction spots from the sample immersed in water were apparently somewhat different, possibly suggesting a modified surface structure. From previous photoemission data [16], recorded by the same group from similarly prepared surfaces, it is concluded that the adsorbed moiety resulting from both preparations is surface hydroxyl, present at coverages of 1.2 ML and 2 ML for preparations (i) and (ii), respectively (the greater than 1 ML coverages arise simply due to fact that besides the binding of OH" species, formed following H2O dissociation, to surface Mg^"^ sites, the H^ counter ions also form hydroxyl species by attachment to oxygen anions in the surface layer (see Fig. 3). NIXSW data were obtained from the MgO(200) diffraction planes, thus providing information about the position of atoms along the surface normal. The locations of both oxygen and magnesium atoms were elucidated using the O KLL and Mg KLL Auger signals, respectively. A method, developed by Woicik et al [17], was employed to extract the surface contribution from the measured NIXSW profiles, and thus determine the surface geometry. For both

205

d200 "^^ Fig. 3. A cross section of the fully hydroxylated MgO(lOO) surface proposed in Ref 14.

surface preparations it was concluded that the perpendicular separation between the oxygen atom of the hydroxyl and the surface layer is 2.13 ± 0.04 A. Assuming, as Liu et al have done, that the hydroxyls are bonded to the surface Mg cations so that the Mg-0 bond Hes along the surface normal, then this distance of 2.13 ± 0.04 A represents the Mg-OH bond length (Fig. 3). Additionally, they found that surface hydroxylation does not induce any vertical shift in the top layer Mg cations. Somewhat surprisingly, at least initially, the derived Mg-OH distance is equal, within experimental uncertainty, to the Mg-0 bond length in the bulk oxide (2.106 A). The authors note, however, that the Mg-O bond lengths in bulk MgO and Mg(0H)2 have been shown to be the same from diffraction studies [18]. Besides the semi-empirical potential calculations discussed above a number of other theoretical methods have also been used to investigate the adsorption geometry of water on MgO(lOO). These include two ab initio calculations, one employing the Car-Parinello approach [19] the other Hartree-Fock methodology [20], performed on isolated water molecules on the surface. Both conclude that adsorption atop a surface magnesium is favourable, with an adsorbatesurface distance of about 2.1 A. They disagree, however, about the orientation of the H-O-H plane, and neither predict the flat geometry suggested by the semi-empirical calculations; a set of Monte Carlo and molecular dynamics simulations [21] does support this flat geometry. Furthermore, neither of these ab initio calculations find that dissociative adsorption is favourable on ideal terrace sites. In contrast, more recent ab initio calculations (e.g. Refs. 22 and 23), which employ larger surface unit cells involving more than one water molecule, indicate that dissociation can occur on the perfect surface to produce hydroxyl species.

206

2.2. MgO(lOO) - C2H2 Only one quantitative structural determination has been performed for acetylene (C2H2) on Mg(lOO), which employed LEED-IV [24]. This work is particularly significant as it represents the first detailed structure determination of a molecular adsorbate, in this case a physisorbed species on an oxide substrate. Data were recorded from a (2x2) overlayer of acetylene, prepared by adsorption onto a cleaved MgO(lOO) surface maintained at 88 K. The (1/2,0) and (0, 1/2) diffraction spots were missing from the LEED pattern at normal incidence, indicating the presence of two orthogonal glide planes. Structure determination involved the same methodology as described in the last subsection for MgO(100)3x2-H2O. The structure predicted by semi-empirical potential calculations [24,25] was used as initial input for the automated tensor LEED structural optimisation. Fig. 4 shows the most favourable structure arising from the theoretical calculations. This has a (2x2) surface unit cell with two glide planes, and contains two acetylene molecules, the centre of mass of each being directly above a surface Mg cation. In the LEED simulations the positions of the carbon atoms and the surface layer Mg and O ions were all optimised (36 structural parameters). The coordinates of the centres of mass and the orientations of the two molecules obtained from the best fit between theory and experiment (Pendry R-factor = 0.14) are displayed in Table 2. Also shown in this table are data from the theoretical calculations and an earlier neutron diffraction experiment [26]. It can be seen that the fully optimised

Fig. 4. The lowest energy structure for the (2x2) acetylene overlayer on Mg(lOO) obtained from semi-empirical potential calculations [24,25].

207

Table 2 Coordinates of the centres of mass and orientations of the two acetylene molecules (1 and 2) in the (2x2) unit cell on MgO(lOO). The values obtained from LEED-IV data [24], neutron diffraction (ND) [26], and semi-empirical potential calculations [24,25] are tabulated. The definitions of x, y, and O are given in Fig. 4. 6 is the polar tilt of the molecule away from the surface plane, and z is with respect to the surface MgO layer. The two calculated z values are for effective charges of ± 1.2 and ± 2, respectively. Molecule/ Method

x(A)

y (A)

z (A)

O

6

1/LEED 1/ND 1/Theory

0.22 ±0.10 ?

-0.16 ±0.10 ?

2.50 ± 0.05 ?

0.00

0.00

2.49/2.39

67° ±10° 45° ±15° 60°

108° ± 5 ° 90° ±15° 90°

2/LEED 2/ND

2.79 ±0.10 7

3.06 ?

2.50 ± 0.05 ?

118° ±10° 135° ±15°

89° ± 5° 90° ± 5°

2/Theory

2.98

2.98

2.49/2.39

120°

90°

LEED-IV structure does not deviate greatly from the theoretically determined structure. The most obvious difference is the tilt of the C-C axis away from parallel to the surface plane (18° ± 5°) for molecule 1. HC-CH bond lengths of 1.2 ± 0.05 A are deduced from the carbon atom positions. The authors note that within experimental error this distance is identical to the C-C bond length of the free molecule (1.18 A), which is as expected for a physisorbed molecule. Surface Mg and O ions were only displaced by amounts less than the experimental uncertainty. 2.3. MgO(lOO) - CO Quantitative structural information pertaining to the orientation of CO molecules adsorbed on MgO(lOO) has been deduced from polarisation infrared spectroscopy (PIRS) [27]. For this study MgO(lOO) surfaces were created via in situ cleavage, and measurements were performed ki a partial pressure of 5x10"9 mbar of CO at various substrate temperatures in the range 32 K to 56 K. IR spectra were acquired in transmission geometry at an angle of incidence of 45° ±1°, with p- and s-polarisation data being obtained alternately. The spectra indicate that the MgO(100)-CO system undergoes a reversible phase transition at approximately 45 K. This is consistent with helium atom scattering (HAS) work, which shows that the surface symmetry changes from c(4x2) to (1x1) as the substrate temperature is increased from below to above 45 K [28]. The higher temperature phase exhibits only one absorption band,

208

detected only with p-polarisation. From this information it was concluded that CO molecules in the (1x1) phase are either oriented perpendicular to the surface plane or are uniformly inclined with random azimuthal orientation. Three absorption features are present in the spectra recorded below 45 K, one being observed with both s- and p-polarisations, the other two with the latter only. Given these data and the HAS results [28], the authors concluded that there are three CO molecules per c(4x2) surface unit cell, and that the molecular axis of one CO molecule lies along the surface normal, whilst the other two are tilted antiparallel to each other. Simulation of the IR spectra produced a tilt angle of 29° away from the surface normal. Theoretical calculations support the conclusion that there are both tilted and upright CO molecules in the c(4x2) phase (e.g. Refs. 29-31). 2.4. MgO(lOO) - CO2 CO2 adsorption on MgO(lOO) has been the subject of both PIRS [32] and C AT-edge NEXAFS [33] measurements. For the PIRS study, preparation of the adsorbate overlayer involved exposure of an appropriately cleaved MgO crystal at 82 K to a CO2 partial pressure of 1x10"^ mbar, which was maintained throughout the measurements. Interestingly, anisotropic broadening of the diffraction spots in the (1x1) clean surface LEED pattern evidenced preferential step orientation, giving rise to a predominantly single domain adsorbate induced (2V2xV2)R45° LEED pattern, at least on freshly cleaved samples. Orientational information was extracted from the dependence of the intensities of the vibrational bands upon the angle of incidence and polarisation of the incoming IR beam. From these data, in combination with LEED, it was concluded that CO2 forms a herring bone structure on the surface with the molecular tilt angle being determined to be 27° ± 10° with respect to the surface plane, and the twist to be 22° ± 20° out of the <001> direction. A contemporary ab initio cluster calculation predicts [34] that an isolated CO2 molecule is adsorbed atop a Mg cation aligned perpendicular to the surface plane, although a flat lying geometry between two neighbouring cations was only slightly less favourable. Overlayer preparation for the C A^-edge NEXAFS study [33] was radically different to that described above. In this work cleaved MgO(lOO) was exposed to 260 Torr of CO2 for 15 min at room temperature. This procedure resulted in the formation of surface CO32-, as evidenced by a sharp resonance in the NEXAFS spectra at 290.2 eV, which was assigned to the Is - 7C* transition of 003^" on the basis of previous work. The angular orientation of the surface CO32- was examined by recording NEXAFS spectra as a function of polar photon incidence angle. The intensity of the 7C* feature varied very little with photon incidence angle, which indicates that either the overlayer is

209

orientationally disordered, or that the tilt angle of the molecules is close to the magic angle [35]. 2.5. MgO(lOO) - Ca Both ion scattering spectroscopy (ISS) with a neutral incident beam [36] and surface x-ray diffraction (SXRD) [37,38] have been applied to the investigation of the structure of Ca segregated onto the MgO(lOO) surface. For the ISS experiment a clean and ordered MgO(lOO) surface was obtained by cleaving in air a MgO single crystal and then repeatedly flash annealing in vacuum to 1073 K. To induce Ca surface segregation, the sample was further annealed up to 1573 K. The LEED pattern was observed to be a clear (1x1) both before and after Ca segregation. A neutral beam of He atoms rather than a beam of He"^ ions was used for the ISS measurements to minimise sample charging problems. Structural information was obtained from the intensity variation with scattering geometry of the spectral peaks due to scattering from Ca and Mg. It was determined, in agreement with theoretical calculations [3942], that Ca substitutes for Mg in the surface layer, and that the surface Mg and O ions are coplanar within experimental error (±0.1 A), whereas Ca is located 0.4 ± 0.1 A above this layer. It is suggested in the paper that both the protrusion of the surface Ca cations, and the Ca surface segregation are driven by the reduction of local strain, which is induced by the ionic radius of Ca2+ being 1.5 times greater than that of Mg2+.

Fig. 5. The optimum structure of the Ca-segregated MgO(lOO) surface determined from SXRD [37,38]. The numerical labels are employed in Table 3.

210

Table 3 z displacements of atoms in MgO(100)(V2xV2)R45°-Ca from analysis of SXRD data [37,38]. The numerical labelling of the atoms corresponds to that in Fig. 5. Both absolute displacements (Az (A)) and shifts relative to the MgO bulk nearest neighbour distance (Az/d) are listed.

l:Ca

2:Mg

Az/d

0.3 ±0.10

-0.032 ± 0.066

Az(A)

0.63 ±0.03

-0.066 ±0.14

3:0

4:0

0 (fixed) 0 (fixed) ±4 ±4 0 ±9

0 ±9

5:Mg

6:Mg

7:0

8:0

0.095 ±0.014

-0.2 ±0.08

-0.2 ±0.14

0.2 ±0.06

0.20 ±0.03

-0.42 ±0.18

-0.42 ±0.32

0.42 ±0.12

The SXRD study [37,38] was performed by Renaud's group. For these measurements the segregation was induced by annealing a sample at 1773 1873 K in air for several hours - the first step in a procedure developed for preparing high quality (100) surfaces of MgO. To remove both carbon contamination and oxygen vacancies, the MgO(100)-Ca surface was further annealed at approximately 1173 K for 30 min in 1x10-4 Torr of O2 in the UHV measurement chamber. In contrast to the ISS work in which the LEED pattern remained (1x1), diffuse, half-order spots were observed, corresponding to a (V2xV2)R45° surface unit cell. This difference probably arises because there was significantly more Ca on the surface in the SXRD study, on account of the significantly higher sample anneal temperature. The surface coverage of Ca was, in fact, estimated to be of the order of 1 ML from Auger electron spectroscopy. Surface structure was elucidated by fitting the (11/) and (20/) crystal truncation rods. Here it was assumed that (i) Ca is in substitutional sites, and (ii) Ca is present only in the topmost atomic plane. The best fit to the experimental data was obtained for a structure having the observed (V2xV2)R45° surface unit cell, in which every other surface Mg cation is replaced by a Ca cation. A schematic diagram of the structure is displayed in Fig. 5, and the vertical displacements of the atoms in the first two layers are listed in Table 3. Qualitatively similar atomic positions were obtained from theoretical calculations performed on the (V2xV2)R45° unit cell, which was found to be energetically favourable [39-42]. 2.6. MgO(lOO) - Ag The atomic scale structure of the MgO(100)/Ag interface has been examined experimentally by SEXAFS [43] and SXRD [37,44-46]. For the SEXAFS experiment an air-cleaved substrate was cleaned in situ by annealing

211

Table 4 Results of best fits to SEXAFS modulations, recorded at normal and grazing incidence, from MgO(100)-Ag at 1 ML equivalent coverage [43]. Also listed are the theoretically expected effective coordination numbers (N*theo) ^^^ ^be adsorption geometry discussed in the text. N*expt (N*theo) Ag-Ag Ag-0 Ag-Ag Ag-0

R (A)

Normal Incidence 8.3 ± 1 (8.5) 3.02 ±0.05 1 ± 1 (0.25) 2.55 ± 0.05 Grazing Incidence 7.3 ± 1 (7) 3.03 ±0.05 2±1(1) 2.53 ±0.05

at 1200 K for 10-20 min in an oxygen ambient of 5x10'4 Pa, followed by cooling at the same oxygen pressure to room temperature. Silver deposition and measurements were undertaken at 300 K. The estimated coverage of silver was 1 ML equivalent (i.e. if continuous over the surface the film would be one atom thick). Ag Ljjj edge SEXAFS data were collected at normal and grazing (15°) photon incidence angles to the surface plane. The results of the best fits to the SEXAFS spectra are Usted in Table 4. The Ag-Ag distance in the overlayer is significantly longer than the nearest neighbour distance in bulk silver (2.89 A). It is, however, within the experimental uncertainty, identical to the lateral separation (2.98 A) of equivalent sites on the MgO(lOO) surface. As to the adsorption site of the silver atoms, the effective coordination numbers (N*exp) ^^^ consistent with adsorption atop a surface ion, fitting favouring oxygen anions over magnesium cations. It should be noted that the authors mention that, on the basis of other work [47], the 1 ML equivalent Ag film is not actually one atomic layer thick, but rather it consists of mostly 2 ML thick islands covering about 60% of the surface area. They took account of this morphology in their structural determination. An SXRD study of this system investigated both the interfacial geometry, and details of the film morphology from submonolayer coverage up to many monolayers thickness [37,44-46]. Their preparation of the MgO(lOO) surface involved firstly, as described in the last subsection, annealing a polished sample in air at 1773 K for 3 h, and then in situ Ar"^ bombardment whilst annealing the sample at 1823 K. The resulting defected surface was improved by oxygen annealing. Silver was deposited at room temperature. Scattered x-ray intensity measured during scans along {hhl = 0.1), and (hOl ==0.1) demonstrated that up to 0.5 ML equivalent coverage the Ag film has, on average, its bulk lattice

212

parameter. This demonstrates that there is only weak coupling between Ag and the MgO(lOO) surface. Between 0.5 ML equivalent and 1 ML equivalent (and beyond) the islands begin to become strained by interaction with the MgO(lOO) substrate. Details about the adsorption site of on site Ag ( i.e. Ag atoms that sit on a lattice having the same in-plane lattice parameter as the substrate) in the first layer, which comprises only a small fraction of the total amount deposited, and the interfacial distance were obtained from analysis of (11/) and (20/) crystal truncation rod scans. It was determined, in agreement with the SEXAFS study [43], that Ag preferentially sits atop oxygen anions. The interfacial distance was found to be on average 2.52 ±0.1 A, again consistent with the SEXAFS result. Ab inito calculations of the Mg(100)/Ag interface agree that the most probable adsorption geometry is Ag atop a surface oxygen [48-51]. However, a variety of Ag-0 bond lengths, ranging from 2.34 A to 2.70 A, are obtained. The most recent and sophisticated calculation provides a value of 2.55 A [51]. 2.7. MgO(lOO) - Pd The Renaud group have also investigated details of the MgO(100)/Pd interfacial structure using SXRD [37,52-54]. Prior to Pd deposition the substrate was cleaned and ordered using the method described above for the MgO(100)-Ag SXRD study. Diffraction measurements were made following Pd deposition at room temperature for a wide range of coverages. At 1 ML equivalent coverage it was concluded from analysis of the (20/) and (31/) crystal truncation rods that 50 % of the Pd film is in registry with the substrate, and that in registry first layer Pd atoms are located above oxygen anions, as was found for silver, at a distance of 2.22 ± 0.02 A. This adsorption site contradicts a surface electron energy loss fine structure (SEELFS) study which concluded that for 3-D Pd clusters of diameter 2 nm on MgO(lOO), first layer Pd is located atop Mg cations [55]. The SXRD results are, however, in rather good agreement with several recent ab initio calculations which show that adsorption of Pd atop a surface layer oxygen is favourable, and predict Pd-O bondlengths ranging from 2.11 - 2.23 A [51,56-58]. 2.8. MgO(lOO) - Ni The MgO(100)-Ni interface was the subject of another SXRD study [59]. Surface preparation replicated that described above and SXRD data were recorded for Ni coverages from 0.2 to 125 ML equivalent. From simulations of the (11/) crystal truncation rod it was again concluded that part of the Ni film adsorbs on site, and that the epitaxial site is above a surface oxygen. For coverages up to 1 ML equivalent the distance between the MgO surface and

213

first Ni layer is approximately 1.82 A. Theoretical calculations are once more in rather good agreement with the experimental results [e.g. Refs. 51 and 60]. 2.9. MgO(lOO) - Fe Urano and Kanaji have utilised LEED-IV to examine the geometric structure of an approximately 1 ML equivalent film of Fe on MgO(lOO) [61]. The MgO(lOO) substrate was prepared by cleavage in air followed by annealing up to 1073 K in a partial pressure of oxygen (lxlO"3 Pa) inside the UHV measurement chamber. Iron film growth was performed with the sample surface at room temperature. A sharp 1x1 LEED pattern was observed at a film thickness of around 1 ML, which the authors interpret as evidence for the film being pseudomorphic. LEED-IV curves were acquired with an electron beam incidence angle of 6.1°. To determine the Fe adsorption site and height of the Fe above the surface layer, experimental data were compared with simulated data, calculated for three different adsorption sites: atop oxygen, atop magnesium and bridge. It should be noted that the position of the substrate ions were not optimised, but rather remained fixed in their bulk terminated positions. From this analysis it was determined that Fe adsorbs above a surface oxygen anion at a distance of about 2.0 A. A theoretical calculation for this system also finds that adsorption atop an oxygen is energetically preferred, but at a rather larger interfacial spacing of 2.3 A [62]. In contrast to the pseudomorphic relationship proposed in the LEED study, SXRD data [37,63] from the MgO(100)-Fe system indicate that for a 1 ML equivalent film the Fe lattice parameter is approximately 2.89 A, which is close to that of bulk Fe. For the latter study, however, a rather different preparative procedure was employed. Firstly, the MgO substrate was cleaned in solvents ex situ and only heated to 633 K in UHV. Then the sample was maintained at 633 K, whilst Fe was sputtered onto it, using a planar magnetron sputter gun. 3. NICKEL OXroE - NiO Nickel oxide, like MgO, usually adopts the relatively simple rocksalt lattice. The natural cleavage plane of NiO is (100), and studies have shown that the resulting surfaces are of high quality, relaxing only slightly away from the ideal bulk terminated (100) surface (see Fig. 1). Structural determinations of adsorbates have been performed on both this surface and the polar (111) surface. To circumvent surface charging problems almost all of these studies have been performed on highly oriented NiO thin films.

214

3.1. Ni(lOO) - NO N K-QdgQ NEXAFS [64] and scanned-energy mode N Is photoelectron diffraction (PhD) [65,66] have been employed to investigate the adsorption geometry of NO on NiO(lOO). Both of these studies utiUsed a NiO(lOO) thin film rather than the surface of a single crystal. Film preparation involved oxidation of a Ni(lOO) surface at elevated temperature, a procedure used extensively by Freund's group, who performed the NEXAFS measurements. Such NiO(lOO) thin films are known to contain a high density of surface defects, which could drastically affect adsorption properties. However, it has been clearly demonstrated that for NO the adsorption behaviour is dominated by ideal (100) surface sites [64,67]. The N ^-edge NEXAFS data were recorded from a NO saturated substrate, by monitoring the N KLL Auger yield as a function of photon energy. Spectra were recorded at a range of photon incidence angles between normal and grazing. The angular dependence of the intensity of the leading 7i* resonance was extracted and fitted with the appropriate trigonometric expression to determine the angular orientation of the N-O bond axis. The angle deduced was 45°. The PhD investigation [65,66] was performed on a NiO(lOO) thin film dosed with a saturation exposure of NO at 135 K. A partial pressure of NO of approximately 7x10"9 mbar was maintained in the experimental chamber during PhD data collection. This was done to overcome problems with photon beam induced NO dissociation/desorption (for more details see Ref 66). Two peaks, separated by approximately 4.4 eV, were observed, as had been done previously [64], in the N Is core level region. Comparing the diffraction displayed by each of these two peaks confirmed Freund et aFs assertion that the two peaks arise due to two different core hole states of the same surface species [64]. Quantitative elucidation of the geometry of this adsorbate involved a two-step process. Initially, a visual inspection of the PhD spectra was carried out to ascertain the approximate position of the N atom relative to its nearest neighbours below. Next, this geometry was used as an initial guess for the generation of simulated PhD data via electron multiple scattering calculations, which were compared to the experimental data set. Agreement between experimental and theoretical data was iteratively improved by varying structural parameters until the best fit, and therefore the optimised structure, was obtained (R-factor = 0.09: this R-factor is defined differently to the Pendry LEED Rfactor, see Ref 68 for details). NO is found to adsorb N atom down atop a Ni atom, and in agreement with the NEXAFS data [64], the N-O bond axis is tilted away from the surface normal. This optimised adsorbate geometry is shown ia Fig. 6 and the values of the optimised parameters are listed in Table 5. Interestingly, although theoretical calculations [64,69,70] favour this adsorption

215

Fig. 6. Schematic of the optimum NO adsorption geometry on NiO(lOO) determined from PhD [65,66]. The geometrical parameters varied during the structural optimisation are displayed. For the adsorbate, the darker shaded circle is nitrogen, and the lighter oxygen.

Table 5 Parameter values obtained from the best fit between experimental N Is PhD data and theoretical simulations for NiO(100)-NO [65,66]. The definitions of the first five parameters are given in Fig. 6. and are the mean square vibrational amplitudes of the nitrogen atom parallel and perpendicular to the surface, respectively. The positive error for Y^Q is replaced by an asterisk due to the fact that all N-O bond lengths greater than the optimum lie within the estimated error. The double asterisk next to the error for 02, the N - 0 tilt, indicates that the error given is only for N-0 bond lengths <1.43 A. Parameter

Value

^NiN (^)

1.88 ±0.02 3 + 3/- 8 1.12+*/-0.15 59 + 31/-17** 2.07 ± 0.04 (1.8+ 4.2/-1.8) X 10-2

eiO ^NO (^) 02 0 cii2 (A) (A2) (A2)

(3.8 +1.9) X 10-3

216

geometry, the Ni-N bond length obtained from such work is significantly longer, of the order of 0.2 A, than that obtained experimentally. 3.2. NiO(lOO) - CO CO adsorption on NiO(lOO) is the subject of another NEXAFS study by Freund's group [71]. ANiO(lOO) thin film, similar to those used for the NO studies above, was employed. NEXAFS data were recorded as a function of photon incidence angle in the region of the C A^-edge following dosing of CO below 100 K. Normalised C AT-edge NEXAFS spectra are displayed in Fig. 7. It is interesting to note the weak intensity of the a* feature (at - 305 eV) in comparison to the 7i* resonance (at ~ 287 eV) at all incidence angles. This is very different from NEXAFS spectra recorded following adsorption of CO on metals, where the a* feature is comparable to the 7C*. The authors indicate that this is due to the weak nature of the surface-adsorbate bond. From analysis of the intensity variation of the 7i* resonance it was concluded that the molecular C K-edge NEXAFS Ni{100)/NiO(100)-CO

c

< 0)

c

0

•D CD

280 300 ' 320 Photon Energy (eV) Fig. 7. Carbon C X-edge NEXAFS spectra of CO on a NiO(lOO) thin film recorded as a function of polar photon incidence angle, 9 (from Ref 71). 0 is defined as the angle between the incoming photon beam and the surface plane.

217

axis of the CO molecule is, within ± 15°, perpendicular to the surface plane. Complementary angle-resolved photoelectron spectroscopy data are consistent with the CO molecule being adsorbed C atom down [71]. Theoretical calculations provide support for this experimentally determined orientation ( e.g. Refs. 69 and 72). In addition, they predict that the CO adsorbs atop a surface Ni cation. This has yet to be corroborated experimentally. 3.3. NiO(lOO) - H2S The local coordination geometry of atomic sulfur on NiO(lOO) has been investigated with S K-edge SEXAFS [73]. Data were obtained at room temperature at normal and grazing photon incidence by monitoring the S Ka fluorescence yield. The surface was prepared by exposure of a single crystal NiO(lOO) sample at 570 Kto 15000 L of H2S. Previously, it had been found that such a preparation gave rise to a change in the LEED pattern, which was ascribed to the formation of a Ni(100)c(2x2)-S surface phase on top of the oxide substrate [74]. In the SEXAFS study a c(2x2) pattern was not observed. The authors suggest that this difference may arise because of the formation of

[100]

Fig. 8. Structural model of Ni(100)c(2x2)-S raft onNiO(lOO) consistent with SEXAFS data [73]. The overlayer raft is shown above an azimuthally aligned NiO(lOO) substrate. In the raft, S atoms are depicted as dark grey spheres, and Ni atoms are shown as light grey spheres. For the NiO(lOO) substrate, Ni atoms are black, and oxygen atoms are white. The overlayer c(2x2) and (1x1) unit cells are indicated. The separation of overlayer and substrate is merely for viewing purposes.

218

smaller surface domains. Analysis of the SEXAFS data evidenced only a single shell of Ni scatterers at a distance of 2.21 ± 0.02 A. From a comparison of the experimental effective coordination numbers with those calculated for various S adsorption sites on Ni(lOO), Ni(llO) and Ni(lll), it was concluded that S occupies the four-fold hollow site of an 11 ± 4% laterally expanded Ni(lOO) raft. Fig. 8 shows a structural model of the Ni(100)c(2x2)-S overlayer consistent with the SEXAFS data. It was not possible to estimate the thickness of the Ni raft from the SEXAFS data due to problems concerning Debye-Waller like effects at the measurement temperature. 3.4. N i O ( l l l ) - N O Freund et al have performed a N A^-edge NEXAFS study of NO adsorbed NiO(lll) [75]. Analogous to their adsorption studies on NiO(lOO) discussed above, they recorded the data from a highly oriented thin film. A (111) oriented oxide surface was obtained by annealing Ni(lll) in a partial pressure of oxygen. This procedure produces a NiO(lll)(lxl) surface, on which a significant concentration of -OH is present. Annealing this surface to 600 K desorbs the -OH, and causes the surface to form a (2x2) superstructure, which is generally believed to be due to the presence of an octopolar reconstruction, a phase consisting of an array of tetrahedral protrusions with (100) nano-faces [76]. NO was adsorbed on the NiO(lll)(lxl)-OH surface. The N iC-edge spectra indicate the presence of molecular NO on the surface, but the intensity of the 71* resonance shows little dependence on the photon incidence angle, suggesting a tilt angle of around the magic angle (54.7°). The authors explain this angular orientation on the basis that patches of the octopolar reconstruction exist even on the -OH covered NiO(lll) surface. They propose that NO adsorbs on the (100) nanofaceted surface in a bent geometry similar to that for

(b)

o

Fig. 9. Simple schematic diagrams of the adsorption geometries of (a) NO and (b) CO on the octopolar reconstruction of NiO(l 11), proposed on the basis of NEXAFS results [75].

219

NO on NiO(lOO) [64-66], and that the NO rotates around the Ni-N bond axis (see Fig. 9a). The orientational disorder resulting from this rotation causes the NEXAFS resonances to be independent of the incidence angle of the linearly polarised x-rays. 3.5. NiO(lll) - CO The angular orientation of CO on NiO(lll)(lxl)-OH has also been elucidated by Freund's group, employing C K-edge NEXAFS [75]. The polarisation dependence of the 7C*/a* resonances in the NEXAFS spectra are consistent with a tilt angle of the C-O bond axis of about 46° away from the surface normal. Analogous to their proposition for NO, the authors suggest that CO is adsorbed on the (100) nanofaceted octopolar reconstruction, with its orientation mimicking that found on NiO(lOO) [71]. Their proposed adsorption geometry is depicted in Fig. 9b. This geometry gives a tilt angle for CO of 54.7° away from the NiO(lll) surface normal, which is within the uncertainty of the experimental value. 4. TITANIUM DIOXIDE - Ti02 Three polymorphs of titanium dioxide occur naturally, namely rutile, anatase and brookite, but surface scientists have concentrated almost entirely on the

[001]

Bridging

Fig. 10. Representations of the atomic structure of the Ti02(ll 0)1x1 surface. Top is apian view, bottom is a side view. The larger (smaller) circles correspond to oxygen atoms (titanium atoms).

220

rutile form. In fact, as pointed out in the introduction, the (110) surface of rutile Ti02 is considered to be a prototypical metal oxide surface. Consequently, many adsorbate studies have been performed on this surface, including a number of quantitative structural determinations. The (110) surface commonly exhibits either a (1x1) termination or a reduced (1x2) reconstruction, which has been shown to consist of added rows of TixOy (see, for example, Ref 77). All the work discussed in this review has involved adsorption on the (1x1) phase, a model of which is displayed in Fig. 10. It consists of rows of bridging oxygens which lie above a plane of oxygen and five-fold coordinated Ti4+ atoms, as confirmed by SXRD [78]. In contrast, other low Miller index surfaces of Ti02 have not attracted so much attention in terms of adsorbate structure determinations. Only two such studies have been reported, both involving Ti02( 100). Fig. 11 shows simple schematic diagrams of the (1x1) and (1x3) phases of this surface, both of which will be mentioned below. The (1x3) reconstruction is known to consist of (110) microfacets from previous work (see Ref. 79 and Refs. therein), whereas the displayed (1x1) structure is merely that expected on the basis of Tasker's rule [80]. Ti

o J Bridging O

[001] , f «

[010]

[001] [010] Fig. 11. Models of Ti02(l00)1x1 (top), and TiO2(100)lx3 (bottom).

Ti

221

4.1.TiO2(110)-HCOOH On Ti02( 110)1x1 it is known that formic acid dissociates to form surface formate species, and at approximately 0.5 ML (where 1 ML corresponds to one formate per exposed Ti4+) coverage an ordered 2x1 overlayer is formed [8183]. The adsorption geometry of the formate in this ordered overlay er has been investigated by PhD/x-ray photoelectron diffraction (XPD) [84-86], reflection absorption infra-red spectroscopy (RAIRS) [87], and C K-edgQ NEXAFS [88]. The PhD/XPD study, performed by Chambers et al, provides the most detailed information about the majority species (see below). For this work, prior to formic acid exposure at room temperature, the Ti02( 110)1x1 surface was prepared by cycles of Ar"*" sputtering and annealing in an oxygen plasma at 900 K. XPD data from both the O Is (formate and substrate oxygens) and C Is (formate carbon) core levels were recorded. C Is data exhibit no diffraction features, indicating that formate is bound via both oxygens in a bidentate geometry. The diffraction pattern obtained from the formate oxygen component of the O Is feature is shown in Fig. 12a. Two intense features can be observed, one at 9 = 30°, (|) = 0^ the other at 9 = 30°, (|) = 180° (see Fig. for definition of 0 and (|)). These arise due to forward scattering by the carbon atom in the formate species. From the location of these maxima in the diffraction pattern it can be concluded that the molecular plane of the formate is aligned with the [001] azimuth (see Fig. 10), i.e., it lies parallel to the rows of bridging oxygens. Additionally, the O-C-0 bond angle was estimated to be 126° ± 4°. The azimuthal alignment of the formate provides circumstantial evidence that the formate bonds to two neighbouring titanium cations in a

[001]

30 60 90 60 30

en

[001]

Fig. 12. (a) Formate O Is XPD pattern obtained from Ti02(ll0)2x1-[HCOO]", and (b) a sketch of the local adsorption geometry [84-86].

222

bridging bidentate configuration, as depicted in Fig. 12b. The authors tested this suggestion with PhD, in both scanned-energy and scanned-angle modes, using the O Is peak of the formate. In order to restrict the amount of experimental data required they first performed Hartree-Fock calculations to get an insight into the most probable geometry. The energy minimised structure obtained was in accord with the above hypothesis. In detail, the calculation predicted a Ti-O formate bond length of 2.05 A, a separation of the two formate oxygens of 2.28 A, and a 0-C-O bond angle of 128.7°. Armed with this knowledge they recorded O Is PhD spectra. Analysis of these data, using electron scattering code to simulate the experimental data and assuming the bidentate bridging site, produced a vertical distance between the Ti cation and the formate oxygen of 2.1 ± 0.1 A; the vertical height was the only structural parameter optimised. This agrees well with the bond length emerging from the Hartree-Fock calculations. Other theoretical studies have also concluded that the bridging bidentate geometry is the most energetically favourable bonding configuration for formate on Ti02(l 10)1x1 [89-93]. The RAIRS study of this system [87] has probed the orientation of formate on Ti02( 110)1x1, the substrate being prepared via sputter/UHV anneal cycles with a final anneal in 1x10"" mbar of oxygen to minimise surface oxygen vacancies, p-polarised RAIRS measurements were recorded as a fianction of azimuthal incidence angle. Two bands were observed, one absorption the second transmission, which were assigned to the antisymmetric and symmetric vibrational modes of formate, respectively. These data evidence the presence of not one but two surface formate species (A and B). A, which is the majority species, is oriented such that its molecular plane is in the [001] azimuth, which

[001] TQJ.

'HXH .CCP

[110] • Fig. 13. Structural model for surface fonnate adsorbed at bridging oxygen vacancies on Ti02(l 10)1x1 [87,89].

223

is consistent with the orientation of formate determined by Chambers et al [84-86]. The molecular plane ofB, the minority species, is perpendicular to A. It is proposed that B is adsorbed at oxygen vacancies, as its concentration appears to be linked to their surface concentration. Fig. 13 shows the adsorption geometry of 5, proposed on the basis of ab initio Hartree-Fock calculations by Wang et al for adsorption of formate on a bridging oxygen vacancy [89]. C ^-edge NEXAFS data [88] also provide evidence of a second perpendicularly aligned formate species. These spectra were recorded from a Ti02(l 10) sample that had been cleaned in situ by cycles of Ar"^ bombardment and annealing in UHV up to 900 K, followed by an anneal at 700 K in 1x10"6 mbar of O2. A (2x1) overlayer of formate was formed by dosing formic acid at room temperature. Three peaks appear in the NEXAFS spectra, at 288.7 eV, 292 eV, and 300 eV, which were assigned to resonances of formate on the basis of previous work [94]. The orientation of the formate was determined from the variation in intensity of the 7C* resonance at 288.7 eV with x-ray incidence angle. Spectra were recorded with the electric vector of the photons in both the [001] and [110] azimuths. From analysis of these data it was found that the molecular plane of formate is tilted away from the surface normal by 20° ± 10° and twisted out of the [001] azimuth by 29° ± 10°. Given the PhD/XPD [84-86] and RAIRS [87] resuhs, it was concluded that the apparent twist out of the high symmetry direction most likely arises from an averaging over majority and minority formate species {A and B above). 4.2. TiO2(110) - H3CCOOH Exposure of Ti02( 110)1x1 to acetic acid (H3CCOOH) also leads to a surface carboxylate moiety, in this case acetate, which forms an ordered (2x1) overlayer at saturation [8,88,95,96]. Details about the angular geometry of the acetate species have been elucidated by both electron stimulated desorption ion angular distribution (ESDIAD) [8] and C i^-edge NEXAFS [88]. In the ESDIAD work, conducted by Guo et al, data were recorded from an oxygen annealed Ti02( 110)1x1 surface exposed to sufficient acetic acid to produce a (2x1) LEED pattern at room temperature. A typical ESDIAD pattern obtained from such a surface, which arises from H~^ desorption, consists of two separate components. There is a lobe peaked at normal emission surrounded by a concentric annulus. The central peak is assigned to H^ desorption from surface hydroxyls arising from the dissociation of acetic acid. The off-normal ring of emission is attributed to hydrogens from the methyl group (-CH3) of acetate. From the symmetry of this ring the authors concluded that acetate is oriented upright on the surface. They argue, convincingly, that a ring, rather

224

than discrete off-normal lobes, is expected due to free rotation of the methyl group around the C-C axis. For the NEXAFS study [88] of this system substrate preparation followed that described above for the NEXAFS measurements from Ti02( 110)2x1[HCOO]". The angular geometry of the acetate was determined from the dependence of the leading acetate 7C* resonance on photon incidence angle. In agreement with the ESDIAD results it was found that the molecular plane of acetate lies along the surface normal (i.e. tilt angle away from surface normal is Qo _^ 50^ jj^g twist angle of the adsorbate's molecular plane is 26° ± 5° out of of the [001] azimuth. 4.3. TiO2(110) - H5C2COOH The orientation of a (2x1) overlayer of propanoate ([H5C2COO]") on Ti02( 110)1x1 has also been determined in Ref. 88, employing NEXAFS at the C i^-edge. The procedure for surface preparation duplicated that for the other two TiO2(110)2xl-carboxylate NEXAFS investigations (see 4.1. and 4.2.). A tilt away from the surface normal of 14° ± 10°, and a twist out the [001] azimuth of 39° ± 10° were found for the carboxylate plane from analysis of the 7C* resonance intensity as a function of measurement geometry. This orientation is rather similar to those obtained from NEXAFS for (2x1) overlayers of formate and acetate [88], indicating that the bond geometry is essentially independent of the alkyl chain length. The significant twist of all three carboxylates out of the high symmetry [001] azimuth probably arises from the presence of two perpendicularly aUgned moieties on the Ti02( 110)1x1 surface, as discussed in subsection 4.1 for formate. 4.4. TiO2(110) - H5C6COOH The adsorption of benzoic acid (H5C6COOH) on Ti02(l 10)1x1 echoes that of formic, acetic, and propanoic acid on the same surface, i.e. at room temperature benzoic acid reacts to produce an overlayer of benzoate, displaying a (2x1) LEED pattern at saturation coverage [95-97]. Guo et al have also examined this system with ESDIAD [97]. A typical ESDIAD pattern of H^ ions recorded from a (2x1) overlayer is shown in Fig. 14. It exhibits a central bright spot along with four low intensity off normal features aligned with the principal azimuths of the surface. Guo et al interpret this pattern in terms of an upright benzoate moiety with the phenyl ring azimuthally aligned either parallel or perpendicular to the bridging oxygen rows. The central peak of intensity is suggested to arise due to the desorption of H~^ from both surface hydroxyls and position 4 of the phenyl ring (see inset in Fig. 14). Off-normal features stem from the ejection of H"^ from positions 3 and 5 of the phenyl ring. The

225

Fig. 14. H+ ESDIAD image from TiO2(110)2xl-benzoate [97]. Also displayed is a schematic of the benzoate anion.

azimuthal alignment of the phenyl ring is ascribed to inter-adsorbate hydrogen bonding. Complementary STM images provide evidence for this interaction [95,97]. They show dimerised, and even trimerised [001] rows of benzoates. This was explained in terms of neighbouring rows having perpendicularly aligned phenyl rings, allowing bonding between a hydrogen atom of one ring to the 7C system of the other. 4.5. TiO2(110) - bi-isonicotinic acid The orientation on Ti02( 110)1x1 of bi-isonicotinic acid (2,2'-bipyridine4,4'-dicarboxylic acid, see Fig. 15), a ligand present in a number of photochemically important organometallic dyes, has been investigated with N K-QdgQ ISfEXAFS [98]. Measurements, as a function of photon incidence angle, were recorded from two TiO2(110) samples mounted such that one had the electric vector of the x-ray beam parallel to the bridging oxygen rows whilst for the other it was perpendicular. The preparation of the samples involved in situ Ar+ ion bombardment and annealing in 1.6xl0"6 mbar of oxygen at 903 K. A submonolayer coverage of adsorbate was obtained by deposition at a substrate temperature of 473 K. 0 1s photoemission data were interpreted as indicating that for such a submonolayer coverage the two -COOH groups are deprotonated and that all four oxygens are equivalent. From the angular dependence of the leading N ^-edge 7i* resonance it was determined that the molecules are oriented with a tilt angle of 25° with respect to the surface normal and twisted 44° out of the [001] azimuth. It is, however, not possible from these data to distinguish between an adsorption geometry with the two rings of the molecule coplanar, and one in which they are twisted relative to one another. Complementary periodic INDO calculations, presented in the same

226

Fig. 15. A schematic diagram of the theoretically calculated lowest energy adsorption structure for bi-isonicotinate on Ti02( 110)1x1. The geometry is consistent with NEXAFS data [98]. In the adsorbate, the light spheres in the rings are nitrogens, the dark ones are carbons, the smallest spheres are hydrogens, and the largest spheres are oxygen atoms.

paper, were performed for several possible adsorption geometries. Fig. 15 shows the lowest total energy adsorption structure, which agrees well with the experimental data. Substrate-adsorbate bonding appears to be very similar to that for the carboxylates discussed in the previous four subsections. 4.6. TiO2(110)-Na Hird and Armstrong have employed O"^ ion beam (10 keV) scattering to determine the adsorption site of Na on Ti02( 110)1x1 [99]. Their in situ substrate preparation consisted of several cycles of argon ion bombardment and annealing in vacuum at 873 K for 2 min. Measurements were recorded from a submonolayer coverage of sodium; the authors state that a more quantitative estimate of the coverage was not possible. The reflection high energy electron diffraction (RHEED) pattern remained (1x1) after Na deposition for the entire coverage range explored. Details about the adsorption site were obtained by measuring the yield of O"^ ions scattered by Ti atoms as a function of both ion incidence and emission angle. Two previously proposed adsorption sites for sodium were considered [100], the so called between and adjacent sites, which are depicted in Fig. 16. Comparison of the experimental data with calculations, simulating the angular variation in ion yield due to atoms shadowing/blocking the ingoing/outgoing trajectory, clearly favours the adjacent site. Furthermore, it is concluded that the bridging oxygens shift away from their relaxed clean

227

Between Site

Adjacent Site

[001] ( 1 M. [ 1 1 0 ]^ Fig. 16. Two possible adsorption sites for Na on Ti02(l 10)1x1, proposed in Ref. 100.

surface locations back up towards their bulk terminated positions. This experimental work contradicts Hartree-Fock/molecular dynamics calculations in Ref. 100 which predict that sodium occupies the between site. 4.7. TiO2(110) - Na/C02 The orientation of surface carbonate, resulting from CO2 exposure to a 0.7 ML Na precovered Ti02(l 10)1x1 surface, has been investigated with NEXAFS at the C if-edge [101]. Surface cleaning/ordering adhered to the normal prescription of Ar"^ bombardment and annealing in vacuum (900 K), completed by a final anneal (700 K) in 1x10-6 mbar of O2, with subsequent cooling in O2 to 500 K. Preparation of the coadsorption system involved Na deposition followed by exposure to 1000 L of CO2, all performed at a substrate temperature of 300 K. The authors note that for exposures up to 1000 L there was no evidence of CO2 sticking onto a similarly prepared Na free surface. A series of NEXAFS spectra were recorded over a range of polar photon incidence angles with the electric vector of the linearly polarised x-rays aligned along either the [1 10] or [001] azimuths of the substrate. The variation in intensity of the lowest energy resonance, which arises from a Is - 7t* transition, was utilised to determine the carbonate orientation. It was found that the molecular plane of the carbonate is twisted 32° ± 5° out of the [001] azimuth with a tilt of 46° ± 5° away from the surface normal. Several possible adsorption geometries suggested by the authors are shown in Fig. 17.

228

^ B

Carbonate Oxygen

^U

In-plane Oxygen

([

) Bridging Oxygen

^

Sodium adatom

9

Carbon atom

#

Titanium atom

[001]

(a)

(b) ^'

-"— [110]

Fig. 17. Models suggested on the basis of C-^ edge NEXAFS data [101] of CO^'^'/Na on Ti02( 110)1x1. (a) Two possible models where carbonate is formed by CO2 incorporating bridging O species (shown by a bold line border), (b) Two possible models where carbonate bonds in between the bridging oxygen atoms and through Na adatoms. The first involves one Na atom in the complex, while the second involves two. The Na atoms involved in the carbonate complex are shown by a bold line border.

4.8. TiO2(110) - (NH4)6Mo7024.4H20 Polarisation-dependent total-reflection fluorescence (PTRP) EXAFS has been employed to determine the local geometry of molybdenum oxides on TiO2(110) in the submonolayer regime [102,103]. This structural study is unique as regards this review in that the data were recorded not in a UHV chamber but rather under atmospheric conditions. Surface preparation consisted of annealing the sample in air at 823 K for 2 hours. Adsorption was achieved by impregnating the sample with an ultra-pure aqueous solution of (NH4)6Mo7024.4H20, followed by oxidation at 773 K for 3 hours. The Mo coverage after this procedure was found to be 0.2 ML from photoemission. PTRF-EXAFS spectra were collected for three different orientations of the surface relative to the electric vector of the incoming x-ray beam i.e. E//[110], E//[001], and E//[l 10]. Data analysis was carried out by comparing theoretically simulated spectra generated for many different trial structures with the experimental data. The best fit between experiment and theory was obtained for a Mo edge-shared dimer structure. Both the internal structure of the Mo dimer and the dimer - substrate bonding were determined. It was found that the Mo - Mo dimer axis lies parallel to the [110] direction of TiO2(110), that the Mo dimer is centred on the bridging row oxygens such that the two Mo

229

i^Mo

cu O

©Ti

Fig. 18. Plan and side views of the structure of molybdenum oxide dimers on Ti02( 110)1x1 [102,103]. The numerical labels are employed in Table 6.

Table 6 Bond lengths for optimised Mo dimer structure on Ti02( 110)1x1 [102,103]. The numerical labelling of the atoms corresponds to that in Fig. 18. Atoms Mol - Mo2 Mol - 0 5 Mol - 0 7 Mol - 0 8 Mol - 0 9 Mol-OlO Mol - 012 Mo2 - 0 3 Mo2 - 0 4 Mo2 - 0 5 Mo2 - 0 6 Mo2 - 0 7 Mo2-011

Bond length (A) 3.35 ± 0.08 2.32 ± 0.08 1.92 ±0.08 1.79 ±0.05 1.78 ±0.08 1.69 ±0.08 2.20 ± 0.09 1.79 ±0.05 1.69 ±0.08 1.92 ±0.08 1.78 ±0.08 2.32 ± 0.08 2.20 ± 0.09

230

atoms bond to two bridging oxygens, and that these bridging oxygens are significantly shifted from their clean surface positions. A schematic of the optimised structure is displayed in Fig. 18, and the values of the corresponding interatomic distances are listed in Table 6. 4.9. TiO2(110) - {Rh(CO)2Cl}2 {Rh(CO)2Cl}2 adsorption onto Ti02(l 10)1x1 at 300 K is dissociative, giving rise to a surface rhodium gem-dicarbonyl (Rh(CO)2) species. The azimuthal alignment of this Rh(C0)2 has been elucidated using RAIRS by Hayden et al [104]. The clean substrate was prepared by cycles of Ar"^ bombardment at 800 K, annealing at 1000 K in vacuum, and O2 dosing whilst cooling to 400 K. After the adsorption of 0.3 ML of Rh(CO)2, RAIRS spectra were recorded with both s-polarised and p-polarised light for several different azimuthal incidence angles. Orientational information was extracted fi*om the variation in intensity of the Vasym(C-O) band. For the p-polarised spectra, the intensity of this band exhibits a maximum when the incoming radiation is parallel to the [1 10] azimuth, and a minimum when it is parallel to [001]. From these data it was concluded that the majority of Rh(CO)2 is adsorbed with its CO molecules aligned along the [110] azimuth. Two possible adsorption configurations suggested by the authors are displayed in Fig. 19. They note that there is a minority of differently aligned species, perhaps due to adsorption at defects, mirroring the results of the RAIRS study performed by the same group on Ti02(l 10>[HCOO]- [87].

[001]

[110] • Fig. 19. Two suggested adsorption geometries for rhodium gem-dicarbonyl (Rh(C0)2) on Ti02(l 10)1x1 [104].

231

4.10 TiO2(100) - SO2 SO2, chemisorbed on either the (1x1) or (1x3) phases of TiO2(100), can undergo the following reaction via thermal activation [105]: SO2 -> SO32- -^ SO42-. Raza et al have performed S AT-edge NEXAFS measurements to obtain the surface orientation of the reactant, intermediate, and product of this reaction [105]. Surfaces were prepared by Ar"^ sputtering and annealing, with either a final anneal in 1x10"" mbar of O2, or a 10 min anneal in vacuum at 1100 K to produce the (1x1) and (1x3) surfaces, respectively. SO2 exposure was carried out at 110 K to achieve -0.5 ML coverage (where 1 ML is defined as occupation of all of the O and Ti surface atom sites by sulfur-containing species). To investigate molecular orientations, NEXAFS spectra were recorded over a range of polar photon incidence angles with the x-ray electric vector in the [001] and [010] sample azimuths. These measurements were performed at three substrate temperatures, namely 130 K, 200 K, and 500 K, in

® Ti

Fig. 20. Models of TiO2(100)lx3 (top) and TiO2(100)lxl (bottom) along with the proposed SO2 adsorption sites corresponding to, from left to right, chemisorbed SO2, 803^"", and SO42- -like species [105].

232

order to determine the orientations of SO2, SO32- and SO42-, respectively. For SO2 from analysis of the 7i* resonance it was found that the C2 molecular axis is oriented 20° ± 2° away from the (100) surface normal on the (1x3) surface, and 26° ± 3° on (1x1), with no azimuthal alignment on either surface. It should be noted that at 110 K SO2 displayed no specific orientation on either surface. Similar analysis for SO32-, which adopts a trigonal-based pyramidal structure, again indicates no azimuthal alignment and tilts of the C3 molecular axis of 24° ± 3° and 25° ± 3° away from the surface normal on the (1x3) and (1x1) surfaces, respectively. Data for S04^" from either surface at 500 K show little photon incidence angle dependence. This is consistent with a sulfate species having T j symmetry, although there is evidence from the spectral features of some distortion away from this highly symmetric configuration. This distortion is more pronounced for the S04^" already present on the surfaces at lower temperatures, specifically between 130 K and 220 K. As the results from the (1x1) and (1x3) surfaces are so similar, the authors use these structural data to argue that the reaction of SO2 with Ti02 is driven by local geometry effects rather than longer range surface structure. Proposed adsorption structures for the various surface species on the two surfaces are shown in Fig. 20. 4.11. TiO2(100) - K The local geometry of the K adsorption site in TiO2(100)c(2x2)K has been probed using K i^-edge SEXAFS [106]. To form the c(2x2) overlayer 0.5 ML of K was deposited onto a stoichiometric TiO2(100)lxl surface, cleaned as described above. SEXAFS measurements were performed at normal and C(2x2)

atop

K inclined

atop bridge

'°^°1

Fig. 21. Schematic of TiO2(100)c(2x2)K derived from SEXAFS data [106], with K atoms bonded to two bridging oxygens (left). The altemative K adsorption sites considered are also shown (right).

233

grazing x-ray beam incidence with the electric vector aligned to the [001] azimuth. Successful modelling of the data required a single shell of oxygen neighbours at 2.62 ± 0.03 A, the analysis rejecting models which contained K or Ti backscatterers. Adsorption site determination entailed a comparison of the experimentally determined oxygen effective coordination numbers with ones calculated for four plausible sites: angled bridge, atop bridge, atop, and Kinclined (see Fig. 21). From this process it was concluded that the angled bridge is the K adsorption site, in which the K-O2 plane is at 45° to the surface normal. It is noted in the paper that the K-0 bond length is close to the sum of the appropriate ionic radii. The structure predicted by a recent ab initio Hartree-Fock calculation is in excellent agreement with the experimental data [107]. 5. ALUMINIUM OXIDE - AI2O3 A number of different bulk phases of AI2O3 exist, including a-Al203 (corundum) and Y-AI2O3. Quantitative structure determinations of adsorbates have been performed on the (0001) face of an a-Al203 single crystal, and a (0001) oriented thin film of y-like AI2O3. a-Al2O3(0001) displays a variety of surface periodicities [37]. Of interest, as regards this review, are the (1x1) and (V31xV31)R±9'^ unit cells, both of which have been investigated by SXRD [37]. For the (1x1) surface it was concluded that the surface is terminated by a single-aluminium layer, whilst a model consisting of a tiling of domains bearing a close resemblance to that of two metal Al(l 11) planes separated by a hexagonal network of domain walls is proposed for the (V31xV31)R±9° termination. The y-like Al2O3(0001) thin film, grown on NiAl(llO) via preferential oxidation, exhibits a complex LEED pattern [108], and its atomic scale structure is the subject of some debate [109]. 5.1. a-Al2O3(0001) - Cu Gota et al have employed SEXAFS at the Cu ^-edge to probe the structures of a-Al2O3(0001)lxl-Cu and a-Al2O3(0001) (V31xV31)R±9°-Cu, following deposition of 0.5 ML equivalent of Cu [110]. Preparation of the surfaces involved Ar+ bombardment, with subsequent annealing at 1073 K in 5x10-7 Torr of O2, and at 1673 K in UHV for the (1x1) and (V31xV31)R±9° terminations, respectively. SEXAFS spectra were obtained at normal photon incidence. On both surfaces the majority of Cu is present in the form of small 3D clusters, and best fits to the SEXAFS data have a significant contribution

234

Table 7 Results of best fits to SEXAFS modulations, recorded at normal incidence, from Al2O3(0001)lxl-Cuand Al2O3(0001)(V31xV31)R±9°-Cu at 0.5 ML equivalent coverage of Cu [110]. N* is the effective coordination number of a shell of scatterers, R is its distance, and Aa^ the mean square relative displacement. N* (±10%) Cu-Cu Cu-Cu Cu-Al

0.7 4.3 2.0

Cu-Cu Cu-Al

4.7 5.7

R(A) (± 0.02 A) (1x1) 2.0 2.37 3.16 (V31xV31)R±9° 2.42 3.24

Aa2 (A2) (±0.5x10-3) 0.010 0.010 0.010 0.012 0.012

from Cu, although information about the interface was also extracted. Specifically, including a shell of Al scatterers was found to significantly improve the fits. It should be noted that no evidence for oxygen scatterers was found. Table 7 lists the structural parameters obtained from analysis of the SEXAFS. The authors interpret the rather long Cu-Al bond distances as possibly being a consequence of the very weak interaction between Cu and the AI2O3 surfaces The relatively high effective coordination number of Al for the (V31xV31)R±9° reconstruction (5.7) is ascribed to Cu adsorption at defects. 5.2. y-like Al2O3(0001) - di-tert-butyl nitroxide The orientation of di-tert-butyl nitroxide (CgHigNO), the structure of which is illustrated in Fig. 22, on a thin film of (0001) oriented y-like AI2O3 has been deduced from N AT-edge NEXAFS measurements [111]. The film was grown on a clean and well-ordered NiAl(llO) substrate, via oxygen exposure and annealing (see Ref. 108 for details). NEXAFS spectra were recorded from a monolayer of di-tert-butyl nitroxide, which had been dosed at a substrate temperature of 100 K. No 71* resonances were observed in these spectra, although such a resonance was present in multilayer spectra. This difference was explained as being due to electron donation from a surface oxygen to the adsorbate's nitrogen atom. Given the lack of TT* resonances, molecular orientation information was obtained from the variation in intensity of the a*(N-C) and a*(N-0) resonances in the spectra. A tilt angle of about 70° for the N-O bond axis away from the surface normal was concluded. From complementary electron spin resonance (ESR) measurements, it is known that

235

Fig. 22. Ball and stick model of di-tert-butyl nitroxide.

the NEXAFS spectra actually arise from two differently bound di-tert-butyl nitroxide species. However, the majority species comprises more than 90% of the adsorbate coverage, and so it is presumed that the contribution of the minority species to the NEXAFS spectra is not significant and that the tilt angle extracted is essentially that of the majority species. 6. CHROMIUM OXIDE - Cr203 Cr203 has the same bulk crystal structure as a-Al203, namely corundum. Of its several low Miller index surfaces only one, (0001), has been employed for adsorbate structural determinations so far. To overcome sample charging problems a thin film has been utilised for these studies, rather than a single crystal. The surface structure of this (0001) oriented thin film has been investigated by LEED-IV [112]. Simulations of the experimental data evidence a chromium terminated surface with large vertical interlayer relaxations, reaching down five or six layers. 6.1. Cr2O3(0001) - CO2 Freund and coworkers have used C ^-edge NEXAFS to examine the angular geometry of the CO32- moiety resulting from CO2 adsorption on Cr2O3(0001) [113]. As indicated above, the substrate was a highly oriented

236

thin film rather than a single crystal. It was grown by first annealing the sample at 500 K in a partial pressure of oxygen for 3 min and then annealing in vacuum at 1000 K to remove excess oxygen. CO2 was adsorbed with the substrate at approximately 100 K. Two NEXAFS spectra were recorded, one at normal photon incidence the other at grazing incidence, to elucidate the orientation of the CO32-. For these measurements the sample was maintained at approximately 170 K to avoid physisorbed CO2. Analysis of the carbonate 7C* resonance at 290.2 eV, which is significantly more intense at normal incidence, indicated that the molecular plane of the CO32- is approximately parallel to the surface normal. 6.2. Cr2O3(0001) - NO NEXAFS, at the N iC-edge, has been performed by Freund's group to obtain the tilt angle of the N-0 bond axis relative to the Cr2O3(0001) surface [114]. A thin film, prepared as above, was employed. NO dosing and NEXAFS measurements were carried out at substrate temperatures of between 100 K and 105 K. The variation in the relative intensities of the NO 71* and a* resonances with polar photon incidence angle was used to obtain the molecular orientation. Fitting to an appropriate trigonometric function [35] provides an angle of 50° between the N-0 bond axis and the surface normal. The authors note that there is a significant error for this tilt angle. 7. IRON OXIDE - FexOy There are three stable bulk forms of iron oxide: FcxO, Fe304, and Fe203. Single crystal faces of each of these three phases have been the subject of surface studies. However, adsorbate geometry determination has only been performed on (111) oriented thin films of FeO and Fe304 grown on Pt(lll). Details of the structure of both of these films have been revealed by the employment of a range of surface science techniques (see Ref 115 and Refs. therein). The FeO (111) film is comprised of a single bilayer of iron and oxygen on the Pt(lll) substrate, with oxygen being uppermost. It has a lateral lattice constant of 3.11 A (the lattice constant in bulk FeO is 3.04 A), and an ironoxygen interlayer spacing of 0.68 A, which is much reduced compared to the (111) interlayer separation in bulk FeO (1.25 A). The Fe304 film is substantially thicker, of the order of 100 A - 200 A. At its surface 0.25 ML of iron cations sit upon a close packed oxygen layer, giving rise to a hexagonal surface unit cell of side 5.92 A

237

Fig. 23. Schematic diagram of stryrene. Only the carbon atoms are shown.

7.1. FeO(lll) - H5C6CHCH2 Wiihn et al have elucidated the angular geometry of styrene (H5C6CHCH2: see Fig. 23 for schematic diagram) on a FeO(lll) thin film using C X-edge NEXAFS [116]. The methodology for substrate preparation involved cycles of iron deposition onto a clean and ordered Pt(lll) surface at room temperature, followed by oxidation for 2 min at approximately 1000 K in a partial pressure of oxygen (10"6 mbar). Adsorbate overlayers of coverage 0.6 ML and 1 ML (where 1 ML is defined as the saturation coverage of styrene on FeO at 200 K) were produced by exposure to 0.7 L and 1 L of styrene, respectively, at 200 K. NEXAFS spectra were recorded at normal and grazing photon incidence for both coverages. Comparison with multilayer/gas phase NEXAFS data, along with results from other techniques, demonstrated that the styrene molecule remains intact upon adsorption. From analysis of the variation in intensity of the strongest 7C* resonance (285 eV), it was concluded that the angles between the surface and the molecular plane of the styrene are 41° and 43° for coverages of 0.6 ML and 1 ML, respectively (the ethylenic and benzenoid fragments of the styrene molecule are believed to be coplanar). An error of ±2° is attributed to these tilt angles by the authors. 7.2. Fe304(lll) - H5C6CHCH2 The orientation of styrene on Fe304(lll) is also derived from C i^-edge NEXAFS in Ref 116. A thin film of (111) oriented Fe304 was prepared using a procedure identical to that described in the last subsection, except that more deposition/oxidation cycles were required. NEXAFS spectra were obtained in the submonolayer regime for styrene coverages of 0.4 ML and 0.8 ML, with the substrate at 240 K and 225 K, respectively. At the lower coverage the tilt of the styrene plane away from the surface was found to be 28° ± 2°. This angle increases to 42° ± 2° at 0.8 ML coverage. On the basis of TPD data [115, 116], this change in geometry is attributed to there being more than one adsorption state for styrene in the submonolayer regime. It is suggested that initially styrene binds via its TT system to surface iron cations in a near flat

238

geometry, then once these iron sites are saturated styrene bonds more weakly to other surface sites giving a tilted geometry. It should be noted that on FeO(lll) ( subsection 7.1) there are no iron cations in the surface layer, and only the weakly bound, more tilted styrene species is observed. 8. ZINC OXIDE - ZnO ZnO adopts the wurtzite structure in which Zn and O are both tetrahedrally coordinated to their counter ions. The O-terminated (0001) and Zn-terminated (0001) basal faces, as well as the non-polar (1010) prism face have all been the subject of adsorbate structure determinations. As regards their clean structures, LEED-IV was utiUsed as a probe over twenty years ago [117,118], and very recently they have been reexamined with SXRD [119-121]. From these measurements it was found that neither polar face relaxes greatly, a feature ascribed to surface enhanced covalency [121]. For the (1010) face, SXRD [120] indicates only small displacements of the surface Zn and O away from their bulk terminated positions. 8.1. ZnO(lOlO) - HCOOH Davis et al have determined the orientation of [HCOO]" on ZnO(lOlO) with C AT-edge NEXAFS [122]. For these measurements the substrate was prepared by cycles of Ar"^ bombardment at 800 K, followed by annealing for 10 min at 800 K in 1x10"6 mbar O2 with subsequent cooling in O2 to 550 K. The formate overlayer was formed by exposure of the substrate at 295 K to 8 L of formic acid, to give a coverage of about 0.2 ML. The NEXAFS data indicate that formate is oriented with its molecular axis tilted towards upright, and that there is no azimuthal alignment. Quantitative analysis, employing the leading 71* resonance, provides a tilt of the formate C2 axis of about 25° away from the surface normal. A possible bond geometry is shown in Fig. 24, in which formate is coordinated in a bidentate configuration to one Zn atom. This structure is consistent with the lack of azimuthal ordering. A recent energy minimised calculation of formate on ZnO(10 10) predicts that a bidentate configuration is more stable than monodentate [123]. However, the suggested structure has formate bridging two Zn atoms along the [1120] direction. Unless the experimental data is very strongly affected by dynamical effects, such ordering would give rise to a clear azimuthal polarisation dependence, which is not apparent. 8.2. ZnO(lOTO) - CO2 Exposure of ZnO(10 10) to CO2 leads to a surface CO32- species. The

239

Fig. 24. Model of ZnO(lOTO) surface, with the bond geometries of formate and carbonate proposed in Ref 122. The tilts out of high symmetry sites have been neglected, and for formate only one azimuthal orientation is shown.

orientation of this CO32- has been determined, using C AT-edge NEXAFS [122]. For these measurements the substrate was prepared as described above (subsection 8.1), and exposed to 45 L of CO2 as it was cooled to the measurement temperature of 180 K. This achieved a coverage of about 0.2 ML. NEXAFS spectra were recorded at a number of photon incidence angles with the x-ray electric vector in either the [000T] or [1120] azimuth The angular dependence of the NEXAFS data indicates that the plane of CO32- lies in the [000 1 ] azimuth, tilted away from the surface normal by approximately 15°. Fig. 24 shows an adsorption geometry suggested in Ref. 122. 8.3.ZnO(1010)-C6H6 The angular geometry of the prototypical aromatic molecule benzene (C6H6) on ZnO(lOTO) has been investigated with C A'-edge NEXAFS [124]. The substrate was cleaned and ordered as outlined above (subsection 8.1), and the overlayer was formed by exposure to 5 L of benzene at the measurement temperature of 180 K. The coverage of benzene was about 0.3 ML. From the intensity variation with photon incidence angle of the leading 71* resonance in the NEXAFS spectra, it was determined that the benzene ring is tilted by 10°

240

away from the surface plane, i.e. benzene lies more or less flat on the surface. This geometry mirrors that found on metal surfaces, which arises due to benzene coupling with the surface through its n system [35,125]. 8.4. ZnO(lOTO) - C5H5N NEXAFS, both at the C ^-edge and N ^-edge, has been used to examine the orientation of another aromatic molecule, pyridine (C5H5N), on ZnO(10 10) [124]. A preparative procedure similar to that used for the studies discussed in the previous three subsections was performed for the substrate. The pyridine adlayer, of coverage approximately 0.1 ML, was formed by exposure to 1 L at the measurement temperature of 295 K. NEXAFS data were recorded at a series of polar photon incidence angles, with the electric vector of the linearly

290 310 Photon Energy (eV) Fig. 25. C AT-edge NEXAFS spectra of ZnO(lOTO) after exposure to 1 L of pyridine at the measurement temperature of 295 K [124]. The peak labels are discussed in Ref. 124. The inset shows the experimental variation in the intensity of the 7i* feature (peaks al and all) with angle of incidence in the two measurement azimuths (data points). This is compared with the best fit to the data (solid lines).

241

polarised photons either parallel or perpendicular to the surface Zn-O dimer rows (see Fig. 24). The NEXAFS spectra show that pyridine stands up on the surface with the ring plane in the [000 1 ] azimuth. Two C K-edgQ spectra, recorded at normal and grazing incidence with the electric vector in [1120] azimuth, are displayed in Fig. 25. The inset in this figure shows the intensity variation of the strongest 7i* feature (a doublet, with components at 285.0 eV and 285.5 eV) with angle of photon incidence in the two measurement azimuths, along with the best fits to these data using an appropriate trigonometric expression [35]. These fits correspond to a tilt of the molecular plane towards the surface of 28° and a 27° twist out of the [11 20] azimuth. A consistent result was obtained from the N ^-edge data. The authors note that this apparent distortion out of a high symmetry geometry should be viewed with some caution, since dynamical effects are known to significantly influence NEXAFS results obtained at room temperature.

8.5. ZnO(lOTO) - C6N3H5, C7N2H6, C7N2H6, C7N3H7 The orientation on ZnO(lOlO) of benzotriazole (C6N3H5) and related molecules (indazole (C7N2H6), benzimidazole (C7N2H6), and 1methylbenzotriazole (C7N3H7)), which are of interest in connection with corrosion protection, has been studied using C and N K-QdgQ NEXAFS [126]. The substrate was prepared as above and the gas phase molecules used to dose

Benzimidazole

1 -Methyl-Benzotriazole

Fig. 26. Ball and stick models of the adsorbates studied in Ref 126.

242

ZnO(10 10) were obtained by evaporation of the solids. Ball and stick models of these molecules are displayed in Fig. 26. At submonolayer coverage, benzotriazole is found to adsorb in an upright geometry with the molecular plane within 35° of perpendicular to the substrate, as indicated by the polarisation dependence of the 7i* resonances. Seven slightly different models of the bond geometry are consistent with the data; two of the most likely models are shown in Fig. 27. Indazole (C7H6N2) was found to be oriented similarly. Benzimidazole (C7H6N2) and 1-methyl benzotriazole (C7H7N3) were found not to be have a preferred orientation at sub-monolayer coverage. 8.6. Z n O ( 0 0 0 1 ) - C O Thornton and coworkers have employed C A^-edge NEXAFS measurements to investigate CO adsorption at 130 K on ZnO(OOOT) [127,128]. The substrate was prepared by cycles of Ar"^ bombardment and annealing in vacuum to about 750 K. This was followed by a final anneal in IxlO"^ mbar O2 and cooling in this ambient to 550 K. CO adsorption gave rise to two K* resonances in the C ^-edge NEXAFS spectra at 287.7 ± 0.2 eV and 290.4 ± 0.2 eV. These were assigned to molecular CO, which is metastable under the prevailing

[1120]

Fig. 27. Model depicting two NEXAFS compatible adsorption geometries for benzotriazole onZnO(10T0)[126].

243

experimental conditions, and CO32-, respectively. XPS/x-ray absorption step edge measurements indicated that the coverage of CO32- was 0.1 ± 0.05 ML, whilst that of CO was 0.04 ML at most. From the dependence of the CO3271* intensity on photon incidence angle it was deduced that the molecular plane of CO32- is inclined away from the surface normal by 32° ± 20°. The authors state that the error bar is large due to the limited dataset, and also uncertainties in normalisation of the data arising from the variation in CO surface concentration. It should be noted that CO was also dosed onto a ZnO(000 1) surface precovered with approximately 0.55 ML of Cu. In this case only one intense 7C* resonance, at 287.7 ± 0.2 eV, was observed. This was again attributed to adsorbed molecular CO. The polarisation dependence of this resonance indicated that the CO molecular axis is tilted by 17° ± 10° away from the surface normal. 8.7. ZnO(OOOT) - CO2 Besides CO adsorption, Ref. 127 also presents C K-edge NEXAFS spectra obtained following exposure at 130 K of ZnO(OOOT) to CO2. Surface CO32again forms, as evidenced by a 7i* resonance located at 290.4 ± 0.2 eV. The estimated CO32- coverage was 0.1 i 0.05 ML, identical to that of the CO32species formed following CO adsorption, as discussed in the last subsection. Analysis of the intensity variation of the 7C* resonance with measurement geometry produced a value of 25° ± 5° for the tilt angle of C03^" away from perpendicular. This value is similar to that for CO32- produced from CO exposure. 8.8. ZnO(OOOT) - HCOOH The orientation of [HCOO]" on ZnO(000 1) has been investigated using C AT-edge NEXAFS [129]. Prior to the measurements, the sample was cleaned in situ by cycles of Ar+ bombardment and annealing in UHV to between 800 K and 1000 K, and ending with an anneal in a partial pressure of O2 (lxl0"6 mbar). Surface formate was created by exposure of the substrate to 20 L of formic acid at 110 K and heating to 340 K. This procedure gave a formate coverage of 0.2 ± 0.05 ML. Two NEXAFS spectra were recorded, one at normal incidence and the other at grazing. From quantitative analysis it was concluded that the tilt angle of the molecular plane of the formate is 55° ± 10° with respect to the surface normal. The authors speculate that the formate may be located at step edges, on the basis that earlier work has demonstrated that formate can only be located at defects on this surface [130]. It must be noted, however, that currently the actual identity of the defects remains unknown.

244

NEXAFS spectra were also acquired from [HCOO]' adsorbed onto a ZnO(000 1) surface, which had previously been exposed to Cu vapour at 140 K such that approximately half of its surface area was covered by 2D Cu islands. The formic acid was again dosed at a substrate temperature of 110 K, but in this case no annealing was necessary to dissociate molecularly adsorbed formic acid. The tilt of the formate in this case was 32° ± 5°, which is close to the magic angle [35]. It is suggested in the paper that this result is due either to buckling of the 2D Cu islands giving rise to a molecular tilt of about 32°, or randomly oriented formate. These two possibilities are indistinguishable by NEXAFS. 8.9. ZnO(000T)-C5H5N Hovel et al have investigated the adsorption of C5H5N (pyridine) on the O terminated basal ZnO face with several techniques, including NEXAFS at the C i^-edge, to determine its angular geometry [131]. The in situ substrate preparation involved Ar"*" bombarding with the sample at 650 K, heating in vacuum at 800 K, and frequent anneals in IxlO"^ mbar of O2, with subsequent cooling to 550 K in the O2 ambient. An adsorbate overlayer was prepared by exposing the surface to 1 L pyridine at the measurement temperature of 110 K. Adsorption was concluded to be molecular in the monolayer regime, and an angle of 66° ± 5° between the molecular plane of pyridine and the surface was derived from the angular dependence of the NEXAFS spectra. 8.10.ZiiO(000T)-K The structure of a p(2x2) overlayer of K on the O terminated basal face of ZnO has been studied with Auger and fluorescence yield polarisation dependent SEXAFS at the K A^-edge [132]. Substrate preparation involved cycles of Ar"^ bombardment and annealing in vacuum to 1100 K. A final anneal was Table 8 The effective coordination numbers resulting from single shell fits to K X-edge SEXAFS data recorded from ZnO(000 I)2x2-K [132]. Also listed are the calculated effective coordination numbers for the three-fold hollow, bridge and atop adsorption sites. Effective coordination numbers Angle of Incidence

Experimental

3-fold hollow

Bridge

Atop

90° 20°

3.8 ±1.5 4.7 ±1.5

2.2 4.4

1.1 3.5

0.0 2.6

245

performed in O2, with subsequent cooling in O2 to 500 K. The p(2x2)K overlayer was formed by evaporating K onto the substrate at room temperature and annealing to 900 K. SEXAFS data were recorded at normal and grazing photon incidence with the substrate maintained at a temperature of 125 K. Successful modelling of the grazing incidence data required a single shell of oxygen neighbours at 2.70 ± 0.04 A, the analysis routine rejecting models containing K or Zn scattering atoms. The K adsorption site was extracted by comparing the experimental effective coordination numbers with those calculated for high symmetry sites (see Table 8). From this comparison it was concluded that K adsorbed in a three-fold hollow site. On this surface there are two different hollow sites, the open and cave (see Fig. 28). On the basis of a lack of any Zn scatterer contribution to the SEXAFS oscillations, the authors argue that of the two sites the cave site is more probable. This is consistent with the prediction of Madelung potential calculations [133,134]. Models of the structure are shown in Fig. 28. [1120] [1010]

M v;i>^^^^^^ :.:mm'm^^m. ^-^f^'n^^m 'im^isw's^^^^,;,,.,.,,;

Fig. 28. Ball and stick models of ZnO(000 1 )2x2-K derived from SEXAFS data, with K atoms bound to three oxygen atoms in a three-fold cave site [132]. The K atoms form the comers of the p(2x2) unit cell outlined. The alternative K adsorption sites considered in modelling the SEXAFS are also shown. The bridge and atop sites are shown in the top and side view, with the open hollow site shown only in the top view.

246

c 3

7C*(CO) [0001]^

•9

e°

<^ Jt*(C032-)

c

I 1

_c

\ ^ ,/

)^

1

B

—1 -J

O •D

CD

\

.

.

•

-

,.._^

-.

"

^^ ,. _ ^_ ___ .

! ' . ''V,.''

90

_^

.- 50

• - ^ ' ~ ^ ' ^ ^ "

CO

"5 E o 2

- 20 _ ^ ^ -

f 'V./ T 290

• • " " ' " • " "

' " " '

"' I" 300

r

320

310

Photon Energy (eV) Fig. 29. C ^-edge NEXAFS spectra, recorded by Gutierrez-Sosa et al [135], of the ZnO(OOOl) surface following a 1 L exposure of CO at a substrate temperature of 110 K.

8.11. ZnO(OOOl) - CO In a paper which describes a C AT-edge NEXAFS study of various Ci^2 oxygenates on the Zn terminated surface [135], the orientation of surface species arising from exposure of the surface to CO was explored. Cycles of Ar"^ sputtering and UHV annealing to 800 K were performed to prepare the surface. ISTEXAFS spectra were recorded following a 1 L exposure of CO at the measurement temperature of 110 K. The resulting ]S[EXAJFS spectra, displayed in Fig. 29, exhibit two features in what is typically the 71* region, which are attributed to the presence of both molecular CO and CO32- on the surface. Fitting the intensity variation with photon incidence angle of each of these two resonances to equations detailed by Stohr [35] shows that the CO bond axis lies along the surface normal (± 10°), whereas the molecular plane of CO32- is canted over at 65° ± 10°. 8.12. ZnO(OOOl) - CO2 Ref. 135 also describes a determination of the angular geometry of a CO32species on ZnO(OOOl), appearing after exposure of the clean surface to 1 L of CO2 at 270 K. The coverage of carbonate was estimated to be 0.3 ± 0.05 ML. The polarisation dependence of the 7C* resonance, located at 290.4 ± 0.5 eV,

247

indicates an almost flat lying geometry, similar to that of CO32- formed from CO (see last subsection). Quantitatively, the angle between the plane of the molecule and the surface normal is 68° ± 5°. 8.13. ZnO(OOOl) - HCOOH Formate, formed following dissociative adsorption of formic acid, is a another species examined by C ^-edge NEXAFS in Ref. 135. Following clean surface preparation, carried out as outlined in subsection 8.11, the sample was maintained at 400 K whilst formic acid was dosed, and latterly flashed to 450 K to remove any molecularly adsorbed formic acid. In this case the NEXAFS spectra show the molecule not to have a preferred orientation on the surface. This is ascribed to formate adsorbing in a monodentate configuration, which is the bonding arrangement deduced from HREELS data [136]. It should be noted that analysis of the NEXAFS data was complicated by formate desorption during the measurements. 8.14. ZnO(OOOl) - CH3OH The C-O bond length in surface methoxide on ZnO(OOOl) has been elucidated by Solomon and coworkers using C K-edge NEXAFS data [137]. The surface was cleaned firstly by Ar"^ bombardment with the sample at 750 K, and then annealing in vacuum at this temperature. Subsequent exposure to methanol at 130 K produced a submonolayer coverage of methoxide. The C i;r-edge NEXAFS spectrum acquired from this surface exhibits only a single broad feature peaked at 295.5 ± 0.2 eV. The authors assigned this peak to a a* resonance. On the basis of an expression formulated by Stohr et al [35], often described as bond length with a ruler, the C-O bond length was calculated from this peak position. A value of 1.39 ± 0.05 A was obtained. We note that there is uncertainty in the x-ray absorption community about the validity of this methodology of obtaining numerical values for bond lengths [138]. 8.15. ZnO(OOOl) - C^YL^ C iT-edge NEXAFS has been used to determine the angular geometry of benzene on the Zn-terminated basal face [139]. The overlayer was formed by exposure of a surface, prepared by Ar"^ bombardment and annealing (800 K) cycles, to benzene at 173 K to give a coverage of 0.1 ML. NEXAFS data were acquired at a number of polar photon incidence angles, allowing the orientation of the adsorbate to be deduced from the variation in intensity of the 71* resonance at 285.5 eV. The results indicate that benzene has a tilt angle of the ring plane with respect to the surface of 11° ± 5°. The authors suggest that this small tilt angle away from the surface plane is due to a minority site

248

Fig. 30. Models of the geometries of adsorbed benzene and phenoxy on ZnO(OOOl) deduced from NEXAFS data. Zn atoms are depicted as small light spheres, and substrate 0 atoms as larger dark spheres.

contribution from molecules associated with defect sites. Such an hypothesis has already been proposed to explain similar tilt angles found for benzene on metal surfaces [140]. A model of benzene bonding with its molecular plane parallel to the surface is depicted in Fig. 30. 8.16. ZnO(OOOl) - C6H5OH The orientation of phenoxy (C6H5O") on Zn(OOOl) was also examined by C i^-edge NEXAFS in Ref. 139. A phenoxy adlayer was generated by exposing the substrate, cleaned as outlined in the last subsection, to phenol at 163 K and heating to the measurement temperature of 380 K, in order to remove any physisorbed molecules. This produced a coverage of 0.17 ML. Quantitative analysis of the NEXAFS spectra indicate a tilt angle of the ring plane from the surface of 35° ± 5°. Fig. 30. illustrates an adsorption geometry, suggested in Ref. 139, for phenoxy on ZnO(OOOl). In this model it is assumed that the oxygen atom in the phenoxy moiety is located atop a Zn, giving a Zn-O-C angle of about 125°, which is close to that expected in a metal alkoxide. 8.17. ZnO(OOOl) - C5H5N Besides their work on pyridine on ZnO(000 1), which is described in

249

subsection 8.9, Hovel et al have also investigated the adsorption of the same molecule on the Zn terminated polar face of ZnO [131]. Again they employed C AT-edge NEXAFS to provide details of the adsorption geometry. A surface suitable for pyridine exposure was produced by Ar"^ bombarding with the sample at 650 K, heating in vacuum at 800 K, and frequent anneals in 1x10"6 mbar of O2, with subsequent cooling to 550 K in the O2 ambient. NEXAFS data were obtained from such a surface, once it had been dosed with 15 L of pyridine at 320 K. The results point to a geometry in which the molecular plane is tilted up from the surface by 71° ± 5°. This angular configuration is similar to that obtained for pyridine on the oxygen terminated polar face, as well as on the non-polar (1010) surface of ZnO [124]. 9. COPPER (I) OXIDE - Cu20 Cu20, the final oxide considered in this review, has the cuprite structure. Oxygen anions are located at the centre and corners of the bulk cubic unit and Cu cations reside in four of the eight tetrahedral holes. There have been only a few studies performed on single crystal surfaces of this oxide, with only a single quantitative structural determination of an adsorbate. This, as described in the next subsection, concerns methoxide on Cu20(l 11). The (111) surface is a nonpolar surface, which has been found to exhibit either a (1x1) or (V3xV3)R30° surface unit cell depending on preparation conditions [4]. 9.1. Cu20(lll) - CH3OH Solomon's group have studied the adsorption of methanol on Cu20(l 11) as well as on ZnO(OOOl) (see ZnO section above), including a determination of the C-0 bond length in surface methoxide from C iC-edge NEXAFS data [137]. In situ preparation of the clean surface entailed cycles of Ar"^ sputtering at 900 K and annealing in vacuum at this same temperature. NEXAFS spectra were recorded from a surface that had been exposed to methanol at 130 K to produce a submonolayer coverage of methoxide. Similar to the ZnO(OOOl)methoxide system, a single broad a* resonance (294.8 ± 0.2 eV) was the only feature apparent. Using the bond length with a ruler method [35] the authors derived a C-O bond distance of 1.41 ± 0.05 A. ACKNOWLEDGEMENTS The authors thank Mrs. Christine Martyniuk, Mr. Tim Ashworth, and Mrs. Ela Michelangeli for their help in preparing the camera ready chapter, and Dr. Aurora Gutierrez-Sosa for proofreading the manuscript.

250

REFERENCES [I] [2] [3] [4] [5]

[6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23]

D.P. Woodruff, T.A. Delchar, Modern Techniques in Surface Science (2nd ed.), Cambridge University Press, Cambridge, 1994. J.C. Vickerman (Ed.), Surface Analysis - The Principal Techniques, John Wiley & Sons, Chichester, 1997. P.R.Watson, M.A. Van Hove and K. Hermann, NIST Surface Structure Database, version 2: NIST Standard Reference Database 42 , NIST, Gaithersburg MD, 1996. V.E. Henrich and P. A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, Cambridge, 1994. H. Kuhlenbeck, G. Odorfer, R. Jaeger, G. Illing, M. Menges, Th. Mull, H.-J. Freund, M. Pohlchen, V. Staemmler, S. Witzel, C. Scharfschwerdt, K. Wennemann, T. Liedtke and M. Neumann, Phys. Rev. B, 43 (1991) 1969. M. Li, W. Hebenstreit, U. Diebold, M.A. Henderson and D.R. Jennison, Faraday Discuss., 114(1999)245. M. Li, W. Hebenstreit, U. Diebold, A.M. Tyryshkin, M.K. Bowman, G.G. Dunham and M.A. Henderson, J. Phys. Chem., B 104 (2000) 4944. Q. Guo, I. Cocks and E.M. Williams, J. Chem. Phys., 106 (1997) 2924. D. Ferry, S. Picaud, P.N.M. Hoang, C. Giradet, L. Giordano, B. Demiirdjian and J. Suzanne, Surf Sci., 409 (1998) 101. D. Ferry, P.N.M. Hoang, J. Suzanne, J.-P. Biberian and M.A. Van Hove, Phys. Rev. Lett, 78(1997)4237. S.C. Parker, N.H. de Leeuw and S.E. Redfern, Faraday Discuss., 114 (1999) 381. S.C. Street, C. Xu and D.W. Goodman, Ann. Rev. Phys. Chem., 48 (1997) 43. R. Wichtendal, M. Rodriguez-Rodrigo, U. Hartel, H. Kuhlenbeck, H.-J. Freund and G.A. Somorjai, Phys. Stat. Sol. A, 173 (1999) 93. P. Liu, T. Kendelewicz, E.J. Nelson and G.E. Brown, Jr., Surf Sci., 415 (1998) 156. M.A. Van Hove, W. Moritz, H. Over, P.J. Rous, A. Wander, A. Barbieri, N. Materer, U. Starke and G.A. Somorjai, Surf Sci. Rep., 19 (1993) 191. P. Liu, T. Kendelewicz, E.J. Nelson, G.E. Brown, Jr. and G.A. Parks, Surf Sci., 412/413 (1998)287. J.C. Woicik, T. Kendelewicz, K. Miyano, P.L. Cowan, M. Richter, B.A. Karlin, C.E. Bouldin, P. Pianetta and W.E. Spicer, J. Vac. Sci. Technol., A 10 (1992) 2041. V.F. Zigan and R. Rothbauer, Neues Jb. Miner. Mh, (1967) 137. W. Langel and M. Parrinello, J. Chem. Phys., 103 (1995) 504. C.A. Scamehom, A.C. Hess and M.I. McCarthy, J. Chem. Phys., 99 (1993) 2786. M.I. McCarthy, G.K. Schenter, C.A. Scamehorn and J.B. Nicholas, J. Phys. Chem., 100 (1996) 16989. L. Giordano, J. Goniakowski and J. Suzanne, Phys. Rev. Lett., 81 (1998) 1271. L. Delle Site, A. Alavi, and R.M. Lynden-Bell, J. Chem. Phys., 113 (2000) 3344.

251

[24] D. Ferry, P.N.M. Hoang, J. Suzanne, J.-P. Biberian and M.A. Van Hove, Phys. Rev. Lett, 78(1997)4237. [25] D. Ferry, J. Suzanne, P.N.M. Hoang and C. Giradet, Surf. Sci., 375 (1997) 315. [26] J.P. Coulomb, Y. Lahrer, M. Trabelsi, and I. Mirebeau, Mol. Phys., 81 (1994) 1259. [27] J. Heidberg, M. Kandel, D. Meine and U. Wildt, Surf. Sci., 331-333 (1995) 1467. [28] R. Gerlach, A. Glebov, G. Lange, J.P. Toennies and H. Weiss, Surf. Sci., 331-333 (1995) 1490. [29] C. Minot, M.A. Van Hove and J.-P. Biberian, Surf Sci., 346 (1996) 283. [30] K. Jug and G. Geudtner, J. Chem. Phys., 105 (1996) 5285. [31] A.K. Sallabi and D.B. Jack, J. Chem. Phys., 112 (2000) 5133. [32] J. Heidberg, D. Meine and B. Redlich, J Electron Spectrosc. Relat. Phen., 64/65 (1993) 599. [33] X. Carrier, C.S. Doyle, T. Kendelewicz and G.E. Brown Jr., Surf. Rev. Lett., 6 (1999) 1237. [34] G. Pacchioni, Surf Sci., 281 (1993) 207. [35] J. Stohr, R. Gomer (Ed.), NEXAFS Spectroscopy, Springer Series in Surface Science Vol. 25, Springer, Berlin, 1992. [36] R. Souda, T. Aizawa and Y. Ishizawa and C. Oshima, J. Vac. Sci. Technol., A 8 (1990) 3218. [37] G. Renaud, Surf. Sci. Rep., 32 (1998) 1. [38] O. Robach, G. Renaud and A. Barbier, Surf. Sci., 401 (1998) 227. [39] P.Masri, P.W. tasker, J.P. Hoare and J.H. Harding, Surf. Sci., 173 (1986) 209. [40] P. Masri and P.W. Tasker, Surf. Sci., 149 (1985) 209. [41] A.M. Stoneham and P.W. Tasker, in: J. Nowotny and L.C. Dufour (Eds.), Surface and Near-Surface Chemistry of Oxide Materials, Elsevier, Amsterdam, 1988. [42] P.W. Tasker, Adv. Ceram., 10 (1988) 176. [43] A.M. Flank, R. Delaunay, P. Lagarde, M. Pompa and J. Jupille, Phys. Rev. B, 53 (1996)R1737. [44] P. Guenard, G. Renaud and B. Villette, Physica B, 221 (1996) 205. [45] O. Robach, G. Renaud, A. Barbier and P. Guenard, Surf. Rev. Lett., 5 (1997) 359. [46] O. Robach, G. Renaud and A. Barbier, Phys. Rev. B, 60 (1999) 5858. [47] F. Didier, Ph.D. Thesis, Orsay, 1994. [48] C. Li, R. Wu, A.J. Freeman and C.L. Fu, Phys. Rev. B, 48 (1993) 8317. [49] U. Schonberger, O.K. Anderson and M. Methfessel, Acta Metall. Mater., 40 (1992) SI. [50] J.R. Smith, T. Hong and D.J. Srolovitz, Phys. Rev. Lett., 72 (1994) 4021. [51] A.V. Matveev, K.M. Neyman, LV. Yudanov and N. Rosch, Surf. Sci., 426 (1999) 123. [52] G. Renaud and A. Barbier, App. Surf. Sci., 142 (1999) 14. [53] G. Renaud and A. Barbier, Surf Sci., 433-435 (1999) 142. [54] G. Renaud, A. Barbier and O. Robach, Phys. Rev. B, 60 (1999) 5872. [55] C. Goyhenex and C.R. Henry, J. Electron Spectrosc. Relat. Phen., 61 (1992) 65.

252 [56] A. Stirling, I. Gunji, A. Endou, Y. Oumi, M. Kubo and A. Miyamoto, J. Chem. Soc. Faraday Trans., 93 (1997) 1175. [57] I. Yudanov, G. Pacchioni, K. Neyman and N. Rosch, J. Phys. Chem. B, 101 (1997) 2786. [58] J. Goniakowski, Phys. Rev. B, 57 (1998) 1935. [59] A. Barbier, G. Renaud and O. Robach, J. Appl. Phys., 84 (1998) 4259. [60] G. Pachionni and N. Rosch, J. Chem. Phys., 104 (1996) 7329. [61] T. Urano and T. Kanaji, J. Phys. Soc. Jpn., 57 (1988) 3403. [62] C. Li and A.J. Freeman, Phys. Rev. B, 43 (1991) 780. [63] B.M. Lairson, A.P. Payne, S. Brennan, N.M. Rensing, B.J. Daniels and B.M. Clemens, J. Appl. Phys., 78 (1995) 4449. [64] H. Kuhlenbeck, G. Odorfer, G. Tiling, M. Menges, Th. Mull, H.-J. Freund, M. Pohlchen, V. Staemmler, S. Witzel, C. Scharhschwerdt, K. Wennemann, T. Liedtke and M. Neumann, Phys. Rev. B, 43 (1991) 1969. [65] R. Lindsay, P. Baumgartel, R. Terborg, O. Schaff, A.M. Bradshaw and D.P.Woodruff, Surf. Sci., 425 (1999) L401. [66] M. Polcik, R. Lindsay, P. Baumgartel, R. Terborg, O. Schaff, A.M. Bradshavs^, R.L. Toomes and D.P. Woodruff, Faraday Discuss., 114 (1999) 141. [67] M. Baumer, D. Cappus, G. Illing, H. Kuhlenbeck and H.-J. Freund, J. Vac. Sci. Technol.,A10 (1992) 2407. [68] R. Dippel, K.-U. Weiss, K.-M. Schindler, P. Gardner, V. Fritzsche, A.M. Bradshaw, M.C. Asensio, X.M. Hu, D.P. Woodruff and A.R. Gonzalez-Elipe, Chem. Phys. Lett., 199(1992)625. [69] V. Staemmler, in: H.-J. Freund and E. Umbach (Eds.), Adsorption on Ordered Surfaces of Ionic Solids and Thin Films, Springer-Verlag, Berlin, 1993. [70] L.G.M. Pettersson, Theor. Chim. Acta 87 (1994) 293. [71] D. Cappus, J. Klinkmann, H. Kuhlenbeck and H.-J. Freund, Surf. Sci., 325 (1995) L421. [72] G. Pacchioni, K.M. Neyman and N. R0sch, J. Electron Spectrosc. Relat. Phenom., 69 (1994) 13. [73] A.P. Woodhead, S.P. Harte, S.A. Haycock, C.A. Muryn, P.L. Wincott, V.R.Dhanak and G. Thornton, Surf. Sci., 420 (1999) L138. [74] A. Stienbrunn, P. Dumas and J.C. Colson, Surf Sci., 74 (1978) 201. [75] M. Schonnenbeck, D. Cappus, J. Klinkmann, H.-J. Freund, L.G.M. Pettersson and P.S. Bagus, Surf Sci., 347 (1996) 337. [76] D. Cappus, M. HaBel, E. Neuhaus, M. Heber, F. Rohr and H.-J. Freund, Surf Sci., 337 (1995)268. [77] C.L. Pang, S.A. Haycock, H. Raza, P.W. Murray, G. Thornton, O. Gtilseren, R. James and D.W. Bullet, Phys. Rev. B, 58 (1998) 1586.

253

[78] G. Charlton, P.B. Howes, C.L. Nicklin, P. Steadman, J.S.G. Taylor, C.A. Muryn, S.P. Harte, J. Mercer, R. McGrath, D. Norman, T.S. Turner and G. Thornton, Phys. Rev. Lett, 78(1997)495. [79] H. Zajonz, H.L. Meyerheim, T. Gloege, W. Moritz and D. Wolf, Surf. Sci., 398 (1998) 369. [80] B.L. Maschhoff, J.-M. Pan and T.E. Madey, Surf Sci., 259 (1991) 190. [81] H. Onishi, T. Aruga, C. Egawa and Y. Iwasawa, J. Catal., 146 (1994) 557. [82] H. Onishi and Y. Iwasawa, Chem. Phys. Lett., 226 (1994) 111. [83] Y. Yamaguchi, H. Onishi and Y. Iwasawa, J. Chem. Soc. Faraday Trans., 91 (1995) 1663. [84] S.A. Chambers, S. Thevuthasan, Y.J. Kim, G.S. Herman, Z. Wang, E. Tober, R. Ynzunza, J. Morals, C.H.F. Peden, K. Ferris and C.S. Fadley, Chem. Phys. Lett., 267 (1997)51. [85] S.A. Chambers, M.A. Henderson, Y.J. Kim and S. Thevuthasan, Surf Rev. Lett., 5 (1998)381. [86] S. Thevuthasan, G.S. Herman, Y.J. Kim, S.A. Chambers, C.H.F. Peden, Z. Wang, R.X. Ynzunza, E.D. Tober, J. Morals and C.S. Fadley, Surf. Sci., 401 (1998) 261. [87] B.E. Hayden, A. King and M.A. Newton, J. Phys. Chem., B103 (1999) 203. [88] A. Gutierrez-Sosa, P. Martinez-Escolano, H. Raza, R. Lindsay, P.L. Wincott and G. Thornton, accepted Surf. Sci. [89] L.-Q. Wang, K.F. Ferris, A.N. Schultz, D.R. Baer and M.N. Engelhard, Surf. Sci., 380 (1997)352. [90] J. Ahdjoudji and C. Minot, Catal. Lett., 46 (1997) 83. [91] S.P. Bates, G. Kresse and M.J. Gillan, Surf. Sci., 409 (1998) 336. [92] P. Kackell and K. Terakura, Appl. Surf. Sci., 166 (2000) 370. [93] P. Kackell and K. Terakura, Surf. Sci., 461 (2000) 191. [94] R. Davis, J.F. Walsh, C.A. Muryn, G. Thornton, V.R. Dhanak and K.C. Prince, Surf. Sci., 298 (1993) L196. [95] Q. Guo and E.M. Williams, Surf. Sci., 433 (1999) 322. [96] P.J. Hardman and G. Thornton, unpublished results. [97] Q. Guo, I. Cocks and E.M. Williams, Surf. Sci., 393 (1997) 1. [98] L. Patthey, H. Rensmo, P. Persson, K. Westermark, L. Vayssieres, A. Stashans, A. Petersson, P.A. Brtihwiler, H. Siegbahn, S. Lunell and M. Martensson, J. Chem. Phys., 110(1999)5913. [99] B. Hird and R.A. Armstrong, Surf. Sci., 431 (1999) L570. [100] M.A. San Miguel, C.J. Calzado and J.F. Sanz, Surf. Sci., 409 (1998) 92. [101] A. Gutierrez-Sosa, J.F. Walsh, R. Lindsay, P.L. Wincott and G. Thornton, Surf. Sci., 433-435 (1999) 538. [102] W.-J. Chun, K. Asakura and Y. Iwasawa, Chem. Phys. Lett., 288 (1998) 868. [103] W.-J. Chun, K. Asakura and Y. Iwasawa, J. Phys. Chem., 102 (1998) 9006.

254

[104] B.E. Hayden, A. King and M.A. Newton, Chem. Phys. Lett, 269 (1997) 485. [105] H. Raza, S.P. Harte, C.A. Muryn, P.L. Wincott, G. Thornton, R. Casanova and A. Rodriguez, Surf. Sci., 366 (1996) 519. [106] K. Prabhakaran, D. Purdie, R. Casanova, C.A. Muryn, P.J. Hardman, P.L. Wincott and G. Thornton, Phys. Rev. B, 45 (1992) 6969. [107] J. Muscat, N.M. Harrison, and G. Thornton, Phys. Rev. B, 59 (1999) 15457. [108] R.M. Jaeger, H. Kuhlenbeck, H.-J. Freund, M. Wuttig, W. Hoffmann, R. Franchy and H. Ibach, Surf Sci., 259 (1991) 235. [109] D.R. Jennison, C. Verdozzi, P.A. Schultz and M.P. Sears, Phys. Rev. B, 59 (1999) R15605. [110] S. Gota, M. Gautier-Soyer, L. Douillard, J.P. Durand and P. Le Fevre, Surf Sci., 325354(1996)1046. [Ill]U.J. Katter, T. Hill, T. Risse, H. Schlienz, M. Beckendorf, T. Kluner, H. Hamann and H.-J. Freund, J. Phys. Chem. B, 101 (1997) 552. [112]F. Rohr, M. Baiimer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Miiller, L. Hammer and K. Heinz, Surf Sci., 372 (1997) L291. [113]H. Kuhlenbeck, C. Xu, B. Dillmann, M. HaBel, B. Adam, D. Ehrlich, S. Wohlrab, H.-J. Freund, U.A. Ditzinger, H. Neddermayer, M. Neuber and M. Neumann, Ber. Bunsenges. Phys. Chem., 96 (1992) 15. [114] C. Xu, H. Hassel, H. Kuhlenbeck and H.-J. Freund, Surf Sci., 258 (1991) 23. [115] Sh. K. Shaikhutdinov, Y. Joseph, C. Kuhrs, W. Ranke and W. Weiss, Faraday Discuss., 114(1999)363. [116]M. Wuhn, Y. Joseph, P.S. Bagus, A. Niklewski, R. Puttner, S. Reiss, W. Weiss, M. Martins, G. Kaindl and Ch. Woll, J. Phys. Chem B, 104 (2000) 7694. [117] C.B. Duke and A.R. Lubinsky, Surf Sci., 50 (1975) 605. [118] C.B. Duke, R.J. Meyer, A. Paton and P. Mark, Phys. Rev. B, 18 (1978) 4225. [119] N. Jedrecy, M. Sauvage-Somkin, and R. Pinchaux, Appl. Surf Sci., 162/163 (2000) 69. [120] N. Jedrecy, S. Gallini, M. Sauvage-Simkin and R Pinchaux, Surf Sci., 460 (2000) 136. [121] A. Wander, F. Schedin, P. Steadman, A. Norris, R. McGrath, T.S. Turner, G. Thornton and N.M. Harrison, submitted to Phys. Rev. Lett. [122] R. Davis, J.F. Walsh, C.A. Muryn, G. Thornton, V.R Dhanak and K.C. Prince, Surf Sci., 298 (1993) L196. [123] P. Persson and L. Ojamae, Chem. Phys. Lett. 321 (2000) 302. [124] J.F. Walsh, R. Davis, C.A. Muryn, G. Thornton, V.R. Dhanak and K.C. Prince, Phys. Rev. B, 48 (1993) 14749. [125] F.P. Netzer and M.G. Ramsey, Crit. Rev. Solid State Mater. Sci., 17 (1992) 397. [126] J.F. Walsh, H.S. Dhariwal, A. Gutierrez-Sosa, R. Lindsay, G. Thornton and R.J. Oldman, Nucl. Instrum. and Meth. B, 97 (1995) 392. [127] A. Gutierrez-Sosa, S. Crook, S. Haq, R. Lindsay, A. Ludviksson, S. Parker, C.T. Campbell and G. Thornton, Faraday Discuss., 205 (1996) 355.

255

[128] R. Lindsay, A. Gutierrez-Sosa, G. Thornton, A. Ludviksson, S. Parker and C.T. Campbell, Surf. Sci., 439 (1999) 131. [129] A. Gutierrez-Sosa, T. Evans, A.P. Woodhead, R. Lindsay, C.A. Muryn, G. Thornton, J. Yoshihara, S.C. Parker, C.T. Campbell and R.J. Oldman, submitted to Surf. Sci. [130] M.A. Barteau, Chem. Rev., 96 (1996) 1413. [131] St. Hovel, C. Kolczewski, M. Wuhn, J. Albers, K. Weiss, V. Staemmler and Ch. Woll, J. Chem. Phys., 112 (2000) 3909. [132] D. Purdie, C.A. Muryn, S. Crook, P.L. Wincott, G. Thornton, and D.A. Fischer, Surf. Sci., 290 (1993) L680. [133] R. Leysen, B.J. Hopkins and P.A. Taylor, J. Phys. C, 8 (1992) 907. [134] P.A. Taylor and B.J. Hopkins, J. Phys. C, 11 (1978) L643. [135] A. Gutierrez-Sosa, T.M. Evans, G. Thornton, S.C. Parker and C.T. Campbell, in preparation. [136] W.T. Petrie and J.M. Vohs, Surf Sci., 245 (1991) 315. [137] P.M. Jomes, J.A. May, J. B. Reitz and E.L Solomon, J. Am. Chem. Soc, 120 (1998) 1506. [138] B. Kempgens, H.M. Koppe, A. Kivimaki, M. Neeb, K. Maier, U. Hergenhahn and A.M. Bradshaw, Phys. Rev. Lett., 79 (1997) 35. [139] A. Gutierrez-Sosa, T.M. Evans, S.C. Parker, C.T. Campbell and G. Thornton, submitted to J. Phys Chem B. [140] A.C. Liu, J. Stohr, C M . Friend and R.J. Madix, Surf Sci., 235 (1990) 107.

256

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

Chapter 6

Atomic Structure of Oxide Surfaces by Surface X-ray Scattering G. Renaud and A. Barbier CEA-Grenoble, Departement de Recherche Fondamentale sur la Matiere Condensee, SP2M, 17, rue des Martyrs, 38054 Grenoble Cedex 9, France. 1. INTRODUCTION As amply shown in the different chapters of the present book, oxide surfaces [1-10] are involved in various technologically important areas such as composites, protective coatings, thin film technology, electronic as well as nuclear combustible and waste packaging, heterogeneous catalysis, gas sensors and the glass industry. Metal/oxide interfaces where either the metal or the oxide has magnetic properties are of growing interest, since they are already implemented in modem hard drive read heads and may find applications in future devices for magnetic memory elements. As a matter of fact, the electrical, mechanical, chemical or thermal properties of many technologically important devices are often intimately related to the atomic structure of the oxide surface. It is now a widely accepted idea [11] that scientific computing will become a major research field in the future, since it will allow predicting the properties of materials before they are elaborated, thus allowing the development of new technological materials with specific properties. However, in the case of oxide surfaces, the theories are still under development [12-18]. One of the major objectives of experiments on oxide surfaces is thus to determine some parameters that allow testing the theoretical models. Many questions need to be answered, such as: How can the surface be reproducibly prepared? Is the surface stoichiometric or does it have vacancies? Are there steps, regular or not, on the surface? Is the surface free of any contaminant and how to remove them? What is the equilibrium structure and morphology of the surface: is it a simple truncation of the bulk, or is it relaxed or reconstructed? How is it affected by defects or by different surface treatments, for instance annealing in oxygen partial pressure or ion sputtering? What are the different reconstructions undergone by an oxide surface upon reduction by heating in UHV? Are polar surfaces (according to the definition of Tasker [19]) unstable or can they be

257

Stabilised for instance by a reconstruction? How does the surface evolve upon exposure to different gases (H2, CO, NO, H2O, CO2, NO2 etc) since oxides are involved in many important gas sensors; does the surface dissociate water? Are there general trends between similar surfaces? More and more studies are devoted to oxide surfaces prepared under "clean" UHV conditions. A few model surfaces (Ti02, a-Al203, MgO, NiO, ZnO, Fe203, Fe304, Cr203 and SrTi03) have been selected both by theoreticians and by experimentalists. Despite a fast growing interest, often little is known of their exact atomic structure, mainly because of the difficulty to quantitatively characterise insulator with the usual surface science techniques based on electron beams or tunnelling effects. Alternative techniques exist, like the scattering of beams of neutral atoms, but they are only sensitive to the top atomic layer of the surface, and thus can not probe relaxations within the subsurface layers nor buried interfaces. One way of avoiding charging effects is to render the oxide conducting or semi-conducting, either by heating or doping. Another way, which has recently been used to a large extend, is to prepare ultrathin (less than 4 monolayers) oxide films on conducting substrates, thus allowing the use of all surface science techniques. However, these thin films often have defects such as vacancies, interstitials, dislocations, domain walls, different phases and variants, grain and sub-grain boundaries, which may completely change their electronic and thus also their macroscopic properties. It is thus often mandatory to work on single crystals either to preserve given properties, such as a large gap and a high resistivity, or to control the type and the density of defects which may play a dominant role on the surface properties, for instance catalytic. Because the Grazing Incidence X-ray Scattering (GIXS) technique [20-22] does not suffer any limitation when applied to insulators, it has been more and more used in the recent years to characterise single-crystal oxide surfaces [23]. This probe has several advantages with respect to electron based surface techniques. X-rays interact weakly with matter, so that a simple quantitative analysis based on a single scattering (kinematic) calculation can be performed, while multiple-scattering effects dominate electron diffraction. X-rays penetrate deeply in matter, enabling the study of atomic displacements deep below the surface. The scattering cross section is very well known for all atoms. X-ray diffraction peaks can be measured with very high resolution and over a large intensity range, thus enabling detailed lineshape analyses that are not accessible with other diffraction techniques. GIXS also benefits from the huge amount of work performed in conventional 3D crystallography, and hence allows the determination of the atomic structure of surfaces and interfaces with high accuracy. The grazing incidence geometry can drastically reduce the x-ray penetration in matter, down to -25 A, and in turn the unwanted bulk elastic and

258

inelastic scattering with respect to the measured surface or interface elastic scattering. The large intensity of today's third generation synchrotron radiation sources allows getting measurable diffraction from as low as 0.01 monolayer (ML) of material. In this chapter, we first recall the basis of X-ray scattering under grazing incidence. The diffraction by a surface is next considered, with methods to quantitatively determine the surface atomic structure (roughness, relaxation or reconstruction). Experimental conditions are next discussed, in particular the sample requirements to obtain reliable quantitative measurements. Several recent determinations of the atomic structure, relaxation or reconstructions of clean oxide surfaces will next be described. A quick overview of the characterisation of diverse oxide thin films by GIXS will then be given before concluding on the present and future possibilities of the technique applied to the field of oxide surfaces.

2. X-RAY SCATTERING BY SURFACES 2.1. Grazing incidence X-rays. The grazing incidence X-ray scattering geometry, shown in Fig. 1, is identical to the three dimensional case, except that, in order to decrease the bulk scattering contributions, the incident X-ray beam, of wavevector k., is kept at a glancing incident angle a^ with respect to the surface. The scattered beam, of wavevector k^ is detected at an angle a^ with respect to the sample surface and at an in-plane angle 20 with respect to the transmitted beam. The momentum transfer Q=kf-k., is often decomposed into two components, Q,, and Qi, respectively parallel and perpendicular to the surface. When a. and a^ are very small, Q-Q,/, the scattering plane is nearly parallel to the surface, and diffracting net planes are perpendicular to it. The scattering geometry being defined by the incident beam and detector directions, one has

Fig. 1. Grazing incidence X-ray scattering geometry. All notations are defined in the text.

259

only to rotate the sample about its surface normal to bring these net planes into diffraction condition, which occurs when they make an angle 0 with respect to both the incident and the scattered beam. In this way, the long-range periodicity parallel to the surface is probed. Since the incident angle is small, it is necessary to consider the effects of refraction at the surface. Because the refractive index of matter for X-rays is slightly less than unity, when a. is smaller than a critical value a^ (typically of a few 0.1°), the beam is totally externally reflected, and only an evanescent wave, which decays over tens of angstroms, is present below the surface. When a. is larger than the critical angle for total external reflection, the transmitted wave propagates. Due to time microreversibility, identical refractive effects occur as a function of the exit angle a^ When a.<
(Q.ai )S^^ {Q.a, )sl^ {Q.a,)

(1)

where A is a constant and 5wCQ.aj)=EexpfiQ.aj.nA;=l2,3 n=0

and F(Q) =

^

fj expf/Q.rjj

(2)

j unit cell

F(Q) is the structure factor, and r. are the atomic positions within the unit cell; fj is the scattering factor (or atomic form factor) of atom j . The intensity is thus the product of the structure factor, which only depends on the atomic positions within the unit cell, by the form factor, related to the shape of the crystal. In the limit of large A^, the S^ function tends to a periodic array of Dirac delta functions with Q spacing of 27i/a, i.e. the intensity is nonzero only if Q.ai = 2;r/i, Q.a2 = 27dc and Q.a3=2;r/, with h, k, I integers. This expresses that Q is a vector of the reciprocal lattice of basic vectors b^, h^ and

260

bj, i.e. Q = hbj + kb^ + Ihy When this Laue condition is fulfilled, the intensity is given by: / « , = AF^,^NlNlNl

with F,,, -

X /;• ^ x p N [hxj+kyj+lzj)\ exp-M^- (3) j unit cell

The summation extends over all atoms of the unit cell; f., x.^ y.y z., M. are respectively the scattering factor, fractional co-ordinates in the unit cell and Debye-Waller factor of atom j . 2.3. Diffraction by a surface Consider now a quasi-two dimensional crystal of finite thickness. The basic cell vector a3 perpendicular to the surface is chosen equal to this thickness. This crystal is handled by setting N^= I. The diffraction is then still sharply peaked in both directions parallel to the surface, but the Laue condition on Q^ (= Q_i) is relaxed, and the intensity is continuous in the out-of-plane direction: the reciprocal space is made of rods perpendicular to the surface plane. If we still define I by Q.aj = 2 Til, I is now taken as a continuous variable since intensity is present for non-integer values of I. The intensity is now given by: li^=AF,lN^Nl

(4)

The intensity variation along the rod (i.e. as a function of Q^ or Z) is solely contained in the structure factor; it is thus related to the z-co-ordinates of the atoms within the unit-cell of this quasi-two dimensional crystal. In general, the rod modulation period gives the thickness of the distorted layer and the modulation amplitude is related to the magnitude of the normal atomic displacements. This is the case of a reconstructed surface, for which rods are found for fractional order values of h and /:, i.e. outside scattering from the bulk. Consider now a perfect crystal truncated by a sharp surface (or semi-infinite crystal). It can be obtained by the product of a step function describing the electron density variation perpendicular to the surface, and an infinite lattice. The diffraction pattern is then the convolution of the 3D reciprocal lattice with the Fourier transform of the step function. An infinity of Fourier components is necessary to build this latter, so that there remains non zero intensity in between Bragg peaks as a function of /: the reciprocal space is made of rods of intensity, called crystal truncation rods (CTR), extending perpendicular to the surface, and connecting bulk Bragg peaks [24, 25]. The intensity variation as a function of Qs (or Qj^ or /) is found by stopping the summation at ns = 0 in Eq. (1) and (2), yielding:

261

'

2 sin^{N^a^Q^/2)sin^{N2a2Q2/^) ' sin\a^Q,/2) sm\a2Q2/2) 2sm\Q^a^)

Bragg peaks are found for integer values of /, but there remains some intensity in between, even when / is a multiple of half-integer, i.e. when successive net planes scatter out-of phase. At these anti-node positions, I^J^ and I^^ have comparable magnitudes: the intensity diffracted by the semi-infinite lattice is of the order of the intensity diffracted by a single monolayer. If the surface is rough on an atomic scale, the step function has to be replaced by a less abrupt function, which needs fewer Fourier components to be built, so that the intensity of the CTR's between Bragg peaks is smaller. As the roughness increases, the intensity between Bragg peaks is reduced, and the result for an infinite 3D crystal is approached. The intensity variation along CTR's is particularly sensitive to the difference between the bulk and surface structures. Let us take for instance a surface whose last interplanar distance is b instead of as. This produces a large asymmetry of the CTR intensity around Bragg peaks. The larger the maximum value of the perpendicular momentum transfer, the larger the interference term and thus the asymmetry, and the better the accuracy on this relaxation. Hence, the measurement of CTR's allows the determination of the atomic structure of a surface. The above expressions of the scattered intensity show that, in general, the reflections have a finite width, which is related to the finite size of the crystal. In order to deduce the structure factor, containing the required structural information at the atomic level, the pertinent measurement is to integrate the intensity of each reflection. Several corrections have next to be applied, which may be found in recent references [26-29]. The determination of the atomic structure of a reconstruction requires the quantitative measurement of as many allowed reflections as possible. Given the structure factors, standard Fourier methods of crystallography, such as Patterson function or electron-density difference function, are used. The experimental Patterson function is the Fourier transform of the experimental intensities, which is directly the electron density-density autocorrelation function within the unit cell. Practically, a peak in the Patterson map means that the vector joining the origin to this peak is an interatomic vector of the atomic structure. Different techniques may be combined to analyse the Patterson map. On the basis of a set of interatomic vectors obtained from the Patterson map, a trial structure can be derived and model structure factor amplitudes calculated and compared with experiment. This is in general followed by a least-squares minimisation of the difference between the calculated and measured structure factors. Of help in the process of structure determination may be the difference Fourier map, which is

262

a difference between the calculated electron density for the model structure, and the electron density obtained by a Fourier transform of the experimental structure factors, but by assigning them the phase calculated from the model. When the model is close to the real structure, this approximation may be valid, and yields a direct picture of excess or missing electron density in the unit cell. In the very recent years, so-called "Direct Methods" have been applied to surface crystallography [30]. They use the strongest reflections in order to determine the possible phase shifts between the different measured reflections, thus trying to unravel the long-known problem of missing phase in crystallography. They might replace the more traditional methods in the future if they are able to prove their exactness in very different cases of complicated structures. 2.4 Instrumental considerations Numerous experimental setups to perform GIXS were recently described in the open literature [31-34]. We just recall that in almost all cases, a high brilliance synchrotron radiation X-ray beam is needed, especially from 3^^ generation synchrotron radiation sources like the ESRF in Grenoble (France), and they always couple a large UHV chamber with standard sample preparation with a heavy duty high precision goniometer that allows the orientation of the sample and the detector. 3. SURFACES OF OXIDE SINGLE CRYSTALS 3.1 Specific sample preparation and requirements One of the most important requirements to perform GIXS is the quality of the sample. It should be a good single crystal, at least over a thickness of several tens of nanometers below the surface. The surface flatness (i.e. the width of the angular distribution of surface normal over long length-scale parallel to the surface) should be much smaller than the critical angle for total external reflection, i.e. typically smaller than 0.01°, and the surface roughness should not exceed a few angstroms. These conditions are not trivial, and are often the limiting factor for successful experiments. For metal or for semi-conductors, procedures to prepare surfaces of high quality are well documented, for instance by ion sputtering and annealing cycles or by deposition and annealing of a buffer layer. This is often not the case for the surfaces of oxides, which in addition are not always stoichiometric: specific preparation procedures have thus to be developed. Only a few oxide surfaces have been quantitatively investigated by GIXS: among them the surfaces of sapphire a-Al2O3(0001), MgO(OOl), rutile TiO2(110), NiO(lll), ZnO(OOOl) -O and -Zn, Cr2O3(0001), SrTiOsCOOl). The

263

two first compounds are very light scatterers, with an average number of electrons per atoms of 10. Therefore, in order to get measurable CTR's, the intensity needs to be concentrated over a very narrow angular range along these CTR's. This requires two conditions: first a starting single crystal of high crystalline quality, with a very small mosaic spread, and second a very flat surface on the length scale of the coherence length of the X-ray beam, typically 1 ]Lim. In many cases, but not always, the crystalline quality and the surface flatness can be improved by an annealing in air or under partial oxygen pressure at high temperature (~1500°C for a-Al203 and MgO(OOl)). The rocking curve full width at half maximum then decreases from -0.03° down to 0.0025° in the case of sapphire, and from -0.01° down to 0.001° in the case of MgO, thus yielding a ten-fold enhancement of the peak intensity along the CTR's, while leading to a negligible rms roughness. However, in some cases, the high temperature anneal has drawbacks like stoichiometry changes or surface segregation of bulk impurities. The latter phenomenon can be minimised by limiting the duration of the annealing to a few hours. However, on some surfaces, segregation can not be avoided, as for MgO(OOl), with Ca segregation. In this case, performing ion bombardment at high temperature can solve the problem. In addition, if we wish to get a high accuracy on the atomic coordinates, it is necessary to measure the CTR's over an extended range of Qi, which requires an X-ray beam of large enough energy (typically between 15 and 25 keV). At these energies, as soon as the incident angle is larger than the critical angle for total external reflection, a large background scattering is present that overcomes the surface scattering. Consequently, the experiment must be performed under very stringent conditions, with the incident angle kept equal or below the critical angle for total external reflection. Since in this region the amplitude of the wavefield varies very rapidly with the incident angle, the diffractometer as well as the sample alignment have to be of the topmost quality. As soon as the atomic species become heavier, which is the case of most oxides, the above constrains somewhat relax. However, because the oxygen ions are much lighter scatterers than the metal ones, their positions will be determined with a lesser accuracy. 3.2.

MgO(OOl) surface [35] The MgO(OOl) surface has been the object of numerous studies because it is widely used as a substrate for the epitaxial growth of metals [36-38] and as a model support for dispersed small catalytic particles. It is also used as a substrate to grow high-temperature superconductors because of its close lattice match to YBa2Cu307.x and its low chemical reactivity. Getting a precise knowledge of the MgO(OOl) surface atomic structure is also important because this surface has been chosen as a model system for

264

testing calculations on ionic oxides [39]. In particular, the top plane relaxation and the differential relaxation between anions and cations (rumpling) have been extensively studied, both theoretically [18, 40-42] and experimentally [43-45]. These relaxations were always found extremely small. As shown below, when it is performed on a surface of very high crystalline quality, GIXS is well suited for studying such small deviations from the ideally truncated surface. For MgO(OOl), the procedure for preparing surfaces of high quality necessarily involves a first step of annealing at high temperature (1500°C1600°C). However, this annealing also results in a strong surface segregation of bulk impurities, mainly calcium. A procedure was thus developed in order to remove the Ca surface contamination while keeping the MgO surface as perfect as possible. For this purpose, the surface was etched by Ar"^ bombardment at 1550°C, which is a temperature high enough to allow the surface to reorder faster than it disorders, and to keep its smoothness. After this treatment, oxygen or magnesium vacancies could be expected on the surface. In order to restore the surface stoichiometry, the sample was annealed for 15 minutes at 700°C in a partial oxygen pressure of 10"* mbar. The surface cleanness was checked using AES, and no remaining impurities were found, to the level of 1% of a monolayer. After this preparation, the samples were never exposed to air, in order to avoid the well-known attack of the surface by water vapour [46]. The GIXS measurements on this clean surface, were aimed at the determination of the roughness, the relaxation p and the rumpling e, defined according to: p = 1/2 (ei H- 62) and 8 = Bi - 82, where 81 and 82 are respectively the fractional displacements of the surface anions and cations, expressed in percentage of the bulk interplanar distance (2.106 A) perpendicular to the surface. For the MgO(OOl) surface, there are two non-equivalent CTR's: (i) « strong » ones with h and k even with intensities proportional to the square of the sum of the atomic form factors of O and Mg, and (ii) « weak » ones, with h and k odd with intensities proportional to the square of the difference of the form factors. When h and k have a different parity, the CTR's are forbidden by symmetry. The « strong » (20f) CTR was mostly used to determine the r.m.s. roughness and the « weak » (1H) one to determine the surface relaxation. Both CTR's, measured by rocking the sample, were everywhere above background, with a very small width (-0.01°), always resolution limited, whatever the experimental resolution, which was a confirmation of the high crystalline quality, and a first indication of a small surface roughness. The Lorentzian shape (FWHM 0.01°) at the in plane anti-Bragg location (1 1 0.05) indicates an exponentially decaying height-height correlation function with a terrace length of -6000 A.

265 10 F

(110

Exp. data

3

(a)

(0 _ + 3 % rumpling no relaxation 2% relaxation no rum pling

1 \10

F

J

I

I

t

I

I

-L.

13 03

10'

0

1

2

i ( r . l . u . of M g O )

3

Fig. 2. Modulus of the structure factor of the (11/) and (20/) CTR's of the clean MgO(OOl) surface. For the (11/) CTR, the continuous line is the best result of a simultaneous fit of the (20/) and (11/) data. The measured (20 /) CTR is fitted with an rms roughness of 2.4 A. The (20/) CTR calculated for a perfectly flat surface is also shown (thin continuous line) for comparison.

Figure 2 shows the (20^) and (IH) integrated CTR's. Figure 2a illustrates the sensitivity of the (11^) CTR to rumpling and relaxation, which is obtained only if the roughness is small enough, and when the CTR is measured over an extended range. Both CTR's were simultaneously fitted with four parameters: an overall scale factor, the relaxation and rumpling in the top plane, as well as the r.m.s. roughness. All step height possibilities had to be introduced, which indicates that most steps are presumably only one atomic plane high, i.e. 2.1 A. The normalised chi-squared agreement factor x^ of 1.1 ±0.1 was very close to the ideal value of 1, which shows that no new parameter, such as atomic relaxations of deeper atoms, could be added. All substrates prepared according to our new procedure yielded the same roughness value of 2.4 ±0.1 A, which is also the value determined on the Casegregated surfaces. Because the relaxation and rumpling are both very small, slightly different values were found on the different substrates. The average values of p = (-0.56 ± 0.35) %, and E = (1.07 ± 0.5) % are thus given, with the error bar estimated from the uncertainties of each fit, and from the different values obtained. In the original paper [35], these values are compared to previous experimental and theoretical determinations. Many early shell model calculations and several experiments yielded much too large rumpling values. In most cases, this can be attributed to an inadequate substrate preparation, i.e.

266

exposure to air before introduction in the UHV chamber. Most other theoretical or experimental results are close to the present ones. In summary, MgO(OOl) surfaces can be prepared with a very high quality and are ideally suited to perform GIXS measurements. This offers the opportunity to investigate the atomic structure and morphology of metal/MgO interfaces by this technique, during in situ deposition in UHV by molecular beam epitaxy, from the very early stages of sub-monolayer deposition, up to fairly thick metallic layers [23]. 3.3. a-Al2O3(0001) -(1x1) and reconstructed surfaces [47-49] The (0001) surface of sapphire (a-alumina, corundum) is one of the most widely used substrate for the growth of metal, semi-conductor or hightemperature superconductor thin films. It is also used as a substrate in silicon on sapphire (SOS) technology. Moreover, its initial state is known to play a role on the overlayer properties [50]. Despite many theoretical calculations of the a-Al2O3(0001) surface structure and relaxation [19, 51-55], the nature (Al or O) of the terminating plane of the unreconstructed surface was still an open topic because of the lack of experimental results. The single Al terminated surface is favoured by electrostatic considerations [19] as well as surface energy calculations. Indeed, for an Al termination no dipole moment is left across the surface and only the longer, and thus the weaker, anion-cation bonds are broken. However, the (1x1) structure had also been experimentally observed on alumina surfaces heated in an oxygen-rich atmosphere and could thus be suspected to be oxygen terminated. As regards the Al-terminated surface, large relaxations have been predicted by pair-potential calculations. More recently, ab initio calculations, by the density functional theory combined with pseudopotential techniques predicted very large relaxation of the last atomic plane (-87%), while HartreeFock calculations yielded smaller, although still sizeable, relaxation (-40%). A tight binding, total-energy method also predicted large out-of-plane relaxations of the Al planes and in-plane displacements of the oxygen atoms. The aim of the GIXS study was to determine experimentally, for the first time, the nature of the terminating plane and the relaxations of the first few atomic planes below the surface by quantitative measurement and analysis of the CTR's. The experiments were performed on a first sample using the LURE W21 beamline and diffractometer [33], and on a second sample using the ESRF ID3 surface diffraction setup [32]. The samples were first annealed in air for three hours at 1500°C, resulting in a surface with wide, atomically flat terraces, and a good near-surface crystalline quality. Both data sets yielded exactly the same conclusions. During the second, more precise experiment, eight CTR's were measured over an extended range of perpendicular momentum transfer

267

Fig. 3. (10/) CTR or the a- Al2O3(0001)-(lxl) surface. Experimental (solid circles) and bestfit models for each possible termination: single Al layer (thick solid line), double Al layer (dashed line) and oxygen terminated surfaces (dotted line). The logarithm of the structure factor is reported as a function of the out-of-plane momentum transfer in reciprocal lattice units of AI2O3.

from 0 up to 7.2 A'\ and corresponding to 873 non-equivalent reflections. The (10^) CTR is reported in Fig. 3. Bulk sapphire has rhombohedral symmetry, which is usually treated as hexagonal (space group/?3c), with 30 atoms (six AI2O3 units) per primitive unit cell. The lattice parameters {a=b=4.1570 A, c= 12.9877 A) and the internal coordinates (x=0.3063, z=0.3522) are taken from ref. [56]. The bulk unit cell consists of an alternated stacking, along the c-axis, of two Al^"^ planes (twelve in the unit cell) with one atom per plane, and one oxygen plane (six in the unit cell) with three O^' ions arranged with a threefold symmetry. From a crystallographic point of view, no two planes among the 18 of the unit cell are equivalent, under any translation. For each possible chemical termination (oxygen, single Al or double Al) there are thus six crystallographically non-equivalent terminations. Fortunately, only two of these yield non-equivalent diffraction patterns. The (1x1) surface has p3 symmetry with the threefold axis lying on the Al atoms. Hence, only their co-ordinate along the c-axis can vary, whereas oxygen can also move parallel to the surface plane as long as they keep the threefold symmetry around the Al sites. Because of this symmetry, they must also remain coplanar. The surface model included displacements of three AI2O3 units plus those of the terminating planes (15, 16 or 17 parameters depending on the termination). Because of the extremely high sensitivity of CTR's to atomic positions, deeper atomic displacements can

268

enhance the fit quality even with very small values but they are not significant for the structure. Al^"^ and O^' atomic scattering factors were used for all atoms although theoretical arguments as well as an experimental Auger electron spectroscopy study [57] suggests that surface atoms may have a different charge. This should be of little importance especially at large momentum transfer. Debye-Waller parameters were taken to be isotropic and set to the bulk values for all atoms. Least square fitting and visual inspection (Fig. 3) clearly allowed ruling out the oxygen and double Al terminations. The surface is thus terminated with a single Al layer, which yielded x^=0.92. The surface structure is schematically shown in Fig. 4. The top two planes undergo large displacements from their bulk positions (0.34 A and 0.23 A respectively). The top Al layer moves inwards so that the interplanar spacing with the nearest neighbour oxygen layer is reduced by 51%. The underlying oxygen atoms shift mainly parallel to the surface plane and are repelled from the first layer Al sites, moving almost radially towards the second layer. This is an almost bond length conservative motion: the bond length is only 4.5% (±2.5%) shorter than the bulk nearest neighbour value. Displacements below the first two planes are smaller: relaxations are +16%, -29% and +20% for the next three interplanar spacings. These results were compared with theoretical calculations. The single Al termination was always the predicted one, because it is the only termination that is « auto-compensated » (both charge neutral and chemically stable) and with no electric dipole moment across the surface. Regarding the surface relaxations, all theoretical studies agree with the large negative relaxation of the top Al layer. Godin et al. [55] suggested a rehybridisation of the top Al atoms from the fourfold co-ordinated arrangement (which may be thought as a sp^ configuration) to a nearly perfect sp^ configuration. More generally, all authors agree on the stronger bonding between Al and oxygen atoms nearby the surface, which is the general trend of the GIXS results, bond lengths being shortened by 4.5% to 6.1% in the first four planes. The results are also in very good agreement with very recent theoretical [58], and experimental [59] results. To our knowledge, this is the first study of the relaxations of an oxide surface by GIXS, and the first experimental determination of the relaxation and termination of the a-Al2O3(0001) surface. Another determination has been performed very recently by combining time-of-flight scattering and recoiling spectrometry with LEED and classical ion trajectory simulations [59], with essentially the same results.

269

Fig. 4. Illustration of the sapphire (0001) surface: truncated bulk and best-fit model obtained from CTR's measurements. The two top planes are strongly shifted with respect to their bulk positions in a nearly bond length conservative displacement.

When the a-Al^OgCOOOl) surface is heated at high temperature in UHV, several reconstructions appear: (V3XV3)R30° around 1100°C, (2V3 x 2V3)R30° around 1150°C, (3V3x3V3)R30° around 1250°C and finally (V31XV31)R±9° around 1350°C. Although their electronic structure and symmetry v^ere wellcharacterised [60-62], their atomic structure remained essentially unknown. The (V3ixV3i)R±9° reconstruction is of particular interest because it has been reported to help epitaxy and enhance adhesion in some cases [63] and because it is unusually stable, even after air exposure. A structural model for this reconstruction had been proposed three decades ago. The LEED pattern was interpreted as the superposition of two reciprocal lattices: that of the hexagonal substrate and that of a nearly cubic overlayer with composition AI2O or AlO, plus the interference pattern because of double diffraction. However, this model remained controversial. This interpretation did not include a supercell formation with atomic relaxations. The aim of the GIXS study was to analyse the a-Al2O3(0001)(V3ixV3i)R ±9° reconstruction in order to get unambiguous answers concerning the presence of a supercell, and ultimately to determine its atomic structure. Clearly, in the case of a reconstruction or of a thin layer, multiple scattering of X-rays is completely negligible. Hence, the GIXS pattern can be fully interpreted using the kinematic theory of diffraction, where only the single scattering events are taken into account. The a-Al2O3(0001) single crystals were first annealed in air at 1500°C for three hours, and next heated to ~1350°C for -20 min. in UHV to obtain the (V3ixV3i)R± 9°reconstruction. Measurements were performed using the LURE

270

W21 beamline and diffractometer [33]. A large number, 366 of which 267 were non-equivalent, of in-plane reflections arising from the reconstruction were measured. All peaks were exactly centred at the expected positions to within 0.001° of azimuthal rotation, which showed that the surface reconstruction is perfectly commensurate with the underlying bulk lattice. Their width and Lorentzian shape indicated an exponential decay in correlations with the decay length of -500 A. Several reconstruction diffraction rods were also measured. The absence of symmetry of the rod intensity with respect to £=0 showed that the reconstruction has the minimal hexagonal symmetry p3. The experimental diffraction pattern (Fig. 5), has six-fold symmetry. Measurable intensity was found at all reciprocal lattice points of the reconstructed unit cell, even far away from bulk Bragg peaks, proving that there indeed is a genuine (Vsi x V3i)R ± 9° supercell formation with atomic relaxations. The diffraction pattern was shown to be qualitatively very similar to that predicted [64, 65] in the case of rotational epitaxy of an hexagonal overlayer, which is expanded and rotated with respect to an ideal overlayer R in perfect registry. The main peaks (hatched in Fig. 5) correspond to the first-order approximation, called "parent" phase, of the adsorbed structure. Their locations yield the expansion, 10.62%, and rotation, 2.361°, applied to the R phase to obtain this rigid hexagonal "parent" phase. The other diffraction peaks are satellites corresponding to the static distortions of this parent phase, and possibly to additional disorder. Figure 6 shows the experimental paircorrelation (Patterson) function. Most Patterson peaks have a nearly perfect hexagonal arrangement.

0 1

3

5

7

9

11 13 15 17 19 21 23 25

h

Fig. 5 Experimental diffraction pattern. The radii of the right-hand halves of the open circles are proportional to the experimental structure factors, while the left-hand open circles are calculated from the best model {j}-\.T). Black disks Bragg and CTRs reflections.

271

The positions of these peaks can be directly constructed by a rotation of -1.4° followed by a small expansion of the projected atomic positions of an FCC(lll) stacking on top of the oxygen HCP(OOOl) stacking of the underlying bulk lattice. The reconstructed structure was interpreted as a tiling of domains bearing a close resemblance to that of two metal Al(l 11) planes, separated by a hexagonal network of domain walls. In the middle of domains, the overlayers are well ordered, with a lattice parameter very close to that of metallic Al (expansion of 4 % with respect to the registered state), and a small rotation (-1.4° with respect to the R state) with the epitaxial relationships: (lll)Al//(0001)Al2O3 and [TlO]Al//(R1.4°)[1120]AlA- In the domain walls, large expansion and rotation, and even loss of honeycomb network topology were found. The observed structure was interpreted in the spirit of rotational epitaxy with non-linear distortions. We have shown just above that the unreconstructed a- Al2O3(0001) surface is terminated by an Al layer with 1/3 compact packing. Hence, starting from this surface, removing the two last O planes would leave 5 Al layers with 1/3 compact packing occupancy at the surface, which is the observed 5/3 filling ratio. It was then suggested that the reconstruction be obtained after evaporation of the two upper oxygen layers of the unreconstructed surface. The physical origin of the ~4 % expansion in the domains is clear, since the overlayer is very close to bulk Al and registry. The observed atomic structure was shown to be consistent with previous studies of the (V31XV31)R± 9° reconstruction [66], which yielded an Al enrichment, intermediate oxidation states of surface aluminium atoms and a reduced surface band gap.

Fig. 6. Experimental Patterson map of the a-Al2O3(0001)-(V31xV31)R± 9° reconstruction in the whole reconstructed unit cell. Lines are only guides to locate the 3-fold axes and the centred 2-fold axis of the Patterson p6 symmetry. Bold lines delimit the asymmetric unit cell of the Patterson map.

272

It is also consistent with the observation [67] of a (Vsi X V3i)R ± 9°reconstruction during the first stage of Al deposition (between 0.4 and 2.5 Al(lll) ML coverage) on an a- Al2O3(0001) surface with (1x1) structure, followed by Al(lll) domain growth for larger coverage. Thus, a fundamental question was opened concerning the process and dynamics of this reconstruction formation by different routes: reduction or Al deposition. This motivated a structural study of the intermediate reconstructions. The (2V3x2V3)R30°and (3^3 X3V3)R30° reconstructions were investigated by GIXS at the ESRF, on the ID3 and BM32 beamlines, under experimental conditions very similar to those of the (VM x V31)R ± 9° study. Both reconstructions can be prepared in a very well defined state, with large domain sizes, and very similar data were obtained in both cases. The experimental Patterson map of the (3V3x3V3)R30° reconstruction is shown in Fig. 7. Out-of plane rod measurements showed that, as in the previous case, the thickness of the reconstruction is limited to one or two atomic planes. Although the analysis has not yet been performed, a qualitative comparison with the diffraction pattern and Patterson map of the (VM x V31)R ± 9° reconstruction shows that the structures are likely to have the same origin, excepted that no rotation is involved in the case of the (3V3x3V3)R30° reconstruction. More precisely, the structure of the (3V3X3V3)R30° reconstruction is likely to consist of an overlayer with hexagonal symmetry, made of one or several planes close to compact planes, and with a lattice parameter slightly larger (-12%) than that of sapphire, such that, along the [110] directions of sapphire, the two lattices coincide every 9 (110) dspacing of sapphire, and 8 (110) d-spacing of the overlayer, yielding the (3V3 X3V3)R30° superlative unit cell of -25 A periodicity.

Fig. 7. Experimental Patterson map of the a-Al2O3(0001)- (3V3 x 3V3)R30° reconstruction in the whole reconstructed unit cell. Bold lines delimit the asymmetric unit cell of the Patterson map.

273

Such a rigid overlayer would yield only the strongest peaks of the diffraction pattern. All the other weaker peaks would correspond to harmonics in Fourier decomposition, arising from small displacements of the atomic positions with respect to the average "parent" rigid lattice. A quantitative analysis with modelling of the compact overlayer is required to get a more detailed picture of the structure. 3.4.

TiO2(110)-(lxl) and (001)-(lx3) surfaces The surfaces of rutile Ti02 have been the subject of intense research because of their photo-catalytic properties for the dissociation of water. The hydroxylation rate on the surface and the kinetics of the reaction were shown to depend strongly on the surface stoichiometry and detailed atomic structure. In addition, like the two above surfaces of sapphire and magnesium oxide, rutile titanium dioxide surfaces stand as model metal oxide surfaces. Their atomic structure is thus of fundamental interest. The structure of the TiOaCHO) surface has been thoroughly studied recently [1, 68-71]. This is the most stable surface of rutile Ti02: only after extensive heating (14 h) at elevated temperatures (900 K) this surface reconstructs to form a (1x2) surface. A recent GIXS study of the CTR's allowed the determination of the structural relaxations of the unreconstructed surface [68]. The surface was prepared in situ by repeated cycles of Ar^ sputtering and annealing at 1100 K, followed by an anneal at 900 K and cooling to 550 K in 110"^ mbar O2 to restore the surface stoichiometry. No surface contamination could by detected by AES. Five CTR's were measured from 0 up to 4.8 A"\ and fitted using 23 fitting parameters including 13 atomic displacements. The final %^ value was 3.8, which is quite large, compared to the ideal value of 1. This could indicate either that a more accurate model could be found, or that the experimental data have additional uncertainties, with respect to the one assigned to the measured structure factors. A possibility recently suggested by theoretical calculations [2] could be that the top Ti atom should be assigned a very large and anisotropic Debye-Waller factor corresponding to a soft mode of vibration. The displacements are schematically shown in Fig. 8. The main relaxations involve the top layer six-fold co-ordinated Ti atoms moving out of the surface and the top layer fivefold co-ordinated Ti atoms moving towards the surface. This creates a rumpling of 0.3 ±0.1 A of the top layer, with a period 6.5 A, the unit cell length along the [110J direction. This rumpling is repeated in the next plane of titanium atoms, with amplitude about half that in the top Ti plane. The relaxation induced modification to the bond length ranges from 11.7 % contraction to 9.3 % expansion. The relaxations deduced from a recent detailed

274

Fig. 8. Real space model of bulk-terminated TiO2(110)-(lxl). Small black circles represent Ti and large grey circles O. The arrows indicate the direction of the relaxations. The atom types varied in the structural refinement are shown. Only the in-plane oxygen atoms were allowed to move laterally in the fit. On symmetry grounds the reminder of the atoms were constrained to move perpendicular to the surface. (Reprinted from Ref. [68]).

theoretical calculation [73] w^ere compared to the experimental ones. There is good agreement in the Ti atom positions, with some significant discrepancy for the relaxation of O atoms. Worse agreement was obtained with other two recent calculations [74, 75]. Contrary to the (110) face, the clean (100) rutile surface is unstable, and facets upon annealing to form (110) and other low index faces [76, 77]. Annealing in UHV the sputter-cleaned (100) surface at 873 K results in the formation of a (1x3) reconstruction. The atomic structure of this reconstruction has been first studied by a combined GIXS and LEED study in 1992 [78], which provided the basis of a structural model, with formation of {110} micro facets. In the first study [78], because of a restricted experimental geometry, only a small portion of the reciprocal space was accessible. A total of 19 in-plane surface diffraction peaks were recorded, and a Patterson map was used to propose a model consisting primarily of (110) planes. The Patterson function is dominated by Ti-Ti pairs because Ti is a much stronger scatterer than O. Since the experimental uncertainty was fairly large, and a significant background was present compared to the surface scattering, no attempt was made to locate O atoms from the Patterson map; a composite electron density map was rather used, together with the maximum entropy method. The final in-plane atomic positions were determined by least squares refinement. The determination of the perpendicular atomic positions was done by performing an LEED-//V experiment including a full dynamical calculation of I-V curve profiles. The micro-facets model can be explained if the most stable surface is formed by breaking the minimum number of cation-anion bonds. Recent photoelectron diffraction (PED) [79] and STM

275

[80, 81] studies have confirmed the micro-facets model, but contradicted the original conclusion that the surface is terminated by oxygen layer. The uncertainty on these first measurements as well as the small number of measured reflections could have led to a misleading interpretation. Nevertheless, these same in-plane data were used more recently [82] to re-visit the structure by using a Direct method. The proposed structure does not show (110) microfacets, but rather some "missing rows" of octahedra, and contains edge- and face-sharing octahedra similar to known defect structures in Tin02n-i. A new, more detailed GIXS investigation [83] has thus been performed recently, including that time out of plane rod data, to re-examine this reconstruction. A total of 464 reflections were measured of which 229 were non-equivalent: 22 in plane, 131 fractional order superlattice and 86 CTR's reflections. The r.m.s. roughness deduced from the fit is 5.6 A, which is small enough to allow measurement of the complete CTR's thanks to the large scattering power of Ti. As noted by the authors, and this is a very general remark, the combined use of reconstruction rods and CTR's in the refinement enhances the reliability of the final structure model. Indeed, because both the reconstruction and bulk scattering interfere along the CTR's, the CTR's present a specific phase contrast and are more sensitive to surface relaxations than the reconstruction rods. Figure 9 shows schematically the final structure in a projection along [001]. Strong lateral and vertical relaxations of the Ti and O atoms in the top layer were found. The co-ordination numbers of the atoms on the surface facets were found to differ considerably from those derived by Zschack et al [78].

-nterst.tia! sue \

Fig. 9: Final structure model of the rutile (100)-(lx3) surface projected along [001]. An arrow indicates the origin of the superstructure cell. The titanium atoms labelled Til-Ti5 and A, B are co-ordinated by oxygen as follows: Til threefold; Ti2 fivefold; Ti3 bridge site; Ti4 sixfold; Ti5 sixfold; A fivefold (interstitial site) and B sixfold (trigonal prismatic). (Reprinted from [83]).

276

For instance, the Til atom has a three-fold co-ordination by oxygen on the facet ridges, as opposed to bulk co-ordination. In the final model, a surface octahedral interstitial site of the O lattice was found to be occupied by a surface titanium atom, with 40% occupancy per (1x3) cell. Partial occupancies of 60% were also found for the Ol and Ti5 surface atoms. The resulting stoichiometry for this surface structure is TiOi.68, which is equivalent to a 15.5% oxygen deficiency on the surface relative to the bulk. Hence, the refined model can be described as the formation of strongly distorted {110} micro facets on the surface with oxygen defects and a partial occupancy of an interstitial site. Relaxations are found down to 9 A below the topmost layer. The different coordinations found for Ti might explain part of the photo-catalytic properties of this surface. 3.5.

NiO(lll)-p(2x2) reconstructions The bulk structure of rock-salt oxides (NiO, MgO, CoO, MnO) has alternating cationic and anionic sheets along the [111] direction, the simple truncated surfaces have divergent electrostatic energy, in theory making them highly unstable [3]. Thus, polar rock-salt surfaces were long believed to be unstable, according to Tasker [84] and to an early experimental evidence of (100) facetting on MgO(lll) [85]. Nevertheless these materials are increasingly used in technological applications because of their unique properties in catalysis [86], as highly correlated materials [87] and magnetism [88-92]. The recent prediction that a particular p(2x2) "octopolar" surface reconstruction (Fig. 10), which cancels the divergence of the electric field in the crystal, may stabilise the surface [93] and the existence of (111) facets on small

Fig. 10: Schematic drawing of the Ni-terminated octopolar reconstruction of NiO(l 11). Small circles stand for Ni atoms and the large ones for O atoms. The first and second planes are respectively 75% and 25% vacant. The arrows indicate the basis vectors of the p(2x2) surface lattice mesh. The symmetry related radial relaxations 5js and ^^^ are respectively represented around apex atoms labeled 1 and 2 by dashed and dotted arrows and applies to the second oxygen and the third nickel layers that are patterned. (Right) Side view of the octopolar reconstruction and of the two first relaxations ^j and ^2 perpendicular to the surface plane.

277

NiO single crystals [94] motivated further investigations of this surface. Moreover p(2x2) surface reconstructions were observed for NiO(lll) thin films grown on gold and nickel substrates [95, 96]. Early experiments on NiO(lll) also showed complex reconstructions attributed to Si segregation [97]. It was thus interesting to establish the optimal preparation conditions, if they exist, of NiO(l 11), to investigate its stability and reconstruction ability. GIXS is a method of choice with this respect but the situation rapidly appeared to be rather complex and the observation of a p(2x2) reconstruction was only a step toward the understanding of the surface chemistry of NiO(l 11), probably driven by the electrostatic properties of this polar surface. Fortunately, the crystalline quality of NiO(lll) single crystals can be improved by adequate annealing in air. The main steps of the optimal crystal preparation consist of an annealing of the NiO boule at 1850 K for 24 hours in air, then cutting, polishing, and re-annealing at 1600 K for 3 hours polished surface against polished surface, which yields flat and shiny surfaces. Mosaic spread better than 0.02° and typical domain sizes larger than 2000A can be routinely obtained. Such crystals have high crystalline quality up to the surface and are well suited for detailed GIXS investigations [98, 99]. 3.5.1. NiO(lll)'p(2x2) structure after air annealing [98, 100] In the vacuum chamber, air annealed surfaces were immediately p(2x2) reconstructed and GIXS data were recorded after such an ex situ sample preparation because further treatments proved ineffective. Annealing in UHV leads to decomposition and formation of Ni clusters [100]. Annealing under up to 10"^ mbar O2 at 700K removes the residual C contamination but drastically transforms the internal structure of the reconstruction. For single crystals prepared accordingly to our optimised preparation method, the p(2x2) reconstruction obtained after air annealing can be measured quantitatively. However the intensity remains weak. The measurements were carried out on the ID03 beamline at ESRF at 18 keV photon energy and an incidence angle of 0.17°, which is the critical angle for total external reflection at this energy. We use the p(2x2) lattice to index the reciprocal space: the crystallographic basis vectors for the surface unit cell are related to the bulk basis by asurf=[rioLi,. bsurf=[oTlL^, and Csurf= [l 1 iLz,^ • The h and k indexes are chosen to describe the in-plane momentum transfer, Q//, and £ the perpendicular one, Qi. All three indexes are expressed in the reciprocal lattice units (r.l.u.) of NiO(lll). The in plane scattering of the p(2x2) patterns was measured quantitatively by rocking scans at all accessible positions belonging to the

278

reconstruction. A total of 33 non-zero peaks were measured at ^=0.1. The symmetry of the diffraction pattern is P6mm leaving 14 non-equivalent peaks C

(c)

)€)€ '^rj. D

rC

i

# e

t) €

C

#

€

Fig. 11: (a) Experimental Patterson map derived from the in-plane scattering from an air annealed N i O ( l l l ) single crystal; (b) Calculated Patterson map for a relaxed octopolar reconstruction; (c) Comparison between the experimental (right half circles) and calculated (left half circles) structure factors.

with a systematic uncertainty of 12%. The out-of-plane structure was derived from the quantitative measurement of the (20^), (02^) and (22^) crystal truncation rods (CTRs). Reproducing the in and out of plane scattering restricted the solution to the Ni-terminated octopolar model uniquely, plus the following symmetry-related atomic relaxations in each atomic layer p (p=0 for the apex layer of atoms). Note that 3 of the 4 atoms in layer p of the unit cell have symmetry-related vertical displacements C,ps, whereas the independent atom has vertical displacement C,p (see Fig. 10). Likewise, Sps defines a radial displacement of symmetry-related atoms away from the in-plane position of the apex atom (see Fig. 10). The Pps terms define the possible rotational displacements. Detailed fitting reveals that all 5^5 and Pps terms are negligible except 6i=0.117±0.015A. The least-squares refinement converges for a 0.2A r.m.s. roughness, Co=0.06±0.02A, Cis=-0.17±0.02A, C2=0.13±0.0lA, C2s=-0.029±0.007A, C3=0.23±0.12A, C3s=-0.09±0.02A, C4=0.02±0.0lA and C4s=-0.005±0.003A. For the 138 measured structure factors a global %^ of 1.5 is obtained with 9 structural parameters, the roughness and a scale factor. The agreement is good (Fig. 11) and the relaxations extend deeply into the crystal. The smallest bond length in the refined model is 1.9A, that is a 10% contraction, and is located between the last complete layer and the 25% vacant layer of the octopole. To summarise, the theoretically predicted octopolar reconstruction effectively applies to air annealed NiO(lll) in order to overcome the divergence of the electrical field. Interestingly, the observed relaxations remain very limited with respect to the ideal structure.

279

Fig. 12: (a) Experimental Patterson map derived from the in-plane scattering from a NiO(lll) single crystal annealed at 950 K under 10"^ mbar partial oxygen pressure; (b) Calculated Patterson map for a spinel configuration including 7% of octopolar reconstruction; (c) Experimental and calculated (left half circles) structure factors.

3.5.2. NiO(lll)-p(2X2) structure after controlled reduction [98] Although they can be transformed in NiO after oxidation the Ni cluster formation during annealing is unfavourable because they increase the surface roughness [100]. Fortunately they can be avoided by exposing the surface simultaneously to a partial pressure of oxygen of typically 1-10' mbar. The in-plane scattering has been measured at several temperatures up to 950K (Fig. 12). The structure factors continuously evolve indicating that the internal structure of the reconstruction changes during anneals. The final structure, which is expected to be close to the final transformation state, fortunately yields much larger intensities than the octopolar reconstruction and several reconstruction rods and CTRs could be measured in this case. The out-of-plane CTRs still resemble the octopolar reconstruction showing that the number of planes involved remains close to that of the octopolar reconstruction. Thus this structure has been used as a starting point to develop a model able to reproduce the observed structure factors. The starting model needs to be Oterminated (Fig. 13 (1)) and includes a centrifuge rotation, pi, that consists in fact of a displacement along the three equivalent <100> directions (arrows in Fig. 13 (2)), a positioning of the apex atom on top of one particular Ni atom as shown by the full arrow in Fig. 13 (3); the resulting position corresponds to Fig. 13 (4). The fit of the reconstruction rods, which are highly sensitive to the outof-plane stacking in the reconstructed unit cell, fortunately converge then only for a very realistic stacking (Fig. 13). The three symmetry non-equivalent Ni atoms of the second layer adopt a geometrical configuration in which the interatomic distances are close to the bulk values, one Ni atom don't move (^i), another goes upwards (^3) and a third goes downwards (^2)- At the same time, the O atom (^o) moves away from the surface at the right position to obtain the

280

tiJf£). tki^^^V-*t ^3

Fig. 13: (Left) Structural transformation steps needed to obtain the spinel configuration starting from the O-terminated octopolar reconstruction. Large circles (resp. small) stand for O (resp. Ni) atoms. (1) O-terminated Octopolar reconstruction; (2) Rotation and centrifuge motion of the Ni atoms; (3-4) translation of an O atom on top of a Ni atom; (5) global [010]/3 shift of the reconstructed layer with respect to the bulk. The final situation corresponds to a spinel configuration. (Right) Out of plane positions of the atoms.

expected Ni-0 bond length. Reproducing all the CTRs becomes only possible after a global shift of the reconstruction with respect to the bulk, p2, along the <010> direction and equal to 2/3 of the reconstruction lattice parameter (Fig. 13 (5)). In the final configuration the up-gone Ni atom sits directly on top of an O atom whereas the down-gone Ni atom rests at a 3-fold site that offers the necessary space. The final registry of the reconstruction is very realistic although the structure is now very different from the initial octopole and is amazingly close to a spinel structure. The best fit yields a global y^ of 1.85 over 119 structure factors with pi= 0.84±0.05 A, p2= 1.95±0.09A, Ci= +0±0.02 A, ^3= +0.24±0.02 A, ^2= -0.29±0.03 A and Co= +0.76±0.05 A. Combining this O-terminated structural model with the Ni-terminated octopolar reconstruction obtained after the room temperature annealing allows reproducing easily the scattering observed at the intermediate temperature. The best fits for 300K, 503K, 723K and 950K were obtained with 100%, 82%, 62% and 7% of octopolar reconstruction and in-plane %^'s of 1.3, 0.7, 0.6 and 1.3, respectively. When this combination is used to fit the 950 K data the global y^ reduces from 1.85 to 1.7 over the 119 structure factors. The two proposed structures are thus sufficient to completely reproduce the data at all temperatures. The annealed crystal adopts the configuration of a Ni304(lll) surface over the four last planes although the Ni304 spinel bulk phase does not exist. Thus this layer cannot extend toward the bulk and is likely to be on the origin of the

281

protective effect of the spinel-like surface layer. In this configuration the crystal is reduced (1:3 for 0:Ni) compared to the octopolar reconstruction (3:1 for 0:Ni) and the divergence of the electrostatic potential is avoided too. The Patterson maps observed for the Ni304 spinel-like surface structure is identical to that of a 00364(111) surface spinel layer (see § 3.6), strongly supporting this structural model. The NiO-^Ni304 surface transformation is fully reversible by air exposure, switching the crystal back to the octopolar reconstructed state. Moreover, once the Ni304 spinel-like surface layer is formed, no decomposition with apparition of Ni clusters could be observed showing that the surface chemistry has strongly changed and that the surface has enhanced stability. The transformation in vacuum is second order like with a transition temperature of 956 K. Dosing the surfaces with up to lO^L with lO'^mbar partial H2O pressure in the chamber tested the eventual role of OH' groups. No effect could be observed. More oxidising gazes like NO and NO2 used in similar conditions were found unable to transform the spinel configuration back to the octopolar one indicating the necessity for high pressures. The stability of the reconstruction is attributed to the high quality of the NiO(lll) surfaces. They contain almost no defects, which are believed to be mandatory to initiate reaction with these gazes. 3.5.3. NiO(lll)-p(2x2) comparison and conclusion On the one hand, the divergence of the electrostatic potential leads to various p(2x2) reconstructions in order to heal the polarity of the surface. From our studies on NiO(lll) it clearly appears that the reconstruction scheme proposed to stabilise polar surfaces is correct. On the other hand NiO(lll) behaves like most oxides: it is almost insensitive to contamination or air exposure, it may decompose through UHV annealing and it may exhibit a surface layer that prevents decomposition. The structural models proposed for the p(2x2) reconstructions of NiO(lll) are fully coherent with these properties and are based on two states: the Ni-terminated octopolar reconstruction and the more stable reduced O-terminated Ni304(l 1 l)-like spinel configuration. 3.6.

CoO(lll) [101] and MnO(lll) CoO, MnO and NiO could a priori be expected to have very similar behaviours. Indeed, they share the same structure and nearly the same lattice parameter (acube(NiO)=4.177A, acube(CoO)=4.26A and acube(MnO)=4.44A). In addition, the atoms in the (111) pure Co, Ni or Mn planes are spin uncompensated i.e. the individual cation (111) ferromagnetic planes have a mesh with a net magnetic moment [102, 103]. The AF ordering is along the [111] direction, exchange-moderated by the pure oxygen planes [104]. The Neel

282

CoQB

CoQR

Itf

CoQR

^^;^

CoQR

2 3 4 5 /i[r.l.u. ofCoO(lll)]

CoO^

[00^] 2 4 L[r.l.u. MnO(111)]

Fig. 14: (a) In-plane scan along the [hOO] direction for a polished CoO(lll) single crystal. The scattering positions of CoO and C03O4 lattices are indicated. The most intense Bragg peak of CoO (/i=3) was avoided. B and R indexes stand for bulk and rod contributions respectively, (b) Reflectivity of a MnO(lll) single crystal. The positions corresponding to MnO and Mn304 lattices are indicated.

temperature increases from MnO to NiO (TN(MnO)=118 K, T N ( C O O ) = 292 K and TN(NiO)=523 K). Mixed Ni and Co oxide or multilayers combine almost linearly the advantages of each of the simple AF oxide: high anisotropy and high working temperature [105-110]. Since p(2x2) reconstructions were found to stabilise the (111) surface plane of polar NiO(l 11) it was interesting to know how far this result may be generalised to similar polar oxide surfaces or if other stabilisation processes exist. We have thus investigated C o O ( l l l ) and MnO(l 11) surfaces by GIXS. C o O ( l l l ) and M n O ( l l l ) polished single crystals of purity 99.9% were provided by Crystal GmBH (Berlin, Germany). The samples were cut and aligned to better than 0.1° prior polishing. After polishing, samples with a surface mosaic spread of about 0.8° for CoO and 1.6° for MnO were obtained. A triangular unit cell was chosen to describe the 3-fold symmetry of the (111) surface plane. The unit vectors are related to the cubic ones by: as=[il0]/2, bs=[110]/2, Cs=[lll]. The h and k indexes describe the in-plane momentum transfer and the £ index the perpendicular one, expressed in reciprocal lattice units (r.l.u.). The [hOO] and [OkO] in-plane directions are defined in such a way that a Bragg peak is found at the (1,0,1) position in the reciprocal space. Contrarily to NiO, due to the existence of many possible oxidation states, a direct annealing of C o O ( l l l ) and M n O ( l l l ) single crystals does not improve them but modifies the stoichiometry and destroys the surface polish. The samples were thus investigated after polishing. For both CoO and MnO additional peaks were systematically found in the reciprocal space (Fig. 14) at

283

positions corresponding to a new crystallographic structure. The additional structure is cubic, presents the extinction from spinel phases (Fd3m group) and has lattice parameters slightly smaller than twice the one of CoO and MnO respectively. The small width in h and k shows that the in-plane lattice parameters are very well defined. The surface mosaic spread of the spinel structures is of the order of the single crystals one's, 0.8° and 1.6° respectively. The width of out-of-plane Bragg peaks yields a crude estimate of the thickness of this new layer, --50 A in both cases. The additional phase for CoO (resp. MnO) has been identified as the spinel €0304(111) (resp. Mn304(lll)) oxide with a fully relaxed lattice parameter of 8.08 A (resp. 8.42 A). For CoO, integrated intensities of all in-plane peaks were quantitatively measured at ^=0.1. The data were of F6mm symmetry. A set of 34 nonequivalent reflections has been deduced from a total of 150 measured peaks. The experimental diffraction pattern could indeed be well simulated on the basis of the spinel €0304(111) unit cell with the epitaxial relationship: Co304(lll)//CoO(lll) and Co3O4[100]//CoO[100], providing the use of a surface terminated by the 3 metallic layers (like Fig. 13-right without the top O atom). The agreement is good (x^= 1.17 and R= 0.176) and the Patterson map is very close (if restricted to the same peaks) to the one observed for the NiO(l 11) single crystal after annealing under partial oxygen pressure. The behaviour of the intensities of the Bragg peaks and diffraction rods of the rock-salt and spinel structures allows to conclude that the CoO(lll) (resp. MnO(l 11)) polished surface is stabilised by the presence of a spinel €0304(111) (resp. Mn304(lll)) surface layer. Only the CoO(lll) surface has been investigated in more details. Since air annealing turned out to be unable to restore good surfaces we attempted to remove the C03O4 layer and to improve the crystallographic quality of the CoO surface by annealing in UHV or under O2 and Ar"^ bombardment. Annealing in UHV leads to the formation of small metallic Co clusters on the surface, which is similar to the NiO case. Unfortunately, while metallic Co is formed, the quantity of the spinel remains constant. A post annealing at 800 K and 10'^ Tore O2 totally oxidised the Co clusters in a few minutes but increases the intensity of the C03O4 peaks. Moreover increasing the oxidation time leads also to the development of the spinel. Ar"" bombardments (U=800 V, p(Ar'')=2xlO'^ Torr, drain current -lOjLiA) at different substrate temperatures proved also ineffective. Again metallic, mainly FCC (same stacking as CoO(lll), Co clusters, in the form of flat islands (presence of CTRs) form with a poor crystallographic quality (mosaic spread of 2.6°). This result is similar to the one obtained for the growth of Co on the polar NiO(lll) surface [111], evidencing the strong influence of polar oxide surface on metallic films. In oxygen, the Co clusters rapidly oxidises in C03O4, definitively showing that €0304(111) has a much

284

lower surface energy than CoO(l 11). This behaviour probably explains why Co crystals become almost impossible to clean once they have been oxidised [112]. Finally, unlike NiO(lll), which is stabilised by a p(2x2) reconstruction, the CoO(lll) and MnO(lll) surfaces are stabilised by spinel structures, in which the problem of the diverging electrostatic potential is also avoided. On the one hand, this behaviour may be the major reason why Nii.xCOxO compounds exhibit spinel structures when x>0.6 [113]. On the other hand C03O4 and Mn304 are bulk compounds whereas Ni304 does not exist in the bulk regime, which might explain our observations on the NiO(lll). In particular the octopolar reconstruction is probably of much higher surface energy than the spinel (111) structure and might only be observed for crystals adopting only the rock-salt structure. Thus the NiO(lll) surface has to make a compromise, depending on the experimental conditions, between surface reconstruction and spinel structure although this one cannot propagate toward the bulk whereas CoO and MnO simply adopt in the surface region their more stable (at the surface) oxidation state. 3.7.

ZnO(OOOl) and ZnO (lOlo) Zincite, ZnO is an important oxide in catalysis, for instance for the synthesis of different alcools; its is one of the most interesting substrates for the growth of Gallium Aluminium Nitride large gap semiconductors, with applications as blue diodes and lasers. It is itself a large gap (3.3 eV) material, intermediate between fully ionic materials and semiconductors. Its wurtzite structure does not have a centre of inversion, and hence ZnO has two polar surfaces perpendicular to the c-axis, The Zn-terminated ZnO(OOOl) and the Oterminated ZnO(000i) surfaces, which are unstable according to electrostatic arguments alone. It also has a stable, cleavage face; the ZnO (1010) surface. All three surfaces have recently been studied by GIXS at LURE [114, 115], and the ZnO(000i) -O surface has also been investigated by another group [116] using the ID03 station of the ESRF. The unit cell of the unreconstructed ZnO (1010) surface is fairly simple, mainly composed of a ZnO dimer. Previous ab initio calculations [117] and LEED [118] analyses concluded to a buckled dimer with significant inward relaxation of more than 0.25 A, with the oxygen pointing outward. LEED concluded to an expanded dimer distance, while theoretical studies concluded to a shortened one. The surface was prepared by cycles of sputtering followed by annealing up to 930°C. Ten CTRs, of which the six more reliable were analysed, were measured at an energy of 15 keV. Although the flatness of the surface as deduced from the fits was small, the average terrace size (50 to 200 A) deduced from the CTR widths at anti-Bragg conditions was very small. This

285

induced an unusually large width of the rods, with the impossibility to fully integrate them, even with large detector slits. Moreover, close to Bragg peaks, a contribution from the bulk was present in addition to the surface contribution. These experimental limitations yield particularly large error bars, and rather poor fits close to Bragg peaks. Fractional occupancies in the first two layers were found mandatory to get a satisfactory fit to the data. They correspond to a large density of vacancies resulting from the ion bombardment process. A satisfactory fit (%^=3.1) was obtained for a bulk-terminated surface with fractional occupancies, further improved (x^=2.6) by allowing small displacements both parallel and perpendicular to the surface for the Zn (AXzn=0.05±0.02 A, AZzn=-0.06±0.02 A) and the O (AXo=-0.04±0.06 A, AZo=0.12±0.16 A) ions, and occupancy factors of 0.77±0.02 and 0.90±0.04 in the first and second layers respectively. The dimer bound length of 1.9 A is 4.5 % shorter than in the bulk, which is consistent with theoretical prediction, at variance with the very small relaxation found here, much smaller than all those predicted by ab initio studies. The very large uncertainty on the displacements of the O ions is due to their much smaller scattering factor as compared to Zn ions. This study highlights the difficulties to perform reliable GIXS measurements on surfaces of poor crystalline quality and with a small average terrace length. It would be highly desirable to find procedures to prepare the surface with a higher degree of order. In the LURE study on the polar faces [114], both ZnO(0001)-Zn and ZnO(000i) -O surfaces were investigated. They were prepared by repeated cycles of ion bombardment and annealing up to 1100 K, yielding again a high density of steps, with small terraces widths between 7 and 18 nm. The steps were found of bilayer height, which implies the presence of two non-equivalent terminations, resulting in a strong damping of the effect of relaxation or nonstoichiometry. The ZnO(0001)-Zn surface was found nearly bulk terminated, with only a small outward relaxation of 0.05 A, and a partial occupancy of 0.75 in the top Zn layer. The fit of the 3 non-equivalent CTRs, (1,0), (1,1) and (2,0), with a systematic error bar of 10%, yielded a x^ value of 3.3. Relaxations in deeper layers did not improve the fit. For the ZnO (OOOlj-O surface, the best fit, corresponding to x^=3.4, was obtained for a large inward relaxation (-0.33 A) of the topmost O plane and an outward relaxation (+0.08 A) of the underlying Zn plane, plus fractional occupancies of 1.3 and 0.7, respectively for the first bilayer O and Zn sites. The occupancy factor larger than 1 for the O reflects the presence of heavier species, either in substitutional positions, or adsorbed on top. In the ID03 study [116], only the O termination was investigated. The surface was again prepared by repeated cycles of sputtering and annealing, but up to a slightly higher temperature of 1200K. X-rays of 17 keV were used.

286

under total external reflection conditions. Five CTRs were measured over an extended range of perpendicular momentum transfer values, and analysed with 16 parameters, 4 vertical displacements in the top two bilayers, 8 Debye-Waller parameters, plus a scale factor and the fraction of the surface adopting the fitted model. The site occupancies were fixed at 1. The best fit (5^^=1.9) was obtained for negligible displacement of the subsurface layer, and, in accordance with the LURE study, for a large inward relaxation of -0.22 ± 0.05 A for the top O ion. The experimental results are in very good agreement with an ab initio calculation described in the same paper, which takes into account for the first time the macroscopic electric field. Both studies thus conclude that these polar surfaces are stabilised by a significant charge transfer leading to metallic surface layers, which may have important implications in gas sensing applications. 3.8.SrTiO3(001) Like the MgO(OOl) surface, SrTiO3(001) is a very commonly used substrate for the growth of high-Tc superconductors, YBa2Cu307.x in particular. SrTiOs is also a model photo-catalyst for the dissociation of water. The surface exhibits oxygen and strontium defects, which are believed to act as centres for catalytic reactions. The (001) surface can terminate with two types of plane, SrO, or Ti02, and the stoichiometry of each plane can be varied according to the preparation conditions. Depending on the degree of surface reduction, several reconstructions have been observed: (2x1), (2x2), c(2x2), (V5xV5)R26.6°, and c(6x2). Substantial relaxation has been theoretically predicted [119] and deduced fromLEED measurements [120]. Reconstruction of the SrO terminated surface to a ferroelectric monolayer below 100 K [121] is also predicted and observed by LEED, which has driven many recent experimental [122] and theoretical [123, 124] investigations. For these reasons, several groups have investigated the SrTiO3(001) by the GIXS technique. A recent paper [125] describes a GIXS study that allowed determination of the surface termination. Chemically treated surfaces were found to be terminated by a TiOi plane, and annealed surfaces were mostly terminated with a Ti02 plane, with a small portion covered by a SrO plane. More recent measurements were performed on the SrTiO3(001)-(lxl) surface [126] using the ID3 diffractometer of the ESRF. The surface was prepared in situ by repeated cycles of Ar"^ sputtering and annealing at 900 K. This was followed by an anneal at 900 K in 1-10"^ mbar O2 to restore the surface stoichiometry and prevent the formation of reconstructions induced by oxygen vacancies. Three CTR's were quantitatively measured and fitted. Ten atomic displacements, 6 Debye-Waller parameters, the scale factor, the roughness, and the surface fraction were fitted. The best fit was achieved with a

287

SrO

Fig. 15 : Diagram showing the atomic displacements for both surface terminations of SrTiO3(001)-(lxl) at room temperature. The black circles represent Ti, the open circles O and the shaded circles Sr. (Reprinted from [126]).

surface that is 22% SrO and 78% Ti02 terminated, in agreement with a coaxial impact-collision ion scattering spectroscopy study of a similarly prepared surface [127]. Figure 15 schematically shows the vertical atomic displacements deduced from the best fit, for the two kinds of terminations. An important result is that, for both terminations, the surface atoms do not undergo any lateral displacements, thus ruling out the possibility of a ferroelectric distortion, which is consistent with the most recent theoretical calculation [123]. Negligible relaxation for Ti is found on the TiO termination, which is consistent with recent IVIEIS results [122], but at variance with the rather large relaxation predicted by theoretical calculations [121, 124]. There is good agreement with theory for the Sr relaxation of the first and third layer on the SrO termination, but poor agreement for the Ti relaxation in the second layer, for which a large inward relaxation is found. Because the scattering power of oxygen ions is much smaller than that of titanium ions, the error bars on the oxygen relaxation are large, on the order of the displacements themselves. The c(6x2) reconstruction, which is obtained by a combination of annealing under oxygen and UHV, has been shown remarkably long-range ordered and stable in O2 or in ambient air [128]. It has been characterised by LEED and STM] and very recently by GIXS [129]. Measurements of in-plane reconstruction rods from this surface have been recorded, but are still under analysis. An original GIXS study was performed on the buried interface of a SrTiOs bicrystal, prepared by fusing two 18.4° miscut crystals (corresponding to a (103) orientation of the normal to the interface) at elevated temperature and pressure [130]. CTR's were quantitatively measured. Modelling showed that the interface region is very narrow, with a thickness of -11 A, and has a significantly changed atomic structure.

288

3.9.

Cr2O3(0001)-(lxl) [131] The Cr2O3(0001) surface has been recently investigated by LEED [132, 133] as well as by GIXS [131], mostly because of its magnetic applications. Cr203 is a ferromagnet with a corundum-type structure (like a-Al203). The LEED investigation and a number of theoretical studies [133, 134] concluded to a surface structure very similar to that reported for sapphire a-Al2O3(0001) (see § 3.3) : the surface was found to be terminated by a single metal ion layer (1/3 of a compact occupancy), with significant inward relaxation of the top layer metal ions. At variance with sapphire, no lateral shifts of the top O^" ions were found. Although this structure yielded a fairly good fit to the CTRs data, the authors propose a different solution with Cr interstitial, yielding an even better fit. The polished single crystal was first annealed in air at 1500 K for 12 h, leading to large flat terraces, and next UHV cleaned by ion bombardment followed by a 5 min. long annealing at 1200 K. Considering only the three usual possible terminations (single O layer, double Cr layer and single Cr layer), the best fit of the measured CTRs (x^=2.53) corresponded clearly to a single Crlayer termination, with a -52% inward relaxation, very close to the values of 60%, -58% and -50% respectively deduced from LEED [132, 133] and from molecular dynamics [133] and Hartree-Fock [134] calculations. Although the measured CTRs were quite satisfactorily reproduced on the basis of this "standard" structure, there was still room for further refinement, as judged from graphical agreement. This was achieved (x^=1.51) by assigning only a 2/3 occupancy (instead of 1) to the top Cr^"^ ions, and thus a random arrangement of these ions, and assuring charge neutrality by adding the lacking 1/3 Cr^^ in the available (distorted) octahedral site (which is empty in the bulk) within the first Cr ^ double layer directly below the top-layer site. The comparison between the experimental (10^) and (30^) CTRs and those calculated for these two models is shown in Fig. 16. The top-layer Cr^^ relaxation was found to be very much smaller (-6%) than that of the "standard" model, which is attributed to a reduced in plane Coulomb repulsion. This model was proposed to yield a simple explanation for a disorder-to-order and an order-to-order phase transitions that appear at low temperature between a (1x1) and a (V3xV3)-R30° structures. Because the calculated CTRs for these two structures are not so different, the LEED data as well as the calculations should be re-examined with this new model. It should also be noted that very large values were found for the atomic displacement factors (Debye-Waller parameters) for the top three atoms: 633=20 A^, 8 A^ and 2.5 A^, respectively for the top Cr^^, the top O^' and the interstitial Cr^"*^ions, which should certainly also be accounted for in new investigations.

289 (3 0L)

(1 0 L)

-10

-8

-6

-4

-2

0

2

4

6

-

6

q^ (rec. tatt. units)

-

4

-

2

0

2

4

6

q^ (rec. latt. units)

Fig. 16. Experimental and calculated CTRs of Cr2O3(0001)-(lxl). Solid lines: interstitial model; dashed lines: single-layer model.

4. STRUCTURE OF THIN OXIDES FILMS Some metal oxide materials have catalytic and magnetic properties, which make them attractive for many applications. We may quote for instance the development of new magnetic recording media and read heads for increased areal information density. Non magnetic oxide thin films may also act as effective tunnel barriers in tunnelling magnetoresistance devices that are promising candidates for future magnetic random access memories, the socalled MRAMs [135]. The synthesis of magnetic and non magnetic oxide thin films is thus a field that has developed very fast in the last years, and many of the films obtained by different growth techniques were characterised by GIXS. It is not our purpose to make a complete list of the works that have been performed, but rather to give relevant examples. 4.1. NiO(lll)-p(2x2) thin film structure NiO(lll) thin films are more adapted to many techniques and were thus investigated well before the single crystals, which are good insulators at room temperature [136]. The very intriguing predictions [93] stating that this polar surface could be stable through the Ni-terminated p(2x2) octopolar reconstruction [93] motivated the growth of NiO(lll) films. Indeed, a p(2x2) LEED pattern was reported for 550 K grown films on Au(lll) [95] and for room temperature grown films on Ni(lll) [96]. Surprisingly, on NiO/Ni(lll) films, the pattern disappeared after water adsorption, presumably due to hydroxyl adsorption and could be restored through annealing. However, this result stands in contradiction with a previous LEED work on the oxidation of Ni(lll) where only c(2x2) and (7x7) reconstructions were reported [137]. From another point of view, only the reconstruction pattern has been reported and not the structure of the reconstruction itself. It was thus interesting to know if these films, grown in out-of-equilibrium conditions, adopt the oxidised or reduced NiO(lll)-p(2x2) surface reconstruction to achieve stabilisation. Since the

290

reported LEED patterns for NiO/Ni(lll) showed poor order and the NiO/Au(lll) excellent contrast, we have chosen to investigate the NiO/Au(lll) interface by GIXS [98]. The GIXS experiments were performed on the ID03 surface diffraction beamline at the ESRF in ultra high vacuum conditions (10"^^ mbar). The thin NiO film was measured at 17 keV and 0.9° (=3xac) incidence angle because the bulk scattering from the Au(lll) substrate was weak in the regions of interest. Similarly to the single crystal studies the crystallographic basis vectors for the surface unit cell describe the triangular p(2x2) lattice of the reconstruction. The NiO(lll) thin film was prepared in situ on Au(lll) according to [95, 96]. The surface used for the growth had a mosaicity of 0.052° and a domain size of 2600A exhibiting the herringbone reconstruction [138]. During deposition the substrate was held at 615 K and the Ni was evaporated from an electronbombarded Ni rod in a 2-10'^ mbar partial pressure of O2. Perfect 2D growth occurred from 3 to 8 monolayers before 3D crystallites formed. The quantitative analysis was performed for a 5 ML thick layer. The NiO(l 11) thin film was in good epitaxy, of surprisingly good crystalline quality, completely relaxed and p(2x2) reconstructed with 0.106° mosaicity, 550A domain size, intense throughout the accessible region of reciprocal space and of perfect Lorentzian shape. The in-plane scattering of the p(2x2) pattern was measured quantitatively by rocking scans at all accessible positions belonging to the reconstruction in reciprocal space. 59 non-zero in-plane peaks of P3ml symmetry were measured at ^=0.3, leaving 32 non-equivalent peaks (error level 11%). The different symmetries between the single crystal and the thin film cannot be explained by the small £ difference but indicates rather a different p(2x2) structure. A direct comparison of the Patterson maps for the thin film and the single crystal is reported in Fig. 17a showing that the structures are unlike. Moreover, several diffraction rods representing a total of 322 non-zero structure factors were measured and their periodicity along £ clearly indicates a 3 atomic layer-thick reconstruction. Effectively, the octopolar reconstruction was unable to reproduce the in-plane data, regardless of the relaxations or the termination (X^>8). A coherent juxtaposition of half Ni- and half O-terminated domains, with same orientation, separated by single-steps yields the experimental Patterson map. The structure was refined, as for the single crystal, using the same relaxations plus a domain fraction, the roughness and a scale factor. The symmetry-related atomic relaxations in each atomic layer receive in addition to the layer number (1 for the apex layer of atoms) an additional S index: C^ps and C,p represent the vertical relaxations.

291

(b)

€ ©

O C

•

X FilnNv / V/ \

Fig. 17: (a) Comparison of the Patterson maps of a 5 ML NiO(lll) film on A u ( l l l ) and of the signal crystal (SC). (b) Comparison between the in-plane experimental and calculated structure factors for a model with Ni- ad 0-terminated octopoles separated by a single step; (right) experimental and (left) calculated.

Likewise, 5^^ defines a radial displacement of symmetry-related atoms away from the in-plane position of the apex atom. To avoid too many parameters we have preferred an approach with identical relaxations in the two domains. Thus ^1 is the perpendicular relaxation of both apex atoms, and so on (see Fig. 10right for example). The best stable solution remained essentially unchanged when we allowed independent relaxations in the two domains. Only three relaxations were non-negligible. The best agreement was obtained for 5is=0.096±0.008A, 52s=0.078±0.007A and Ci=-0.103±0.028A, with 50% Niand 50% O-terminated domains, a 1 A r.m.s. roughness and with a global x^ of 1.4 close to the ideal value of 1. The agreement is excellent up to large momentum transfers. The comparison for the in-plane data is given in Fig. 17b. Importantly, the single step belongs to the model itself and will thus not influence the coherence length. It may be located anywhere and does not necessarily form large continuous step edges. This might explain why STM investigations, although for thinner layers, have mainly reported double steps, since they define really the step edges [95]. The presence of both terminations might be explained by the nonequilibrium growth conditions: the growing surface necessarily passes through non-equilibrium structures in order to fill the incomplete lower layers. A perfect Ni-terminated octopolar domain can be transformed to O-termination via incorporation of a half-monolayer of nickel and a half-monolayer of oxygen atoms. An alternative model, able to explain the observed GIXS data based on a direct method approach has been proposed using initially the same data set [139]. The resulting surface model is 3 layer thick, Ni-terminated and mainly build with O and Ni-terminated octopolar blocks stacked in each other in such a way that the O plane is complete (3 basis atoms from the top octopole and the

292

apex atom from the Ni octopole below). The model is however nonstoichiometric, presents atomic vacancies in the third layer and needs the presence of electron holes in the complete O layer. The resulting model is thus single phased but still based on 50%-50% O and Ni-terminated octopoles. Although this last model may explain better the previous STM results (obtained on thinner layers only) it remains difficult to understand how it is compatible with a 2D growth process. The agreement with the data are however similar. The stability of this surface against hydroxylation was tested by extensively dosing the surface with up to 10^ Langmuir of H2O at 3-10'^ mbar partial pressure at room temperature. Similarly to the single crystal case, no structural effect could be detected. This stands in contradiction with the NiO(l 11) films of poor structural quality made by oxidation of Ni(lll) [132, 140], which were found unstable against hydroxylation. We thus suggest that the reactivity of NiO(lll) against water takes mainly place at defects (like for MgO), which are almost absent on our surfaces and thin films. Obviously, NiO(lll) is terminated and stabilised by p(2x2) reconstructions with different internal configurations. Importantly the electrostatic criterion is always fulfilled showing that it is very important when polar surfaces are considered. Interestingly, the octopolar reconstruction prediction and the observation of the p(2x2) surface cell were far not enough to understand the polar NiO(lll) surface. Note also that oxide thin films may not exhibit the same surface structure as their bulk counterpart. The studies carried out on thin oxide films may thus only be carefully extrapolated to the bulk surface. 4.2. AI2O3 layer on NiAI The oxidation of NiAl(lOO) and NiAl(l 10) surfaces yields well ordered thin oxide layers which have been thoroughly investigated as supports for model catalysts, by a very large variety of surface structural tools [141]. Oxidation of NiAl(lOO) has been shown to yield a (2x1) superstructure consisting in a few atomic plane thick 0-AI2O3 layer with two different types of domains, while oxidation of the (110) face yields an incommensurate 7-AI2O3 like oxide layer, consisting of two in-plane rotational domains. Both thin films have recently been investigated by GIXS by two different groups, using the ID32 ESRF setup for the (100) face [142], and the ID03 one for the (110) face [143]. Measurements of the extended reflectivity, (00^) rod, CTRs and (2x1) reconstruction rods were performed on the NiAl(lOO) face. Precise measurements together with a detailed analysis yielded a structure made of the superposition of a 0.76 nm-thick 0-AI2O3 like two-dimensional layer, and 6AI2O3 islands of 2 nm average thickness. In these islands, which are attributed to the Stranski-Krastanov growth mode, the Al ion positions at the surface and near the interface are strongly modified, both parallel and perpendicular to the

293

surface, with some of them even switching from original octahedral sites to tetrahedral co-ordinated positions and vice versa. The oxide layer is commensurate, as suggested by the strong modulations of the CTRs induced by the oxidation. Quantitative analysis showed that only a small part of the surface is covered with ordered alumina in a (2x1) superstructure. The major part of the surface is either not oxidised, or covered by a commensurate oxide layer. The results on the NiAl(llO) face are not yet published, and hence only preliminary results are available [143]. A 6 A-thick oxide was formed, as the result of a 30 min.-long annealing at 950°C under 2-10'^ Pa of O2. This thickness was determined by X-ray reflectivity measurements. Two different incommensurate structures were observed, one of which had been previously observed by SPA-LEED. The other, which had much stronger reflections, had never been reported previously, and was indeed invisible on the LEED patterns. This new structure is still under analysis. 4.3. FciOa and Fe304 layers A combined GIXS, AFM and TEM study of Fe203 and Cr203 thin films (-1000 A thickness) deposited on silicon wafers is described in [144]. X-ray reflectivity was used to estimate the changes in density of the films, and GIXS to determine the epitaxial relationships. GIXS was also used to follow the ion beam mixing effect at the Fe203/a-Al203 interface, the -500 A thick Fe203 layers being deposited by pulsed laser ablation [145]. Sub-stoichiometric phases were evidenced. 4.4.

CY2O2, layers GIXS was combined with XPS to characterise thin oxide films formed by heating at 873 K in air FeCr alloys with different concentrations [146]. The main crystallographic structure of the oxide film was found of corundum type structures (Fe203 and Cr203) although it depends on the bulk chromium concentration. The oxidation of Cr(llO) surfaces is of both fundamental and applied interest for effective surface corrosion protection [147]. In addition, thin oxide films of Cr(llO) have recently proven effective as model systems for the study of heterogeneous catalytic reactions at solid state surfaces [148]. The oxidation of Cr(llO) surfaces has thus been studied by many techniques. However, open questions were left regarding the exact oxide film thickness, the oxide growth laws and the relation of the oxide microstructure to the oxidation behaviour. The oxidation process was thus characterised by GIXS [149-151] as a function of temperature, oxygen exposure and time. The single crystal Cr(llO) films were prepared by MBE on a Nb(110)/Al2O3(ll2 0) buffer system, with thickness ranging from 100 to 1000 A. Oxidation was performed with typical

294

oxygen partial pressure of 5-10'^ mbar, in the 400-800K temperature range. In situ X-ray reflectivity was performed to deduce the thickness of the growing oxide layer, its defect density and the roughness of the oxide surface and the metal/oxide interface for different times and temperatures. Below 725 K, the oxidation was found to proceed by a layer-by-layer growth without roughening of the metal/oxide interface. Fairly thin (a few 10 A) oxide layers were obtained, with a smaller density than bulk Cr203. The oxide film thickness in quasi-equilibrium was found to increase linearly with the temperature, and the oxide surface and metal/oxide interface remained smooth on the order of 2 to 4 A. After a fast decrease of the oxide thickness in the early growth stage, saturation of the thickness was observed. It was found to depend markedly on the oxidation temperature, indicative of an oxidation process that is controlled by the Cr-ion diffusion. Above 725 K, the oxide surface was found to roughen, and cavities were found to form at the metal/oxide interface. In-plane and out-of plane GIXS measurements allowed the identification of the oxide structure as orthorhombic a-CraOs with the (0001) axis parallel to the Cr(llO) axis, Cr2O3[1120]//Cr[n0] and Cr2O3[li00]//Cr[001], and with six-fold symmetry. Good epitaxy occurs in the Cr[001] direction because the misfit is less than 0.5%. Annealing in UHV at higher temperature (850 K) was found to significantly improve the oxide crystalline quality, with a decrease of the mosaic distribution from 5.1° to 3.7°. The same group also investigated the thermal oxidation of a 13 A thick (llO)-oriented Fe film grown epitaxially on Cr(llO) [151], by X-ray reflectivity, GIXS and AES measurements. The final oxide layer was larger than expected when only the Fe layer had been oxidised. Indeed, an additional Cr2O3(0001) layer was found to form on top of the Feoxide layer, providing evidence that the Cr-ion is the diffusing species during oxidation. A 22 A thick Cr layer grown on top of the Fe(l 10) one was found to transform into a 46 A thick Cr2O3(0001) film protecting the Fe layer against oxidation, thus showing that Fe-ion diffusion through the Cr-oxide layer is more difficult than through the Fe-oxide layer because of the smaller defect density in the former. 5. CONCLUSIONS AND FUTURE OUTLOOK The aim of this chapter was to give an overview of the structural information that has been gained at the atomic level by use of the GIXS technique on many oxide surfaces. This technique is powerful when not hampered by experimental limitations. One major limitation is the need for single crystals and surfaces of very high crystalline quality. GIXS also often requires synchrotron radiation, which means that a limited amount of time is devoted to a particular study, so that all pertinent parameters (such as for

295

instance a substrate temperature or defect density) can not be systematically varied. Only a very small number of oxide surfaces have been investigated so far by GIXS. However, the number of studies is expected to increase steadily, since more and more teams are involved in this field, and exciting studies still have to be attempted in different directions. Moreover, the number of fields in which oxides are involved is growing; they are no longer only used as substrates. Catalytic, magnetic or corrosion protection properties may be thoroughly used in future devices. With this respect, the precise atomic structure of oxide surfaces and interfaces will become increasingly important. For instance, the investigation of the reactivity of the surface oxide ions, and the adsorption of different gaseous species on oxide surfaces is expected to develop. Investigations of stepped oxide surfaces and the effect of the miscut of vicinal substrates on the structure of the overlayer still have to be done, as well as the in situ investigation of reactive metal/oxide interfaces or of the metal oxidation process. More and more research is devoted to the growth of oxide thin films of high quality. GIXS has been demonstrated on metal surfaces to be a very powerful probe of the growth mode during homoepitaxy, and will almost certainly be used to investigate the homoepitaxial or heteroepitaxial growth of oxides on different substrates, metal, semiconductors or ceramics. Importantly the ultrathin limit imposed by the charge build-up may then be overcome easily. The advent of highly intense X-ray sources also makes possible microGIXS studies with lateral resolution by scanning the incident X-ray beam or the in situ investigation of fast kinetic processes involved at interfaces. Finally, note that GIXS is one of the very seldom surface probes that do not necessarily require a UHV environment. Investigation of the growth by different liquid or vapour phase epitaxy techniques is foreseen in the near future, together with investigation of surfaces under gaseous pressure between UHV and high pressure and of oxide/liquid interfaces, reactive or not. ACKNOWLEDGMENTS We first would like to thank all the colleagues who have been strongly involved in the work from our team described here : B. Villette, P. Guenard, O. Robach, C. Mocuta, M. Noblet and O. Ulrich, as well as our external collaborators, A. Stierle, K. F. Peters, H. Kuhlenbeck, B. Richter, J. Jupille and M. Gautier-Soyer. We are greatly indebted to Professor D.P. Woodruff, for having invited us to write this chapter.

296

REFERENCES [1] V.E. Henrich and P.A. Cox, The surface science of metal oxides (Cambridge University Press, Cambridge, UK, 1994). 2] V.E.Henrich, The surfaces of metal oxides. Rep. Prog. Phys. 48 (1985) 1481. 3] C. Noguera, Physics and chemistry at oxide surfaces (Cambridge University Press, Cambridge, UK, 1996). 4] J. Nowotny and Louis-Claude Dufour, Surface and near-surface chemistry of oxide materials (Elsevier Science Publishers B.V., 1988). 5] E.A. Colboum, Surf. Sci. Rep. 15 (1992) 281. 6] Metal-Ceramic Interfaces, Acta-Scripta Metall. Proc. 4 (edited by M. Riihle, A. G. Evans, M. F. Ashby, J. P. Hirth, Pergamon Press, 1990). 7] Proceedings of the International Symposium on Metal-Ceramic Interfaces, Acta. Metall. Mater. S40 (1992). 8] J. Nowotny, Science of Ceramic Interfaces (Elsevier Science Publisher B.V., Amsterdam, The Netheriands, 1991). 9] H.Bialas, K.Heneka, Vacuum 45 (1994) 79. 10] F.Reniers, M.P.Delplancke, A.Asskali, V.Rooryck, O.Van Sinay, Applied Surf. Sci. 92 (1996) 35. 11] G. Pacchioni, in Chemisorption and Reactivity on Supported Culsters and Thin Films, eds. R.U. Lambert and G. Pacchioni (Kluwer Academic publishers. The Netherlands, 1997), 395. 12] M.W. Finnis, Acta Metall. Mater. 40 (1992) S25. 13] A. M. Stoneham, P. W. Tasker, J. Phys. C18 (1985) L543. 14] J. R. Smith, T. Hong, D. J. Srolovitz, Phys. Rev. Lett. 72 (1994) 4021. 15] U. Schonberger, O. K. Andersen, M. Methfessel, Acta Metall. Mater. 40 (1992) SI. 16] P. Blochl et al., in Metal-Ceramic interfaces, (edited by M. Rhtile, A.G. Evans, M.F. Ashby and J.P. Hirth, Pergamon Press, Oxford, 1990). 17] D.M. Duffy, J.H. Harding and A.M. Stoneham, Acta Metall. Mater. 40 (1992) S l l . 18] L. Spiess, Surf. Rev. Lett. 3 (1996) 1365. 19] P.W. Tasker, Adv. Ceram. 10, 176 (1984). 20] R. Feidenhans'l, Surf. Sci. Rep. 10 (1989) 105. 21] I.K. Robinson and D.J. Tweet, Rep. Prog. Phys. 55 (1992) 599. 22] S. Dietrich and A. Haase, Phys. Rep. 260 (1995) 1. 23] G. Renaud, Surf. Sci. Rep. 32 (1998) 1-90. 24] S.R. Andrews and R.A. Cowley, J. Phys. C18 (1985) 6247. 25] I.K. Robinson, Phys. Rev. B 33 (1986) 3830. 26] E. Vlieg, J. Appl. Cryst. 30 (1997) 532. 27] E.D. Specht and F.J. Walker, J. Appl. Cryst. 26 (1993) 166. 28] O. Robach, Y. Garreau, K. Aid and M.B. Veron-Jolliot, J. Appl. Cryst. 33 (2000) 10061018. [29] N. Jedrecy, J. Appl. Cryst, in press.

297 [30] X. Torrelles, J. Rius, C. Miravitlles and S. Ferrer, Surf. Sci. 423 (1999) 338 ; L.D. Marks, Phys. Rev. B 60 (1999) 2771. [31 P. H. Fuoss and I.K. Robinson, Nucl. Instrum. Methods 222 (1984) 171. [32 S. Ferrer and F. Comin, Rev. Sci. Instrum. 66 (1995) 1674. [33 G. Renaud, B. Villette and P. Guenard, Nucl. Instr. Meth. B95 (1995) 422. [34: R. Baudoing-Savois, G. Renaud, M. De Santis, A. Barbier, O. Robach, P. Taunier , P. Jeantet, O. Ulrich , J.P. Roux, M.C. Saint-Lager, A. Barski, O. Geaymond, G. Berard, P. Dolle M. Noblet and A. Mougin, Nucl. Inst. Meth. In Phys. Res. B. 149 (1999) 213. [35 0 . Robach, G. Renaud and A. Barbier, Surf. Sci. 401 (1998) 227. [36 Y.C. Lee, P. Tong and P.A. Montano, Surf. Sci. 181 (1987) 559. [37: C. Li, R. Wu, A.J. Freeman and C.L Fu, Phys. Rev. B48, (1993) 8317. [38: P. Guenard, G. Renaud and B. Villette, Physica B 221 (1996) 205. [39 G. Bordier and C. Noguera, Phys. Rev. B44 (1991) 6361. [40 J. Goniakowski and C. Noguera, Surf. Sci. 323 (1995) 129. [41 G. Pacchioni, T. Mierva and P.S. Bagus, Surf. Sci. 275 (1992) 450. [42 J. P. LaFemina and C.B. Duke, J. Vac. Sci. Technol. A9 (1991) 1847. [43 H. Nakamatsu, A. Sudo and S. Kawai, Surf. Sci. 194 (1988) 265. [44 D.L. Blanchard, D.L. Lessor, J.P. LaFemina, D.R. Baer, W.K. Ford and T. Guo, J. Vac. Sci. Technol. A9 (1991) 1814. [45 J.B. Zhou, H.C. Lu, T. Gustafsson and P. Haberle, Surf. Sci. 302 (1994) 350. [46 C. Duriez, C. Chapon, C.R. Henry and J.M. Rickard, Surf. Sci. 230 (1990) 123. [47 P. Guenard, G . Renaud, A. Barbier and M. Gautier-Soyer, Mat. Res. Soc. Symp. Proc. 437 (1996) 15. [48 P. Guenard, G. Renaud, A. Barbier and M. Gautier-Soyer, Surf. Rev. Lett. 5 (1998) 321. [49 G.Renaud, B. Villette, I. Vilfan and A. Bourret, Phys. Rev. Lett. 73 (1994) 1825. [50 M. Gautier, J.P. Duraud, L. Pham Van, Surf. Sci. Lett. 249 (1991) L327. [51 J. Guo, D.E. ElUs, D.J. Lam, Phys. Rev. B 45 (1992) 13647. [52: W.C. Mackrodt, R.J. Davey, S.N. Black, R. Docherty, J. Cryst. Growth 80 (1987) 441. [53: 1. Manassidis, M.J. Gillan, J. Am. Ceram. Soc. 77, 335 (1994) ; Surf. Sci. Lett. 285 (1993) L517. [54: M. Causa, R.Dovesi, C. Pisani, C. Roetti, Surf. Sci. 215 (1989) 259. [55 T.J. Godin, J.P. LaFemina, Phys. Rev. B 49 (1994) 7691. [56 A. Kirfel, K. Eichhom, Acta Cryst. A 46 (1990) 271. [57 G.C. Ndubuisi, J.Liu, J.M. Cowley, Mic. Res. Tech. 20 (1992) 439. [58 V.E. Puchin, J.D. Gale, A.L. Shluger, E.A. Kotomin, J. Giinster, M. Brause and V. Kempter, Surf. Sci. 370 (1997) 190. [59: J. Ahn and J.W. Rabalais, Surf. Sci. 388 (1997) 121. [6o: T.M. French and G.A. Somorjai, J. Phys. Chem. 74 (1970) 12. [61 S. Baik, D.E. Fowler, J.M. Blakeley, R. Raj, J. Am. Ceram. Soc. 68 (1985) 281. [62: M. Gautier, J.P. Duraud, L. Pham Van and M.J. Guittet, Surf. Sci. 250 (1991) 71. [63 M. Gautier, J.P. Duraud and L. Pham Van, Surf. Sci. Lett., 249 (1991) L327.

298 [64] A. D. Novaco and J. P. McTague, Phys. Rev. Lett. 38 (1977) 1286; J. P. McTague and A. D. Novaco, Phys. Rev. B 19 (1979) 5299. 65] H. Shiba, J. Phys. Soc. Japan, 48 (1980) 211. 66] E. Gillet and B. Ealet, Surf. Sci., 273 (1992) 427. 67] M. Vermeersch, R. Sporlcen, Ph. Lambin and R. Caudano, Surf. Sci. 235 (1990) 5. 68] G. Charlton et al. Phys. Rev. Lett. 78 (1997) 495. 69] P. W. Murray, P.M. Leibsle, C.A. Muryn, H.J. Fisher, C.F.J. Fhpse and G. Thornton, Phys . Rev. Lett. 72 (1994) 689. 70] P. W. Murray, N.G. Condon and G. Thornton, Phys. Rev. B 51 (1995) 10989. 71] H. Onishi and Y. Iwasawa, Phys. Rev. Lett. 76 (1996) 791. 72] N.M. Harisson, X.-G. Wang, J. Muscat and M. Scheffler, Faraday Discuss. 114 (1999) 305. 73] M. Ramamoorthy, D. Vanderbilt, andd R.D. King-Smith, Phys. Rev. B 49 (1994) 16721. 74] D. Vogtenhuber, R. Podloucky, A. Neckel, S.G. Steinnemann, and A.J. Freeman, Phys. Rev. B 49 (1994) 2099. 75] P. Reinhardt and B.A. Hess, Phys. Rev. B 50 (1994) 12015. 76] Y.W. Chung, W.J. Lo and G.A. Somorjai, Surf. Sci. 64 (1977) 588. 77] L.E. Firment, Surf. Sci. 116 (1980) 205. 78] P. Zschack, J.B. Cohen and Y.W. Chung, Surf. Sci. 262 (1992) 395. 79] P.J. Hardman, N.S. Prakash, C.A. Muryn, G .N. Raikar, A.G. Thomas, A.F. Prime, G. Thornton and R.J. Blake, Phys. Rev. B 47 (1993) 16056. 80] P.W. Murray, F.M. Leibsle, H.J. Fisher, C.F.J. Flipse, C.A. Muryn and G. Thornton, Phys. Rev. B 46 (1992) 12877. 81] G.W. Clark and L.L. Kesmodel, Ultramicroscopy 41 (1992) 77. 82] E. Landree, L.D. Marks, P. Zschack and C.J. Gilmore, Surf. Sci. 408 (1998) 300. 83] H. Zajonz, H.L. Meyerheim, T.Gloege, W. Moritz and D. Wolf, Surf. Sci. 398 (1998) 369. 84] P.W. Tasker, J. Phys. C: Solid State Phys. 12 (1979) 4977. 85] V.E. Henrich, Surf. Sci. 57 (1976) 385. 86] D. Cappus, C. Xu, D. Ehrlich, B. Dillmann, C.A.Jr. Ventrice, K. Al-Shamery, H. Kuhlenbeck, and H.J. Freund, Chem. Phys. 177 (1993) 533. 87] M. Finazzi, N.B. Brookes and F.M.F. de Groot, Phys. Rev. B 59 (1999) 9933. 88] W.H. Meiklejohn, J. Appl. Phys. 33 (1962) 1328. 89] J.P. Hill, C.C. Kao and D.F. McMorrow, Phys. Rev. B 55 (1997) R8662. 90] S. Soeya, S. Nakamura, T. Imagawa and S.Narishige, J. Appl. Phys. 77 (1995) 5838. 91] V. Fernandez, C. Vettier, F. de Bergevin, C. Giles and W. Neubeck, Phys. Rev. B 57 (1998) 7870. 92] G. Ju, A.V. Nurmikko, R.F.C. Farrow, R.F. Marks, M.J. Carey and B. A. Gumey, Phys. Rev. B 58 (1998) Rl 1857. 93] D. Wolf, Phys. Rev. Lett. 68 (1992) 3315. 94] J.M. Cowley, Surf. Sci. 114 (1982) 587.

299

[95] C.A.Jr. Ventrice, Th. Bertrams, H. Hannemann, A. Brodde and H. Neddermeyer, Phys. Rev. B 49 (1994) 5773. [96] F. Rohr, K. Wirth, J. Libuda, D. Cappus, M. Baumer and H.J. Freund, Surf. Sci. Lett. 315 (1994) L977. [97] N. Floquet and L.C. Dufour, Surf. Sci. 126 (1983) 543. [98] A. Barbier, C.Mocuta, H.Kuhlenbeck, K.F.Peters, B.Richter and G.Renaud, Phys. Rev. Lett. 84 (2000) 2897. [99] A. Barbier, C.Mocuta and G.Renaud, Phys. Rev. B In press, - (2000). [IOO] A. Barbier and G.Renaud, Surf. Sci. Lett. 392 (1997) L15. [I011 C. Mocuta, A.Barbier and G.Renaud, Appl. Surf. Sci. 162-163 (2000) 56. [I021 W. L. Roth, J. Appl. Phys. 31 (1960) 2000. [lo31 T. Shishidou, T. Jo, J. Phys. SOC.Jpn. 67 (1998) 2637. [I041 P.W. Anderson, Phys. Rev. 79 (1950) 350. [I051 T.R. McGuire and T.S. Plaskett, J.Appl.Phys. 75 (1994) 6537. [I061 R. P. Michel, A. Chaiken, and C. T. Wang, J. Appl. Phys. 81 (1997) 5374. [I071 M. Tan, H.-C. Tong, S.-I. Tan, and R. Rottmayer, J. Appl. Phys. 79 (1996) 5012. [lo81 M.J. Carey and A.E. Berkowitz, Appl. Phys. Lett. 60 (1992) 3060. [ 1091 M.J. Carey and A.E. Berkowitz, J. Appl. Phys. 73 (1993) 6892. [I101 M.J. Carey, A.E. Berkowitz, J.A. Borchers, and R.W. Erwin, Phys. Rev. B 47 (1993) 9952. [ I1 11 C. Mocuta, A.Barbier, G.Renaud and B.Dieny, Thin Solid Films. 336 (1999) 160. [112] A. Barbier, P. Ohresser, V. Da Costa, B. Carrikre and J.-P. Deville, Surf. Sci. 405 (1998) 298. [113] M.W. Nydegger, G. Couderc, and M.A. Langell, Appl.Surf.Sci. 147 (1999) 58. [I 141 N. Jedrecy, M. Sauvage-Simkin and R. Pinchaux, Appl. Surf. Sci. 162-163 (2000) 69. [ l 151 N. Jedrecy, S. Gallini, M. Sauvage-Simlun and R. Pinchaux, Surf. Sci. 460 (2000) 136. [116] A. Wander, F. Schedin, P. Steadman, A. Norris, R. McGrath, T.S. Turner, G. Thomton and N.M. Harrison, submitted. [ 1171 A. Filippetti, V. Fiorentini, G. Cappellini and A. Bosin, Phys . Rev. B 59 (1999) 8026. [118] C.B. Duke, R.J. Meyer, A. Paton andP. Mark, Phys. Rev. B 18 (1978) 4225. [I191 W.C. Mackrodt, Phys. Chem. Minerals 15 (1988) 228. [ 1201 N. Bickel, G. Schmidt, K. Heinz and K. Muller, Phys. Rev. Lett. 62 (1989) 2009. [121] V. Ravikumar, D. Wolf and V.P. Dravid, Phys. Rev. Lett. 74 (1995) 960 [122] A. Ikeda, T. Nishimura, T. Morishita and Y. Kido, Surf. Sci. 433-435 (1999) 520. [123] J. Padilla and D. Vanderbilt, Surf. Sci. 418 (1998) 64. [124] Z.-Q. Li, J.-L. Zhu, C.Q. Wu, Z. Tang and Y. Kawazoe, Phys. Rev. B 58 (1998) 8075. [I251 Z. Liu, S. Ogota and T. Morishita, Proceedings of the SPIE, 2696 (1996) 627. [126] G. Charlton, S. Brennan, C.A. Muryn, R. McGrath, D. Norman, T.S. Turner and G. Thornton, Surf. Sci. 457 (2000),L376-L380. [I271 M. Yoshimoto, T. Maeda, K. Shimozono, H. Koinuma, M. Shinohara, 0. Ishiyama and F. Ohtani, Appl. Phys. Lett. 65 (1994) 3197.

300

[128] A. Jiang and J. Zegenhagen, Surf. Sci. 367 (1996) LA2. [ 1291 J. Zegenhagen, personal communication. [130] A. Kazimirov, J. Zegenhagen, I. Denk, J. Maier, D.-M. Smilgies and R. Feidenhans’l, Surf. Sci. 352 (1996) 875. [131] Th. Gloege, H.L. Meyerheim, W. Moritz and D. Wolf, Surf. Sci. 441 (1999) L917L923. [I321 F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Muller, L. Hammer and K. Heinz, Surf. Sci. 372 (1997) L291. [I331 F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Muller, L. Hammer and K. Heinz, Surf. Sci. 389 (1997) 391. [134] C. Rehbein, N.M. Harrison and A. Wander, Phys. Rev. B 54 (1996) 14066. [135] J.S. Moodera, L.R. Kinder, T.M. Wong and R. Meservey, Phys. Rev. Lett. 75 (1998) 3273. [ 1361 J.E. Keem, J.M. Honig and L.L. Van Zandt, Phil. Mag. B 37 (1978) 537. [137] W.D. Wang, N.J. Wu and P.A. Thiel, Chem. Phys. 92 (1990) 2025. [138] A.R. Sandy, S.G.J. Mochrie, D.M. Zehner, K.G. Huang and Doon Gibbs, Phys. Rev. B 43 (1991) 4667. [139] N. Erdman, 0. Warschkow, D.E. Ellis, and L.D. Marks, Surf. Sci. (2000) in press [140] N. Kitakatsu, V.Maurice and P.Marcus, Surf. Sci. 41 1 (1998) 215. [141] See e.g. M. Baumer and H.-J. Freund, Prog. Surf. Sci. 61, 127 (1999). [142] A. Stierle, V. Formoso, F. Comin, G. Schmitz and R. Franchy, Physica B 283, 208 (2000), and submitted. [143]. Bikondoa, F. Wendler, X. Torrelles, H. Isem, W. Moritz and G.R. Castro, unpublished. [144] T. Kosaka, S. Suzuki, M. Saito; Y. Waseda, E. Matsubara, K. Sadamori and E. Aoyagi, Thin Solid Films 289 (1996) 1. [145] G. Marest, S.B. Ogale, B. Hannoyer, S. Joshi, A. Benyagoub and N. Moncoffre, J. Appl. Phys. 80 (1996) 2228. [ 1461 T. Kosaka, S. Suzuki, H. Inoue, M. Saito, Y. Waseda and E. Matsubara, Appl. Surf. Sci. 103 (1996) 55. [147] A. Atkinson, Rev. Mod. Phys. 57 (1985) 437. [148] C. Xu, M. Hassel, H. Kuhlenbeck and H.-J. Freund, Surf. Sci. 258 (1991) 23. [149] A. Stierle, P. Bodeker and H. Zabel, Surf. Sci. 327 (1995) 9. [150] A. Stierle and H. Zabel, Europhys. Lett. 37 (1997) 365. [151] A. Stierle and H. Zabel, Surf. Sci. 385 (1997) 167.

Oxide Surfaces D.P. Woodruff, editor ©2001 Elsevier Science B.V. All rights reserved.

301

Chapter 7

Structure of thin epitaxial oxide films and their surfaces Scott A. Chambers Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, PO Box 999, MS K8-93, Richland, Washington, USA. 1. INTRODUCTION Metal oxides possess many properties that make them unique and potentially important for a range of technologies. No other class of materials exhibits such a wide range of behavior: band gaps spanning the visible and UV; electronic properties ranging from superconducting to metallic to semiconducting to insulating; magnetic properties ranging from ferromagnetic to antiferromagnetic; and dielectric properties ranging from low-k to ferroelectric and piezoelectric. The next century will require new materials systems encompassing these properties, often in artificially-structured combinations, and fabricated with a high degree of control. For example, it is clear that UV-based optoelectronics, magnetic tunnel junction and ferroelectric nonvolatile memory, spin quantum devices, and highly effective photocatalytic materials will be of great importance in information, medical, and environmental applications. Metal oxides and their associated surfaces will play important roles in many of these functions. The rich chemical, electronic, optical and magnetic behavior of metal oxides, along with superior thermal stability, uniquely suit these materials for a number of applications. The richness in properties of metal oxides is derived from the inherent chemical and physical complexities of these materials. These complexities include multiple valences, complex phase diagrams, and vastly different properties among isostructural oxides. The inherent complexity of bulk metal oxides also makes the study of oxide surfaces a difficult undertaking. Substantive issues include surface stoichiometry and termination, geometric and electronic structure, and the role of defects on surface properties. Accordingly, activity in oxide surface science has exhibited exponential growth over the past decade as the fascinating scientific issues and varied technological importance of oxide surfaces become more apparent to the international surface science community. Much of the

302

work thus far has been conducted on readily available single-crystal oxide surfaces, including MgO, a-Al203, TiOi and SrTiOs. These materials are in many ways prototypical in that they provide a fertile test bed for fundamental investigations, and many interesting facets of oxide surfaces have been discovered as a result. However, these materials do not span sufficiently far across the range of physically and chemically important properties, or far enough into the realm of technologically important oxides, to sustain a healthy scientific enterprise over the long term. For instance, none of the commonly studied oxides listed above is magnetic. Yet, magnetic oxides represent a very important class of materials for fundamental materials physics and a variety of technological applications. In order to generate a more comprehensive range of oxide surfaces, oxide thin film growth has become a vigorous activity in a number of laboratories. The materials science challenges associated with growing and characterizing epitaxial oxide films with excellent crystallographic and surface morphological properties are formidable. Yet, without sufficient definition of the surface, highly controlled experiments in surface geometric and electronic structure and chemical reactivity are not possible. One can in principle synthesize any oxide material of interest, and is limited in composition only by the ability to generate and control atomic beams in a clean vacuum environment. It is, however, essential to balance competing kinetic and thermodynamic forces that drive the chemistry and mass transport associated with film formation, particularly when the metal of interest exhibits a range of oxidation states. In one form or another, molecular beam epitaxy (MBE) has been the method of choice for this enterprise because of its relative simplicity and high degree of flexibility in varying the film composition. In contrast, control over crystalline structure and surface morphology has proved to be much more difficult to attain due to the wide range of structures and lattice parameters exhibited by oxides. As a class of materials, oxides exhibit a much more diverse range of structures than do metals and semiconductors. Therefore, lattice-matched heteroepitaxy, which results in the highest quality films, is rarely achievable. One must often settle for a less-thanperfect lattice match and, in some cases, differences in crystal symmetry between the substrate and film material. Detailed surface and film structural and morphological characterization becomes essential in order to determine the outcome of the growth experiment. The purpose of this chapter is to review recent progress in surface structural characterization of select oxide films, with particular emphasis on systems that are sensitive to some of the problems listed above. By "thin", we mean a few monolayers up to -1000 A. We take the approach of addressing the issues of oxidation chemistry, either during or after film formation, and lattice and symmetry mismatch with the substrate. We examine representative case studies that illustrate the materials issues listed above. It is assumed that the reader is

303

familiar with the techniques of surface and thin-film structure determination; no attempt is made to review the fundamentals of these methods. The reader is also referred to more comprehensive reviews of oxide film growth that have recently appeared in the literature [1,2]. 2. OXIDATION CHEMISTRY AND ITS EFFECT ON SURFACE AND FILM STRUCTURE Metal oxides constitute an extremely diverse class of materials, with virtually all types of electronic, magnetic and chemical behavior represented. The simplest oxides are those whose cation valence orbitals are of s or p character, and whose cations have only one stable oxidation state [3,4]. More complex behavior occurs in transition-metal oxides (TMOs), whose cation valence orbitals have d symmetry [3]. The simplest TMOs are those whose d shells are either empty (cf) or completely full (d^^), such as Ti02 and CU2O. The broadest range of behavior occurs for TMOs whose d orbitals are partially filled ((/). An important property of TMOs is that the transition-metal cations generally have more than one stable oxidation state. For vanadium oxides, for example, V2O5, VO2, V2O3 and VO are all stable bulk compounds, as are several other phases that have mixed cation valence states. This range in valence state leads not only to a multitude of bulk compounds with diverse properties, but also to very interesting surface electronic, magnetic and chemical properties. However, the range of oxidation states exhibited by transition metals also creates significant challenges in the MBE growth of the associated oxides. Synthesizing the most highly oxidized form of a metal by MBE is relatively simple, provided the oxygen source is of sufficient oxidizing power to fully oxidize the metal. The choice of oxygen source is oxide MBE has been discussed in detail elsewhere [2]. However, reaching the intermediate oxidation states while still producing a crystalline film requires careful control over the absolute fluxes of both metal and oxygen [5]. The fashion in which the metal is oxidized, either during or after film growth, can have some surprising effects on surface properties. We illustrate this result with two examples - the stable oxides on iron and chromium. 2.1. Fe304(lll) and a-FCiOaCOOOl) on P t ( l l l ) and a-A^OaCOOOl) The oxides of iron are of broad interest because of their importance in such diverse fields as corrosion, catalysis, geochemistry and magnetism. Despite their different structures and metal oxidation states, the oxides of iron - FeO, Fe304, and a- and y-Fe203 ~ have in common a close-packed plane of oxygen anions in the (111) orientation in which the nearest-neighbor distances are rather similar (within --5%). The various phases differ in the distribution of Fe within the cation planes that lay between the oxygen planes. FeO is rocksalt, whereas

304

Fe304 and a-Fe203 are inverse spinel and spinel, respectively, and a-FeaOa is corundum. As a result, it is possible to nucleate and grow the basal plane of any of these phases on a suitably lattice matched substrate, such as Pt(lll) or aAlsOsCOOOl). Indeed, FeO(lll) and Fe304(lll) have been grown and extensively studied on P t ( l l l ) by reactive evaporation of Fe in an O2 background [6-18], and a-FcaOs has been grown and investigated by oxygen-plasma-assisted (OPA) [5, 19-21] and N02-assisted MBE [22] on aAl2O3(0001). Molecular oxygen is not of sufficient oxidizing power to oxidize Fe metal atoms arriving at the surface to Fe203 at ordinary UHV-compatible O2 pressures. As a result, Fe304 was found to be the "terminal product" for the asgrown film on Pt(l 11). However, Wang et al. [23, 24] claim to have converted Fe304(lll) films grown on Pt(lll) into fully stoichiometric a-Fe2O3(0001) by high-temperature (~800°C) laser heating in a rather high (--10"^ torr) O2 partial pressure after growth. Indeed, by most evidences, it appears that nominally pure a-Fe2O3(0001) films several hundred A thick were produced in this way. However, a curious result arose pertaining to the surface termination, a key structural parameter for any compound material. Wang et al. reported scanning tunneling microscopy (STM) images consistent with both Fe and O terminations being present, with the relative amounts of the two terminations depending on the oxidation conditions. Surfaces consisting of-100% 0-terminated terraces appeared to result from the most oxidizing conditions [24]. Furthermore, the stability of the 0-terminated surface was predicted by spin density functional theory (spin DFT) calculations to be sufficient that it should remain in tact once formed [23]. The spin-DFT calculation also predicted a set of interlayer relaxations for this termination. In addition, these calculations predicted that the Fe-terminated surface is most stable, with interlayer relaxations in reasonable agreement with those generated experimentally by x-ray photoelectron diffraction (XPD) for a-Fe2O3(0001) films grown on a-Al2O3(0001) by OPAMBE [25]. The 0-termination in not expected to be stable based on the qualitative criterion that stable surfaces of compound materials are those with a vanishing perpendicular surface dipole moment. This condition is equivalent to that of an autocompensated surface, in which the surface forms in such a way that all anion- (cation-) dangling derived dangling bonds are completely full (empty) [26, 27]. The only termination of a-Fe2O3(0001) that is autocompensated (e.g. nonpolar) is that consisting of 1/3 ML of Fe. We illustrate these results in Figures 1-4. We show in Fig. la a side view of Feterminated (and autocompensated) a-Fe2O3(0001) with interlayer relaxations measured by XPD and predicted by both spin-DFT and molecular mechanics [28]. Fig. lb shows the O-terminated surface, which is not autocompensated, but predicted to be stable by spin-DFT. Based on the measured and calculated

305

Fig. 1 Fe-terminated (a) and O-terminated (b) ct-Fe2O3(0001) surface structures. The interlayer relaxations shown in (a) were measured by XPD and predicted by two independent theoretical methods. Those shown in (b) are predicted by spin-DFT theory, but have not been confirmed experimentally.

interlayer spacings described above, the irreducible step height is expected to be -2.3A for a surface consisting entirely of Fe-terminated or O-terminated terraces. Likewise, the minimum height difference between O- and Feterminated terraces is expected to be -l.SA (~0.5A) if the upper terraces are O(Fe-) terminated. Wang et al. measured with STM a minimum step height of ~1 A between what was assigned, based on atomic corrugation, to be (upper) Oterminated terraces and (lower) Fe-terminated terraces. A representative image is shown in Fig. 2. Although the --lA step height does not match that expected if both terminations are present, its presence, along with the different atomic corrugations on the two kinds of terraces, was taken to be substantial evidence for the stability of the O termination coexisting with the Fe termination. However, we could not form the O termination by OPA-MBE under any circumstances appropriate to that experiment, including stopping the film growth by closing the Fe shutter and allowing the specimen to remain at growth temperature (400°C) and then cool down to near room temperature in the oxygen

306

20 40 line scan direction (A)

Fig. 2 STM image of an Fe304(l 11) film grown on Pt(l 11) after laser annealing to 800°C in -10"^ torr of O2 to convert the film to a-Fe2O3(0001). The two visible terraces have been interpreted as being due to Fe- (left) and O- (right) terminations.

plasma [29]. This treatment should produce the 0-terminated surface if indeed this surface is stable. However, STM images, obtained by doping the film with Fe(II) in the middle of the growth to increase conductivity, revealed an irreducible step height of ~2A. This result indicates the presence a single termination. Furthermore, XPD established that this surface was Fe-terminated, as shown in Figs. 3 and 4. Fig. 3 shows a comparison of experimental angular distributions obtained at low emission angles (to maximize surface sensitivity) for two surface preparations. These preparations were designed to maximize the probability of formation of 0-terminated (a) and Fe-terminated (b) terraces. In preparation a, the surface was allowing to cool down from the growth temperature to near room temperature while being continuously exposed to the oxygen plasma. Preparation b involved annealing a previously grown specimen at temperatures of ~600°C for a few minutes in molecular oxygen at a chamber pressure of 1 x 10'^ torr. This procedure was used to clean MBE-grown a-Fe2O3(0001) used in an Fe 3p XPD experiment conducted at the Berkeley Advanced Light Source after a through-air transfer, and was found to result in an Fe-terminated surface [25]. Experiment is also compared with theory for the two terminations in Fig. 3, using the optimized coordinates shown in Fig. 1. Visual comparison of theory

307

20 40 60 80 100 Azimuthal Angle, <|) (degrees)

120

Fig. 3 Shallow take-off angle (8 - relative to the surface) XPD 01s azimuthal scans for epitaxial a-Fe2O3(0001) on a-AlaOsCOOOl) grown by OPA-MBE. The surface was prepared two different ways to maximize the likelihood of achieving an O termination (a), and an Fe termination (b). Also shown are single scattering simulations for the two terminations. Qualitatively, agreement is superior for the Fe termination for both methods of preparation.

308

and experiment suggests that the Fe-teraiinated surface gives the best agreement with experiment, a conclusion further established by the R-factor analysis shown in Fig. 4. How can these disparate results be resolved? While not completely satisfactory, the argument can be made that stacking faults at the aFe2O3(0001)/Pt(lll) interface form during the high-temperature anneal in oxygen required to convert Fe304 to a-Fe203. If present, these stacking faults, would give the appearance of the simultaneous presence of both Fe- and O-

20

40

60

80

100

Percentage of Surface with Fe Termination

Fig. 4 R-factor analysis, utilizing R5, for the data shown in Fig. 3. All experimental scans for both surface preparations agree quantitatively better with the Fe-termination simulation.

309

terminated domains. Oxidation of Fe304(lll) to a-Fe2O3(0001) requires extensive oxygen diffusion throughout the film. Oxidation might resuh in stacking faults if this process leads to the breaking of Fe-Pt bonds at the Fe oxide/Pt(lll) interface [16] and the subsequent creation of 0-Pt bonds within select regions of the film. Stacking faults would exist at boundaries between domains of oxide characterized by Fe-Pt and 0-Pt interfacial bonds, and the height difference at the surface would be '-I A. However, this model does not provide an explanation for the different surface corrugations measured on the different domains, as described earlier [23, 24]. The larger (smaller) corrugations were interpreted as being associated with the Fe- (0-) terminated domains. It was suggested that the smaller corrugation on 0-terminated domains is a result of Fe-derived states that dominate the tunneling profiles and are partially screened by the terminal O layer. The stacking fault model would predict the same corrugation on both kinds of domains, because the domains would be of the same atomic termination. In order to fully reconcile these results, it will be necessary to perform a definitive structural determination on the allegedly 0-terminated surface. To this end, preliminary LEED I-V analysis has been carried out on 0-stabilized aFe2O3(0001)/Pt(lll), but has not produced satisfactory agreement between experiment and any theoretical structure simulated thus far [30]. This troubling result, together with the fact that ~500A thick films were sufficiently conductive to easily perform STM (pure a-Fe203 is a 2.2 eV wide bandgap semiconductor) suggests that the film prepared in this way is highly defected. However, further research is needed to fully clarify the issue. Nevertheless, this example amply illustrates the effect that film preparation procedures can have on the resulting surface structures for oxides of a metal that exhibits multiple oxidation states. 2.2. Cr02 on TiOjCllO) & (100), and a-CriOs on a-Al2O3(0001) Cr is another transition metal that exhibits multiple oxidation states. CrOa is purported to be a half-metallic ferromagnet, in that one spin subband is metallic whereas the other is semiconducting [31-33]. In contrast, a-Cr203 is antiferromagnetic. In addition, the oxides of Cr exhibit a wide range of physical and chemical properties. The most stable oxide is a-Cr203, which exhibits a corundum structure. Cr02, a rutile, is less stable and spontaneously decomposes to a-Cr203 at 280°C. Cr304, a tetragonally distorted spinel, is stable only above -^1600°C. Finally, Cr03 is orthorhombic, and melts at 195°C and boils at 250°C. This wide range of structures and thermal stabilities makes epitaxial growth a challenge, particularly for the less stable phases. For instance, if one wants to grow Cr02 by MBE, one must deal with its tendency to decompose to Cr203 at a temperature substantially less than that at which good oxide epitaxy is generally expected to occur, as well as the high volatility of Cr03. In order to prevent

310 decomposition to CY2O3, one might try growing under oxygen rich conditions. However, doing so is likely to produce at least some CrOs, which is extremely volatile and may desorb so fast that a substantial fraction of the incident Cr does not remain on the substrate surface long enough to convert to Cr02. Cr and O atomic fluxes in a 1:2 ratio, along a substrate temperature of less than ~250°C, might be expected to generate Cr02 films, although the structural quality may suffer due to the rather low growth temperature. Preliminary attempts in our laboratory to grow epitaxial CrOi on TiOiCUO) by OPA-MBE have been uniformly unsuccessful, despite the reasonably small lattice mismatch (Aa/a = 1.25% along [001] and -3.56% along [110]). Quasi-epitaxial a-CraOs consistently formed over a wide range of growth conditions, and the final thickness was consistently lower than that expected on the basis of the Cr flux, indicating high desorption rates for what was presumably a CrOs intermediate. However, Gupta and coworkers [34] have been able to grow epitaxial Cr02 on TiO2(100) using atmospheric pressure chemical vapor deposition using a CrOs precursor and O2 as the carrier gas at a substrate temperature of 400°C. A high incident flux of CrOs is apparently needed to drive the decomposition of CrOs to Cr02 on the surface while excess O2 is required to prevent reduction to Cr203. In contrast, it is relatively easy to grow a-Cr203 on closely lattice matched substrates such as P t ( l l l ) (Aa/a - +2.46%) and a-Al2O3(0001) {Aa/a = +3.36%). Diebold and coworkers have grown thin Cr oxide films on Pt(l 11) by evaporating Cr metal in a O2 ambient [35-37]. These authors used a combination of STM, low-energy electron diffraction (LEED), ion scattering spectroscopy (ISS), ultraviolet photoemission spectroscopy (UPS) and electronic structure calculations to investigate the structure, composition and electronic structure of the evolving film. We have successfully grown a-Cr203 on aAl2O3(0001) by OPA-MBE [38]. In order to explore the surface structure, we have prepared the surface in the same ways as those described in connection with the study of the a-Fe2O3(0001) surface (methods a and b discussed in Section 2.1). In contrast to a-Fe2O3(0001), the a-Cr2O3(0001) surface shows evidence of an 0-stabilized surface when surface preparation a was used. It is known from detailed studies of O2 chemisorption on a-Cr2O3(0001), prepared by oxidation of Cr(l 10), that molecular adsorption at 90K leads to O2 dissociation and the formation of surface chromyl species (Cr=0) at temperatures > 415K [39]. We find a similar result when surface preparation a is used. We show in Fig. 5 high-resolution Ols core-level spectra for a-Cr2O3(0001) surfaces prepared according to methods a and b. These are designated 0/ a-Cr2O3(0001) and a-Cr2O3(0001), respectively. Spectra were measured at normal (6 = 90°) and grazing (0 = 10°) emission. The spectra for the surface prepared by method

311

c

a-Cr2O3(0001)

O/a-Cr2O3(0001)

01s

01s

, lattice O

13

CO CO

c

CD

9 = 90

*-• _c c o

chemisorbed O 1/3 ML

CO CO

A = 0.85 eV

E CD

o o

JO

CL

I . . . . I . . . . I.

532

530

528 532 Binding Energy (eV)

530

528

Fig. 5 High-resolution O Is spectra for Cr-terminated a-Cr2O3(0001) at normal (0 = 90°) and grazing (0 = 10°) emission (left), and for chromyl-termintated a-CraOsCOOOl) (right).

b, which should be Cr terminated, show a singlet at both emission angles, which is assigned to lattice oxygen. No surface core-level shift is visible, despite the presence of 0-derived dangling bonds at the surface. In contrast, the surface prepared according to method a shows a second feature shifted to higher binding energy by 0.85 eV, which is enhanced at grazing emission, suggesting surface segregation. The binding energy shift is too small to be associated with surface OH, resulting from dissociative chemisorption of water [40], which is consistently seen for a-Fe2O3(0001) surfaces prepared by both methods [25, 29]. However, a 0.85 eV shift is consistent with a terminal oxygen layer. Assuming a simple continuum attenuation model and an electron attenuation length of 27 ± 2A, the intensity ratio of chemisorbed oxygen to lattice oxygen intensities measured at 6 = 10"^ reveals a surface fractional oxygen coverage of 0.33 ± 0.03. The nominal Cr termination for the autocompensated surface is 1/3 ML. If each terminal Cr is converted to a chromyl species, the surface O coverage would also be 1/3 ML, consistent with the spectral data. It is expected that addition of this terminal layer of oxygen will modify the surface structure. The Cr-terminated surface has been investigated by LEED I-V analysis and found to be strongly inwardly relaxed, as is the Fe-terminated aFe2O3(0001) surface. Rohr et al. [41, 42] measured inward relaxations of-3 8%, -21%, -25%, and +11% of the associated bulk values. The structure of the chromyl-terminated surface has not been determined.

312

We emphasize that the oxygen termination resulting from the formation of chromyl species is fundamentally different from that shown in Fig. lb. The latter is a bulk terminated surface, with relaxations, and is not autocompensated. In contrast, the former consists of a single chemisorbed layer of oxygen added to the Cr-terminated surface. Surface autocompensation of the Cr-terminated surface cannot be violated by the presence of the chromyl layer, because neutral oxygen is added to an already autocompensated surface. In a formal sense, the problem can be thought of as follows. The Cr-terminated surface is autocompensated in that each surface Cr has three dangling bonds (DBs) with 0.5 e" each, whereas each second-layer O has 1.5 e" in one DB. There is 1 (3) Cr (O) atoms per surface unit cell. Therefore, transfer of 1.5 e' from Cr to O DBs will completely empty for former and fill the latter. If 1/3 ML of O is now added to the surface to form chromyl bonds, all four electrons in each double bond can be thought of as being donated by the oxygen, leaving 2 e' to occupy the O DB. No more detailed electron counting analysis can be carried out unless the rehybridization of the surface Cr t2g orbitals with the chromyl O s and p orbitals is understood in detail. However, it is sufficient to state that the addition of a neutral species to an autocompensated surface cannot remove the autocompensated state of that surface. The changes in surface structure brought about by the formation of the chromyl surface layer have a marked effect on the high-resolution Cr 2p spectrum. We show in Fig. 6 Cr Ip^n spectra for the Cr- and chromylterminated surfaces at G = 90° and 10°. These spectra are strongly multiplet split as a result of 2/7 core-hole interaction with the unpaired d electrons in the crystal-field-split valence band. Of most interest are the differences between the surface- and bulk-sensitive spectra. The differences are primarily in the relative intensities in the case of the Cr-terminated surface. However, there is a substantial binding energy shift for the highest-energy multiplet in the more surface-sensitive spectrum for the chromyl-terminated surface. These spectra can be predicted using atomic multiplet theory, guided by Cr Ls-edge x-ray absorption spectra, which exhibit better peak resolution than XPS due to lifetime effects. The differences between the spectra obtained at 6 = 90° and 10° can be traced to changes in the crystal field strength at the surface relative to the bulk, and to the break in local symmetry at cation sites resulting from formation and relaxation of the surface. We are currently carrying out this analysis for both aFe2O3(0001) and a-CraOsCOOOl), and the results will be reported at a later date.

313 a-Cr2O3(0001)

b/a-Cr2O3(0001) Cr 2p3/2

Cr 2p3/2 CO

0 = 90°

e = 90-

c o CO

E


580

576

572 580 Binding Energy (eV)

576

572

Fig. 6 High-resolution Cr 2py2 spectra for Cr-terminated a-Cr2O3(0001) at normal (9 = 90°) and grazing (0 = 10°) emission (left), and for chromyl-termintated a-Cr2O3(0001) (right).

3. LATTICE MISMATCH AND RESIDUAL STRAIN FIELDS Lattice mismatch results in the accumulation of strain energy up to the point that the total strain energy exceeds the energy associated with some structural transformation in the film. At this point, known as the critical thickness, it becomes energetically more favorable for the film to undergo the structural change and relax, than for the film to remain strained. The most common structural transformations are: (1) the transition from a flat but biaxially strained film to an at least partially relaxed array of nanocrystalline islands, and, (2) the formation of misfit dislocations throughout the film. These processes have been extensively studied in technologically important lattice mismatched semiconductor systems such as GexSii.x/Si(001) and InxGai.xAs/GaAs(001), but have not been investigated for oxides. For a given heteroepitaxial system, it might be expected that the critical thickness would be higher when the epitaxial film is in lateral tension compared to the case when it is in compression. The ostensible reason for this behavior is that the interatomic potential curve is highly anharmonic, and rises much more steeply on the repulsive portion below the equilibrium interatomic separation than above it. However, the anharmonicity is small for typical displacements from equilibrium that accompany strained-layer epitaxy (typically a few precent), so in reality, the difference should not be so great. Consider, for instance, the Si/Ge system, for which the lattice mismatch is 4% (asF5.43A and aoe = 5.65A). The critical

314

thickness for Ge on Si(OOl) and Si on Ge(OOl) is 4-6 ML [43, 44]. The tetragonal distortions perpendicular to the interface have been measured by XPD for both systems below their respective critical thicknesses and found to be less than that predicted by classical elastic theory by -O.lA [45]. Similarly symmetric behavior is expected for oxides. However, very few systems have been investigated in which the same material has been grown in both tension and compression on different substrates. Likewise, few oxide superlattice systems have been grown in which A on B can be compared structurally to B on A. One stable oxide, a-Cr2O3(0001), has been grown in lateral compression on a-Al2O3(0001), and in lateral tension on a-Fe2O3(0001). In addition, (a-Fe203/ a-Cr203) superlattices have been grown on a-Al2O3(0001) [38]. The structural results are very surprising, as discussed below. 3.1. (a-Fe203/a-Cr203)„ on a-AUOaCOOOl) All three of these oxides exhibit a corundum-like structure. The in-plane lattice parameters (a) are 4.76A, 4.92A, and 5.03A for a-Al203, a-Cr203 and aFe203, respectively. The out-of-plane lattice parameters (c) for the three oxides are 12.99A, 13.7A, and 13.7A, respectively. We have used reflection highenergy electron diffraction (RHEED) and transmission electron microscopy (TEM) to determine lattice distortions during and after growth of superlattices of these materials [38]. We show in Fig. 7 RHEED streak spacings during the growth of a (a-Fe203/a-Cr203) superlattice on a-Al2O3(0001). The streak spacing is inversely proportional to the in-plane lattice parameter at the film surface. The streak spacings expected for the fully relaxed materials at room temperature are also shown as dashed lines in the figure. In response to compressive in-plane mismatch, the first a-Cr203 film begins to relax almost immediately after growth is initiated, but retains some residual in-plane strain after growth (1000 seconds, which results in 120A of a-Cr203). In contrast, the first a-Fe203 layer relaxes immediately. The apparent excursion to an in-plane lattice parameter that is only - 9 1 % that of the substrate as the first ML grows has been interpreted as being due to low-angle facets created by buckling on a compressionally stressed wetting layer [46]. Once relaxation is complete after a few ML, the a-Fe203 layer maintains an in-plane lattice parameter that is - 1 % larger than the room temperature value for bulk a-Fe203, presumably as a result of thermal expansion at the growth temperature, ~500°C. Then, quite surprisingly, the second a-Cr203 layer retains the in-plane lattice parameter of a-Fe203, as do all subsequent a-Cr203 layers, even when grown to a thickness of several hundred A. RHEED streak spacing measurements carried out after the growth of each layer with the sample at room temperature (not shown) confirm that all layers of both a-Cr203 and a-Fe203 exhibit the in-plane lattice

315

I • > ' I ' I ' I ' I « 1 ' } ' I ' I ' { ' I M

a-CraOs

c Q.

Fig. 7

90

I I

\ i

1000

a-FaaOs I }

i

I I

i i

' I ' i • I • I ' I ' t '

a-CraOs

} ) I, t

) 1 1 .1 ,) t »

I M

a-FeaOs i

i

i

I I

2000 4000 Time (seconds)

i„, ,1 „ J

,1

'

I » I ' { ' II

a-CraOs i

t

6000

1 1

1 1 }

I I » B

8000

RHEED streak spacing during a-Cr203/a-Fe203 superlattice growth on a-

AbOsCOOOl). The 30 keV beam was aligned parallel to [11 00].

parameter of bulk a-FeaOsCOOOl). Such a large critical thickness for a-Cr203 is probably a result of the fact that a-Cr203 and a-Fe203 have the same out-ofplane lattice parameter, despite the 3,36% in-plane lattice mismatch. The excellent out-of-plane lattice match presumably reduces the strain energy per layer so that strained films can be maintained to much large thicknesses than would ordinarily be expected. In order to probe the out-of-plane lattice parameter, we turn to TEM. We show in Fig. 8 a selected area diffraction pattern obtained in cross section with the beam aligned along [101 0]. The beam was incident on two periods of a 120A a-Cr2O3/140A a-Fe203 superlattice capped with a thick a-Cr203 epitaxial film, along with a portion of the underlying substrate. The individual diffracted beams are split, indicating multiple lattice parameters. A scan along the (0012) beam, which is shown at the bottom of the figure, reveals three distinct lattice parameters, which are assigned to the substrate, fully relaxed a-Fe203, and partially relaxed a-Cr203. The c/a ratio of epitaxial a-Cr203 is 2.70 compared to 2.78 in the bulk. This example illustrates the unusual structures that can be achieved by strained-layer epitaxy of oxides. The resulting electronic, magnetic and optical properties of artificially structured oxide layers and their respective surfaces may prove to be quite interesting, and potentially useful in nextgeneration device technology.

316

(0012) a-Fe2O3(0001) a-Cr2O3(0001) a-Al2O3(0001) C=13.7A , C=13.5A I c=13.0A

Z

3

distance across (0012) beam (arb. units) Fig. 8 TEM selected area diffraction pattern for the a-Cr203/a-Fe203 superlattice grown on a-AbOsCOOOl). The primary beam was aligned along [lOl 0].

4. CRYSTAL SYMMETRY MISMATCH The limited number of high quality substrates suitable for oxide epitaxy, together with the wide range of structural properties exhibited by oxides, may require the use of a substrate with a different crystal symmetry than the film material. If the crystal symmetries are sufficiently different, antiphase boundaries (ABPs) may result during nucleation of the initial monolayers. Such APBs tend to be very stable and thus typically become permanently ingrained in the film structure. The question we address here is the effect that APBs in the bulk of the film have on the surface structure.

317

4.1. Fe304 on MgO(OOl) Magnetite, Fe304, is a conductive inverse spinel at temperatures above 123K, the so-called Verwey transition. Magnetite is also ferrimagnetic. Fe(III) at tetrahedral (A) sites are of one spin orientation whereas the alternating Fe(II) and Fe(III) at octahedral (B) sites are of opposite spin orientation. In addition, the magnetic moments of Fe(II) and Fe(III) at B sites are of different magnitude. Conductivity above the Verwey transition results by electron hopping between B-site Fe(II) and Fe(III) cations. The mobile electron is the spin-down 6^^ d electron formally associated with Fe(II) cations. Since the mobile electron is of a single spin orientation, magnetite is further purported to be a half-metallic ferromagnet [47]. If found to be true, this property should result in 100% spin polarization at the Fermi level. However, preliminary results based on spinpolarized valence band photoemission for epitaxial Fe304 on MgO(OOl) reveal spin polarizations at the Fermi level of-30% [48]. Fe3O4(001) has been grown epitaxially on MgO(OOl) using a number of growth methods, including OPA-MBE [5, 19, 49, 50], N02-assisted MBE [5154], and reactive sputter deposition [55, 56]. Although these two materials exhibit different crystal symmetries (Fe304 - inverse spinel, space group Fd3m; MgO - rocksalt, space group Fm3m), and have lattice parameters that differ by nearly a factor of two (Fe304 - 8.396A; MgO - 4.212A), their respective oxygen sublattices are nearly identical. Continuity of the anion sublattice across the interface has been shown to be a major stabilizing factor in the heteroepitaxy of at least partially ionic materials. The spinel structure differs from rocksalt in that there are inequivalent cation layers (A and B) that alternate along [001], resulting in a lower cation density per layer and a correspondingly larger lattice parameter. As a consequence, it is possible in principle to nucleate twodimensional "rafts" of Fe304 by aligning the oxygen sublattices of the rafts with that of the substrate such that APBs naturally arise upon raft coalescence. This situation is depicted in Fig. 9a. One raft of epitaxial Fe304 is displaced by (l/4)ao<110> relative to the other. Alignment of the oxygen sublattice of each raft with that of the substrate is the same in each case. However, there is an APB within the A-site iron sublattice. Margulies et al. [56] have generated evidence for APBs in epitaxial Fe304 grown on MgO(OOl) by sputtering. These defects leave a fingerprint in plan view TEM dark field images constructed from the 220 reflection. One such image is shown in Fig. 10. From an analysis of this image, the antiphase domain size was suggested to range from 50 to lOOOA, with an average size of ~275A. Presumably, the APBs formed during nucleation of the first full monolayer and propagated throughout the film. The presence of these ABPs provides a plausible explanation for the anomalous magnetic properties of these films, including the large saturation magnetization fields [55, 56].

318

Fe (A site) Fe (B site) (a)"

(b)

(1/4)ao<110>

[•

*

!-•

•

\m

•

1

Fig. 9 Two epitaxial rafts of (lxl)-Fe3O4(001) on MgO(OOl) displaced by (l/4)ao<110> relative to one another (a). Two rafts of (V2xV2)R45°- Fe3O4(001) on MgO(OOl) displaced by (l/4)ao relative to one another which show only the tetrahedral Fe(III) terminal layer.

It would seem at first glance that these defects should be visible at the surface by atomically resolved scanning tunneling microscopy, similar to what is seen in Fig. 9a. However, a complication arises due to the tendency of the surface to reconstruct in order to achieve autocompensation. For the surface terminated on the Fe A-type layer, the surface eliminates 1/2 ML tetrahedral Fe(III) and forms a (V2xV2)R45° reconstruction that is seen in LEED and RHEED [19, 50, 57]. We have obtained atomically resolved STM images of this surface in order to image the tetrahedral Fe(III) surface net, and have carried out scanned-angle XPD measurements in order to determine the interlayer relaxations [57]. A representative STM image is shown in Fig. 11.

319

Fig. 10 Plan view dark field TEM image based on a 220 reflection and selected area diffraction pattern for 500A Fe304 on MgO(OOl). Arrows indicate the location of intersection points of sets of three APBs.

• - tet. Fe(IH) O - Qot Fe{(ll) ^ ' 0€t. Fa(n) ' tet. O

Fig. 11 Filled-state STM image of (V2xV2)R45°- Fe3O4(001) on MgO(OOl) grown by OPAMBE.

320

The tetrahedral Fe(III) cations appear as bright spots, and dominate the image; the underlying octahedral iron cations are not visible. In this and all other images we obtained, there is no evidence for APBs on the surface. The anticipated effect of APBs on the STM image is illustrated in Fig. 9b. Here, we show the same two rafts of Fe304 as shown in Fig. 9a, except that the underlying B-type layer has been removed, and the (V2xV2)R45° reconstruction has been formed by removing one half of the tetrahedral Fe(III) cations in the A-type laver. This picture is thus a facsimile of the STM image for the A-type (V2xV2)R45° reconstruction in the vicinity of an APB. The effect of the APBs on the image should be readily visible. However, there is no evidence for such APBs in our images, or in those published by other groups using MBE-grown material [53]. It is possible that these APBs do not form in MBE grown Fe304. However, the only way to be sure is to perform the TEM measurement shown in Fig. 10 on an OPA-MBE grown sample, and that has not yet been done. The scanned-angle XPD structural analysis we carried out on this surface is not expected to be sensitive to the presence of APBs at the concentrations reported by Margulies et al, if such APBs are present at all. We show in Fig. 12 a summary of the experiment/theory comparison based on Fe 2p3/2 azimuthal scans at shallow emission angles, and in Fig. 13 the different structural models that were tested [57]. These models include: (1) the optimized, reconstructed Atype surface, (2) the optimized, reconstructed B-type surface, (3) a preliminary model from an unpublished surface x-ray scattering experiment on bulk Fe3O4(001) [58], and, (4) the structure predicted by a molecular dynamics calculation [59]. The best overall agreement between theory and experiment for the as-grown surface occurs for the A-type surface with interlayer spacings of -14%, -57%, -19%, and +29% of the respective bulk values for the first four interlayer spacings, respectively. A recent low-energy ion scattering experiment carried out on N02-assisted MBE grown Fe304 film on MgO(OOl) also revealed an inwardly relaxed A-type termintion for the as-grown film, although the magnitude of the surface-layer relaxation was different [60]. 5. CONCLUSIONS The range of metal oxidation states and crystal structures exhibited by metal oxides, together with the limited number of readily available epi-quality substrates, presents several challenges for the heteroepitaxial growth of oxides. Gaining control over oxidation chemistry in two dimensions as surface diffusion, condensation and crystallization occur at the growth front is essential in order to obtain the desired phase. Errors here can have a pronounced effect on the resulting oxide surface structures. Likewise, the inevitable deviations from zero lattice and symmetry mismatch can have some surprising effects on

321

0

20 40 60 80 Azimutha! Angle, <^ (degrees)

Fig. 12 Experimental (dots) and calculated (lines) Fe Ip^/i XPD azimuthal scans for (V2xV2)R45°- Fe3O4(001) on MgO(OOl) grown by OPA-MBE. The letters adjacent to the theory curves correspond to the structural models shown in Fig. 13.

322

[010]

H'

(a)

[110] [100]

• ^ © O

tetrahedral Fe termination

o

-tet. Fe(lll)

-oct Fe(lll) -oct. Fe(ll) -tet.O - tet. O vacancy

octahedral Fe + tetraliedral O termination interstitial tet. Fe(lll),

x-ray crystal truncation rod determination (d)

[110]

molecular dynamics prediction

Fig. 13 Structural models used to analyze the XPD scans shown in Fig. 13. These include the optimized, reconstructed A-type surface (a), the optimized, reconstructed B-type surface (b), a preliminary model from an unpublished surface x-ray scattering experiment on bulk Fe3O4(001) (c), and the structure predicted by a molecular dynamics calculation (d).

surface structure, all of which must be adequately considered in order to generate structurally and morphologically well-defined surfaces. This chapter illustrates these concepts with a specific set of examples available from the limited scientific literature on the subject. It is no overstatement that oxide epitaxy is in its infancy compared to semiconductor epitaxy. A very wide range of oxide material structures with diverse and unexplored properties have yet to be grown and characterized.

323

ACKNOWLEDGEMENTS The writing of this chapter, and much of the research described therein, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Division of Materials Science. The research was performed in the Environmental Molecular Sciences Laboratory, a national scientific user facility sponsored by the U.S. Department of Energy's Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory. REFERENCES [I] R. Franchy, Surf Sci. Rep. 38 (2000) 195. [2] S.A. Chambers, Surf Sci. Rep. 39 (2000) 105. [3] P. A. Cox, Transition Metal Oxides: An Introduction to Their Electronic Structure and Properties, (Clarendon Press, Oxford, 1992). [4] V.E. Henrich and P.A. Cox, The Surface Science of Metal Oxides, (Cambridge University Press, 1994). [5] Y. Gao, Y.J. Kim and S.A. Chambers, J. Mater. Res. 13 (1998) 2003. [6] H.C. Galloway, J.J. Benitez and M. Salmeron, Surf. Sci. 298 (1993) 127. [7] H.C. Galloway, J.J. Benitez and M. Salmeron, J. Vac. Sci. Technol. A12 (1994) 2302. [8] W. Weiss, A.B. Boffa, J.C. Dunphy, H.C. Galloway, M.B. Salmeron and G.A. Somorjai, Springer Ser. Surf. Sci. 33 (1993) 221. [9] W. Weiss and G.A. Somorjai, J. Vac. Sci. Technol. Al 1 (1993) 2138. [10] T. Schedel-Niedrig, W. Weiss and R. Schlogl, Phys. Rev. B52 (1995) 17449. [II] H.C. Galloway, P. Sautet and M. Salmeron, Phys. Rev. B54 (1996) Rl 1145. [12] Y.J. Kim, C. Westphal, R.X. Ynzunza, H.C. Galloway, M. Salmeron, M.A. Van Hove and C.S. Fadley, Phys. Rev. B55 (1997) R13448. [13] W. Weiss, Surf. Sci. 377-379 (1997) 943. [14] Y.Q. Cai, M. Ritter, W. Weiss and A.M. Bradshaw, Phys. Rev. B58 (1998) 5043. [15] M. Ritter, W. Ranke and W. Weiss, Phys. Rev. B57 (1998) 7240. [16] Y.J. Kim, C. Westphal, R.X. Ynzunza, Z. Wang, H.C. Galloway, M. Salmeron, M.A. Van Hove and C.S. Fadley, Surf Sci. 416 (1998) 68. [17] M. Ritter and W. Weiss, Surf Sci. 432 (1999) 81. [18] W. Weiss and M. Ritter, Phys. Rev. B59 (1999) 5201. [19] Y.J. Kim, Y. Gao and S.A. Chambers, Surf. Sci. 371 (1997) 358. [20] Y. Gao, Y.J. Kim, S.A. Chambers and G. Bai, J. Vac. Sci. Technol. A15 (1997) 332. [21] Y. Gao and S.A. Chambers, J. Cryst. Growth 174 (1997) 446. [22] T. Fujii, D. Alders, F.C. Voogt, T. Hibma, B.T. Thole and G.A. Sawatzky, Surf Sci. 366(1996)579. [23] X.-G. Wang, W. Weiss, S.K. Shaikhutdinov, M. Ritter, M. Peterson, F. Wagner, R. Schlogl and M. Scheffler, Phys. Rev. Lett. 81 (1998) 1038. [24] S.K. Shaikhutdinov and W. Weiss, Surf. Sci. 432 (1999) L627. [25] S. Thevuthasan, Y.J. Kim, S.I. Yi, S.A. Chambers, J. Morals, R. Denecke, C.S. Fadley, P. Liu, T. Kendelewicz and G.E. Brown, Jr., Surf Sci. 425 (1999) 276. [26] M.D. Pashley, Phys. Rev. B40 (1989) 10481.

324

[27] J.P. LaFemina, Crit. Rev. Surf. Chem. 3 (1994) 297. [28] E. Wasserman, J.R. Rustad, A.R. Felmy, B.P. Hay and J.W. Halley, Surf. Sci. 385 (1997)217. [29] S.A. Chambers and S.I. Yi, Surf. Sci. 439 (1999) L785. [30] W. Weiss, private communication (1999) [31] K. Schwarz, J. Phys. F. 16 (1986) L211. [32] K.P. Kamper, W. Schmitt, G. Guntherod, R.J. Gambino and R. Ruf, Phys. Rev. Lett. 59(1987)2788. [33] R.J. Soulen, J.M. Byers, M.S. Osofsky, B. Nadgomy, T. Ambrose, S.F. Cheng, P.R. Broussard, C.T. Tanaka, J. Nowak, J.S. Moodera, A. Barry and J.M.D. Coey, Science 282 (1998) 85. [34] X.W. Li, A. Gupta, T.R. McGuire, P.R. Duncombe and G. Xiao, J. Appl. Phys. 85 (1999)5585. [35] P.S. Robbert, H. Geisler, C.A. Ventrice, Jr., J. van Ek, S. Chaturvedi, J.A. Rodriguez, M. Kuhn and U. Diebold, J. Vac. Sci. Technol. A16 (1998) 990. [36] L. Zhang, M. Kuhn and U. Diebold, Surf Sci. 375 (1997) 1. [37] L. Zhang, M. Kuhn and U. Diebold, J. Vac. Sci. Technol. A15 (1997) 1576. [38] S.A. Chambers, Y. Liang and Y. Gao, Phys. Rev. B61 (2000) 13223. [39] B. Dillmann, F. Rohr, O. Seiferth, G. Klivenyi, M. Bender, K. Homann, I.N. Yakovkin, D. Ehrlich, M. Baumer, H. Kuhlenbeck and H.-J. Freund, Faraday Discuss. 105 (1997) 295. [40] P. Liu, T. Kendelewicz, J. G.E. Brown, E.J. Nelson and S.A. Chambers, Surf Sci. 417(1998)53. [41] F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Stammler, S. Muller, L. Hammer and K. Heinz, Surf Sci. 372 (1997) L291. [42] F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Stammler, S. Muller, L. Hammer and K. Heinz, Surf Sci. 389 (1997) 391. [43] J. Bevk, J.P. Mannaerts, L.C. Feldman, B.A. Davidson and A. Ourmadz, App. Phys. Lett. 49 (1986) 286. [44] A.A. Williams, J.M.C. Thornton, J.E. MacDonald, R.G.van Silfhout, J.F.van der Veen, M.S. Finney, A.D. Johnson and C. Norris, Phys. Rev. B43 (1991) 5001. [45] S.A. Chambers and V.A. Loebs, Phys. Rev. B42 (1990) 5109. [46] S.I. Yi, Y. Liang, S. Thevuthasan and S.A. Chambers, Surf Sci. 443 (1999) 212. [47] V.Y. Irkhin and M.I. Katsnel'son, Phys. Usp. 51 (1994) 659. [48] S. Morton, G.D. Waddill, S. Kim, I. Schuller, S.A. Chambers and J.G. Tobin, Oral Presentation at the March Meeting of the American Physical Society, St. Louis, MO (2000) [49] D.M. Lind, S.D. Berry, G. Chem, H. Mathias and L.R. Testardi, Phys. Rev. B45 (1992) 1838. [50] S.A. Chambers and S.A. Joyce, Surf Sci. 420 (1999) 111. [51 ] P.J.H. Bloemen, P.A.A. van der Heijden, R.M. Wolf, J. aan de Stegge, J.T. Kohlhepp, A. Reinders, R.M. Jungblut, P.J. van der Zaag and W.J.M. de Jonge, Mater. Res. Soc. Symp. Proc. 401 (1996) 485. [52] F.C. Voogt, T. Hibma, P. Smulders and L. Niesen, J. Cryst. Growth 174 (1997) 440. [53] J.M. Gaines, P.J.H. Bloemen, J.T. Kohlhepp, C.W.T. BuUe-Lieuwma, R.M. Wolf, A. Reinders, R.M. Jungblut, P.A.A. van der Heijden, J.T.W.M. van Eemeren, J. aan de Stegge and W.J.M. de Jonge, Surf. Sci. 373 (1997) 85. [54] F.C. Voogt, T. Fujii, P.J.M. Smulders, L. Niesen, M.A. James and T. Hibma, Phys. Rev. B60 (1999) 11193.

325

[55] D.T. Margulies, F.T. parker, F.E. Spada, R.S. Goldman, J. Li, R. Sinclair and A.E. Berkowitz, Phys. Rev. B53 (1996) 9175. [56] D.T. Margulies, F.T. Parker, M.L. Rudee, F.E. Spada, J.N. Chapman, P.R. Aitchison and A.E. Berkowitz, Phys. Rev. Lett. 79 (1997) 5162. [57] S.A. Chambers, S. Thevuthasan and S.A. Joyce, Surf. Sci. 450 (2000) L273. [58] W. Moritz, private communication [59] J.R. Rustad, E. Wasserman and A.R. Felmy, Surf. Sci. 432 (1999) L583. [60] A.V. Mijiritskii, M.H. Langelaar and D.O. Boerma, J. Magn. Magn. Mater. 211 (2000) 278.

326

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

Chapter 8

Molecules on well-structured oxide surfaces H.-J. Freund, H. Kuhlenbeck, and T. Risse Fritz Haber Institute of the Max Planck Society, Department of Chemical Physics, Faradayweg 4-6, 14195 Berlin, Germany 1. INTRODUCTION Oxide surfaces under ultrahigh vacuum conditions are thermodynamically not well defined, due to the absence of a gas phase [1]. However, it is possible to kinetically stabilize the surfaces by forming thin films on metallic substrates, or by cleavage of single crystals in vacuum, and by keeping the temperature of the system below the threshold for oxygen vacancy formation. These surfaces can be investigated by surface science techniques. The literature contains books and several review articles on such systems [2-14]. We will focus on some specific aspects of geometric and electronic structure concerning the interaction of molecules with oxide surfaces. We discuss a variety of cases: MgO(lOO) and NiO(lOO) surfaces are examples for bulk terminated non-polar surfaces of rocksalt-type crystals with very moderate relaxations of the interlayer spacings. Cr2O3(0001) and V2O3(0001) surfaces are depolarized polar surfaces of corundum type materials with strong relaxations several layers down into the bulk. TiO2(110), RuO2(110) and V02(l 10) surfaces are intermediate cases. The V2O5(0001) surface is an example for a layered oxide structure. The electronic structure of these surfaces has been investigated by various electron spectroscopic methods resulting in a deeper understanding of the "ideal", i.e. defect "free" surfaces, giving us a reference point to address the question of defect formation and defect structure. Several adsorption studies on probe molecules such as CO, NO, CO2, H2O, H2 and O2 have been undertaken. Here we mention the remarkable differences we find for CO adsorption on rocksalt (100) and corundum (0001) surfaces. While CO stands upright - as expected - with the carbon end pointing to the substrate on NiO(lOO) it lies almost flat on Cr2O3(0001) and V2O3(0001). We show, that while the adsorption energies are similar in both cases, CO as a probe molecule reacts strongly towards differences in the electric field distributions on the surfaces. The experimental observations are complemented with theoretical

327

Rocksalt unit cell

Lattice constants for different rocksalt-type oxides

CaO: CoO: MgO: MnO: NiO:

4.8105 A 4.2667 A 4.2112 A 4.4448 A 4.1684 A

Fig. 1. Structure of rocksalt oxides. Lattice constants from [15].

considerations in order to address and rationalize details of the molecule surface bonding. Another remarkable issue concerns the adsorption and dissociation of molecular oxygen on oxide surfaces. When such surfaces are metal terminated in UHV, the dissociative adsorption of oxygen may lead to oxygen terminations, very different from bulk terminations. This is exemplified for the case of Cr2O3(0001). Special techniques in surface science to study molecular motion in adsorbed layers are a problem that has found little attention in surface science, even though it is of utmost importance. Electron spin resonance is one such method. We discuss diffusion of NO2, rotational motion of ((CH3)3C)2NO and melting of self-assembled layers of spin-labeled fatty acids adsorbed on Al2O3(0001). 2. SURFACES OF OXIDES WITH ROCKSALT STRUCTURE Fig. 1 shows the rocksalt lattice [15]. We will discuss MgO and NiO as limiting cases of oxides, one containing a simple metal ion and the other one a transition metal ion. The (100) surface of such a material represents a non-polar surface, the (111) surface represents a polar oxide surface. Since the lattice constants are very similar for both oxides (IVIgO: 4.21 A, NiO: 4.17 A) [15], we expect the surface structures to be similar. The non-polar surface exhibits a nearly bulk terminated surface as shown in Fig. 2a and it is very similar for both materials. We have put together information from LEED [16-21] and STM [22-25] analysis. There is very small interlayer relaxation and only a small rumpling of the surface atoms, whereby the larger anions move outwards and the small cations very slightly inward. A completely different situation is encountered for the polar (111) surfaces. Due to the divergent surface potential [13] on an ideally, bulk terminated polar surface, the surface reconstructs and exhibits a so

328

(P) :^0,surf ^Me,surf bulk

;g||||j.:.^^;;a^a||j.:;^l:j||^

Surface structural parameters for MgO(100) and NiO(100): Buckling Method Reference Oxide Relaxation IV-LEED <5% -3 .. 0 % R1 NiO 5±2.5 % IV-LEED 1±2% R2 MgO R3 MgO -0.56±0.35 % 1.07±0.5% GIXS

10 15 Distance (^)

20

Buckling: Ao,surf-AMe ,surf Relaxation: Abuik-(Ao,surf + AMe.surf)/2

Fig. 2a. (a): Model for buckled surfaces of oxides with rocksalt structure and some experimental results for the surfaces of NiO(lOO) and MgO(lOO) (Rl: ref. [21], R2: ref. [17], R3: ref. [34]). (p): STM image of a UHV-cleaved NiO(lOO) surface [25]. Vsampie = 1.3 V, Itunnei = 1.0 nA.

called octopolar reconstruction or even a mesoscopic facetting as opposed to the nanofacets in the case of the octopolar reconstruction [26, 27]. The structural and morphological evidences based on LEED [28], GIXD [29-31], STM [32] and electron microscopy [33] are collected in Fig. 2b. The data reported so far suggest that the d-electrons of the transition metal ion have a very small influence on structural parameters. Before wc proceed to a more detailed discussion of the binding of adsorbed molecules, a few remarks concerning the electronic structure of oxide surfaces are appropriate. Early on, Stefan Hufner and coworkers have investigated the electronic structure of transition metal bulk samples and a great deal has been learned [35, 36]. There have also been attempts to investigate the surfaces of these materials with respect to electronic structure, notably in the recent work by Noguera and IVIackrodt [37, 38]. Qualitatively, it was expected that, due to the high ionicity of some compounds, there are pronounced surface effects. Photoelectron spectroscopy has been used to experimentally verify these expectations through the detection of chemically shifted core levels. However, the shifts are not large enough to be detectable due to the relatively complex satellite structure accompanying metal core ionization [36]. Eventually, this was also rationalized via more sophisticated quantitative calculations which showed that there are several compensating contributions rendering the surface fields only slightly different from the bulk. Applying techniques allowing for higher energy resolution have then clearly demonstrated surface effects. In the

329

a) plan view

Structural parameters of the (2x2) reconstruction for NiO(111)/Au(111) and a NiO(111) single crystal.

Sis 528

has

ko

kis

r'

K2S ^38

^

^48

single crystal 0.117±0.015

thin film 0.096±0.008 0.078±0.007

0.06±0.02 -0.17±0.02 0.13±0.01 -0.029±0.007 0.23±0.12 -0.09±0.02 0.02±0.01 -0.005±0.003

-0.103±0.028

-

-

-

Fig. 2b. (a), (p): Models for the bulk-terminated and the (2x2) reconstructed N i O ( l l l ) surface. In both images the surfaces are terminated by nickel atoms, (y): Structural parameters for the (2x2) reconstructions of NiO(lll) and N i O ( l l l ) / A u ( l l l ) determined with grazing mcidence X-ray scattering [31]. The meaning of some of the parameters is schematically illustrated m the left part, ^n means a vertical displacement of the n* layer, and 5n a lateral displacement. If an s is added in the index then displacements correlated by symmetry are meant. For the single crystal surface only nickel termination was observed whereas for the thin film nickel and oxygen termination occur. (5): SEM image of the MgO(l 11) surface after annealing at 1400 K for 1 min [33]. Large facets terminated by (lOO)-type surfaces are visible (s): STM image of a thin NiO(lll) film on A u ( l l l ) recorded with a non-metalhc tip apex [32]. U = -0.3V, I = 0.5 nA. The dark part of the image is due to the (2x2) reconstruction whereas the bright part in the lower left corresponds to Au(l 11).

330

Energy loss [eV]

Fig. 3. Electron energy loss spectra of NiO(lOO) surfaces. Lower panel: clean NiO(lOO) surfaces, UHV-cleaved single crystal. The assignment of the features according to theory is given and supported by spin polarized measurements [44, 45]. Upper panel: adsorbate covered NiO(lOO) films. Lower trace: defects OH saturated, before NO adsorption; upper trace: after adsorption of NO [40].

Merz group [39] and our laboratory [40] electron energy loss spectroscopy (EELS) in the regime of electronic excitations has been used to identify excitations in the surface layer. Fig. 3 shows a set of spectra taken on NiO(lOO) surfaces. The lowest trace has been taken on a clean single crystal. The broad features peaking at 4-5 eV correspond to charge-transfer excitations crossing the band gap of the insulating NiO bulk. In the gap there are narrow features due to excitations within the delectron state manifold of the open shell Ni^^ ions. As the excitation energy within this manifold increases the number of states increases, so that near the charge-transfer band those states overlap and lead to the monotonous increase of intensity in this energy region. IVIost of the optically allowed transitions have been spectroscopically observed by transmission spectroscopy of bulk samples [41-43]. Barbel Fromme and Eberhard Kisker have performed spin-polarized EELS measurements, which allow an assignment of the spin character of all states via the control of spin-polarization in the scattering conditions [44, 45]. The assignment and a spin-polarization measurement have been superimposed on the spectra. The important point here is that there are additional spectroscopic fea-

331

Fig. 4. Correlation of structural data with electronically excited states on NiO(lOO). Lower panel: (left) coordination of a Ni ion in the bulk, (middle) coordination of a Ni ion in the clean (100) surface as well as (right) in the case of adsorbed NO. Upper panel: orbital diagram and total energies from cluster calculations [40, 46].

tures, most pronounced at 0.6 eV excitation energy which are not due to excitations in the bulk but rather in the surface layer. This can be experimentally demonstrated by an adsorption study [40]. Those excitations localized in the surface should be most strongly affected by adsorbed species. The experiment has been performed on a thin film sample because the surface has to be cooled to adsorb an appreciable amount of NO in this case. It is very obvious that the peak at 0.6 eV is influenced. In fact, it is shifted towards the position of a feature originating from an excitation in the bulk. In passing we note that the NiO(lOO) film has been treated with water before the experiments had been performed, in order to saturate the defects with hydroxyls via dissociative adsorption of water. The vibrafional losses caused by the hydroxyls are clearly visible before NO adsorption took place. NO adsorption then induces yet a further vibrational loss at lower loss energy. What is the nature of the surface excitation and why is it different from the bulk excitation? Volker Staemmler and his group have performed ab-initio cluster calculations [40, 46], the result of which can be summarized as follows: Due to the localization of the Ni-d-electrons it is sufficient to consider a single

332

Ni ion within its octahedral coordination sphere if we consider the situation encountered in the bulk (see Fig. 4). The ground state is a ^A2g state with two unpaired d-electrons in the Cg orbitals of the ligand field split set of d-orbitals. The first excited state results from an excitation from the completely filled tig subset into the partly filled Cg subset giving rise to an excitation located near 1 eV. There are many more higher excited states, some of which are assigned according to the work of Fromme et al. [44, 45]. In the surface, however, one of the coordinated oxygen ions is missing and the symmetry of the local Ni^^ site is reduced to C4V. Consequently, the degenerated Cg subset is split. The tig subset is also split, but this effect is not so important. The d-orbital of the former Cg subset pointing along the Ni-O axis has lost part of its destabilizing interaction and, consequently, its energy decreases. The calculation shows that still both orbitals in the former Cg subset are singly occupied after reduction of symmetry. Therefore, the first excited state in a Ni ^ surface ion is at lower energy than in the bulk, as also revealed by the experiment.

•6 0.010

\

0.005 S 0.000

^' ^^

-0.005 -0.010

C-0

Oxide |Ni4s

/ ^ ^ ^

27188 ^Ni3d

2K.

\ \

0 I

c Covalent bond

^5c3^00

4a ^po 3a ^ 2a ® •

,Pauli

.

|0 2p 05+ I C5-

Electrostatic interaction

Fig. 5. Orbital diagram for the bonding of CO to Ni-metal (left) and to Ni-oxide (right).

333

If an NO molecule is now coordinated to the Ni^^ surface ion, the energetic position of the orbital is raised again, (similar to the presence of the sixth oxygen ion) effectively moving the excitation energy back close to the bulk position. It is clear from these results that the surface effect on the excitation energies is of the order of 0.4 eV, and thus the above mentioned lack of evidence from other techniques can be rationalized. There has been an interesting attempt to look with EELS for the surface states at higher excitation energies at ~ 1 eV and 2.1 eV which have not been addressed here but have been mentioned by Freitag et al. [40] and discussed by Fromme et al. [44, 45]. Miiller et al. [47] have varied the thickness of the NiO layer between one and several layers and determined the EELS spectra. This study supports the previous assignment of the peak at 0.6 eV to a surface state but they cannot find evidence for a surface state contribution at 1 eV. The authors argue that this may be due to cross section limitations under the chosen experimental conditions. They do find clear evidence for a surface contribution at2.1eV. At this point we return to the question raised above, on the binding and interaction of molecules with oxide surfaces. Molecules bind to oxides via a bonding mechanism considerably different from metal surfaces. A CO molecule, for example, binds to metals via chemical bonds of varying strength involving charge exchanges [48]. Fig. 5 illustrates the bonding of CO to a Ni- metal atom via the so-called a-donation/ji-back-donation mechanism schematically on the basis of a one electron orbital diagram. The a- and Ti-interactions lead to a relative shift of those a- and 7i-orbitals involved in the bond with respect to those orbitals not involved. The diagram reflects this via the correlation lines. This may be contrasted by the electrostatically dominated interaction between a CO molecule and a Ni ion in nickel-oxide [46, 49]. There is a noticeable a-repulsion between the CO carbon lone pair and the oxide leading to a similar shift of the CO 5o-orbital as in the case of the metal atom. However, there is no or little 7c-back-donation so that the CO 7i-orbitals are not modified [50, 51]. Conceptually, the situation is transparent and one would expect that a detailed calculation reveals the differences quantitatively. However, as it turns out the description by ab-initio calculations is very much involved and today a full account cannot be given [52]. Theoredcally (Table 1) [46, 52-63], the prediction is that CO as well as NO bind very weakly to NiO [52]. The predicted binding energy of CO is of the order of 0.1 eV and it is expected to be similar to CO binding to MgO(lOO), i.e. the influence of the Ni d-electrons should be negligible [52]. To shed light on this problem it was necessary to perform thermal desorption measurements on cleaved single crystal surfaces, being the surfaces with the least number of defects [64]. In Figs. 6 and 7 TDS data for CO and NO on vacuum-cleaved NiO(lOO) are compared with data for thin NiO(lOO) films grown

334

by oxidation of Ni(lOO). At temperatures of 30 K and 56 K multilayer desorption for CO and NO, respectively, shows up. The pronounced features at higher temperatures correspond to desorption of the respective adsorbate at Table 1. Literature data for adsorption of CO and NO on NiO(lOO) and MgO(lOO).

Author

System

Method

Pacchioni and Bagus [53J Klliner [54]

CO/NiO(100)

ab-initio cluster calculation ab-initio cluster calculation, BSSE correction ab-initio cluster calculation, BSSE correction TDS, Redhead IRS, ClausiusClapeyron ab-initio cluster calculation, BSSE correction ab-initio cluster calculation, BSSE correction TDS, Redhead

NO/NiO(100)

Pohlchen and Staemmler [46]

CO/NiO(100)

Cappus et al. [55] Vesecky et al. [56]

CO/NiO(100)/Ni(100) CO/NiO(100)/Ni(100)

Staemmler [57]

NO/NiO(100)

Pohlchen [58]

NO/NiO(100)

Kuhlenbeck et al. [591 Nygren and Pettersson [52]

NO/NiO(100)/Ni(100) andNO/NiO(100) CO/MgO(100)

Chen et al. [60] Neyman et al. [61]

CO/MgO(100) CO/MgO(100)

He et al. [62]

CO/MgO(100)/ Mo(lOO)

Furuyama et al. [63] CO/MgO powder

ab-initio cluster calculation, BSSE correction DFT DFT, BSSE correction IRS, ClausiusClapeyron, TDS, Redhead IRS, ClausiusClapeyron

Adsorption energy (eV) 0.24 ^0

0.03 to 0.1

0.32 0.45 0.1

<0.23

0.52 0.08

0.28 0.11 0.43 0.46 0.15 to 0.17

BSSE: Basis Set Superposition Error. TDS: Thermal Desorption Spectroscopy. DFT: Density Functional Theory. IRS: InfraRed Spectroscopy. Clausius-Clapeyron: evaluation of pressure and temperature dependent IR-intensities with the Clausius-Clapeyron equation. Redhead: evaluation of TDS data with the Redhead equation [651.

335

L

CO on an UHV-cleaved NiO(100) single crystal

1.55

0=0.80Heating rate 1 K/s

n h in

CO on a thin NiO(100) film on Ni(100)

NO on an UHV-cieaved NiO(100) single crystal

1 11

Heating rate 1 K/s

_

0.42^^^ 0.28-^^ 0.17^^^\ 0.11\^\

]

J^^ ^§ \ jf

145 K NO on a 0=1.0^^^ ^ A NIO(100)film 0.78^^^ 1 \ onNiO(IOO) 0.45-^^^ yV \ \ Heating rate 3 K/s

Heating rate 3 K/s

0.34^^^ 02K^^^

0.12\^^yy

0

0

50

100 150 Temperature [K]

200

250

Fig. 6. Thermal desorption spectra of CO on NiO(lOO) cleaved in vacuum (upper part) and CO on a thin NiO(lOO) film grown by oxidation of Ni(lOO) (lower part). The mass spectrometer was set to mass 28 (CO). CO doses are given relative to the dose needed to prepare a monolayer.

0

50

100

^ 150 200 250 Tennperature [K]

1 300

350

400

Fig. 7. Thermal desorption spectra of NO on NiO(lOO) cleaved in vacuum (upper part) and NO on a thin NiO(lOO) film grown by oxidation of Ni(lOO) (lower part). The mass spectrometer was set to mass 30 (NO). NO doses are given relative to the dose needed to prepare a monolayer.

(sub)monolayer coverage. In the case of the CO adsorbate at 34 K desorption of the second layer is found and the states at 45 K and 145 K for CO and NO, respectively, are due to adsorption on defects as concluded from data obtained from ion bombarded surfaces (not shown here). It is obvious that for both adsorbates the thin film data and the data of the cleaved samples agree well, in particular for NiO(lOO) the thin film data are comparable to those from the more perfect surfaces of the cleaved samples. The higher defect density of the thin film surfaces leads to small, but clearly visible additional peaks in the TDS data which show up as shoulders near to the main peak, for example in the NO spectra. Nevertheless, the general shapes of the thin film spectra of both adsorbates are very similar to those of the cleaved samples. For CO the shift of the peak maximum with increasing coverage indicates that at higher coverage repulsive lateral interaction comes into play which may lead to occupation of energetically less favorable sites. This is not the case for the NO adsorbate which may be attributed to smaller lateral interactions and to the higher adsorption energy which makes adsorption more site specific, and thus may inhibit compression of the layer involving site changes.

336 Heating rate 0.2 K/s Vi

'E Z2

— 1.8

xi SL

1— 0=0.8

00 CM

(0

E 0) (0

r" 1 1M 1

o •Q.

o

(0 (U Q

0"

H—0.7

/ ) ( 1^^0.5 %-0.3

i^

it:

in

csi

20

Ead=0.14eV

40

60 Temperature [K]

80

100

Fig. 8. Thermal desorption spectra of CO on MgO(lOO) cleaved in UHV. The mass spectrometer was set to mass 28 (CO). CO doses are given relative to the dose needed for the preparation of a monolayer.

60 80 100 Temperature [K]

Fig. 9. Thermal desorption spectra of NO on MgO(lOO) cleaved in UHV. The mass spectrometer was set to mass 30 (NO). NO doses are given relative to the dose needed for the preparation of a monolayer.

TDS data for NO and CO on vacuum-cleaved MgO(lOO), for comparison, are plotted in Figs. 8 and 9. Multilayer desorption is found at 29 K and 56 K for CO and NO, respectively. The small features around 45 K and 100 K are likely due to defect adsorption since they saturate at rather low coverage. Desorption from layers with small coverage is found at 57 K and 84 K for CO and NO, respectively. The data for CO and NO on NiO(lOO) have been evaluated using the leading edge method as well as a complete analysis. Details of the procedures may be found in [66, 67]. Both methods determine the heat of adsorption as a function of the coverage of molecules already on the surface. The results of the evaluation are shown in Figs. 10 and 11. Both graphs exhibit a trend which is generally to be expected for laterally interacting adsorbate layers: the adsorption energy decreases with increasing coverage. At coverages near to one monolayer the energies converge towards the multilayer values (0.09 eV and 0.18 eV for CO and NO, respectively [68]). At low coverage the lateral interactions are most likely small so that the corresponding adsorption energies may be compared with theoretical results since in the calculations lateral interactions have not been considered. As indicated in Figs. 10 and 11, the low coverage adsorption energies are Table 2. Compilation of low-coverage bonding energies for NO and CO on NiO(lOO) and MgO(lOO) obtained in [10]

CO NO

NiO(lOO) 0.30 eV 0.57 eV

MgO(lOO) 0.14 eV 0.22 eV

337 0.40

NO on NiO(100) cleaved i o •• in UHV: TDS analysis ^ •;*j'>° 0.57(+/-0.04) eV

T

\ \

0.30

iL

CO on NiO(100) cleaved in UHV: TDS analysis

°

0.35

0.25

»u 9^ o u.oU(+/-u.Uo) ev

M°

^

r.,. •i

•

• o 1

° A V

0.0

5

••ii

o

• • ' • ' f 0 Complete analysis 1 K/s Complete analysis 0.2 K/s Leading edge 1 K/s Leading edge 0.2 K/s Leading edge 2 K/s Leading edge 0.1 K/s 0.5 1.0 Relative coverage [e/0monoiayer]

°

• o • 1.5

Fig. 10. Adsorption energy of CO on NiO(lOO) cleaved in vacuum as a function of coverage. The data have been determined from TDS spectra like those shown in Fig. 6 (upper part) using the leading edge method and complete analysis. TDS data taken with heating rates of 0.1, 0.2, 1 and 2 K/s have been used.

0.0

Complete analysis 1 K/s Leading edge 1 K/s Leading edge 0.2 K/s 0.5

1.0

1.5

Relative coverage [©/©monolayer]

Fig. 11. Adsorption energy of NO on NiO(lOO) cleaved in vacuum as a function of coverage. The date have been determined from TDS spectra like those shown in Fig. 7 (upper part) using the leading edge method and complete analysis. TDS data taken with heating rates of 0.2 and 1 K/s have been used.

0.30 and 0.57 eV for CO and NO, respectively. The low coverage adsorption energies for CO and NO on NiO(lOO) and MgO(lOO) are compiled in Table 2. According to theory the interaction of the adsorbates with MgO(lOO) and NiO(lOO) are expected to be similar since the bonding should be mainly electrostatic in nature [52] (the electric fields at the surfaces of NiO(lOO) and MgO(lOO) are similar). However, according to Table 2 the bonding energies are considerably different, with the higher values being obtained for NiO(lOO). Covalent interactions involving the Ni3d electrons may play a role for the adsorbate-substrate interaction which do not show up in the calculations published so far. As far as it concerns the basis set superposition error (ESSE) corrected calculations listed in Table 1 which are expected to yield qualitatively better results as compared to the non-corrected calculations, it appears that the theoretical results for adsorption on MgO(lOO) are in general in line with our experimental results, whereas a similarly favorable comparison can not be made for NiO(lOO). It appears necessary to re-investigate the role of the NiSd electrons in future theoretical studies. The interaction between the molecule and the d-electrons of the substrate system is more obvious in the case of NO interaction with NiO. The molecule is tilted with respect to the surface normal as is experimentally evident from X-ray absorption measurements [59] as well as photoelectron diffraction investigations

338

by the Woodruff group [69]. The tih angle is near 45-60° with the nitrogen end of the molecule pointing towards a Ni ion. The interatomic distance between N and O is near the one known from the free molecule, the separation from the surface Ni ion is determined to be 1.88 + 0.2 A. It was early pointed out by Staemmler and his group that this tilt is a consequence of the interaction of the unpaired electron on the NO moiety with the open shells on the Ni ion. A detailed discussion of the bonding mechanism can be found in ref [57]. On MgO(lOO) preliminary calculations have indicated that the NO molecule is oriented perpendicular to the (100) surface plane. This appears to be reasonable in light of the missing d-electrons of the Mg ion. There are further indications for weak chemisorptive interactions between NO and the NiO(lOO) surface: Although there is only one NO species on the surface, the Nls photoelectron spectrum exhibits a broad band with two maxima. Such phenomena are connected with the dynamics of the screening processes accompanying the formation of the core hole. Detailed studies on screening phenomena have been reported for molecules on metal surfaces, [70-72], and it is well known, that systems with weak chemisorptive interactions exhibit the most complex line shapes. For weak chemisorption systems there is a high probability to find the final hole, both on the adsorbed molecule itself or, through extra molecular screening, on the substrate, giving rise to several final states, and thus to several intense components in the spectrum. It was proposed some time ago that such phenomena are also responsible for the line shapes observed for NO/NiO(100) [59]. The above mentioned recent photoelectron diffraction investigations have clearly demonstrated that the observed components in the Nls spectrum stem from a single species. The adsorption of CO on MgO has been thoroughly investigated by Heidberg and his group using IR-spectroscopy [73] and by Weiss and Toennies using helium atom scattering spectroscopy (HAS) [74]. They have clearly demonstrated that CO develops ordered phases on the cleavage planes and that order and spectroscopic properties depend on the quality of the prepared surfaces. From their experiments the influence of the presence of surface defects on adsorption properties is very obvious but a quantitative evaluation based on the number and the nature of the defects has not been reported. The quantitative evaluation of defects is a well defined but hard to tackle problem for future studies that has to be taken on by our research community. Water adsorption is an example that lends itself to a study of the influence of defects, because at lower coverage the (100) cleavage planes of MgO and NiO do not dissociate water, while the presence of defects does induce water dissociation. Water adsorption on MgO(lOO) is the best studied system so far [64, 75-77]. The groups of Heidberg [75], Weiss and Toennies [76] have studied it with IR, LEED and HAS. While these groups have used surfaces prepared through in situ cleavage, Stimiman et al. [78] have employed bulk single crystals prepared via

339

MgO(100) sputtered 0,3Ar7Mg2kV 6,8, 17LH2O)

196 K 239 K 0.51 eV 0.66 eV (Redhead)| heating rate 0.5 K/s 150

200 250 temperature [K]

300

Fig. 12. Thermal desorption spectra of H2O on UHV-cleaved NiO(lOO) and MgO(lOO). A schematic representation of the c(4x2) structure is included (according to ref. [75]). For comparison a thermal desorption spectrum from MgO(lOO) after creation of defects via sputtering is shown.

sputtering and heating in oxygen. Their TD spectra correspond grossly to the TDS data published by Wichtendahl et al. on in situ cleaved MgO(lOO) surfaces, which are shown in Fig. 12 [64]. Data taken on a thin IVlgO film [79] reveal the essential features, although the peaks appear to be shifted significantly with respect to the temperature-calibrated data of Wichtendahl et al. The assignment of the feature at 196 K was unclear so far, although it was near at hand to assume that it is connected with the phase transition from the c(4x2) to the (3x2)pg phase which occurs at about this temperature and was reported by several groups [75, 76, 79]. It could be shown that the molecule density within the c(4x2) phase is higher than in the (3x2)pg phase (~ 1.3 atoms and -1.0 atoms per magnesium surface ion, respectively) [78] so that a transition between these phases must be accompanied by a desorption peak in TDS. The TDS data of H2O/MgO(100) in combination with data for H2O/NiO(100) taken in our group [80] shed new light upon this question. Fig 12 [80] reveals the data. The feature near 150 K is due to multilayer desorption and shall not be discussed here in more detail. The peak at 196 K, which corresponds to the transition from

340

the c(4x2) phase to the (3x2)pg phase, is observed in the spectra from both surfaces at the same temperature. This observation is a first hint that part of the molecules of the c(4x2) phase belong to the second layer and do not have contact with the oxide surface. The highest temperature peak, which corresponds to desorption of the (3x2)pg phase in the MgO(lOO) case, i.e. the monolayer, is found at slightly different temperatures, i.e. 225 K for NiO(lOO) and 239 K for MgO(lOO), because the direct interaction between water and the Ni and Mg sites, respectively, is different. This conclusion is supported by results of molecular dynamics simulation by the Besan9on group on H2O/MgO(100) [77]. Results are displayed in Fig. 13 [77]. The interesting result is that a monolayer of H2O forms with the H2O molecules oriented with the plane parallel to the oxide surface and with a network of Hbonds within that layer, but fully disconnected from the second and higher layers. This monolayer, which is to be identified with the (3x2)pg phase, completely shields the higher layers from the underlying oxide surface. Therefore, the molecules desorbing during the phase transition from the c(4x2) phase to the (3x2)pg phase may be identified as being part of the second layer, since these molecules do not feel the differences between NiO(lOO) and MgO(lOO) as documented by the data shown in Fig. 13. In closing this chapter we would like to mention that there are several studies in the literature where more complex molecules, which are also of technical interest, have been studied. These include methanol, formic acid and aromatic compounds [81-88]. We shall not dwell on these studies here but merely refer the reader to the literature.

1=150 K

T=300 K

Fig. 13. Simulation of water on MgO(lOO) at coverages higher than one monolayer at 150 K and 300 K, according to ref [77]. a) top and side views of map-shots at 150 K. b) top and side views of map-shots at 300 K. (The first layer molecules are not represented in the top view.) The side views show that the overlayer beyond the monolayer tends to aggregate.

341

3. SURFACES OF OXIDES WITH CORUNDUM STRUCTURE Fig. 14 shows the lattice structure of a non-magnetic system, i.e. a-alumina [15]. We note here in passing that there are antiferromagnetic corundum structures where the metal centers carry different spin-orientations which renders the size of the unit cell considerably larger [89-91]. The surface energies for a structure as shown in Fig. 14 has been calculated for a variety of orientations [93]. Some of these energies are compiled in Table 3. Surfaces with (0001) and (1120) orientations have been investigated with the large data set available for the (0001) orientation. We therefore restrict ourselves Corundum unit cell

Corundum(OOOI) surface

r

|Corundum(0001 )-2Me ICorundum(0001 )-Me

1 ^^

|Corundum(0001)-O

® t

o

^

d mP m ^ * ^ f T

'

#1 w

w

w

Id r f ^ ^ m m\

Lattice parameter for different corundum-type oxides oxide a c

a-Al203 4.7628 a-Cr203 4.954 V2O3 5.105 a-Fe203 5.035

20

13.003 13.584 14.449 13.72

40

line scan direction (A)

Fig. 14. Top: Schematic representation of the corundum unit cell and surface structures with different terminations. Bottom left: Scanning tunneling image of the Fe2O3(0001) surface. The darker area is believed to be metal terminated and the brighter area metal terminated according to corundum(0001)-Me [92]. Bottom right: List of lattice parameters for different oxides with corundum structure [15].

342

Table 3. Recent Surface Energies from First Principles [93]

Surface Orientations 0001 1011 1012 1120

Surface Energy 2.76 Jm"' 2.55 Jm"^ 1.97 Jm"^ 1.86 Jm"^

to a discussion of the latter case. The (0001) surface is an example for the socalled charge neutralized polar surfaces. These surfaces exhibit relatively strong relaxations remnant of the phenomena discussed in connection with the polar (111) rocksalt type surfaces, because if the (0001) surface were terminated between an oxygen and a metal layer, these surfaces would be polar with a diverging surface potential. However, the surface is terminated with only half of a metal layer present. Formally, this corresponds to a surface created by cutting the bulk structure within the buckled metal ion layers. Fig. 15 shows the results of structural determinations for the three related systems Al2O3(0001) [8], Cr2O3(0001) [94, 95] and Fe2O3(0001) [96]. In all cases a stable structure in UHV is the metal ion terminated surface retaining only half of the number of metal ions in the surface as discussed above [92]. The interlayer distances are very strongly relaxed down to several layers below the surface [10]. The perturbation of the structure of oxides due to the presence of the surface is considerably more pronounced than in metals, where the interlayer relaxations are typically of the order of a few percent. The absence of the screening charge in a dielectric material such as an oxide contributes to this efSide view Y-^i

''%

M

"^J

'^'j

"'/•

A

^ < ; ,-*% f^

^

"^^

#

" i , -'-^^ ^ ^

-

A2 A4 A6

^3 A5

AUO 2^3

CfoO 2'^3

Weak lateral relaxation

No lateral relaxation

FesOg

Fe term.

and

O term.

Fig. 15. Experimental data on the structure of corundum type depolarized (0001) surfaces (side and top views).

343

Fig. 16. Low energy electron diffraction diagram of ¥203(0001) grown on Au(l 11) [98]

feet eonsiderably. Another material belonging to this family is V2O3 which can also be prepared as a (0001) surface [97]. The structure of this system has, however, not yet been thoroughly investigated, although the LEED pattern, as shown in Fig. 16, is quite sharp [98]. In this case the V2O3 (0001) surface has been prepared as a thin film on a Au(lll) substrate by evaporation and oxidation. We assume that the surface structure is similar to the related cases discussed above. It has recently been pointed out that oxide structures may not be as rigid as one might think judged on the relatively stiff phonon spectrum in the bulk [99, 100]. In fact, at the surface the phonon spectrum may become soft so that the geometric structure becomes rather flexible, and thus also very much dependent on the presence of adsorbed species [100]. The vibration modes of a clean Cr2O3(0001) surface have been studied under ultrahigh vacuum conditions with high resolution electron energy loss spectroscopy [100]. Fig. 17 shows the spectrum of the clean surface at the bottom. A mode, which is confined to the first few atomic layers of the oxide, was detected, in addition to the Fuchs-Kliewer phonons, at 21.4 meV. This mode shows only a very small isotopic shift when the Cr2O3(0001) film is prepared from ^^62 instead of ^^02. In contrast to the Fuchs-Kliewer phonons which extend deeply into the bulk the 21.4 meV mode is very sensitive towards the presence of adsorbates, as is shown in Fig. 17. This can be seen by its attenuation upon exposure of the surface to CO, which in this case is only weakly adsorbed. Full-potential linearized augmented plane wave calculations [100] show that this mode is an in-phase oscillation of the secondlayer oxygen atoms and the Cr atoms of the two layers below. A schematic representation of the mode is shown in the inset of Fig. 17. Bulk oxide stoichiometrics depend strongly on oxygen pressure, a fact that has been recognized for a long time and we have alluded to above [10]. So do oxide surfaces, structures and stoichiometrics, a fact that has been shown again in a recent study on the Fe2O3(0001) surface by the Scheffler and Schlogl

344

HREELS

Cr2O3(0001)/Cr(110) at 90 K Ep=7.5 eV, FWHM=2.3 meV

CO exposure

1.2 L

0.6 L

.50 clean C r , 0 , 150 Energy [meV]

Fig. 17. Electron energy loss spectra in the vibrational regime of clean and CO-covered Cr2O3(0001)/Cr(110) surface. (Ep! 7.5 eV, specular scattering) The inset shows a schematic representation of the calculated normal mode of a Cr2O3(0001) surface at 21.4 meV, according to ref. [100].

groups [92]. In fact, if a Fe203 single crystalline film is grown in low oxygen pressure, the surface is metal terminated while growth under higher oxygen pressures leads to oxygen termination. This surface would be formally unstable on the basis of electrostatic arguments [10]. However, calculations by the Scheffler group have shown that a strong rearrangement of the electron distribution as well as relaxation between the layers leads to a stabilization of the system. STJVl images by Weiss and co-workers corroborate the coexistence of oxygen and iron terminated layers and thus indicate that stabilization must occur. Of course, there is need for further structural characterization. For the surface of Cr2O3(0001) infrared spectroscopic data suggest a different termination when the clean surface is exposed to oxygen [101] as will be discussed below. The electronic structure of corundum type oxide surfaces have been studied with photoelectron spectroscopy [102, 103] as well as with electron energy loss spectroscopy [104]. Henderson and Chambers [103] have recently reviewed the photoelectron spectroscopic evidences. Still, there seems to be an ambiguity about the oxidation state of the chromium ions in the immediate surface layer. XPS Cr 2p data may be to a large extent reconciled as being due to Cr(III) in the surface, because there is obviously no change in the spectra as a function of

345

emission angle. However, cluster calculations indicate that even though the number of d-electrons on the Cr ions is compatible with Cr^^, the total charge on the Cr ions is closer to Cr^^ because of a considerable hybridization between the chromium and oxygen wave functions [104]. In a previous review we have discussed the assignment of the local d-excitations on chromia surfaces and the reader is referred to this paper for details [7]. We would like to mention, however, a particular aspect of surface charge transfer excitations for the case of chromia(OOOl) which also touches upon the question of the oxidation state of surface chromium ions. The charge transfer states often mark the edges of the band gap in transition metal oxides, and they are also modified by the presence of the surface. Unfortunately, NiO does not show this effect very clearly. The electronic excitations at the chromia(OOOl) surface, on the other hand, are good examples to discuss the phenomenon. Fig. 18 shows the EELS spectrum [105-107] of this surface at low temperature. Again, a thin film has been used. The sharp features in the band gap are excitations within the manifold of d-orbitals. A detailed discussion [107] has shown that the excitations are characteristic of surface Cr ions with three delectrons. However, the Cr ions do not carry a net charge of 3+ as expected (and found for the bulk Cr ions) but rather of 2+ charge due to strong hybridization with the neighboring oxygen atoms. When we now perform EELS measurements after adsorption of CO2, not only the d-excitations are influenced but also the very intense feature near 4.8 eV. Again, on the basis of cluster calculations performed in the Staemmler group [104], it has been possible to assign this intense feature to a surface charge-transfer excitation at the band gap, which is shifted to lower energy as compared with the corresponding excitation in the bulk (see Fig. 18). The decrease in energy is reasonable because the Cr d-orbispecular I detection lE„=100eV|

ind gap

i 20 L CO2 IT=300 K

\ clean 1

2

'

I

4

'

I

6

I

'

I '

I ' I

10 12 14 16

Energy loss [eV] Fig. 18. EELS spectra of the clean and adsorbate covered Cr2O3(0001) surface. Peak assignments are indicated [105-107].

346

6 L ^''O2/Cr,O3(0001) oII o

^V^^n^^^.W^-^-.V.-.

..^.^-v.

415 K

£ 1 0)

o 150 K 90 K 1

800

'

1

'

1

'

1

'

1

'

1

'

r

1000 1200 1400 Wavenumbers [cm'']

Fig. 19. IRAS spectra after adsorption of a saturation coverage of O2 on Cr2O3(0001)/Cr(l 10). The spectrum at the bottom was taken at 90 K. Other spectra are taken at the temperatures indicated [101].

tals have been lowered, the more open surface structure (see Fig. 15) and the coordination of the Cr ions to only three oxygen ions allows for better charge separation than in the bulk. In NiO the effect is less pronounced because the surface Ni ions are still five-fold coordinated. The analysis of the electronic structure of Cr2O3(0001), as discussed so far, has been performed at 90 K. We note that upon increasing the temperature, the structure of the surface changes [107], and we have speculated that these changes are connected with changes in the magnetic structure of the surface. Oxide surface magnetism is a field that needs to be explored in the future. Before we discuss the adsorption and bonding of molecules such as CO and NO to the corundum type surfaces in order to accentuate the differences with respect to the rocksalt type surfaces, we would like to comment on the chemistry of molecular oxygen on the corundum type surfaces. As the (0001) corundum type surfaces are depolarized polar surfaces we always have to define whether we are talking about the metal terminated or the oxygen terminated surface. If not mentioned specifically we always assume metal termination. Chromia when exposed to oxygen leads to a very interesting chromyl terminated surface. What do we know about the chemistry between molecular oxygen and chromia(OOOl) that leads to this final result? Figure 19 shows the appearance of a sharp line in infrared spectra which develops out of a broad feature after heating the surface to room temperature. As has been discussed earlier and also supported by isotopic labeling experiments by Dillmann et al. [101] chromyl groups are formed on the surface and it was suggested that the chromium ions in the surface layer of the clean surface bind oxygen to form such chromium-oxygen double bonds. As is obvious from Fig. 19 there are adsorbed oxygen precursors to the chromyl

347

(TDSl m/e = 34

^^02, ""^Os mixture (50:50) / Cr2O3(0001)/Cr(110) \

r-~JL \r~JX

'_ A _A

rv 100

1x10-"'^ mbar

\h

Z

1x10"^ mbar

I

3

°-^'"

0.1 L

10L

^ /V r^ 100

Temperature [K]

300

rn/e = 32 500

Fig. 20. Thermal desorption spectra of oxygen desorbing from Cr2O3(0001) [101].

formation. The width as well as frequency is indicative of a O2' species being adsorbed on the surface in a more or less upright position [101]. Fig. 20 is indicative of the thermal chemistry going on the surface. The TDS data reveal that around room temperature most molecular oxygen is lost from the surface. This molecular oxygen does not result from a recombinative desorption as can be seen directly from the isotopic labeling experiments. There is basically no isotopic mixing beyond the originally present contaminants. Above room temperature there is a very high temperature peak due to recombinative desorption stemming in all likelihood from the chromyl oxygen present on the surface at elevated temperatures, and observed in the infrared spectra. We will see further below examples of how these chromyl groups influence the chemistry of other adsorbed species on the surface. CO adsorption on corundum type (0001) surfaces is interesting because, opposite to checker board rocksalt type surfaces, on the (0001) surface of Cr203 CO has been shown by angle resolved photoemission to assume a strongly inclined almost flat adsorption geometry. Infrared spectroscopy and recent cluster calculations by Staemmler's group have revealed the reason for this [108]. Briefly, on the basis of TDS data the activation energy for the desorption of CO from chromia(OOOl) is estimated to be about 45 kJ/mol. So the appropriate state may be termed weakly chemisorbed. In addition, more weakly bound CO molecules (28 kJ/mole) which are present on the surface at lower temperature have been detected. Multilayer adsorption of CO does not occur on Cr2O3(0001)/Cr(110) at 90 K. IR-spectra confirm the presence of different CO species on Cr2O3(0001). Fig. 21 shows a series of IRAS-spectra as a function of initial CO dosage at 90 K. Using a CO-exposure of 0.05 L we detect a signal of the CO-stretching vibration at 2178 cm"^ and at 0.1 L a second signal appears at 2165 cm'\ With increasing coverage the high-frequency signal considerably

348

increases and slightly shifts to lower wave numbers. The weak signal at 2165 cm'^ does not gain further intensity. Instead it disappears in the shoulder of the main signal. This state is to be correlated with the chemisorbed state observed in TDS. Finally, at exposures above 3 L a third signal appears at 2136 cm"\ which is due to the more weakly held species. Cluster calculations and TDS measurements agree fairly well as far as the adsorption of CO at the Cr2O3(0001) surface at low coverages is concerned. The chemisorbed species corresponds to an undissociated CO molecule residing in the Os-hollow position with the CO axis strongly tilted against the surface normal, oriented along a line connecting two surface Cr ions and in a hollow position with respect to the O^" ions in the second layer. The more weakly held species corresponds to a shallow local minimum in the potential energy curve, in which CO lies nearly flat on the surface, also between two surface Cr ions, but on top an O^" in the second layer (0-ontop position). In contrast to the adsorption geometry there is some discrepancy concerning the adsorption energies as calculated by the cluster calculations. The values of 45 and 28 kJ/mol estimated for the two TDS peaks from the Redhead formula are considerably larger than the calculated values of about 27 and 15 kJ/mol. As has been pointed out in connection with NiO, it has to be expected that the calculated values are too low since neither adsorbate-induced relaxations of the IRAS I

CO / Cr2O3(0001)/Cr(110)

90 K

2178

1 1900

1 2000

1 2100

1

\

1

1

2200

2300

2400

2500

Energy [cm'']

Fig. 21. FTIR spectra of CO adsorbed on Cr2O3(0001) at 90 K as a function of exposure [108].

349

Cr2O3(0001) surface nor extra cluster van der Waals attractions are accounted for. Whether the inclusion of these effects will lead to a perfect agreement with the measured values is not clear. In any case it seems justified to call the O3hoUow minimum a „chemisorption" though the dominant contribution to the interaction energy has electrostatic origin and is characterized by the situation in which the quadrupole moment of CO fits into the strongly inhomogeneous electrostatic field above the Cr2O3(0001) surface. Good agreement between theory and experiment exists concerning the frequencies shifts of the CO stretching vibration. The calculations predict a blue shift of about +22 cm'^ for the CO in the Os-hollow position, which compares favorably with the observed shift of 30-35 cm'^ in the most prominent IRAS band. For the weakly bound CO in the O-ontop position a small red shift of about -5 cm"^ is obtained; the signal at 2136 cm'^ observed at higher coverages for the weakly bound CO species does indeed exhibit such a small red shift of7 cm"\ However, the ab-initio calculations cannot offer an explanation for the shoulder in the IRAS peak at 2165 cm'^ observed for dosages in the range between 0.1 and 0.8 L (Fig. 21). The frequency itself seems to correspond to that of CO in the Cr-ontop position, but this does not constitute a local minimum and will hence be not occupied. We also might speculate that the shift observed in the main IRAS peak from 2178 cm"^ at low coverage to 2170 cm"^ at high coverage is due to lateral CO interactions, most probably because each CO molecule is slightly squeezed out of the absolute minimum by neighboring CO molecules. The results of the present study should also be interrelated with UPS and NEXAFS experiments which have been performed on the CO/Cr2O3(0001)/Cr(110) system previously [105, 106]. Whereas the present data allow the identification of strongly and weakly bound CO molecules, only the more strongly adsorbed species has been observed in the photoelectron spectroscopic studies. The explanation may be found in the experimental conditions: At the slightly elevated surface temperature (105 K) used in those experiments and because of UV- and X-ray-induced CO-desorption the weakly bound CO species could not be detected. (In fact, the UPS and NEXAFS measurements had to be performed at a background CO pressure of 1 x 10"^ mbar.) A very interesting point is the adsorption geometry of the CO-species on Cr2O3(0001)/Cr(l 10). On the basis of the ARUPS and NEXAFS data [105, 106] the tilt angle between surface plane and molecular axis of the chemisorbed CO was estimated to be about 20°. The present IR-data are consistent with an adsorption model of a strongly tilted CO-chemisorbate, in line with the cluster calculations. On the other hand, the IR-signal of the weakly bound CO-species (at» 2136 cm'^) shows a very small intensity compared to the relative population of states as deduced from TDS (not shown) [108]. Neglecting possible differences in the strength of the dynamic dipoles we can conclude that for the

350

weakly bound CO molecules the tilt angle between surface normal and molecular axis is even larger than for the chemisorbed CO: the weakly bound CO lies almost flat on the surface. This conclusion is confirmed by the cluster calculation as well. It is very interesting to note that CO adsorption on ¥203(0001) seems to show a very much related behavior although the study of orientation of the molecule is not complete yet [98]. Water adsorption has been studied for Fe203 as well as for Cr2O3(0001) surfaces by Henderson and Chambers [103, 109, 110]. In general, both, dissociative adsorption as well as molecular water adsorption has been observed. Fig. 22 shows TDS data for chromia [103]. The low temperature peak corresponds to molecular physisorption, including multilayer formation (peak at 165 K), while the higher temperature peaks are consistent with, partially, recombinative desorption of chemisorbed water from the surface. The two states at 295 K and at 345 K are populated sequentially, starting with the high temperature state at the lowest exposure. Vibrational spectra revealed the presence of isolated hydroxyl groups. These are associated with the peak at 345 K. In fact, Henderson and Chambers [103] proposed that water in the dissociated state is comprised of two distinct hydroxyl groups, i.e. a terminal and a bridging hydroxyl group. Molecularly chemisorbed water desorbs at 295 K. These authors also mention the strong influence of a predosage of the surface by oxygen on the surface

100

200

300 Temperature [K]

400

Fig. 22. Thermal desorption spectra of water desorbing from CriOsCOOOl) after various exposures [103].

351

chemistry. Their observations are in line with results on the adsorption of CO2 described briefly below [111, 112]. Water adsorption on alumina leads to molecular adsorbates [110]. There are some reports on the formation of isolated hydroxyl groups on regular sites [113, 114]. Whether they are the results of dissociative water adsorption is not completely clear, although there is a recent theoretical paper that predicts the dissociation of water on Al2O3(0001) by protonation of a surface oxygen, while the molecule previous to dissociation resides on a cation site [115]. There is no experimental verification of this result as yet. There are, however, explored routes to hydroxylate alumina surfaces either by exposure to atomic hydrogen or by a combination of aluminium deposition and subsequent hydrolyzation with water [116]. Carbondioxide adsorption is another example of molecular adsorption which has been studied quite extensively in the past. While for some time is was generally accepted that CO2 forms carbonates with chromia surfaces very readily upon exposure, it was recently observed that carboxylate species may be formed. TDS spectra indicate [111, 112] that there are more weakly and less weakly bound CO2 species on the surfaces. We have studied the nature of those species by various techniques including infrared spectroscopy. Fig. 23 shows several sets of IR spectra. The pair of sharp bands around 2300 cm"^ can easily be assigned to the more weakly bound CO2 with only slightly distorted structure Identification of adsorbed species

Comparison with microcrystalline samples CO2 on Cr2O3(0001)/Cr(100)

'^COp

1282 1313

"'135]W1342" 190 K

1289

"Isoi^ 342

300 K

12691 ^2002"^

il

250 K 90 K

CO2 on polycrystalline a-chromia

^

^3002^"

15i3|^

90 K

1^89

«^„

825 ° ' ^

-

1242 16251695

2344I2358

D. Scarano, A. Zecchina, Universlta Degli Studi Di Torino

CO2 on O/Cr2O3(0001)/Cr(100) s^T k.

90 K

si

1

+ 4 L CO2 at90K

2352

1289 [2375 T — I — I — I — I — I — I — r 1000 1200 1400 1600 1800 2000 2200 2400 energy [cm'']

T CO2 on oxygen precovered chromia

" ~ ^

D. Scarano, A. Zecchina, Universita Degli Studi Di Torino

0.1 t o r r 0 2 at 100 K

—1—1—1—1—1—1—1—1—I—r1200 1600 2000 energy [cm''']

Fig. 23. IRAS spectra of CO2 adsorbed on Cr2O3(0001) surfaces and on polycrystalline chromia, left panel: IRAS spectra at different surface temperatures and with isotopically labeled CO2 as well as Cr203, right panel: adsorption of CO2 after pre-adsorption of oxygen [112].

352

as compared with the gas phase species. By a combination of isotopic labeUng the adsorbed CO2 (shift of frequencies) as well as the oxide layer (no shift of CO2 bands) we have demonstrated that the single band centered around 1300 cm'^ is due to the presence of a carboxylate species, i.e. a bent anionic CO2 species, and not, as perhaps expected, to a carbonate [105]. The bands between 1610 and 1700 cm"^ are missing because of the applicability of surface selection rules in thin film systems. This means, all nontotally symmetric bands are suppressed in intensity. A quick comparison with CO2 adsorption on chromia microcrystalline material as shown in Fig. 23 indicates the presence of the bands between 1610 and 1700 cm"^ as expected for adsorption on a bulk dielectric material. It is remarkable how similar the thin film data are in comparison with the microcrystalline material. This has been discussed in detail by Adriano Zecchina's group [117]. Also, the response of the two systems with respect to preadsorption of oxygen is very similar. In fact, as shown in Fig. 23, CO2 adsorption in form of the less weakly bound CO2" is fully suppressed on the thin film system and very strongly attenuated for the microcrystalline system. This indicates that CO2 occupies the chromium sites, because we know that oxygen from the gas phase adsorbs on the chromium ions. As we remarked above, ELS [106] and XPS [102] spectra of the Cr2O3(0001) surface have been used to deduce that the Cr-ions in the surface are in a low oxidation state, i.e. Cr^^ as opposed to chromium ions in the near surface and bulk regions. It is therefore not surprising that such a surface provides electrons to adsorbed molecules, leading to electron transfer as, for example, documented by the formation of O2" and CO2". The low valence state of the Cr surface ions also has consequences in other reactions, such as the polymerization of ethene which has been studied on Cr2O3(0001) [118], and in connection with other model studies [119]. A tri-atomic molecule that has so far not found wide attention is NO2, although the molecule is interesting with respect to environmental catalysis [120]. We have studied the interaction of NO2 with alumina films [121]. Fig. 24 shows TDS data for this system: The dosages given in Fig. 24 are expressed in terms of the rise of the base pressure for 10 s at a fixed distance between doser and surface. By multimass TPD measurements it has been checked that NO2 is the only desorbing species. For small coverages a single desorption peak is found at 135 K which saturates at higher exposures. This first desorption peak can be assigned to isolated NO2 molecules physisorbed to the oxide with a binding energy of 38 kJ/mol estimated according to Redhead's analysis [65]. At higher coverages a second desorption feature at nearly 10 K higher temperature appears. The observation of a TPD peak at higher temperature which does not saturate upon increase of exposure is quite unusual compared to the "normal" desorption behavior of, e.g., N02/Au(l 11) published by Bartram and Koel [122]. It indicates an intermolecular interaction energy larger than the binding energy from the

353

TPD

N02/Al203(l 1 l)/NiAl(l 10)

(m=46)

S-IO"'^ mbar A*
LJ!U-_

MO"'' mbar

L--Mr^\»,^Vv^

4-10"^° mbar

100

200 300 Temperature [K]

Fig. 24. Thermal desorption spectra of NO2 desorbing from AI2O3 after exposure to various background pressures at 50 K.

molecule to the substrate. This desorption with higher binding energy may be due to two- and three-dimensional molecular aggregates, respectively. Obviously, there is not a strong interaction of the molecule with the substrate rendering the surface chemistry in this system rather simple. However, due to the radical nature of NO2 this system offers unique possibilities to study molecular motion of adsorbed molecules by applying electron spin resonance (ESR) spectroscopy under UHV conditions [121, 123]. We have developed this technique to be applied under UHV conditions to the systems addressed here and we refer the reader to the literature for technical details of the experiment [123]. Since ESR spectroscopy is a widely applied technique in chemistry we assume that the reader is familiar with the general ideas of ESR spectroscopy, although we realize that in the surface science community this may not be the case [124]. Fig. 25 shows ESR spectra for the system NO2/AI2O3. Before we go into details of the spectra in Fig. 25 we try to explain the characteristics of the recorded ESR spectra. There are two main contributions to the habit of the spectra. First, the unpaired electron resides in a particular electron orbital of the molecule with a specific spatial orientation, which in turn leads to a dependence of the electron spin resonance on the orientation of the molecule in the external magnetic field. In other words, the isotropic g value of the free electron is transformed into an anisotropic g tensor, which gives rise to a splitting of the resonance lines. The hyperfine interaction of the unpaired electron with the nuclear spin ( 1 = 1 ) of the nitrogen atom (^"^N) as the second contribution may be described by the A tensor and finally leads to an appearance of the spectrum, ex-

354

hibiting three well-separated parts. The spectra are presented as the first derivative of the microwave absorption as a function of the external field. In Fig. 25 coverage dependent ESR spectra of NO2 on AUOBCIH) are shown. The main observation is the increasing line width AB of the spectra with increasing coverage. Assuming the line width to be due to dipolar interacfion (AB - 1/r^) of the unpaired spins of neighboring molecules [125], it is obvious from the increase of the line width that the distance of the NO2 molecules decreases. If the sample is annealed at different temperatures, as shown in the lowest spectrum in Fig. 25 and in the inset, a drastic decrease of line width and intensity after annealing to temperatures above 60 K is observed, although temperatures are well below the desorption temperature of 135 K. At about 100 K the decrease of intensity causes the spectrum to vanish below the detection limit within a few minutes. The decreasing of the concentration of NO2 is caused by the reaction of paramagnefic NO2 monomers to the diamagnetic N2O4 dimer [126]. Because at low temperatures the equilibrium of this reaction completely favors the dimer, there must be a kinetic restriction causing the molecule not to dimerize at low tem-

40

80 0.05 ML 0.15 ML

0.5 ML

5

r^A-v^^^ 3.3

3.4

ML

0.5 ML (annealed at 105 K)

3.5 B[kG]

Fig. 25. Electron spin resonance spectra of NO2/AI2O3 adsorbates at different NO2 coverages. The bottom trace was recorded after 0.5 ML coverage had been annealed at 105 K. The inset shows the line widths as a function of temperature [121].

355

peratures. We believe the onset of the reaction is coincident with the onset of translational diffusion of the NO2 monomers on the surface. If the anneahng procedure is repeated or continued over larger time intervals the decrease of the line width stops at a certain value (see the inset in Fig. 25), but the signal intensities continue to decrease below the limit of noise. At very low coverages no decrease of the line width can be observed, because it has reached its lower limiting value even after adsorption. This value may be assumed when intermolecular distances between the monomers become so large that the contribution of the dipolar interaction to the line width can be considered to be small as compared to other contributions. Then the only effect of the dimerization process is decreasing the spectral intensity. We compare computer simulations with the spectra in Fig. 25. The model for fitting these spectra is a three-dimensional random distribution of static molecules yielding a so-called powder pattern. A detailed description of the fitting procedure is given in [127, 128]. The simulations match the experimental spectra except for some small differences. ESR spectra of a submonolayer coverage of NO2 were taken at different temperatures by warming the sample up step by step before recording the spectra. As mentioned above the signal intensity decreases within a few minutes at about 100 K, rendering it impossible to obtain spectra with a reasonable signalto-noise ratio. However, it is possible to deduce an estimate for the diffusion coefficient D from this time behavior. Assuming the dimerization to be a diffusion controlled reaction, the theoretical investigation of Freemann and Doll [129] provides a relation between the rate constant observed and the diffusion coefficient. An evaluation in this framework yields a value of D-10"^^ cm^/s. The estimates based on coverage and spectral intensity versus time are quite crude; however, the calculated diffusion coefficient is remarkably small, compared to those measured, for example, for CO/Ni(100) [130] (10"^ - 10'^ cm^/s). The analysis so far has been restricted to translational motion. While the detailed analysis of the spectra, in particular at low temperature, also reveal details of the other aspects of molecular dynamics, i.e. rotational motion, the low temperature onset of dimerization in the case of NO2 obscures this to some extent [131]. It is therefore favorable to investigate radicals which do not dimerize. A typical example is the molecule DTBN (di-tert-butyl nitroxide) which represents basically an NO molecule carrying two bulky hydrocarbon side chains at the nitrogen end which keeps it sterically from dimerizing (see inset in Fig. 26). The adsorption behavior of this species is relatively complex and has been discussed in detail elsewhere. Briefly, there are three different binding modes of the molecule to the surface, only one of which is ESR active, i.e. still carrying an unpaired electron. The dynamics of this species, which represents about 10% of a monolayer, has been studied with ESR. The molecule is bound to an aluminium cation site via the oxygen lone pair and resides in an environment established by a large fraction of molecules bound with both nitrogen and oxygen

356

3300

B[G]

3400

Fig. 26. Electron spin resonance spectra of a DTBN (structure see inset) adsorbate at various temperatures [131].

end to the surface, being however, non ESR active. Therefore the ESR active molecules are well diluted on the surface and exhibit well resolved ESR line shapes as shown in Fig. 26. Here the temperature dependence of the species has been recorded. Below 200 K a rigid-limit ESR spectrum as shown in Fig. 26 is obtained at 40 K. The parameters gained from this spectrum (g and A tensor components) form the basis for the simulation of spectra measured at higher temperature including the effects of molecular dynamics. Above 200 K changes in line shape as shown in Fig. 26 reveal the onset of molecular dynamics (the simulations (solid lines) are carried out assuming a particular motional model, the "moderate jump" model, which is discussed below.) The line width has dropped to about 12 G, which is known to result from intramolecular interactions manifested as unresolved proton hyperfme splittings [132]. The dipolar interactions are now averaged out by the molecular dynamics. The remarkable intensity behavior (increasing intensity at higher temperature) is caused by the reversible conversion of DTBN molecules from the ESR-inactive to the ESR-active state and is discussed elsewhere [133].

357

Four different models for the molecular dynamics have been tested to simulate the experimental spectra. Brownian rotational diffusion and jump type diffusion [134, 135] have been used for this analysis, both in their pure forms and in two mixed models. Brownian rotational diffusion is characterized by the rotational diffusion constant D and jump type motion by a residence time x. The motions have been assumed to be isotropic. In the "moderate jump" model [135], both Brownian and jump type contributions to the motion are coupled via the condition Dx= \, The dynamics of the adsorbed DTBN molecules have two components: first, the jump contribution, which may be taken to represent the fast switching between the six equivalent positions around an aluminum binding center at the surface as shown in Fig. 27; second, an additional tumbling contribution which cannot be caused from collisions with gas phase molecules, since the experiment takes place under UHV conditions. The high stability of the adsorbate under UHV conditions also rules out a translational diffusion of the DTBN molecules. The tumbling involves two other motional degrees of freedom for the DTBN molecule as a whole: a rotation around the binding axis and a vibration against the surface. If both motions are of low amplitude, they may be reflected in the additional Brownian contribution to the jumping motion. The low-amplitude tumbling is traceable to the bulky tert-butyl groups which interact via van der Waals forces with the surface and restrict both movements to small angles. The rotation around the binding axis can also be identified as illustrated in Fig. 27. The vibration of the molecule against the surface can be seen as a hindered rotation of the molecule around the aluminium center as also shown in Fig. 27. ESR spectroscopy has also been used to study dynamics in more complex systems on oxide surfaces, i.e. self-assembled molecular (SAM) films [136, 137]. Fig. 28 shows ESR spectra taken for spin labeled stearic acid, namely ndoxyl-stearic acid (n-DXSA) immersed onto a thin alumina film. These nitroxides, with an oxazolidinyl ring as paramagnetic group connected to different positions of the aliphatic chain, are well known as paramagnetic probes in the study of natural and synthetic membranes. The spin labeled molecules are pres-

a) jump type motion

\\

...^A,:;^...

^^^4|^^vibration b) Brownian type motions

^ ^ #

fJ o ^ r o t a t i ^ n

Fig. 27. Schematic representation of motions of DTBN on AI2O3.

358

(b)

370 K

300 K

190 K

60 K

3300

B[G]

3400

3300

B[G1

3400

Fig. 28. Electron spin resonance spectra of 5-doxyl-stearic acid (a) and 16-doxyl-stearic acid (b) on AI2O3 as a function of temperature [136, 137].

ent only as 5 % of a monolayer and are diluted with non spin labeled stearic acid molecules. Fig. 28a shows the measured temperature dependence of a SAM where the spin label is close (n = 5) to the anchoring carboxyl group of the chain to the oxide surface, while in Fig. 28b the corresponding spectra for a film with the spin label in n = 16 position, i.e. near the end of the aliphatic chain close to the vacuum interface are presented. The line shape of the spectra in Fig. 28a for temperatures up to 200 K remains unchanged whereas the intensities decrease according to Curie's law. Increasing the temperature to values higher than 200 K results in a decrease of the line width (see, e.g., the spectra at 60 and 300 K in Fig. 28a). However, the positions of the characteristic spectral features remain unchanged. This behavior may be interpreted in terms of very slow motion of the spin label but it is difficult to understand, because it is known from the literature [138] that the line width of the spectra should increase first before changes in line positions occur. An alternative mechanism for line broadening at low temperatures is inhomogeneous broadening due to unresolved proton superhyperfine interactions. Apart from an increasing line width, this mechanism should change the line shape from Lorentzian to Gaussian [139]. This can be shown to be the case [140]. The solid lines in Fig. 28a represent computer simulations of the spectra assuming a rigid, randomly oriented ensemble of spin labels. Whereas the spectrum at 300 K is computed with Lorentzian lines, the spectra at 200 and 60 K are calculated

359

using Gaussian line shapes. The change from Lorentzian to Gaussian line shape is necessary in order to fit the rigid limit spectra with a single set of g- and Atensors. This indicates that the changes in the line width above 200 K are due to the motion of the protons in the surroundings of the spin label. On subsequent raising of the temperature higher than 315 K, the line shape changes again. But now the changes of the line shape, namely, the growth of a structure between the two maxima of the spectra, are indicative of the motion of spin labels connected to the alkyl chains themselves [141]. Turning now to the spectra in Fig. 28b with the spin label far away from the oxide surface we compare the spectrum recorded at 60 K with the corresponding spectrum in Fig. 28a. There are no pronounced differences. This behavior is expected because both spectra reflect an ensemble of static molecules. The situation changes at higher temperatures: Whereas the spectrum of 5-DXSA at 200 K could be described as a rigid limit spectrum with Gaussian line shape due to unresolved proton superhyperfme interactions, the equivalent spectrum of 16DXSA is a rigid limit spectrum as far as the spin label is concerned, but the line shape is almost Lorentzian. The onset of the protons motion for the 16-DXSA is approximately at 140 K. This is in good agreement with helium atom scattering experiments of Scoles et al. [142] who observed a loss of the ordered structure above 100 K and attributed this effect to the motion of the terminal alkyl groups. The two high-temperature spectra clearly indicate the motion of the spin label itself Compared to the 5-DXSA films the motion is considerably more excited. In an attempt to understand the dynamics in terms of liquid-like motion, the increase of the rotational constants should lead to a shift of the whole structure at the high- and low-field extrema toward the isotropic values. There is still more information in the spectra which we will not dwell upon at this point. In a recent detailed article simulations [137] including the various degrees of freedom of the chains and the constituting groups have been reported. 4. SURFACES OF OXIDES WITH RUTILE STRUCTURE Fig. 29 shows the volume structure of rutile [15]. The sizes of the unit cells of Ti02, VO2 and RUO2 are given in Fig. 29. Probably, the best studied clean oxide surfaces are the TiO2(100) and TiO2(110) surfaces. A STM image of the clean (lxl)TiO2(110) surface taken by Diebold and her group is shown in Fig. 30 [143]. It is noteworthy, that one of the first atomically resolved images of this surface was reported by Thornton and his group [144]. The inset shows a ball and stick model of the surface. There is now accumulating evidence from theoretical modeling of the tunneling conditions, but also from adsorbate studies using molecules which are assumed to bind to the exposed Ti-sites, that the bright rows are representing Ti atoms. Iwasawa and his group [145-148] have successfully used formic acid in such a study, and showed in line with the theoretical

360 Rutile unit cell

Lattice parameter for different rutile-type oxides oxide Cr02 M0O2 RUO2 Sn02 Ti02 VO2 WO2

a 4.41 4.86 4.51 4.74 4.59 4.55 4.86

c 2.91 2.79 3.11 3.19 2.96 2.85 2.77

Fig. 29. Schematic representation of the bulk rutile unit cell (lattice parameters are given in the inset for different materials) [15, 150]

predictions, and counter-intuitive w^ith respect to topological arguments, that the Ti-ions are imaged as bright lines and the oxygen rows as dark lines. Taking the resolvable interatomic distances within the surface layer the values correspond to the structure of the charge neutral truncation of the stoichiometric (110) surface [149]. Interatomic distances normal to the surface, however, are substantially different from the bulk values as is revealed by X-ray scattering experiments [149]. The top layer six-fold coordinated Ti-atoms move outward and the five-fold-coordinated Ti-atoms inward. This leads to a rumpling of 0.3 ±0.1 A. The rumpling repeats itself in the second layer down with an amplitude of about half of

Max. relaxation -11.9%

Fig. 30. Structure of the TiO2(110) surface as determined via STM (left panel) [144] and grazing incidence X-ray scattering (right panel) [149].

361

that in the top layer. Bond length variations range from 11.9 % contraction to 9.3 % expansion. These strong relaxations are not untypical for oxide surfaces and had been theoretically predicted for quite a while [8]. Recently, STM images from the Ru02(l 10) surface, prepared as a thin film on a Ru(OOOl) surface, have been reported [151]. The results are shown in Fig. 31. Here the bright rows are the oxygen atoms as demonstrated by adsorption of CO, which resides between the bright rows, indicating that in the case of Ru02 the contrast is reversed with respect to Ti02. A clear reasoning for the differences has not been given so far. In addition to STM the structure has been analyzed through a LEED I/V study. The surface appears to be basically bulk terminated with smaller relaxations as compared to the Ti02(l 10) case. Over et al. have also started to prepare other low index surfaces and LEED analysis are under way [152]. For V02(l 10) grown epitaxially on a Ti02(l 10) surface, structural data based on photoelectron diffraction are available from the Granozzi group [153-157]. It seems that the V02(l 10) surface is terminated similar to the Ti02(l 10) case. Molecular adsorption has been extensively studied on TiO2(110), but very little on the corresponding Ru02 and VO2 surfaces. Interestingly, there is very scarce recent information on CO and NO adsorption even on Ti02(l 10). CO adsorption has been imaged as mentioned above for RUO2 [151]. The molecules are residing on the Ru ions between the protruding surface oxygen rows and the binding energy is 1.2 eV. The molecules are immobile at 300 K and can be imaged with the STM. A LEED structure determination yielded a C-Ru distance of 1.95 A and a C-O bond length of 1.13 A. Here the bonding is thought to be in accord with the Blyholder model (Fig. 5). Upon desorption carbon dioxide is formed leaving surface oxygen vacancies behind. The oxygen is taken from the

Fig. 3L Left: STM image of RuO2(110)/Ru(0001) (U = -L21 eV, I - 0.46 nA). Right: the same surface after exposure to 10 L CO (U = -1.08 eV, I = 0.46 nA). From ref. [151].

362

lattice and can be replenished by exposing the surface to molecular oxygen. This process of oxygen uptake by dissociating molecular oxygen from the gas phase was well known before from studies on Ti02(l 10), where exactly the same process applies [158]. In the latter case the extra oxygen adatoms are placed on the fivefold coordinated Ti ions. Water adsorption has been studied quite extensively in the past, both on the TiO2(100) and the TiOiCllO) surfaces [159, 160]. There is substantial evidence that water dissociates easily on the TiO2(100) surface leading to the formation of hydroxyl groups [161, 162]. On the TiO2(110) surface studies by Henderson [159] and by Hugenschmidt et al. [163] suggest that only a small fraction (from 1 % to 25 %) of the water molecules dissociates. Fig. 32 shows TDS data from Henderson's work [159]. There are basically three states of water on the surface. The low temperature peak is likely to be associated with thicker ice layers, while the peak at 175 K is due to water molecules in the second layer. The monolayer shows up as the peak shifting from close to 300 K to 270 K as a function of coverage. Henderson [159] also investigated the vibrational properties of adsorbed water allowing the identification of monolayer and second layer water molecules through the shift of the OH stretching frequencies. Interestingly, the data suggest that the protons on the monolayer are not involved in hydrogen-bonding second layer water molecules. Engel and his group [160] have recently corroborated via molecular beam studies, that, in fact, water does not dissociate on the regular TPD of water adsorbed on TiO2(110) at 135 K

O

150

200

250

300

350

400

Temperature [K]

Fig. 32. Thermal desorption spectra of water desorbing from TiO2(110) after various exposures. The inset shows an enlarged plot of traces determined for low water exposures [159].

363

Fig. 33. Schematic representation of the geometry of adsorbed formate on TiOiCl 10) [165].

Ti02(l 10), and that the small fraction of dissociating molecules reside on defect sites. Another widely studied adsorption system on Ti02(l 10) is related to formic acid [164-166]. Upon exposure of the Ti02(l 10) surface to formic acid, strongly bound formate is formed, and this formate has been shown to reside on the Ti ions as already mentioned above. The protons form surface hydroxyl groups with a stretching frequency of 3605 cm"\ Photoelectron diffraction studies [165] revealed the geometry of the bound formate molecule as reproduced in Fig. 33. It binds through both oxygens to pairs of neighboring Ti ions oriented along the (001) rows. The 0-C-O bond angle is 126 ± 4"", and the vertical distance between formate oxygens and surface Ti ions is 2.1 +_0.1 A in good agreement with structure optimizations based on Hartree-Fock total energy minimization.

5. SURFACES OF OXIDES WITH LAYERED STRUCTURES Fig. 34 shows the volume structure of vanadium-pentoxide [15]. The layered structure of the material is quite obvious and its electronic structure has been discussed in detail on the basis of diffraction data and theoretical investigations. V2O5 unit cell

Fig. 34. Schematic representation of the bulk structure of V2O5 including the dimensions of the unit cell [15].

364

Fig. 35. Four constant current STM images of the (001) surface of Na doped V2O5 recorded with similar experimental parameters (U: 2-3 V, I: 0.6-1.0 nA) [167]. Images (a), (b), and (c) have the same scale and orientation. The image in (d) covers a slightly larger area and has a different orientation. The surface repeat unit is indicated by the rectangle in (a). The vertical scale from black to white is in: (a) 2.25 A, (b) 1.5 A, (c) 1 A and (d) 2.25 A.

The material cleaves almost exclusively along the (001) surface, i.e. between the layers which are held by rather weak forces. Fig. 35 displays well resolved STM data reported for this surface [167]. The authors attribute the differences between the images to surface disorder mainly involving the surface vanadyl groups which are expected to be responsible for the bright areas in Figs. 35a, b and d. Electron spectroscopic studies have been undertaken early on by Zhang and Henrich and by the Horn group [167-171]. The latter group, however, has been more interested in bulk than in surface properties. Fig. 36 shows a photoemission spectrum of ¥205(00!) at the top of the lower panel taken in the authors laboratory. It may be interpreted on the basis of calculations by Klaus Hermann and his group [172]. Briefly, the spectrum is dominated by 02p/02s emissions and there is a minor admixture of the V3d wave functions in the panel showing the results of the calculations. There are no features near the Fermi energy which would, if they were present, be characteristic for defects leaving V atoms in lower oxidation states.

365

0 ( 1 ) ^ 0(2) 0(3) 1 ^ ^ Mr§L^ ^^

Theory (V2O5) V16O49H18 Cluster total DOS Vanadium

0(1) 0(2) 0(3)

plan view 4 6 8 Binding energy [eV]

Fig. 36. Photoelectron spectra and schematic representations of structures of various vanadium oxides [98, 173]. For comparison a computed density of states [172] is shown.

The oxygen derived part of the valence band can be separated into contributions from the various types of oxygens constituting the structure as shown in the right part of Fig. 36. The terminal vanadyl oxygen {0(1)} gives rise to the central features whereas the bridging oxygens {0(2) and 0(3)} connecting the vanadyl groups are contributing to the wings of the valence band. We will use this fact to identify oxygen specific reactivity. The spectrum in the panel below V2O5 is that of VO2. VO2(110) has been grown as a thin film on the isostructural rutile(llO) surface as discussed above. The oxygen derived valence band is slightly different from that of V2O5. A characteristic difference is the appearance of a feature close to the Fermi edge in VO2 indicating the presence of vanadium 3d electrons. Its intensity represents to some extent the population of the V3d orbitals. When we turn from VO2 to V2O3 (also discussed above) an even stronger increase in the V3d intensity is observed. In Fig. 36, in the lower panel, the valence band photoemission spectrum of ¥263(0001)/Au(111), is shown. The V2O3 film shows a sharp hexagonal LEED [98] pattern (see Fig. 16) representing the corundum type structure similar to AI2O3, Cr203 and Fe203 (see above). Again, the V2O3 oxygen valence band emission is not very characteristic, as compared with VO2 and V2O5. It is only the change in the near Fermi edge structures that show a characteristic variation from V^^ to V^^.

366 ARUPS hv=40.8 eV T=300 K

2

VO2(110)/

4

6

8

Binding energy [eV]

Fig. 37. Photoelectron spectra of V2O5(001) and VO2(110)/TiO2(110) in comparison with a photoelectron spectrum obtained after dosing atomic hydrogen to a ¥205(001) surface.

Vanadium pentoxide is rather inactive with respect to adsorbates. For example, molecular hydrogen only leads to observable effects in the photoelectron spectrum after lO"^ L exposure, On the other hand, atomic hydrogen leads already at low exposure to recognizable changes in the photoelectron spectrum. The spectrum shown in Fig. 37 in the middle has been obtained after exposing the surface to atomic hydrogen. Opposite to the finding with molecular hydrogen the surface is chemically corroded. Defects form, as we also know from electron energy loss spectroscopy, and concomitantly reduced metal centers show up near the Fermi edge. In fact, the spectrum looks very similar to the one of VO2 which we show for comparison. We have investigated the surface with vibrational spectroscopy and found no indication for the formation of hydroxyls. It is very likely, that during exposure water evolves from the surface. Still the complete absence of OH on the surface is surprising and might point to the formation of hydrogen vanadium oxide bronzes which are well-known to exist and which also have been used as precursors in the preparation of catalysts [174]. 6. SYNOPSIS The present review has been focussed on discussing the geometric and electronic properties of surfaces as related to their ability to adsorb and electronically modify molecules. Still this field is in its infancy for oxide surfaces when compared to vast amount of data available for metal surfaces. It is important to recognize, that the interaction of molecules with oxide surfaces has to be compared with care with the established bonding models for metal surfaces. This fact, although quite obvious is not always properly appreciated. For oxide surfaces electrostatic effects have to be taken into account, especially for the highly ionic materials. For example, even though CO binds to MgO or NiO in the same geometry as on a metal surface, the synergetic o-bonding-Ti-backbonding mechanism (Blyholder model) cannot be appropriately used to understand the

367

bonding. Other oxides, which have more covalent bonding character, may, on the other hand, be properly described using such models. RUO2 seems to be an example. In general, water adsorption has been studied to a large extent due to its importance in geochemical and electrochemical studies. While information for the well ordered and carefully prepared surfaces is accumulating, it becomes more and more important to try to understand the nature of well identified defects on the adsorption properties. In the field of oxide surface chemistry research, it is the belief of the authors, that this field will be an interesting but also a challenging one, if well characterized systems are used. REFERENCES [I] [2] [3] [4]

[5] [6] [7]

[8] [9] [10] [II] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23]

H. Schmalzried, Chemical Kinetics of Solids, (Verlag Chemie, Weinheim, 1995). V.E. Henrich and P.A. Cox, The Surface Science of Metal Oxides, (Cambridge University Press, Cambridge, 1994). C. Xu and D.W. Goodman, in Handbook of Heterogeneous Catalysis, edited by G. Ertl, H. Knozinger, and J. Weitkamp, Vol. 2 (Wiley-VCH, Weinheim, 1997), page 826. S.C. Street and D.W. Goodman, in Growth and Properties of Ultrathin Epitaxial Layers, The Chemical Physics of Solid Surfaces, edited by D. A. King and D. P. Woodruff, Vol. 8 (Elsevier, New York, 1997), page 375. R.J. Lad, Surf Rev. Lett., 2 (1995) 109. H.-J. Freund, H. Kuhlenbeck and V. Staemmler, Rep. Progr. Phys., 59 (1996) 283. H. Kuhlenbeck and H.-J. Freund, in Growth and Properties of Ultrathin Epitaxial Layers, The Chemical Physics of Solid Surfaces, edited by D. A. King and D. P. Woodruff, Vol. 8 (Elsevier, New York, 1997), page 340. G. Renaud, Surf Sci. Rep., 32 (1998) 1. C.A. Ventrice and H. Geisler, in Thin Films: Heteroepitaxial Systems, (Chapter 4), edited by A. W. K. Liu and M. B. Santos (World Scientific, Singapore, 1999). H.-J. Freund, Faraday Disc, 114 (1999) 1. S.A. Chambers, Surf Sci. Rep., 39 (2000) 105. R. Franchy, Surf Sci. Rep., 38 (2000) 195. C. Noguera, Physics and Chemistry at Oxide Surfaces, (Cambridge University Press, Cambridge, 1996). P.W. Tasker, Adv. Ceram., 10 (1984) 176. R.W.G. Wyckoff, Crystal Structures, (Wiley Interscience, New York, 1965). K.O. Legg, M. Prutton and C. Kinniburgh, J. Phys. C: Solid State Phys., 7 (1974) 4236. D.L. Blanchard, D.L. Lessor, J.P. LaFemina, D.R. Baer, F. W.K. and T. Guo, J. Vac. Sci. Technol.A,9(1991)1814. K.H. Rieder, Surf Sci. Lett., 118 (1982) 57. M. Prutton, J.A. Walker, M.R. Welton-Cook, R.C. Felton and J.A. Ramsey, Surf Sci. Lett., 89 (1979) 95. F.P. Netzer and M. Prutton, J. Phys. C: Solid State Phys., 8 (1975) 2401. C.G. Kinniburgh and J.A. Walker, Surf Sci., 63 (1977) 274. M.R. Castell, C. Muggelberg, S.L. Dudarev, A.P. Sutton, G.A.D. Briggs and D.T. Goddard, Appl. Phys. A, 66 (1998) S963. M.R. Castell, P.L. Wincott, N.G. Condon, C. Muggelberg, G. Thornton, S.L. Dudarev, A.P. Sutton and G.A.D. Briggs, Phys. Rev. B, 55 (1997) 7859.

368 [24] H. Hosoi, K. Sueoka, K. Hayakawa and K. Mukasa, Appl. Surf. Sci, 157 (2000) 218. [25] M.R. Castell, S.L. Dudarev, G.A.D. Briggs and A.P. Sutton, Phys. Rev. B, 59 (1999) 7342. [26] C. Noguera, A. Pojani, F. Finocchi and J. Goniakowski, in Chemisorption and Reactivity on Supported Clusters and Thin Films, edited by R. M. Lambert and G. Pacchioni, NATO ASI Series E, Vol. 331 (Kluwer Academic Publishers, Dordrecht, 1997), page 445. [27] D. Wolf, Phys. Rev. Lett, 68 (1992) 3315. [28] F. Rohr, K. Wirth, J. Libuda, D. Cappus, M. Baumer and H.-J. Freund, Surf Sci., 315 (1994) L977. [29] A. Barbier and G. Renaud, Surf Sci., 392 (1997) L15. [30] A. Barbier, in Proceedings of the 1st International Workshop on Oxide Surfaces, (Elmau, Germany, 1999). [31] A. Barbier, C. Mocuta, H. Kuhlenbeck, K.F. Peters, B. Richter and G. Renaud, Phys. Rev. Lett., 84 (2000) 2897. [32] C.A. Ventrice Jr., T. Bertrams, H. Hannemann, A. Brodde and H. Neddermeyer, Phys. Rev., B49 (1994) 5773. [33] V.E. Henrich, Surf Sci., 57 (1976) 385. [34] O. Robach, G. Renaud and A. Barbier, Surf Sci., 401 (1998) 227. [35] S. Hufner, P. Steiner, I. Sander, M. Neumann and S. Witzel, Z. Phys. B, 83 (1991) 185. [36] S. Hufner, Photoelectron Spectroscopy: Principles and Applications, (Springer-Verlag, Berlin, 1995). [37] C. Noguera and W.C. Mackrodt, J. Phys. Condens. Matter, 12 (2000) 2163. [38] W.C. Mackrodt and C. Noguera, Surf Sci., 457 (2000) L386. [39] A. GorschlUter and H. Merz, Phys. Rev. B, 49 (1994) 17293. [40] A. Freitag, V. Staemmler, D. Cappus, C.A. Ventrice Jr., K. Al-Shamery, H. Kuhlenbeck and H.-J. Freund, Chem. Phys. Lett., 210 (1993) 10. [41] R. Newman and R.M. Chrenko, Phys. Rev., 114 (1959) 1507. [42] R.J. Powell, Report 5220-1 Stanford Electronics Laboratory, 1967. [43] D. Adler and J. Feinleib, Phys. Rev. B, 2 (1970) 3112. [44] B. Fromme, M. Moller, T. Anschutz, C. Bethke and E. Kisker, Phys. Rev. Lett., 77 (1996)1548. [45] B. Fromme, C. Koch, R. Deussen and E. Kisker, Phys. Rev. Lett., 75 (1995) 693. [46] M. Pohlchen and V. Staemmler, J. Chem. Phys., 97 (1992) 2583. [47] F. Muller, R. de Masi, P. Steiner, D. Reinicke, M. Stadtfeld and S. Hufner, Surf Sci., 459(2000) 161. [48] M. Henzler and W. Gopel, Oberflachenphysik des Festkorpers, (Teubner-Verlag, Stuttgart, 1991). [49] K.M. Neyman, G. Pacchioni and N. Rosch, in Recent Developments and Applications of Modem Density Functional Theory and Computational Chemistry, edited by J. M. Seminario (Elsevier, Amsterdam, 1996), page 569. [50] H.-J. Freund, Angew. Chem. Int. Ed. Engl., 36 (1997) 452. [51] M. Baumer, J. Libuda and H.-J. Freund, in Chemisorption and Reactivity on Supported Clusters and Thin Films, edited by R. M. Lambert and G. Pacchioni, NATO ASI Series E, Vol. 331 (Kluwer Academic Press, Dordrecht, 1997), page 61. [52] M.A. Nygren and L.G.M. Pettersson, J. Chem. Phys., 105 (1996) 9339. [53] G. Pacchioni and P.S. Bagus, in Adsorption on Ordered Surfaces of Ionic SoHds and Thin Films, edited by H.-J. Freund and E. Umbach, Springer Series in Surface Science, Vol. 33 (Springer Verlag, Berlin, 1993), page 180.

369 [54] T. Kluner, private communication. [55] D. Cappus, J. Klinkmann, H. Kuhlenbeck and H.-J. Freund, Surf. Sci., 325 (1995) L 421. [56] S.M. Vesecky, X. Xu and D.W. Goodman, J. Vac. Sci. Technol. A, 12 (1994) 2114. [57] V. Staemmler, in Adsorption on Ordered Surfaces of Ionic Solids and Thin Films, edited by H.-J. Freund and E. Umbach, Springer Series in Surface Science, Vol. 33 (Springer Verlag, Berlin, 1993), page 169. [58] M. Pohlchen, PhD Thesis, Ruhr-Universitat Bochum, 1992. [59] H. Kuhlenbeck, G. Odorfer, R. Jaeger, G. Illing, M. Menges, T. Mull, H.-J. Freund, M. Pohlchen, V. Staemmler, S. Witzel, C. Scharfschwerdt, K. Wennemann, T. Liedtke and M. Neumann, Phys. Rev. B, 43 (1991) 1969. [60] L. Chen, R. Wu, N. Kioussis and Q. Zhang, Chem. Phys. Lett., 290 (1998) 255. [61] K.M. Neyman, S.P. Ruzankin and N. Rosch, Chem. Phys. Lett., 246 (1995) 546. [62] J.-W. He, C.A. Estrada, J.S. Comeille, M.-C. Wu and D.W. Goodman, Surf Sci., 261 (1992) 164. [63] S. Furuyama, H. Fuji, M, Kawamura and T. Morimoto, J. Phys. Chem., 82 (1978) 1028. [64] R. Wichtendahl, M. Rodriguez-Rodrigo, U. Hartel, H. Kuhlenbeck and H.-J. Freund, phys. Stat. sol. (a), 173 (1999) 93. [65] P.A. Redhead, Vacuum, 12 (1962) 203. [66] H. Pfnur, P. Feulner and D. Menzel, J. Chem. Phys., 79 (1983) 4613. [67] C.N. Chittenden, E.D. Pylant, A.L. Schwaner and J.M. White, Thermal Desorption and Mass Spectrometry, (CRC Press, Boca Raton, 1995). [68] H. Schlichting and D. Menzel, Rev. Sci. Instrum., 64 (1993) 2013. [69] R. Lindsay, P. Baumgartel, R. Terborg, O. Schaff, A.M. Bradshaw and D.P. Woodruff, Surf. Sci., 425 (1999) L401. [70] K. Hermann and P.S. Bagus, Phys. Rev. B, 15 (1977) 4195. [71] H.-J. Freund and E.W. Plummer, Phys. Rev. B, 23 (1981) 4859. [72] E. Umbach, Surf. Sci., 117 (1982) 482. [73] J. Heidberg, M. Kandel, D. Meine and U. Wildt, Surf Sci., 333 (1995) 1467. [74] R. Gerlach, A. Glebov, G. Lange, J.P. Toennies and H. Weiss, Surf. Sci., 331-333 (1995) 1490. [75] J. Heidberg, B. Redlich and D. Wetter, Ber. Bunsenges. Phys. Chem., 99 (1995) 1333. [76] D. Ferry, A. Glebov, V. Senz, J. Suzanne, J.P. Toennies and H. Weiss, J. Chem. Phys., 105 (1996) 1697. [77] A. Marmier, P.N.M. Hoang, S. Picaud, C. Girardet and R.M. Lynden-Bell, J. Chem. Phys., 109 (1988) 3245. [78] M.J. Stimiman, C. Huang, R.C. Smith, J.A. Joyce and B.D. Kay, J. Chem. Phys., 105 (1996)1295. [79] C. Xu and D.W. Goodman, Chem. Phys. Lett., L65 (1997) 341. [80] R. Wichtendahl, PhD-Thesis, Freie Universitat Berlin, 1999. [81] C M . Truong, M.-C. Wu and D.W. Goodman, J. Am. Chem. Soc, 115 (1993) 3647. [82] C M . Truong, M.-C. Wu and D.W. Goodman, J. Chem. Phys., 97 (1992) 9447. [83] M.-C. Wu, C M . Truong and D.W. Goodman, J. Phys. Chem., 97 (1993) 4182. [84] S.C Street, Q. Guo, C Xu and D.W. Goodman, J. Phys. Chem., 100 (1996) 17599. [85] A. Kara, S. Durukanoglu and T.S. Rahman, Phys. Rev. B, 53 (1996) 15489. [86] K. Domen, N. Akamatsu, H. Yamamoto, A. Wada and C Hirose, Surf Sci., 283 (1993) 468. [87] A. Bandara, J. Kubota, A. Wada, K. Domen and C Hirose, Surf. Sci. Lett., 364 (1996) L580.

370

[88] A. Bandara, J. Kubota, A. Wada, K. Domen and C. Hirose, J. Phys. Chem. B, 101 (1997)361. [89] W.L. Roth, J. Appl. Phys., 31 (1960) 2000. [90] H.P. Rooksby, Acta Cryst., 1 (1948) 226. [91] S. Greenwald and J.S. Smart, Nature, 166 (1950) 523. [92] X.-G. Wang, W. Weiss, S.K. Shaikhoutdinov, M. Ritter, M. Petersen, F. Wagner, R. Schlogl and M. Scheffler, Phys. Rev. Lett., 81 (1998) 1038. [93] I. Manassidis and M.J. Gillan, J. Am. Ceram. Soc., 77 (1994) 335. [94] F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Muller, L. Hammer and K. Heinz, Surf. Sci., 372 (1997) L 291. [95] F. Rohr, M. Baumer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Muller, L. Hammer and K. Heinz, Surf. Sci., 389 (1997) 391. [96] S.K. Shaikhoutdinov and W. Weiss, Surf Sci., 432 (1999) L627. [97] K.B. Lewis, S.T. Oyama and G.A. Somorjai, Surf. Sci., 233 (1990) 75. [98] A.-C. Dupuis, PhD-Thesis, in preparation. [99] N.M. Harrison, X.-G. Wang, J. Muscat and M. Scheffler, Faraday Disc, 114 (1999) 305. [100] K. Wolter, D. Scarano, J. Fritsch, H. Kuhlenbeck, A. Zecchina and H.-J. Freund, Chem. Phys. Lett, 105(2000)295. [101] B. Dillmann, F. Rohr, O. Seiferth, G. KHvenyi, M. Bender, K. Homann, I.N. Yakovkin, D. Ehrlich, M. Baumer, H. Kuhlenbeck and H.-J. Freund, Faraday Disc. 105, (1996) 295. [102] C. Xu, M. HaBel, H. Kuhlenbeck and H.-J. Freund, Surf Sci., 258 (1991) 23. [103] M.A. Henderson and S.A. Chambers, Surf. Sci., 449 (2000) 135. [104] J.A. Mejias, V. Staemmler and H.-J. Freund, J. Phys. Cond. Matter, 11 (1999) 7881. [105] H. Kuhlenbeck, C. Xu, B. Dillmann, M. HaBel, B. Adam, D. Ehrlich, S. Wohlrab, H.-J. Freund, U.A. Ditzinger, H. Neddermeyer, M. Neuber and M. Neumann, Ber. Bunsenges. Phys. Chem., 96 (1992) 15. [106] C. Xu, B. Dillmann, H. Kuhlenbeck and H.-J. Freund, Phys. Rev. Lett., 67 (1991) 3551. [107] M. Bender, D. Ehrlich, LN. Yakovkin, F. Rohr, M. Baumer, H. Kuhlenbeck, H.-J. Freund and V. Staemmler, J. Phys. Cond. Matter, 7 (1995) 5289. [108] M. Pykavy, V. Staemmler, O. Seiferth and H.-J. Freund, Surf Sci., (submitted). [109] M.A. Henderson, S.A. Joyce and J.R. Rustad, Surf Sci., 417 (1998) 66. [110] P. Lin, T. Kendelewicz, G.E. Brown, E.J. Nelson and S.A. Chambers, Surf. Sci., 417 (1998)53. [111] D. Ehrlich, PhD-Thesis, Ruhr-Universitat Bochum, 1995. [112]0. Seiferth, K. Wolter, B. Dillmann, G. KHvenyi, H.-J. Freund, D. Scarano and A. Zecchina, Surf. Sci., 421 (1999) 176. [113] V. Coustet and J. Jupille, Surface and Interface Analysis, 22 (1994) 280. [114] V. Coustet and J. Jupille, Surf Sci., 307-309 (1994) 1161. [115] K.C. Haas, W.F. Schneider, A. Curloni and W. Andreoni, Science, 282 (1998) 265. [116] J. Libuda, M. Frank, A. Sandell, S. Andersson, P.A. Briihwiler, M. Baumer, N. Martensson and H.-J. Freund, Surf Sci., 384 (1997) 106. [117] A. Zecchina, D. Scarano, S. Bordiga, G. Ricchiardi, G. Spoto and F. Geobaldo, Catal. Today, 27 (1996) 403. [118] I. Hemmerich, F. Rohr, O. Seiferth, B. Dillmann and H.-J. Freund, Z. Phys. Chem., 202 (1997)31. [119] P.C. Thune and J.W. Niemantsverdriet, Israel. J. Chem., 38 (1998) 385.

371 [120] 12th International Congress on Catalysis, Studies in Surface Science and Catalysis, edited by A. Corma, F.V. Melo, S. Mendioroz and J.L.G. Fierro (Elsevier, Amsterdam, 2000). [121] H. Schlienz, M. Beckendorf, U.J. Katter, T. Risse and H.-J. Freund, Phys. Rev. Lett., 74 (1995)761. [122] M.E. Bartram and B.E. Koel, Surf. Sci., 213 (1989) 137. [123] U.J. Katter, H. Schlienz, M. Beckendorf and H.-J. Freund, Ber. Bunsenges. Phys. Chem.,97(1993)340. [124] Principles of Electron Spin Resonance, edited by N.M. Atherton (Ellis Horwood PTR Prentice Hall, New York, 1993). [125] A. Abragam and B. Bleaney, in Electron Paramagnetic Resonance of Transition Ions (Clarendon Press, Oxford, 1970), page 492. [126] P.O. Gray and A.D. Yoffe, Chem. Rev., 55 (1955) 1069. [127] M. Beckendorf, U.J. Katter, H. Schlienz and H.-J. Freund, J. Phys. Condens. Matter, 5 (1993)5471. [128] M. Beckendorf, PhD-Thesis, Ruhr-Universitat Bochum, 1994. [129] D.L. Freeman and J.D. Doll, J. Chem. Phys., 78 (1983) 6002. [130] B. Roop, S.A. Costello, D.R. MuUins and J.M. White, J. Chem. Phys., 86 (1987) 3003. [131] U.J. Katter, T. Hill, T. Risse, H. Schlienz, M. Beckendorf, T. Kluner, H. Hamann and H.-J. Freund, J. Phys. Chem. B, 101 (1997) 3776. [132] G. Poggi and C.S. Johnson, Jr., J. Magn. Reson., 3 (1970) 436. [133] U.J. Katter, T. Hill, T. Risse, H. Schlienz, M. Beckendorf, T. Kluner, H. Hamann and H.-J. Freund, J. Phys. Chem. B, 101 (1997) 552. [134] P.A. Egelstaff, J. Chem. Phys., 53 (1970) 2590. [135] G.L. Millhauser and J.H. Freed, J. Chem. Phys., 81 (1984) 37. [136] T. Risse, T. Hill, M. Beckendorf, U.J. Katter, H. Schlienz, H. Hamann and H.J. Freund, Langmuir, 12(1996)5512. [137] T. Risse, T. Hill, J. Schmidt, G. Abend, H. Hamann and H.-J. Freund, J. Chem. Phys., 108(1998)8615. [138] J.H. Freed, G.V. Bruno and C.F. Polnaszek, J. Phys. Chem., 75 (1971) 3385. [139] B.L. Bales, in Spin Labeling Theory and Applications: Biological Magnetic Resonance, edited by L. J. Berliner and J. Reuben, Vol. 8 (Plenum Press, New York, 1989), page 77. [140] T. Risse, PhD-Thesis, Ruhr-Universitat Bochum, 1996. [141] M. Ge and J.H. Freed, Biophys. J., 65 (1993) 2106. [142] C.E.D. Chidsey, G.-L. Liu, P. Rowntree and G. Scoles, J. Chem. Phys., 91 (1989). [143] U. Diebold, J.F. Anderson, K.-O. Ng and D. Vanderbilt, Phys. Rev. Lett., 77 (1996) 1322. [144] P.W. Murray, N.G. Condon and G. Thornton, Phys. Rev. B, 51 (1995) 10989. [145] H. Onishi and Y. Iwasawa, Surf. Sci., 313 (1994) L783. [146] H. Onishi and Y. Iwasawa, Chem. Phys. Lett., 226 (1994) 111. [147] H. Onishi, K. Fukui and Y. Iwasawa, Bull. Chem. Soc. Jpn., 68 (1995) 2447. [148] H. Onishi and Y. Iwasawa, Phys. Rev. Lett., 76 (1996) 791. [149] G. Charlton, P.B. Howes, C.L. Nicklin, P. Steadman, J.S.G. Taylor, C.A. Muryn, S.P. Harte, J. Mercer, R. McGrath, D. Norman, T.S. Turner and G. Thornton, Phys. Rev. Lett., 78(1997)495. [150] D.B. McWhan, P.D. Dernier, M. M. Marezio and J.P. Remeika, Phys. Rev. B, 10 (1974).

372

[151] H. Over, Y.D. Kim, A.P. Seitsonen, S. Wendt, E. Lundgren, M. Schmid, P. Varga, A. Morgante and G. Ertl, Science, 287 (2000) 1474. 152] H. Over, private communication. 153] P.J. Moller, Z.S. Li, T. Egebjerg, M. Sambi and G. Granozzi, Surf. Sci., 402-404 (1998) 719. 154] M. Sambi, G. Sangiovanni, G. Granozzi and F. Parmigiani, Phys. Rev. B, 54 (1996) 13464. 155] M. Sambi, E. Pin, G. Sangiovanni, L. Zaratin, G. Granozzi and F. Parmigiani, Surf. Sci., 349 (1996) L169. 156] M. Sambi, G. Sangiovanni, G. Granozzi and F. Parmigiani, Phys. Rev. B, 55 (1997) 7850. 157] M. Sambi, M.D. Negra, G. Granozzi, Z.S. Li, J.H. Jorgensen and P.J. MoUer, Appl. Surf Sci., 142(1999) 146. 158] W.S. Epling, C.H.F. Peden, M.A. Henderson and U. Diebold, Surf. Sci., 412/413 (1998) 333. 159] M.A. Henderson, Surf Sci., 355 (1996) 151. 160] D. Brinkley, M. Dietrich, T. Engel, P. Farrall, G. Gantner, A. Schafer and A. Szuchmacher, Surf Sci., 395 (1998) 292. 161] M.A. Henderson, Surf Sci., 319 (1994) 315. 162] P.B. Smith and S. Bemasek, Surf Sci., 188 (1988) 241. 163] M.B. Hugenschmidt, C. Janicke and C.T. Campbell, Surf Sci., 302 (1994) 329. 164] Y. Iwasawa, Surf Sci., 402-404 (1998) 8. 165] S. Thevuthasan, G.S. Herman, Y.J. Kim, S.A. Chambers, C.H.F. Peden, Z. Wang, R.X. Ynzunza, E.D. Tober, J. Morais and C.S. Fadley, Surf Sci., 401 (1998) 261. 166] S.A. Chambers, M.A. Henderson, Y.J. Kim and S. Thevuthasan, Surf Rev. Lett, 5 (1998)381. 167] R.L. Smith, W. Lu and G.S. Rohrer, Surf Sci., 322 (1995) 293. 168] T. Oshio, Y. Sakai, T. Moriya and S. Ehara, Scanning Microscopy, 7 (1993) 33. 169] R.L. Smith, G.S. Rohrer, K.S. Lee, D.-K. Seo and M.-H. Whangbo, Surf Sci., 367 (1996)87. 170] R.A. Goschke, K. Vey, M. Maier, U. Walter, E. Goering, M. Klemm and S. Horn, Surf Sci., 438 (1996) 305. 171]Z.ZhangandV.E.Henrich, Surf Sci., 321 (1994) 133. 172] K. Hermann, M. Witko, R. Druzinic, A. Chakrabarti, B. Tepper, M. Eisner, A. Gorschluter, H. Kuhlenbeck and H.-J. Freund, J. Electron Spectrosc. Relat. Phenom., 98-99(1999)245. 173] B. Tepper, PhD-Thesis, in preparation. 174] M. Figlarz, Progr. Solid State Chem., 19 (1989) 1.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

373

Chapter 9

Atomic-scale Chemical and Electronic Structure Studies of Well-defined Metal Oxide Surfaces Charles C. Chusuei, Douglas C. Meier and D. Wayne Goodman Department of Chemistry, P.O. Box 30012, Texas A&M University, College Station, TX 77842-3012, USA

1. INTRODUCTION Understanding relationships between atomic-scale chemical and electronic structure with chemical activity at the metal oxide interface are important for tailoring nanostructured materials in applications for chemical sensors, corrosion inhibitors, ceramics, dielectric materials and heterogeneous catalysis. While an impressive array of electron spectroscopic probes have been developed in the past few decades for surface sensitive analysis [1, 2], the utilization of these tools for studying bulk metal oxides is limited. Difficulties arise from the insulating properties of these materials, which result in surface charging and inhomogeneous sample heating (during thermal desorption analysis). However, these problems can be overcome by depositing the metal oxides as thin-films (~50 A thick) on refractory metal substrates via "hot" filament evaporation under elevated O2 pressure (~10"^ Torr). These films retain the bulk chemical properties of the metal oxides while being thin enough to be sufficiently conductive for charged particle probes. This approach used in our laboratory has been extensively described for a variety of oxides including Ti02, Si02, AI2O3 and MgO in several reviews [3-7], comparing the oxide thin-films with single crystal analogs. The techniques used include high resolution electron energy loss (HREELS), ion scattering (ISS), X-ray photoelectron (XPS), Auger electron (AES) and infrared reflection absorption (IRAS) spectroscopies, low energy electron diffraction (LEED) and temperature programmed desorption (TPD). Most recently, we have studied oxide-supported nanosized noble metal particles, linking their activity with

374

electronic structure as well as the chemical properties of other catalytically interesting adsorbates (including D2O, CH3OH, CO and NO) on these surfaces. In this chapter, we focus our discussion on these latest findings. In the opening section on MgO, the use of metastable impact electron spectroscopy (MIES) in conjunction with ultraviolet photoelectron (UPS) spectroscopy and TPD for studying surface defects and the electronic structure of adsorbates at the topmost atomic layers is introduced (section 2). Next, the analysis of nanosized, catalytically active noble metal particles by scanning tunneling microscopy (STM) and (STS) spectroscopy, demonstrating correlations of activity (for CO oxidation) and electronic and morphological structure, are shown. The CO oxidation activity studies are supplemented by IRAS analysis, which provided insight on the nature of CO bonding to the highly dispersed, catalytically active Au/Ti02 surface (section 3). This discussion is followed by a summary of recent advances pertaining to STM imaging of AI2O3 thin films and surface characterization of mixed NiO systems (section 4). Finally, atomicallyresolved in- and ex-situ scanning probe microscopic analysis of nanosized gold clusters deposited on TiO2(110) at elevated pressures (simulating practical catalytic conditions) is presented (section 5). 2. MAGNESIUM OXIDE Magnesium oxide, MgO, has a rocksalt lattice, the simplest metal oxide crystal structure with equal numbers of Mg^^ and 0^~ cations and anions, respectively, and can thus be easily characterized. This well-known structure has been thoroughly studied theoretically [8-10] and experimentally, employing XPS, AES, IRAS, HREELS, LEED and TPD analysis [11-14]. The results of these studies show that thin-film MgO exhibits the same chemical features as single crystal MgO. MIES and UPS are powerful tools for probing the electronic structure of the near-surface. When used properly, information regarding the geometric arrangement of admolecules at the top-most layers can be readily obtained using these techniques. MIESAJPS analyses are particularly suited for analyzing defects on the MgO surface. 2.1. Adsorption of benzene on MgO/Mo(100) studied by MIES/UPS The orientation of molecules (e.g. whether it is parallel or perpendicular with respect to the surface) can be determined by comparing changes in the relative intensities of various molecular orbitals (MOs) during adsorption. Benzene (C^H^) is a particularly interesting system in which to study the effects of anisotropic molecule-molecule and molecule-solid interactions during the growth of molecular crystals. Its MO features measured in the gas-phase (using UPS) are well documented and can be used to compare with MIES

375

(a)

(a) /3e,^

^il|I^Q IVMLIO

UPS

c ZJ

(n)

E^)-^

+-»

CD

B31

O

o

/A

ig

\ II E.

BJ

J

B) 20

18

16

14

12

10

8

6

4

2

0

-2

binding energy (eV)

Fig. 1 Comparison of MIES and UPS spectra (raw data shown) of multilayer benzene adsorbed on Mo(lOO) at 100 K.

[15, 16]. Fig. 1 shows MIE and UP spectra of 4 monolayers (ML) of QH^ adsorbed on a Mo(lOO) single crystal at 100 K [17]. It should be noted that MIES of the CgH^ is dominated by an Auger de-excitation (AD) process since there are no unoccupied states in resonance with the impinging He 2s electron; thus, a reasonable comparison between MIES and UPS data can be performed. The MIES/UPS binding energies (in parantheses) of the Sajg (16.4 eV/16.4 eV), 2bi„ (15.1 eV/ —) and Icig (9.0 eV/8.8 eV) MOs are clearly observed whereas those emanating from the la2u, ^^ig (H-S eV/11.4 eV) and lb2u, Sci^ (13.7 eV/14 eV) remain unresolved. These latter unresolved features are labeled as "Ia2u/3e2g" and "lb2u/3eiu". The measured MIES binding energies are -0.5 eV higher than those measured by Kubota et al. [16] for C^H^ adsorbed on Cu at 150 K; however, their relative peak positions and line intensities match. Compared to gas-phase data, the binding energies of condensed phase C^R^ are ca. 0.4-0.8 eV lower, but can be accounted for by a 0.9 eV shift due to polarization of the neighboring molecules [18]. The binding energies determined by MIES are closer to the gas-phase-measured ionization potentials than those obtained by UPS; this can be explained by a lower polarization in the top-most layers. MIES is more sensitive to the s-

376

derived levels of the MOs; thus, the 3aig band (a convolution of the C 2p and H Is orbitals) is more intense as compared to the corresponding UPS feature [19]. On the other hand, UPS is more sensitive to the 2p-derived levels, owing to the relatively high intensities of the le^g and la2u MOs. The determining factor for the AD of the metastable helium is the overlap between the projectile and target wave functions. To simplify the remaining discussion of the MIES/UPS spectral features, the le^g, Ia2u/3e2g, lb2u/3eiu and Sa^g C^H^ bands are designated as B^, B2, B3 and B4, respectively. Fig. 2 shows waterfall plots of MIE and UP spectra obtained from CgH^ exposed to thin-film MgO on a Mo(lOO) single crystal substrate [17]. The MgO film was --1 ML thick as estimated using the AES inelastic mean free path. The MIES and UPS both have a very high signal-to-noise ratio; raw data is shown. Both of these spectra were acquired simultaneously in-situ. The CgH^ was dosed at 5 x 10"^ Torr background pressure at 100 K with an acquisition time of 180 seconds for each MIES/UPS spectrum. The bottom spectrum in both MIES and UPS waterfall plots are those of the clean MgO(lOO), matching the energy values of the single crystal in previous literature reports [20, 21]. No intensity is observed between the Fermi level (Ep) and 3.8 eV since MgO is an insulator and MIES is an AD process. The O 2p features in both the MIES and UPS emanate from the MgO(lOO) valence band structure of the MgO [20, 22]. The uppermost spectra correspond to condensed C^H^ layers. Unfortunately, the O 2p feature of the MgO overlaps with the Bi band and is not clearly resolvable at the onset of its exposure to the surface. However, after this onset, the B2, B3 and B4 bands from the CgH^ begin to appear. The dotted spectrum (in both the MIES and UPS data sets) provide a demarkation between the underlying MgO and the C^H^ overlayers. As more C^He is exposed to the surface, B3 and B4 monotonically increases while B2 intensity suddenly decreases (denoted by the bold lines between the 18^^ and 24^^ spectra). The B2 band (Ia2u/3e2g) corresponds to the ionization of the two highest occupied 71-levels of C^H^ consistent with previously reported MIES gas-phase data [23]. The C^H^ features are well-developed by the 5* MIES scan. The observed decrease in the B2 intensity by the 18^^ scan is consistent with the C^H^ standing on-edge, approximately perpendicular with respect to the surface. A prior TPD and HREELS study of C^H^ adsorption onto thin-film MgO has shown that the phenyl rings initially adsorb with the rings in a parallel orientation with respect to the surface [24] at low coverages (< 1 ML). As more C^H^ adsorbs, an intermediate metastable phase appears (> 1 ML), having an edge-on perpendicular orientation. Fig. 3 shows another view of the MIES waterfall plot (Fig. 2), showing more clearly changes in the work function as a function of exposure. The interactions of adsorbate with the surface, however, were found to be relatively weak. The constant work

377

8

10 12 14 16 18 20 22 24 26 28 30

8

binding energy (eV)

10 12 14 16 18 20 22 24 26 28 30

binding energy (eV)

Fig. 2 MIES and UPS spectra of benzene adsorbed on an MgO thin film on Mg(lOO). The bottom spectrum is that of the clean MgO(lOO) underlying substrate. The uppermost spectra are those of the highest benzene coverages. exposure time (min.) 59

73

88

103

117 132 1 1 work function

•J£ 3.5 k

20

25 30 spectrum

Fig. 3 Comparison of absolute MIES peaking intensities of the 0(2p) + Icig [0(2p)+Bi] structure and the Ia2u/3e2g (B2) and lb2u/3eiu(B3) benzene bands. The plotted "between" intensities (valleys between the peaks) denote the background intensity.

function v^as indicative of no charge transfer between the benzene and the MgO substrate (Fig 3). Note the drop in the count rate after the 18^^ spectrum for the B2 band (Fig. 3); this corresponds to the emergence of the edge-on C^Hg phase. The "between" lines in the plot denote the valleys in the waterfall features between bands, i.e. between the Bj, B2 and B2, B3 peaks. For simplicity, the B4 band is not shown. We emphasize that this edge-on orientation was detected only at the top-most layer. The corresponding trend in the B2 band was not observable in the UPS, since the technique samples the top

378

several layers; contribution to the B2 UPS signal from the sub-surface layers obscured the intensity decrease signifying the development of the on-edge phase. The edge-on phase is detected earlier for the MgO-covered Mo(lOO) surface as compared to bare transition metal (TM) surfaces. We postulate that the absorption mechanism is analogous to C^H^ adsorption onto TM surfaces, but with the first chemisorbed layer absent. In a study by Xi et al. [25] of CgH^ adsorbed on Cu(lll) , weak molecule-surface interactions were also reported and accounted for in their proposed bilayer structure with the n rings perpendicular to the surface at the top-most C^H^ layers. The MIES results in our investigation provide an additional verification of the edge-on orientation at the top (second) layer. Furthermore, the data reveals additional information in regard to its packing density during adsorption onto the MgO thin-film surface. The continuous decrease in the MIES background (Fig. 3) indicates that even at high C^H^ coverages, there is considerable intensity from the underlying MgO(lOO), suggesting that C^H^ does not absorb on top as contiguous layers, i.e. no formation of completely close-packed layers. 2.2. MIES and TPD analysis of CH3OH and DjO adsorption on MgO/Mo (100) Methanol (CH3OH) has shown to be a key reagent for C-C bond formation in the synthesis of a- and P-unsaturated hydrocarbons and C2-C5 alcohols using MgO catalysts [26]. Despite the promising gains offered by this synthetic route, fundamental issues regarding CH3OH wetting behavior and its geometric orientation during adsorption on the MgO support are not wellunderstood. The characterization of these factors may be a key consideration for tailoring improved activity and selectivity. The initial stages of CH3OH growth onto thin-film MgO was studied and compared with the growth behavior of water (D2O) in order to gain insight into molecular-level adsorption dynamics [21, 27, 28]. TPD data, used to quantitate surface coverage and obtain information regarding molecule-molecule interactions, are compared directly with MIES data, in order to provide complementary insights into the adsorbate electronic structure. Fig. 4 compares MIES with TPD spectra taken from various coverages of CH3OH dosed on MgO/Mo(100) at 100 K [28]. The MgO thin-film preparation was the same as described in section 2.1. Both MIES and TPD spectra are depicted with the same y-offsets to aid comparisons of the two data sets. The MIES peak designations (with measured binding energies in parentheses), M^ (7.1 eV), M2 (9.2 eV) and M3 (11.5 eV) correspond to the 2a''(noi), 7^'(no//) and 6a'(cTco) MOs of methanol, respectively, matching previously reported values obtained from gas phase photoelectron and Penning

379

100

200

300

400

Temperature (K)

0

2

4

6

8

10

12

binding energy (eV)

Fig. 4 TPD and MIES data as a function of CH3OH dosage onto MgO/Mo(100) at 100 K.

ionization electron spectra (PIES) [29]. The bottom MIES spectrum (in bold) shows the clean MgO surface with increasing CH3OH coverages "moving" towards the top of the stackplot. A complete attenuation of the O 2p band (emanating from the MgO substrate) in the MIES is observed in conjunction with the onset of a low-temperature feature in the TPD at 144 K (in bold) at a 4-L coverage. This observation is consistent with the assertion that CH3OH efficiently and completely wets the MgO surface, forming a closed-packed, two-dimensional (2D) structure (a closed CH3OH adlayer) prior to threedimensional (3D) adsorption. We attribute the onset of the 144 K TPD feature to the initial formation of a 3D CH3OH structure. Two distinct hightemperature desorption peaks appear at 275 and 240 K, labeled as "A" and "B", respectively. It has been shown from previous HREELS and MIESAJPS studies that methanol adsorbs non-dissociatively with hydroxyl groups oriented towards the surface [4, 28]. In this context, the presence of the two TPD states is suggestive of repulsive interactions between CH3OH molecules within the first monolayer. Due to the relatively high intensity of B, it seems unlikely that the desorption arises from defect sites on the MgO surface. We estimate, based on a 4-fold surface mesh model [30] and assuming that (1) adsorption occurs at thermodynamic equilibrium conditions and (2) interactions with nearestneighbor sites are independent of adsorbate coverage, that there is a ~2.8

380

kJ/mol interaction energy between adjacent CH3OH molecules on the surface. The low interaction energy is indicative of van der Waals forces type moleculemolecule interactions. The wetting behavior of CH3OH is quite different from that of D2O. In order to illustrate these differences, the use of the O 2p MIES signal from the underlying MgO substrate will be used as a diagnostic tool for detecting imperfections (incompleteness) of adsorbate surface coverage. Fig. 5 shows MIES and TPD data obtained from various D2O exposures to the MgO/Mo(100) surface [28]. The bottom MIES spectrum (in bold) shows the clean surface with increasing D2O exposures plotting in ascending order from bottom-to-top. The upper-most MIES spectrum is that of the MgO surface completely covered by 7 L of D2O. In order to determine the contribution of the MgO substrate in the MIES, the intensities from 2-7eV (in bold) were curvefitted by using a combination of lineshapes from the bottom- (clean MgO) and upper-most spectra where MgO is completely covered. During the curvefitting procedure, the mixture of the two lineshapes were varied until a best fit to the MIES data points were obtained. The Ibj and Sa^ labeled peaks

100

200 300 400 Temperature (K)

0

2 4 6 8 10 12 binding energy (eV)

Fig. 5 TPD and MIES data as a function of D2O dosages onto MgO/Mo(100) at 100 K.

381

denote the corresponding MOs from gas-phase photoelectron spectra [31, 32] and PIES [33] for D2O measured at 13.3 and 9.9 eV binding energies, respectively. A low-temperature desorption state (160 K) in the TPD appears after a 2.5 L-exposure of D2O, the start of a second-layer formation. In contrast to the CH3OH wetting behavior (which had been adsorbed onto the same MgO surface), a substantial MgO signal argues against second-layer adsorption occurring following the formation of a closed-packed layer of water molecules. Even after a 4-L exposure of D2O, the O 2p substrate intensity can readily be seen, signifying an incompletely covered oxide surface. Thus,the D2O wetting behavior on MgO demonstrates a striking contrast to that of CH3OH. Comparative TPD and MIES analysis of D2O adsorption indicates the onset of the physisorbed phase (160 K) before the substrate is covered whereas CH3OH completely wets the surface. 2.3. Surface defect characterization of MgO by MIES Characterization of surface defects on thin-film oxides have a technological importance for promoting activity and selectivity of oxide catalysts. For example, MgO catalysts with a high density of surface defects have been reported to promote the oxidative coupling of methane to ethane and ethylene through a methyl radical intermediate as compared to their relatively defectfree counterparts [34]. Due to its extremely high surface sensitivity, MIES is particularly suited for probing MgO thin-film defects at the near-surface. Fig. 6 shows a stackplot of MIES and UPS acquired from (A) an as-prepared MgO thin-film, (B) and (C) MIES and UPS after a 3kV, 5x10'^ C/cm^ electron beam bombardment (used to create defects) and (D) after O2 exposure [35]. The asprepared film (A) shows an O 2p valence band structure with a band gap of -6.7 eV. Spectra (B) and (C) show an feature at ~2 eV consistent with the formation of F-centers or F-aggregates upon electron bombardment. Prior electron stimulated desorption experiments performed on MgO have shown that oxygen anions and neutrals are preferentially removed, leaving oxygen vacancies on the surface [36]. Furthermore, this energy position coincides with theoretical predictions for F-centers [37, 38]. The new ~2 eV band is stable with time under ultra-high vacuum (UHV) but diminished immediately upon exposure to a few Langmuirs (L) of O2 (D). Previous electron-energy-loss spectroscopy (EELS) studies of single-crystal MgO(lOO) also showed a feature at ~2 eV, which grew upon electron irradiation [11, 39]. This feature was attributed to the formation of F-centers that could be quenched with O2. The appearance of the 2 eV feature in both MIES and UPS suggests that both surface and subsurface states contribute to the electron emission. The formation of small, non-metallic Mg clusters on MgO could be responsible for

382

e" dose:3kVDH0"'C/cm'

2 1 Binding energy [eV\ Fig. 6 Low binding energy region of MIE spectra of the Mg(lOO) film after various treatments: (A) the as-prepared MgOfilm;(B) after a 3kV, 3 x 10"^ C electron beam bombardment; (C) UPS acquired simultaneously with (B); (D) MIE spectrum of (B) after an 8-L O2 exposure. The inset shows a wide binding energy range plot of (B).

this feature in the MIES [21, 38]. However, it is unlikely that this mechanism is responsible for the "F" feature since the formation of these small Mg aggregates under electron bombardment requires (1) a high efficiency for oxygen vacancy formation and (2) a high surface mobility of the created Fcenters. Neither of these conditions exist since the electron beam fluxes were low (< 0.3 |iA) and the bombardment was carried out at low temperatures (ca. 100 K). The UPS exhibits this same feature, implying that there is a significant density of subsurface defects. From the relative intensities of the O 2p and "F" feature in the MIE spectrum (Fig. 6), the surface concentration is estimated to be on the order of 10^^ defects per cm^. Defects can have a significant effect on chemical activity. It is generally recognized that oxide surfaces with few defects are inert whereas those with a high defect density are reactive [40]. Because of its effectiveness as a titrant, NO was chosen to probe defect adsorption sites and characterize the near-

383 1

•

1

• 1 • 1 '

1 ' 11

/^

1

•

1 «

1 • 1 •

:MIES NO/MgO,, .

AJ

H

' /^ 'c CO (0

'

c B

'//^^^B. ^

r/^

^ ^

W/c

OL J 1

L

1

= ^ 1 3 LJ

:

B \17 A

UOM

V^

^=1

307 L ^ 1

20

18

16

14

•

1

12

1

1

10

8

6

Binding energy [eV]

15 10 Binding energy [eV]

5

Fig. 7 (A) MIE spectra of an MgO thin film at 95 K as a function of NO exposure. The Upper-most spectrum corresponds to that of the clean MgO(lOO) surface while the bottommost spectrum is that after a 307 L exposure of NO; (B) MIE spectra of defect-rich MgO thin film at 95 K as a function of NO exposure. The uppermost spectrum corresponds to the defect-rich MgO film and the top-most spectrum to the 307 L-thick NO coverage; (C) UP spectra of NO adsorbed on a defect-rich MgO film at 95 K. The dotted lines denote changes in the work function.

surface regions of the MgO film. It is also an especially relevant probe molecule due to interest in studying surface properties governing NO reduction [41], an industrially important reaction for automotive catalysts. Recent TPD experiments on vacuum-cleaved MgO(lOO) have shown exclusive adsorption ofNO on surface defects at T > 90 K [42]. Fig. 7A shows a waterfall plot of MIE spectra of NO exposed to a low-defect MgO thin film [grown on Mo(lOO)] at 90 K. In order to create a high density of defects, the MgO/Mo(100) film was annealed to 1100 K for ~1 min. and quenched to -100 K; this treatment has been shown in a previous HREELS study to generate defects at the near-surface [43]. The MgO/Mo(100) sample was thus similarly treated and NO was exposed to this surface. Fig. 8 shows a waterfall MIES plot on the defect-rich MgO thin film. A --0.6 eV work function increase upon saturation of the NO coverage (Figs. 7B and 7C) is indicative of charge transfer from the MgO to the adsorbate. There is a -25% attenuation of the O 2p intensity at the onset of the appearance of bands A, B and C, which we assign to the ionization of the 27i, 7a and \n states of N2O, respectively. The presence of N2O was verified by TPD, which showed the characteristic mass spectral cracking pattern for the molecule. Separate MIES/UPS and TPD measurements were carried out independently after N2O adsorption on the MgO/Mo(100) to further verify these peak assignments and preclude the

384 — I

'

1

'

1

'—

M g O ( 1 0 0 ) 95 K

as-prepared (O

c

(D

MgO 0 ( 2 p )

CO LU

with defects -O o

,

50

100

150

200

•

O—^

#—•

as-prepared

250

300

35

NO dose [L]

Fig. 8 Evolution of MgO 0(2p) and N2O 7a as a function of NO coverage. The open and filled symbols denote defect-rich and as-prepared Ti02, respectively.

possibility of N2O impurity in the NO gas. These peak positions match closely with reported gas-phase measured ionization energies for NjO [44]. The attenuation is more pronounced for the high-defect surface (Fig. 7B) where the O 2p peak is attenuated by 80%. The UPS waterfall plot (Fig. 7C) acquired simultaneously with MIES shows only minor changes (A, B and C bands not well-developed) during the adsorption, indicating that a relatively small fraction of the surface is covered with adsorbate. Using integrated TPD peak areas to calibrate the CH3OH coverage on MgO, Gtinster et al. [27] found that the fraction of adsorbate coverage could be quantified (approximately) via attenuation of the O 2p MIES signal. Based on the abatement of the O 2p intensities following NO adsorption on the MgO/Mo(100) systems, we estimate the NO saturation coverage to be 13±2 % and 35±5 % for the low- and high-defect MgO films (Figs. 7A and 7C), respectively [45]. Fig. 8 shows a plot of the MIES O 2p and N2O 7a peak intensities as a function of NO coverage [35]. Saturation occurs at varying coverages, depending upon the extent of thermal quenching. The as-prepared MgO films appear to contain a significant density of extended defects (i.e. steps, kinks and grain boundaries), which increased after thermally quenching the sample. A noticeable

385

broadening of the O 2p MIES peak is observed after the quench. This broadening, accompanied with a decrease in the band gap, is diagnostic of an increased extended defect density [20, 46]. The expansion of the extended defects coincide with the accumulation of N2O and appears to adsorb on the defect sites, perhaps near steps or grain boundaries. We postulate that dissociative adsorption of NO on MgO to produce N2O occurs via the following mechanism: N O ^ NOd -> Nd + 0 ->N20d

(1)

where the subscript "d" signifies a defect site, stimulating the production of N2O. Differences in the MIE spectra between the as-prepared MgO and the quenched MgO is strongly suggestive that defect density is an important factor controlling NO adsorption kinetics. 3. TITANIA Ti02 is an important reducible metal oxide support used for stabilizing nanosized noble metal clusters that catalyze a number of industrially relevant reactions in photocatalysts, chemical sensors and heterogeneous catalysis. These reactions include low-temperature CO oxidation [47-65], hydrogenation and partial oxidation of hydrocarbons [56, 58, 66, 67] and the selective oxidation of propylene to propylene oxide [66, 68]. Rutile TiO2(110) is the most thermodynamically stable form of Ti02 and has thus been the focus of numerous experimental and theoretical studies due to its well known electronic and geometric properties. In addition to using this single crystal, complementary experimental studies have also been performed using thin film Ti02 [69-72], which can be studied with the use of charged particle probes. 3.1. Structural and electronic properties of Au on TiO2(110) Single crystal TiO2(110) has been widely used for obtaining atomicallyresolved images via scanning probe microscopy. Images showing individual atom rows running along the [001] have been reported by various groups [7376]. The n-type semiconductor was found to be amenable for STM and charged particle spectroscopy after cycles of Ar^ bombardment and annealing to 700-1100 K [77]. This property makes the single crystal an ideal support for preparing and surface-analyzing supported metal cluster model catalyst systems. In particular, model planar catalysts consisting of finely dispersed Au and Ag clusters of varying size have been prepared on TiO2(110) for study. Since atomic-scale characterization of the unique physical and chemical properties of these nanosized supported metal cluster systems have been

386

discussed at length elsewhere [6, 7, 78-80], only one selected example will be illustrated in this section with emphasis on a recent landmark work, namely the finding that the onset of catalytic activity for a Au/Ti02 system coincides at a specific cluster size accompanying the appearance of non-metallic admetal clusters [81]. Fig. 9A shows a constant current topographic scanning tunneling micrograph (CCT-STM) of 0.25 monolayer equivalent (MLE) of Au deposited onto single crystal TiO2(110)-(lxl). The clusters were deposited at 300 K, followed by an 850 K anneal for 2 min [81]. The inter-atomic distances of the Ti02 rows are --0.65 nm. Three-dimensional clusters with dimensions of-2.6 nm X 0.7 nm (diam. and height, corresponding to a 2-3 atom thickness) are observed as bright protrusions and preferentially nucleate at the step edges. Fig. 9B shows STS taken at various points on the surface. The ST spectra had been acquired by positioning the STM tungsten tip at a predetermined point (over a cluster) and interrupting the feedback loop. The tunneling current (I) as a function of voltage bias (V) was measured, providing I-V curves measured for a variety of Au cluster sizes on the Ti02 surface. The "plateaus" between the inflection points of the I-V curves along the bias voltage axis (where the slope rapidly changes at the zero tunneling current) denotes the band gap. The STS lineshapes thus show that the electronic character of these deposits vary between that of a metal and non-metal depending upon its size. As the clusters enlarge, they gradually adopt metallic characteristics with an enhanced density of states (DOS) at the Ep, signified by more abrupt changes in the I-V curve for larger clusters. For example, the cluster with the 2.5 nm x 0.7 nm size has a larger band gap than that of the 5.0 nm x 2.5 nm cluster. Smaller Au clusters have non-metallic properties as signified by relatively large band gaps in the STS whereas larger clusters having bulk-like characteristics have virtually no band gap. Intermediate-sized 4.0 nm x 1.5 nm and 3.0 nm x 1.0 nm clusters also follow this cluster-size-band gap dependency. Interestingly, there is a correlation between both the Au cluster size and electronic structure with its catalytic activity for CO oxidation on Au/Ti02. Fig. lOA shows a plot of activity for the oxidation of CO [expressed as (product molecules)/((total atoms on surface sites) x (second)) or turnover frequency (TOF)] on Au clusters supported at 350 K on a Ti02 thin-film surface (prepared via "hot" filament evaporation on a Mo(lOO) substrate [82]) as function of cluster size. The product molecules (CO2) from the 2 CO + O2 -> 2 CO2 reaction was extracted from the reactor with a vacuum syringe, compressed and then analyzed using a gas chromatograph [81]. In a parallel experiment, a one-to-five C0:02 mixture had been reacted on the Au(0.25 MLE)/TiO2(110)-(lxl) surface at 40 Torr [61, 67, 81]. The cluster sizes obtained from STM of equivalent Au coverages on the respective Ti02 surfaces

387

Bias Voltage (V) Fig. 9 (A) A CCT-STM image of 0.25 MLE Au deposited onto Ti02(l lO)-(lxl) prepared just prior to a C0:02 reaction. The sample had been annealed to 850 K for 2 min.; (B) STS data acquired for Au clusters of varying sizes on the Ti02(l 10)(1x1). An STS of the Ti02 substrate, having a wider band gap than the Au clusters, is also shown as a point of reference.

v^ere used to calculate the TOF. For each point in Fig. lOA, a particular Au cluster size was prepared and then subjected to the C0:02 reaction. The activity for the Au/Ti02 catalyst exhibits a maximum 1.90 TOF at an average cluster size of 3.5 nm x 1.0 nm (300 atoms/cluster). Fig. lOB shows a plot of the STS band gaps measured over the same Au cluster size range used in the C0:02 reactions. Maximum activity coincides with the size-dependent metalto-nonmetal transition, showing the first evidence of the reaction being sensitive to the electronic structure of the Au particles. Fig. IOC shows a histogram showing the relative distribution of the Au sizes ranging from 2.04.0 nm in diameter at the maximum catalytic activity conditions, having drastically reduced band gaps of 0.2-0.6 V. It appears that 2-atom-thick clusters (with a 2.5-3.0 nm diam.) are optimally active for CO oxidation. Noteworthy is the fact that maximum activity is also observed at the same Au cluster size (2.9±0.5 nm diam.) on the (practical) high-surface area Ti02 catalyst [53] as in the (model) planar metal oxide, thus illustrating the utility of a surface science approach to obtain insight into "real world" systems.

388 2.20

o

Au clusters with a band gap of 0.2-0.6 V measured by STS

45 30

0

o

15

OH

0 0.0

2.0

4.0

6.0

8.0

10.0

Cluster Diameter (nm) Fig. 10 (A) The activity for CO oxidation at 350 K as a function of Au cluster size supported on TiOjCl lO)-(lxl) assuming a total dispersion of the Au. A 1:5 CO: O2 mixture had been used at a total pressure of 40 Torr. Activity is expressed as (product molecules/ ((total Au atoms) x (second)); (B) cluster band gaps measured by STS as a function of Au cluster size supported on TiO2(110)-(lxl). The band gaps were obtained while the corresponding topographic scan was acquired on various Au coverages ranging form 0.20.4 MLE. ( • ) 2D clusters, (O) 3D clusters, two-atom layers in height, (A) 3D clusters, three-atom layers or greater in height; (C) relative population of Au clusters (two-atom layers in height) that exhibit a band gap of 0.2-0.6 eV, as measured by STS of Au/ TiO2(110)-(lxl).

3.2. Adsorption of CO on Au/TiOj The nature of CO bonding to Ti02 is an important consideration for understanding how CO oxidation reactions are governed, in particular with regard to activity and selectivity. Previous theoretical and experimental work has shown that CO binds weakly onto the Ti02 support. Ab initio molecular orbital calculations indicate that CO absorbs through the carbon atom to the lattice Ti sites on the defect-free TiO2(110) surface with an adsorption energy

389

of 17 kcal/mol (0.73 eV per CO molecule) [83]. A similar result was obtained using a full potential linearized augmented plane-wave (FLAPW) method and along with the local density approximation (LDA); an adsorption energy value of 15 kcal/mol (0.64 eV per CO molecule) was determined [84]. In particular, the presence of metallic clusters supported on the metal oxide could dramatically affect the CO bonding properties. In order to better understand the effects of nanosized Au particles on CO bonding behavior, a study was performed comparing CO adsorption onto single crystal Au(llO) and a (0.5 MLE)Au/Ti02 planar model catalyst. This 0.5 MLE Au coverage corresponds to the formation of a particular mean cluster size in which maximum catalytic activity is known to occur [61]. Studies of CO adsorption on these respective substrates employing IRAS as a function of CO pressure and sample temperature were performed. Au(llO)(1x2) serves as an excellent model for comparison with highly dispersed Au/Ti02 systems since this reconstructed surface has a regular array of (111) facets with step-edges consisting of 6-coordinate atoms for every fourth atomic row [85-88]. This high density of low-coordination sites on the Au(l 10)-(lx2) is comparable to that of practical catalyst systems. IRAS is a useful technique for probing the chemical and physical properties of Au supported metal oxide systems. Coverage dependent absorption intensities of the CO probe molecule [89-91] can be measured in order to obtain isosteric energies of adsorption (Eads) via the Clausius-Clapeyron treatment of the adsorbate absorbance intensities [92]. As expounded by Woodruff et al. [90], relative shifts in frequency can be used to determine the relative position of the CO 2n^ state with respect to the Ep and thus provide insight to its bonding behavior to the metal. The 2n^ level lies relatively close to Ep at the onset of adsorption to the surface. Upon broadening of this band with increased CO coverage, the state can become less or more filled, depending upon whether it lies below (blue shift) or above the Ep (red shift), and can be experimentally measured using the surface vibrational spectroscopy. Eight IRAS data sets were acquired varying the CO coverage at a range of surface temperatures at pressures from 1.3 x 10"^ to 1.0 x 10"^ Torr on Au(110)-(lx2). At 1.3 X 10~^ Torr, the low pressure limit, the CO absorbance (with sample temperatures in parantheses) is a single entity at 2118 cm"^ (170 K) at the onset of adsorption. As the temperature was decreased, the CO coverage increased, resulting in a red shift of this feature to 2108 cm~^ (110 K) when saturation was reached. According to the Blyholder model [93], the red shift can be explained by a chemical effect due to decreased n back-bonding to the substrate concomitant with increased donation of electron density to the substrate upon CO adsorption. Comparable red shifts were observed at the

390

range of pressures used for these data sets. For instance, at the upper pressure limit used (1.0 x 10^ Torr) a frequency of 2116 cm~^ (230 K) was observed at the onset of adsorption, red shifting to 2106 cm"^ (110 K) at the CO saturation coverage (Fig. 11). The adsorbed CO measurements were red-shifted relative to gas-phase CO at 2143 cm"^ [4]. These isobaric data sets are rearranged into isotherms in Fig. 12. Integrated IRAS peak areas correlate linearly with coverage, setting the saturation intensity approximately equal to a 50% CO coverage (0 = 0.5) [94]. A representative set of isosteres are plotted at every 1.6% coverage increment up to and including 35%) (Fig. 13). CO adsorption energies can then be determined by the slopes of these isosteres via the Clausius-Clapeyron relation: d(\nP) d{\IT)

(2)

ads

R

where R is the gas constant. E^ds values are found to be -12 kcal/mol below a 6%

70 1 ' 1 ' 1 i T i T - ^

f"

2106 cm

//

^ >-

.-^60

5%

=>55

2

130 K

F^^ V 45 f ^ <•^'

.•^35

f x-^ . U

170 K

otfi

f

L

c

03

•e

y

<.40 H

150 K

83%

L

2 50 [•

110K

/

,'X '

X -

*

•

f

,y

210 K

>/" i> / 1 _/ / ;§2o / r "^ ••' / / 5 15 r j /y O-'

230 K

c

190 K

1%

16 cm'

~

250 K 2100

2050

2000

Frequency (cm')

Fig. 11 IRAS intensities at various cover -ages (and substrate temperatures) at 1x10"^ Torr of CO adsorbed on Au(l 10) -(1x2).

o -A

>'

m C/) 1 42% ^•

^^

^•+ -J 39% gj. -x J ,>-" 0 J 35% Q-

'< ^ ' • ^ J 32%
•'*

\

o J

(1

J

16% — • •130K

K ^^-A J 13% —- •-+X-" • 1140 50K o* •110% / •160K •170K / _ .... 1— -. 180 K y J 6% - D - • 1 9 0 K /^ •zs*

p

•

y'^

/%

•1

3% 0%

L-kjJ 3 % -19-18-17-16-15-14-13-12-11-10 •

'

•

*

•

i_Ldi 1 1 L J . 1 1 '1

• '

%

•

\

p. -o 1

•

"0—*^' y 0-<^<^*

2150

-^ '••' /

L '' ' 5 rr ^fi'' 0

0% 2200

1 r

45%

r nm n

+

y^

7^'

!E25h

$2%

1 1

. . o » 200 K — A - 210 K - • v - 220 K —O— • 2 3 0 K

InBCO Fig. 12 IRAS adsorption isotherms at varying pressures of CO adsorbed on Au(l 10)-(lx2).

391

17% CO coverage on the Au(110)-(lx2) surface and plateaus to a lower limit of ~9 kcal/mol above a 22% coverage [95]. For comparison, the CO adsorption energy was similarly measured onto a Ti02 thin-film (deposited onto a Mo(lOO) single crystal) covered with a 0.5 MLE of Au evaporated onto it (1 MLE = 1.387 x 10^^ Au atoms/cm^). It has been shown from a previous experiment, calibrating the Au coverage using the AES "break" [96] in conjunction with cluster diameters measured using STM, that Au particles with an average distribution centered at ~30 A in diam. form on the TiOs substrate at the 0.5 MLE coverage [61]. This is the particle size at which the metal-nonmetal transition (measured by STS band gaps) of the Au clusters is observed. The -30 A diam. cluster size gives rise to maximum catalytic activity conditions for CO oxidation. Unlike the CO adsorption behavior on the unreconstructed Au(lll)-(lx2) surface, the vibrational frequency for CO adsorbed on the 0.5 MLE Au/Ti02 remains relatively constant throughout the wide pressure range. For instance, at 190 K an absorption at 2122 cm~^ is observed from pressures as low as 3.2 x 10"^ to 3.2 X 10^ Torr, blue-shifted as compared to the bulk Au surface (Fig. 14). |—i-n—1—1

n

\\ \\

CO coverage J • 1.6%

^ 8.0%

\ \ "\

•

\ \ \ \ \ t

L 1

Q." c

[

\ \

r

\

\ t \ \ \

\ \

\

\ A

\

S'^S \ \

[

\

\

1

\

\

\

\

x\

V

\ \ \ *\

-16 h

\

H

\

U\ \\ t\ \

-17 h

r L

4.0

\

VW M

-15 h

-18

\

15%

2122 cm"'

W

• 21% H + 27% N X 35% W

\

[ ^ VV \ -13

1—•—1—^1 — ^ — 1 — ' —

» 1 »

\

4.5

5.0

\A

\

5.5

y

6.0

\

6.5

7.0

7.5

1000/T(K) Fig. 13 Representative isosteres obtained from a Claussius-Clapeyron data treatment of CO adsorbed on Au( 110)-( 1x2).

8.0

2200

2150

2100

2050

Frequency (cm'^)

Fig. 14 IRAS intensities of CO adsorbed on (0.5 MLE)Au/Ti02 at 190 K at varying pressures.

392

As the substrate temperature is cooled, more CO accumulates on the surface. The integrated peak areas of the IRAS absorbances can then be measured and followed by a Claussius-Clapeyron treatment of these data sets. An E^ds of--12 kcal/mol is thus found at CO coverages of 17% and lower, identical to that of the low coverage of CO on Au(l 10)-(lx2). The indistinguishable isosteric E^ds values for both the Au(110)-(lx2) and (0.5 MLE) Au/Ti02 surfaces is suggestive of similar (if not identical) adsorption dynamics, owing to the electronic character of metallic Au. One difference that stands out in comparing CO adsorption behavior between these surfaces is that the CO absorbance frequency on the 0.5 MLE Au/Ti02 system is blue-shifted by 6 cm~^ with respect to CO adsorbed onto the Au(110)-(lx2) surface and remained constant at 2122 cm"^ with increasing CO coverage up to saturation. We attribute the blue shift to be a result of an increase in the energy level position of the CO 2n^ band (initially above the Fermi level) and/or a decrease in the energy of the 5 a state (with respect to the Fermi edge) as compared to CO adsorbed on the bulk Au substrate. The blue shift is thus suggestive of relatively less n back-bonding of the CO to the (0.5 MLE)Au/Ti02 and stronger surface interactions via the CO 5a state on this surface. The degree to which these specific electronic interfacial interactions contribute to enhanced CO oxidation activity (in the presence of the metal-to-nonmetal transition Au clusters [61,81]) remains an issue for further investigation. 4. THIN-FILM ALUMINA AND MIXED-METAL OXIDES AI2O3, due to its inertness, chemical stability and mechanical strength, is an important ceramic material in catalysis, coatings and microelectronics. In heterogeneous catalysis, AI2O3 has been used to support Ag clusters to promote the partial oxidation of ethylene to ethylene oxide [68, 97]. In order to facilitate the study of metal cluster supported interactions with this oxide surface, a means of obtaining atomically-resolved images of thin-film AI2O3 was devised. The method to be described follows the common theme of using oxide thin-films to overcome insulating properties that would otherwise hamper surface analysis. In addition, our thin-film methodology has recently been extended for studying the interfacial properties of mixed-metal oxide systems (using results from single-component oxide studies as a benchmark) to gain insight into the relationship between catalytic activity and atomic-scale electronic and chemical structure. Two systems, NiO-MgO and NiO-CaO, are studied to gain insight on how their interfacial structure affects NO adsorption.

393

4.1 STM imaging of AI2O3 thin-films STM imaging of AI2O3 was made feasible using a thin-film deposition approach [98]. As a relatively wide band gap material, AI2O3 (Eg = 8.7 eV [97]) cannot be made conductive by doping. To enhance conductivity, thin films (ca. 5-30 A thick) of AI2O3 were deposited on Re(OOOl) at 300 K using "hot" filament evaporation under ambient oxygen pressure (1x10"^ Torr) [99103]. This same methodology was successful for preparing MgO [another wide band gap material (Eg = 7.8 eV)] thin-films for STM analysis [30]. In the case of AI2O3 thin films, Re(OOOl) was chosen as the substrate due the fact that its lattice constant (2.76 A) matches closely with the in-plane O^" lattice constant (2.77 A) of the single crystal AI2O3 [99, 100]. Fig. 15 shows various coverages of AI2O3 film growth on the Re substrate: (A) 0.4, (B) 2.2 and (C) 5.2 monolayer equivalents (MLE). The features imaged correspond to Al cations only; the oxygens are not imaged since direct tunneling into (or out of) the O 2p state is ~3 eV below the Fermi level, which is beyond the operational limit of the microscope (Omicron STM-1). A useful calibration plot comparing the MLE of AI2O3 produced from an Al doser flux used to make the thin film is shown in Fig. 16. The MLE AI2O3 coverage was determined using the integrated volumes measured by STM, assuming that the particles consisted of single-layer-thick structures (Fig. 15A). The flux of the Al evaporator was calibrated via AES of Al deposited onto the Re(OOOl) surface. XPS showed that the AI2O3 film was stoichiometric. The Auger "break" in the plot of Al/Re AES signal versus deposition time was used to assign the completion of the first monolayer. The deposition rate of the oxide film (synthesized via evaporating Al under ~10~^ Torr O2) was directly proportional to the Al doser flux. The 1.0 ML Al coverage on Re based on the metal evaporation rate was defined as 1.4 x 10^^ atoms/cm^. A 0.3 ML Al flux corresponds to a 0.4 MLE coverage of AI2O3 (Fig. 15A). The areas within the unit cell (which can be imaged by STM) can be determined based on the known atomic structures of Al [2] and AI2O3 [104, 105]. One Al atom in an Al unit cell (fee, 4.05 A lattice constant) occupies 7.10 A^ while one Al atom in AI2O3 occupies {[(2 x 2.76 A) X (3 X 2.76 A)/ 4 atoms] =} 11.43 A^. A conversion factor of 1.61 is thus derived to convert Al to AI2O3 coverage (in MLE units). The slope of the least squares fit of the plot (n = 4, r^ = 0.922) shows that a deposition of 1.0 ML of Al produces 1.4 MLE of AI2O3 (Fig. 16). This coverage estimation assumes a layer-by-layer growth mode, which had been observed in a prior ISS study of an epitaxial AI2O3 thin-film deposition on Ta(llO) [101]. It was necessary to increase the bias voltage of the STM tip in order to image thicker coverages of AI2O3 due to diminished conductivity. Voltages from 2.0-5.0 V were needed to obtain stable images of the 0.4-5.2 MLE coverages (Figs. 15A-C). Images could not be obtained beyond a 8.7 MLE coverage. For relatively thick AI2O3

394

coverages, even the LEED diffraction spots become diffuse, suggesting greater disorder as compared to the thinner films [98].

A ^

w™ H »

.|Hi'^%f»^^^-

20

B

0.2

0.4

0.6

ML Al from doser

Fig. 16 Calibration plot showing coverage ofAl203 produced on Re(OOOl) (STM measured) from the Al doser flux (calibrated with AES). H.X''*^^

3

f*

••;

v^-:•^'',

•vxl--^'." V -

^•v'^*>

.

Fig. 15 Epitaxially grown AI2O3 thin-films on a Re(OOOl) substrate: (A) 100 nm x 100 nm STM image (2.0 V, 0.30 nA) of 0.4 MLE AI2O3; (B) 100 nm X 100 nm STM image (3.0 V, 0.50 nA) of 2.2 MLE AI2O3; (C) 200 nm x 200 nm STM image of 5.2 MLE ofAl203 (5.0 V, 0.20 nA).

395

4.2. NO adsorption on MgO-NiO and CaO-NiO NiO, MgO and CaO thin films were deposited on a Mo(lOO) single crystal via "hot" filament evaporation under ambient O2 pressure [13, 106-110]. In preparing the mixed oxide system, a first layer of oxide was deposited (20 MLE-thick, calibrated using the AES "break" point) and followed by deposition of a second oxide layer of a different type. LEED, employed to monitor the long range order during the epitaxial growth of the layered oxide films, showed that the NiO/CaO(100) film had a more diffuse (1x1) pattern as compared to the sharp LEED images obtained from the NiO/MgO(100), indicating that the latter exhibited superior long-range surface order [110]. Also noteworthy is the fact that the NiO/MgO(100) and MgO/NiO(100) showed more highly ordered LEED patterns than on the Mo(lOO) surface. This effect is attributed to a closer lattice matching between the MgO and NiO as compared to either oxide with Mo(lOO). NiO, MgO and CaO all have a simple rocksalt structure, but with differing degrees of lattice mismatch. NiO(lOO) [2.947 A] and MgO(lOO) [2.978 A] have very similar lattice constants [in

200

300

400

500

600 200

300

400

500

600

Kinetic Energy (eV) Fig. 17 ISS intensities of NiO on (A) MgO(lOO) and MgO (B) NiO(lOO) at various overlayer coverages. The oxide overlayers were grown at 300 K and annealed to 500 K.

396

brackets] with a 1% mismatch whereas NiO(lOO) and CaO(lOO) [3.402 A] have a markedly contrasting mismatch (15%). Similar lattice constants facilitate epitaxial growth whereas dissimilar ones lead to a significant strain between the mixed-metal oxide interface, resulting in a relatively poor epitaxy. ISS was performed in order to elucidate the growth mode during the mixed oxide film growth. Fig. 17 shows two ISS stackplots of a NiO film grown on an MgO(lOO) single crystal and a MgO thin-film grown on NiO(lOO). Three peaks are seen and assigned to the scattering of O, Mg and Ni at 273, 357 and 487 eV kinetic energy, respectively. At higher coverages, the substrate cation is attenuated while the overlayer cation increases in intensity. The oxygen signal remains unchanged during the bimetallic oxide formation. A similar series of experiments were performed for a NiO film growth on CaO(lOO) and CaO film growth on NiO(lOO). The same kinetic energy peak positions were observed for Ni, Mg and O along with a peak at 435 eV, signifying Ca being scattered off the surface. An identical trend was observed with the substrate cation decreasing during attenuation, with the overlayer cation increasing during adsorption and the relative oxygen signal remaining constant. Normalized, integrated peak area intensities are plotted as a function overlayer 1.0

.0.8

• •

NiO/MgO(100) MgO/NiO(100)

• A

NIO/CaO(100) CaO/NiO(100)

0) 0.6 •a "x O 0)

I(0 0.4

nD


0.0 1 2 3 4 Overlayer Coverages (ML)

Fig. 18 ISS relative substrate oxide intensities as a function of overlayer coverages for NiO/MgO, MgO/ NiO, NiO/CaO and CaO/NiO. The lines are guides to the eye, not fits.

397

oxide coverage for each of the mixed oxides [NiO on MgO(lOO), MgO on NiO(lOO), CaO on NiO(lOO) and NiO on CaO(lOO)] and shown in Fig. 18 [110]. Normalization was achieved using the relative sensitivities of Ni, Mg and Ca from their respective single-component oxide systems. The NiO/MgO(100) and MgO/NiO(100) mixed oxides exhibit identical film growth behavior. The NiO/CaO(100) and CaO/NiO(100) also exhibit identical film growth behavior, but different from that of the NiO/MgO(100) and MgO/NiO(100) growth. The Ni [or Mg] ISS intensity decreases more rapidly than those of Ni [or Ca] with respect to the CaO [or NiO] coverage. At a 1MLE coverage, for instance, the substrates are -90% covered for NiO/MgO(100) and MgO/NiO(100), but only --65% covered for the CaO/NiO(100) and NiO/CaO(100). These results signify greater wetting properties between NiO and MgO than between NiO and CaO. In addition, it can be concluded that NiO [or MgO] grows in a layer-by-layer fashion on MgO [or NiO(lOO)] within the first monolayer whereas significant 3D clustering occurs within the first monolayer for CaO [or NiO] growth on NiO(100)[orCaO(100)]. The chemical properties of the NiO-CaO and NiO-MgO were probed using NO adsorption followed by TPD analysis [110]. Figs. 19A and 19B show TPD stackplots of varying exposures of CaO and MgO on NiO(lOO), respectively, followed by a saturation exposure of NO. A gradual shift of the NO peak temperature maximum (T^J is observed for the CaO film, from 230 K on clean NiO(lOO) to 213 K on (2 MLE)CaO/Ni(100). A larger shift is observed for NO adsorption on an MgO film, from 230 K on clean NiO(lOO) to 180 K on (2 MLE)MgO/Ni(100). This contrast is more explicitly shown in Fig. 19C. For an identical oxide overlayer coverage, a much larger shift of NO T^ (from the NiO sites) occurs on the MgO/NiO(100) as compared to CaO/NiO(100), signifying different growth modes. The ISS data (Fig. 18) thus shows a layer-by-layer growth mode for MgO on NiO(lOO) in the first monolayer while significant 3D clustering is observed for CaO/NiO(100) even at submonolayer coverages. Due to the greater wetting ability of MgO as compared to CaO, a substantially larger coverage of CaO is needed to completely obscure the effects of the underlying NiO(lOO) substrate. To better illustrate this point, the NO T^^ is plotted as a function of exposed NiO(lOO), determined by the integrated peak areas of the NO TPD (Fig. 20); MgO and CaO both affect NO adsorption similarly when normalized to the area of the NiO(lOO) substrate covered. Coadsorbed MgO and CaO have a destabilizing effect on NO adsorption on NiO(lOO), with the largest effects on NO adsorption occuring at the initial coverages. This behavior is consistent with a long-range, substrate mediated interaction with adsorbed NO. Furthermore, the matching (or mismatching) of lattice constants of the single-component oxides

398

i

NO TPD on MgO/NIO(100) i

MgO

1

I

t

NO from

J

N"'

0 o]

-^•
-—-^ 100

150 200 250 Temperature (K)

300

^ol V"

^^^

V . 0.0

150 200 250 Temperature (K)

100

CaO/NiO(100)

e,v^{ML)

300

1.0

0.5 1.0 1.5 Overlayer Coverages (ML)

Fig. 19 NO TPD acquired after a 6 = 0.10 NO coverage on Ni(lOO) with increasing (A) CaO and (B) MgO coverages; (C) NO TPD desorption temperatures as a function of overlayer coverages. Overlayer Covered Area (%) 0

20

40

60

80

100

230

5^ 220H 0) 3

^210 E o Kc o "S. 2 0 0 k_

o en a> O

• D

o 2

MgO/NiO(100) CaO/NiO(100)

190

180 100

80

60

40

20

Available NiO sites (%)

Fig. 20 Plot of normalized NO desorption temperatures as a function of MgO and CaO overlayer coverages.

2.0

399

in the NiO-CaO and NiO-MgO systems can have a large effect in the growth mode and thermal stability of adsorbed NO. MgO has a closer lattice match to the NiO than CaO that thus more effectively blocks NO adsorption sites on the NiO(lOO). In an IRAS study by Xu et al. [110], the clean NiO(lOO) surface showed a weak intensity NO frequency at 1855 cm"^, which was assigned to surface defects on the NiO [111, 112]. After covering the NiO with 0.2 MLE of MgO followed by a 0.2 L NO adsorption, this peak disappeared signifying NO preferentially blocking the defect sites. The dominant IRAS feature at 1800 and 1820 cm"^ (assigned to multiple bonding sites) at varying coverages of MgO on NiO (0-2 MLE) showed no shift in wavenumber for the 0.2 L adsorbed NO, indicating that charge transfer played a minimal role in NO bonding.

5. HIGH PRESSURE STM STUDIES OF SUPPORTED GOLD CLUSTERS In order to better simulate practical catalyst systems using a surface science approach, considerable strides have been made towards bridging the "materials gap" and the "pressure gap", enabling correlation of data obtained under UHV conditions with those at elevated pressures. Recent efforts have been directed towards in-situ monitoring of atomic-scale surface structure and reaction intermediates at pressures spanning UHV and actual catalytic conditions (10"^^ -10~^ Torr) using STM [113]. In the first example, a variable temperature and pressure STM set-up is used to obtain atomically-resolved images of Au metal clusters adsorbed onto a TiO2(110) single crystal. The breakthrough advancement worth noting is the ability to image specific, targeted areas on the surface in-situ while varying the reaction gas pressure conditions over a wide span, from -10"^ to 4 Torr (section 5.1) [45]. In the second example, a hexagold phosphine-stabilized cluster, Au6(PPh3)6[BF4]2, used as a precursor for a Au/Ti02 high-surface area catalyst preparation, is deposited onto a planar oxide support under ambient pressure conditions (1 atm) and introduced into the UHV-STM system. Atomically resolved images of these clusters as unagglomerated, single-unit entities were obtained, demonstrating the potential of using pre-selected Au cluster cages for creating monodispersed, sizeselected admetal clusters onto planar (or high surface area) metal oxide supports.

400

5.1. Imaging Au on TiOjCllO) at elevated pressures with variable temperature and pressure STM The growth behavior of admetal clusters on metal oxide surfaces at elevated pressure and temperature has been found to be a governing factor in strong metal support interactions (SMSI). In a model catalyst study of Pd nanoparticles supported on TiO2(110), temperature dependent enhancements were found to control re-oxidation around the Pd particles which were interpreted to be a result of spillover of oxygen from the Ti02 support causing encapsulation of Pd (by Ti02 ^t 573 K, 10"^ Torr O2) and its subsequent deactivation [114]. Elevated pressures are known to profoundly alter the surface structure of catalyst systems. Changes in the structural morphology of Rh particles supported on TiO2(110) induced by high pressure CO conditions (ca. 10"^-10"^ Torr at 300 K) has been attributed to disruption of Rh crystallites (1-3 nm in diam.) into smaller units [115]. Following this same investigative approach, the effect of high O2 pressure and temperature (450 K) on the morphology of Au/Ti02(l 10) was investigated using STM. Fig. 21 shows a series of CCT-STM images of a 0.3 MLE Au coverage on TiO2(110) followed by various O2 pressure treatments at elevated temperature [45]. Au clusters decorating the step edges are imaged. Clusters exhibiting pressure-induced morphological changes are circled with a dashed line while unchanged clusters are circled with a solid line. The accompanying bar graph shows the volume for Au cluster " 1 " for each of oxygen pressures. Note that as the O2 pressure is increased, the volume of the Au " 1 " cluster diminishes, particularly in the dashed regions. There is a continuous disappearance of specific Au clusters (within the dashed circles) with increasing O2 exposure and time. Upon prolonged exposure, structures with a steady state 1-2 nm diam. are formed and particularly pronounced in Fig. 2ID. Two protrusions, shown on the right side of each of the CCT-STM images (Figs. 21A-D) within the solid circles, are particularly stable in each of the elevated O2 and temperature treatments. Two tentative explanations are offered for these features. The first scenario is that the Au nanoparticles become encapsulated underneath the Ti02 substrate via its growth about the cluster upon exposure to high O2 pressures. A similar phenomenon was observed with nanosized Pd clusters supported on Ti02 with high O2 pressures inducing the encapsulation [115]. The encapsulation was stimulated via surface segregation of interstitial Ti ions and oxygen spillover onto the Pd admetal. The second possible mechanism involves the disruption of the Au-Au bonds in the clusters to form smaller units. An analogous mechanism was proposed for oxygen-induced sintering of supported Ag nanoclusters on Ti02 [79]. In obtaining these

401

o ilVIAGE

A

Fig. 21 CCT-STM images of a 40 nm x 40 nm area acquired at 450 K at various O2 pressures: (A) PO2 = 0 Torr; (B) PO2 = 10"^ Torr; (C) PO2 = 2 Torr, 90 min.; (D) PO2 = 4 Torr, 150 min.

images, several additional structxiral observations are noted in regard to the underlying TiOsCllO) substrate. At elevated O2 pressures, the (1x2) termination is unstable, even at 300 K. Islands and small clusters oriented along the <001> direction, forming a (1x1) structure, is a result of disruption and removal of missing row atoms. The formation of the (1x1) structure has been reported under similar conditions owing to the segregation of Ti^ ions from the Ti02(l 10) substrate [116, 117]. An important cautionary note to consider in order to effectively obtain atomically-resolved images at elevated pressures (aside from increased susceptibility to acoustic noise) is that relative adhesive interactions of the Au clusters between the STM tip and the TiO2(110) substrate can result in

402

experimental artifacts. The selection of the type of STM tip to be used is an important factor for obtaining artifact-free images. The type of tip that should be used depends on the gas environment under which imaging is performed [118]. For example, at certain imaging conditions, tip-cluster interactions may exceed cluster-support interactions, resulting in either cluster movement or adhesion to the tip. Such modifications to the tip can result in artificially enlarged images. Etched tungsten tips could only be used up to 10~^-10~^ Torr; at higher pressures stable images could not be obtained. Pt/Ir tips were needed to image at atmospheric pressure conditions. This effect can be minimized to some degree by imaging with a 3-6 GQ tunneling resistance. Nevertheless, in this operational mode the onset of high O2 pressure-induced weakening of the cluster-support interaction (for Au clusters 2-5 nm in diam.) can be observed at ca. 10"^ Torr (leading to nanoparticles adhering onto the tip). Subsequent cluster-tip interactions are more pronounced at higher O2 pressures. Also, increasing the sample temperature > 300 K facilitates disruption of Au clustersupport interactions. 5.2. Imaging solution-deposited Au6(PPh3)JBF4]2 deposited TiOjCllO) Iwasawa and co-workers have recently begun employing phosphinestabilized Au complexes, Au(PPh3)(N03) and Au9(PPh3)8(N03)3, for preparing highly dispersed, nanosized Au particles on high surface area and asprecipitated oxide supports for catalytic reactions [119-122]. The use of phosphine-stabilized Au clusters provides a means of controlling monodispersion via synthesis of various-sized Au cluster precursors as starting materials (e.g. AU4, Aun, AU55, ^^c). Upon deposition and calcination on high surface area Ti02, the Au aggregated into 30 nm diam. particles, inhibiting its catalytic properties. More success was obtained on the as-precipitated Ti(0H)4, which resulted in clusters 3 nm in diam. that exhibited high activity for CO oxidation [122]. These Au complexes, while showing promise as precursors for practical catalyst preparations, have yet to be prepared on model metal oxide surfaces, making them amenable for surface sensitive probes and hence atomic-scale surface structure characterization. To this end, a phosphine-stabilized Au cluster, Au6(PPh3)6[BF4]2 (Au^L^) analogous to those used by Iwasawa and co-workers, has recently been deposited onto a single crystal Ti02(l 10) via solution deposition under ambient conditions (1 atm, 300 K) and probed with STM [123]. During initial attempts in depositing the Au^L^ onto the TiO2(110) crystal in CH2CI2 solvent, large agglomerations due to Au cluster aggregation of the complex (50-100 A in diam.) were observed. To avoid this problem, it was necessary to pre-treat the crystal with acetone in order to obtain the finely dispersed complex on the TiO2(110) crystal. Measured surface charge

403

Fig. 22 (A) 100 nm X 100 nm CCT-STM image of a freshly sputter-cleaned Ti02(l 10) surface. (B) 100 nm x 100 nm CCT-STM view of homogene -ously distributed AugLg clusters deposited onto TiO2(110). ( C ) 3 0 0 n m x 3 0 0 nm CCT-STM wide-range view of the Au6L6/Ti02(l 10).

100

150

200

250

300

404

densities and zeta potentials of colloidal mineral oxide suspensions, such as yAI2O3, Ti02 and Si02, have been reported to be strongly influenced by organic solvents [124-126]. A study by Rodrigues et al. [126] has shown that the presence of acetone lowers the net proton surface charge density (G^) of silica colloids as measured by potentiometric titration (at pH = 8). We postulate an analogous phenomenon occurring with the Au6L6/TiO2(110) system enabling the Au^L^ (which has a net charge of+2 without the BF4~ counterions) to better disperse and adhere onto the substrate. Fig. 22 shows a series of reproducible CCT-STM images of (A) a freshly, sputter-cleaned TiO2(110) to serve as a reference point (i.e. a blank), (B) Au^L^ deposited on the Ti02(l 10) surface and (C) a wide-range view of the Au6L6/TiO2(110). All images were obtained ex-situ using an Omicron UHV STM-1 scanner. The hexagold particles were widely and homogeneously distributed over the 1 cm^ TiO2(110) surface area. Multiple areas of the Au6L6/TiO2(110) that were sampled showed the same structural morphology (Fig. 22C). These particles are remarkably stable after prolonged exposure to air for 72 hours. Surface roughening is evident in Figs. 22B and 22C. However, the diagonal TiO2(110) terraces are still apparent. Using the integrated areas of the STM (assuming a one-atom-layer thick coverage), 10% of the TiO2(110) substrate was covered. XPS showed ample signal from the Au 4f core level, indicating its presence, and no signal from either B or F, indicating that the counterions were apparently removed by solvent evaporation. STM line profiles of these clusters showed an average height of 8±1 A (n = 10), which is approximately the same volume as expected for the Au^L^ entity. Experiments are currently under way for optimizing ligand removal in order to prepare these surfaces for various catalytic reactions (e.g. CO oxidation, propylene oxidation, etc.) in hopes of monitoring atomicscale structure of these practical admetal clusters during their performance. 6. SUMMARY Recent significant strides have been made on the development of available techniques that can be used to investigate the interfacial activity of reactions occurring at well-defined metal oxide interfaces. MIES/UPS has been an especially useful technique for elucidating geometric molecular orientation of molecules as they adsorb onto the metal oxide; the high surface sensitivity of the probe (in conjunction with TPD) can be exploited to quantitate near-surface defects and its effects on molecular adsorption. STM and STS has been shown to be particularly effective in monitoring the admetal cluster size and electronic structure effects on model planar oxide catalyst systems, showing a correlation of relative activity with these properties. Further advances have been made in gaining insight on the nature of the CO bond to these highly reactive systems

405

and further extended into the study of mixed metal oxide systems using IRAS, LEED and TPD. In the area of scanning probe microscopy, progress has been made in bridging the pressure gap for obtaining atomic-resolution images at realistic catalytic reaction conditions. These tools will play an important role in the frontier areas of surface science, permitting us to monitor molecularlevel interactions in order to control the size, structure and spatial distribution of adsorbed admetal particles and/or reactants (e.g. NO, CO) on oxides for the rational design of chemically active surfaces. ACKNOWLEDGMENTS We acknowledge with pleasure the support of this work by the Robert A. Welch Foundation, the Department of Energy, Office of Basic Sciences, Division of Chemical Science and the Dow Chemical Company. We also thank Jeffrey A. Stultz for proofreading the manuscript. REFERENCES 1. D. P. Woodruff and T. A. Delchar, Modem Techniques of Surface Science, 2nd ed.; Cambridge University Press: Cambridge, 1994. 2. G. A. Somorjai, Introduction to Surface Chemistry and Catalysis; Wiley: New York, 1994. 3. D. W. Goodman, J. Vac. Sci. Technol. A 14 (1996) 1526. 4. S. C. Street and D. W. Goodman, Annu. Rev. Phys. Chem. 48 (1997) 43. 5. S. C. Street and D. W. Goodman, in: Growth and Properties of Ultrathin Epitaxial Layers; (Elsevier, Amsterdam, 1997) 375. 6. D. R. Rainer and D. W. Goodman, J. Mol. Catal. A 131 (1998) 259. 7. T. P. St. Clair and D. W. Goodman, Top. Cat. 13 (2000) 5. 8. A. J. Martin and H. Bilz, Phys. Rev. B 19 (1979) 6593. 9. G. Lakshmi and F. W. de Wette, Phys. Rev. B 22 (1980) 5009. 10. G. Lakshmi and F. W. de Wette, Phys. Rev. B 23 (1981) 2035. 11. V. E. Henrich, G. Dresselhaus and H. J. Zeiger, Phys. Rev. B 22 (1980) 4764. 12. T. Urano, T. Kanaji and M. Kaburagi, Surf Sci. 134 (1983) 109. 13. M.-C. Wu, J. S. Comeille, C. A. Estrada, J.-W. He and D. W. Goodman, Chem. Phys. Lett. 182(1991)472. 14. J.-W. He, J. S. Comeille, C. A. Estrada, M.-C. Wu and D. W. Goodman, J. Vac. Sci. Technol. A 10 (1992) 2248. 15. E. Lindholm, Faraday Disc. Chem. Soc. 54 (1972) 200. 16. H. Kubota, T. Hirooka, T. Fukuyama, T. Kondow, K. Kuchitsu and A. J. Yencha, J. Electron Spectrosc. Relat. Phenom. 23 (1981) 417. 17. J. Giinster, G. Liu, V. Kempter and D. W. Goodman, Surf Sci. 415 (1998) 303. 18. K. Y. Yu, J. C. McMenamin and W. E. Spicer, Surf Sci. 50 (1975) 149. 19. S. Hiifner, Photoelectron Spectroscopy; Springer-Verlag: Berlin, 1995.

406 20. D. Ochs, W. Maus-Friedrichs, M. Brause, J. Giinster, V. Kempter, V. Puchin, A. Shluger and L. Kantorovich, Surf. Sci. 365 (1996) 557. 21. J. Gunster, G. Liu, V. Kempter and D. W. Goodman, J. Vac. Sci. Technol. B 16 (1998) 996. 22. L. H. Tjeng, A. R. Vos and G. A. Sawatzky, Surf. Sci. 235 (1990) 269. 23. S. Masuda, M. Aoyama, K.Ohno and Y. Harada, Phys. Rev. Lett. 65 (1990) 3257. 24. S. C. Street, Q. Guo, C. Xu and D. W. Goodman, J. Phys. Chem. 100 (1996) 17599. 25. M. Xi, M. X. Yang, S. K. Jo and B. E. Bent, J. Chem. Phys. 101 (1994) 9122. 26. W. Ueda, H. Ohowa, K. Iwasaki, T. Kuwabara, T. Ohshida and Y. Morikawa, Stud. Surf Sci. Catal. 90 (1994) 35. 27. J. Gunster, G. Liu, J. Stultz and D. W. Goodman, J. Chem. Phys. 110 (1999) 2558. 28. J. Gunster, G. Liu, J. Stultz, S. Krischok and D. W. Goodman, J. Phys. Chem. B 104 (2000) 5738. 29. H. Yamakado, M. Yamauchi, S. Hoshino and K. Ohno, J. Phys. Chem. 99 (1995) 17093. 30. M. C. Gallagher, M. S. Fyfield, J. P. Cowin and S. A. Joyce, Surf. Sci. 339 (1995) L909. 31. D. W. Turner, C. Baker, A. D. Baker and C. R. Brundle, Molecular Photoelectron Spectroscopy; John Wiley and Sons, Ltd.: London, 1970. 32. M. S. Banna, B. H. McQuaide, R. Malutzki and V. Schmidt, J. Chem. Phys. 84 (1986) 4739. 33. V. Cermak and A. J. Yencha, J. Electron Spectrosc. Relat. Phenom. 11 (1977) 67. 34. J. H. Lunsford, Catal. Today 6 (1990) 235. 35. A. Kolmakov, J. Stultz and D. W. Goodman, J. Phys. Chem. B, in press. 36. T. Gotoh, Y. Fukunaga and S. Takagi, Surf Sci. 357-358 (1996) 690. 37. L. N. Kantorovich, J. M. Holender and M. J. Gillan, Surf Sci. 343 (1995) 221. 38. L. N. Kantorovich, A. L. Shluger, P. V. Sushko, J. Gunster, P. Stracke, D. W. Goodman and V. Kempter, Faraday Discuss. 114 (1999) 173. 39. P. R. Underhill and T. E. Gallon, Solid State Comm. 43 (1982) 9. 40. V. E. Henrich and P. A. Cox, The Surface Science of Metal Oxides; Cambridge Univ. Press: New York, 1994. 41. X. Zhang, A. B. Walters and M. A. Vannice, J. Catal. 146 (1994) 568. 42. R. Wichtendahl, M. Rodriguez-Rodrigo, U. Hartel, H. Kuhlenbeck and H.-J. Freund, Phys. Stat. Sol. (a) 173 (1999) 93. 43. M.-C. Wu, C. M. Truong and D. W. Goodman, Phys. Rev. B 46 (1992) 12688. 44. K. Kimura, S. Katsumata, Y. Achiba, T. Yamazaki and S. Iwata, Handbook of Hel Photoelectron Spectra of Fundamental Organic Molecules; Halsted Press: New York, 1981. 45. A. Kolmakov and D. W. Goodman, Catal. Lett., in press. 46. L. N. Kantorovich, A. L. Shluger, P. V. Sushko and A. M. Stoneham, Surf Sci. 444 (2000)31. 47. M. Haruta, S. Tsubota, T. Kobayashi, H, Kageyama, M. J. Genet and B. Delmon, J. Catal. 144(1993)175. 48. S. D. Lin, M. Bollinger and M. A. Vannice, Catal. Lett. 17 (1993) 245. 49. F. Boccuzzi, A. Chiorino, S. Tsubota and M. Haruta, J. Phys. Chem. 100 (1996) 3625. 50. M. A. Bollinger and M. A. Vannice, Appl. Cat. B 8 (1996) 417. 51. N. W. Cant and N. J. Ossipoff, Catal. Today 36 (1997) 125. 52. K. Fukushima, G. H. Takaoka, J. Matsuo and I. Yamada, Jpn. J. Appl. Phys., Part I 36 (1997)813.

407

53. G. R. Bamwenda, S. Tsubota, T. Nakamura and M. Haruta, Catal. Lett. 44 (1997) 83. 54. S. Minico, S. Scire, C. Crisafulli, A. M. Visco and S. Galvagno, Catal. Lett. 47 (1997) 273. 55. Z. M. Liu and M. A. Vannice, Catal. Lett. 43 (1997) 51. 56. M. Haruta, Catal. Today 36 (1997) 153. 57. Y. lizuka, H. Fujiki, N. Yamauchi, T. Chijiiwa, S. Arai, S. Tsubota and M. Haruta, Catal. Today 36 (1997) 115. 58. M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Azuma and M. Haruta, Catal. Lett. 51 (1998) 53. 59. M. A. P. Dekkers, M. J. Lippits and B. E. Nieuwenhuys, Catal. Lett. 56 (1998) 195. 60. S. Tsubota, T. Nakamura, K. Tanaka and M. Haruta, Catal. Lett. 56 (1998) 131. 61. M. Valden, S. Pak, X. Lai and D. W. Goodman, Catal. Lett. 56 (1998) 7. 62. J.-D. Grunwaldt and A. Baiker, J. Phys. Chem. B 103 (1999) 1002. 63. J.-D. Grunwaldt, C. Kiener, C. Wogerbauer and A. Baiker, J. Catal. 181 (1999) 223. 64. E. D. Park and J. S. Lee, J. Catal. 186 (1999) 1. 65. H. Liu, A. I. Kozlov, A. P. Kozlova, T. Shido, K. Asakura and Y. Iwasawa, J. Catal. 185 (1999) 252. 66. T. Hayashi, K. Tanaka and M. Haruta, J. Catal. 178 (1998) 566. 67. M. Valden and D. W. Goodman, Isr. J. Chem. 38 (1998) 285. 68. M. Haruta, in: 3rd World Congress on Oxidation Catalysis; (Elsevier Science, Amsterdam, 1997) 123. 69. D. Martin, F. Creuzet, J. Jupille, Y. Borensztein and P. Gadenne, Surf. Sci. 377-379 (1997) 958. 70. D. Abriou, D. Gagnot, J. Jupille and F. Creuzet, Surf. Rev. Lett. 5 (1998) 387. 71. K. Luo, T. P. St. Clair, X. Lai and D. W. Goodman, J. Phys. Chem. B 104 (2000) 3050. 72. Q. Guo, W. S. Oh and D. W. Goodman, Surf Sci. 437 (1999) 49. 73. D. Novak, E. Garfunkel and T. Gustafsson, Phys. Rev. B 50 (1994) 5000. 74. S. Fischer, A. W. Munz, K.-D. Schierbaum and W. Gopel, Surf. Sci. 337 (1995) 17. 75. A. Szabo and T. Engel, Surf Sci. 329 (1995) 241. 76. P. W. Murray, N. G. Condon and G. Thornton, Phys. Rev. B 51 (1995) 10989. 77. C. Xu, X. Lai, G. W. Zajac and D. W. Goodman, Phys. Rev. B 56 (1997) 13464. 78. X. Lai, T. P. St. Clair, M. Valden and D. W. Goodman, Prog. Surf. Sci. 59 (1998) 25. 79. X. Lai, T. P. St. Clair and D. W. Goodman, Faraday Discuss. 114 (1999) 279. 80. C. C. Chusuei, X. Lai, K. Luo and D. W. Goodman, Top. Cat., in press. 81. M. Valden, X. Lai and D. W. Goodman, Science 281 (1998) 1647. 82. W. S. Oh, C. Xu, D. Y. Kim and D. W. Goodman, J. Vac. Sci. Technol. A 15 (1997) 1710. 83. H. Kobayashi and M. Yamaguchi, Surf Sci. 214 (1989) 466. 84. Z. Yang, R. Wu, D. W. Goodman and Q. Zhang, J. Chem. Phys., submitted. 85. W. Moritz and D. Wolf, Surf Sci. 88 (1979) L29. 86. M. Garofalo, E. Tosatti and F. Ercolessi, Surf Sci. 188 (1987) 321. 87. G. Binnig, H. Rohrer, C. Gerber and E. Weibel, Surf Sci. 131 (1983) L379. 88. C. Hofner and J. W. Rabalais, Surf Sci. 400 (1998) 189. 89. P. Hollins and J. Pritchard, Surf Sci. 89 (1979) 486. 90. D. P. Woodruff, B. E. Hayden, K. Prince and A. M. Bradshaw, Surf Sci. 123 (1982) 397. 91. P. Hollins, K. J. Davies and J. Pritchard, Surf Sci. 138 (1984) 75. 92. C. M. Truong, J. A. Rodriguez and D. W. Goodman, Surf Sci. 271 (1992) L385.

408

93. G. Blyholder, J. Phys. Chem. 68 (1964) 2772. 94. R. Ryberg, Surf. Sci. 114 (1982) 627. 95. D. C. Meier, V. Bukhtiyarov and D. W. Goodman, submitted. 96. L. C. Feldman and J. W. Mayer, in: Fundamentals of Surface and Thin Film Analysis; (North-Holland, New York, 1986) Chapter 6. 97. B. Ealet, M. H. Elyakhloufi, E. Gillet and M. Ricci, Thin Solid Films 250 (1994) 92. 98. X. Lai, C. C. Chusuei, K. Luo, Q. Guo and D. W. Goodman, Chem. Phys. Lett. 330 (2000) 226. 99. Y. Wu, E. Garfunkel and T. E. Madey, J. Vac. Sci. Technol. A 14 (1996) 2554. 100. Y. Wu, E. Garfunkel and T. E. Madey, Surf. Sci. 365 (1996) 337. 101. P. J. Chen and D. W. Goodman, Surf Sci. Lett. 312 (1994) L767. 102. M.-C. Wu and D. W. Goodman, J. Phys. Chem. 98 (1994) 9874. 103. C. Xu, X. Lai and D. W. Goodman, Faraday Discuss. 105 (1996) 247. 104. D. R. Jennison, C. Verdozzi, P. A. Schultz and M. P. Sears, Phys. Rev. B 59 (1999) R15605. 105. C. Verdozzi, D. R. Jennison, P. A. Schultz and M. P. Sears, Phys. Rev. Lett. 84 (1999) 799. 106. J.-W. He, C. A. Estrada, J. S. Comeille, M.-C. Wu and D. W. Goodman, Surf Sci. 261 (1992) 164. 107. C. M. Truong, M.-C. Wu and D. W. Goodman, J. Chem. Phys. 97 (1992) 9447. 108. M.-C. Wu, C. M. Truong and D. W. Goodman, J. Phys. Chem. 97 (1993) 4182. 109. J. S. Comeille, J.-W. He and D. W. Goodman, Surf Sci. 306 (1994) 269. 110. C. Xu, W. S. Oh and D. W. Goodman, J. Phys. Chem. B, in press. 111. E. E. Platero, S. Coluccia and A. Zecchina, Surf Sci. 171 (1986) 465. 112. E. E. Platero, B. Fubini and A. Zecchina, Surf Sci. 179 (1987) 404. 113. G. A. Somorjai, Appl. Surf Sci. 121/122 (1997) 1. 114. R. A. Bennett, P. Stone and M. Bowker, Faraday Discuss. 114 (1999) 267. 115. A. Berko, G. Menesi and F. Solymosi, J. Phys. Chem. 100 (1996) 17732. 116. M. Li, Wilhelm, Hebenstreit, L. Gross, U. Diebold, M. A. Henderson, D. R. Jennison, P. A. Schultz and M. P. Sears, Surf Sci. 437 (1999) 173. 117. R. A. Bennett, P. Stone, N. J. Price and M. Bowker, Phys, Rev. Lett. 82 (1999) 3831. 118. J. A. Jensen, K. B. Rider, Y. chen, M. Salmeron and G. A. Somorjai, J. Vac. Sci. Technol. B 17 (1999) 1080. 119. Y. Yuan, K. Asakura, H. Wan, K. Tsai and Y. Iwasawa, Catal. Lett. 42 (1996) 15. 120. Y. Yuan, K. Asakura, H. Wan, K. Tsai and Y. Iwasawa, Chem. Lett. (1996) 755. 121. Y. Yuan, A. P. Kozlova, K. Asakura, H. Wan, K. Tsai and Y. Iwasawa, J. Catal. 170 (1997) 191. 122. Y. Yuan, K. Asakura, A. P. Kozlova, H. Wan, K. Tsai and Y. Iwasawa, Catal. Today 44(1998)333. 123. C. C. Chusuei, X. Lai, K. A. Davis, E. K. Bowers, J. P. Fackler and D. W. Goodman, submitted. 124. M. E. Labib and R. Williams, J. Colloid Interface Sci. 97 (1983) 356. 125. M. Kosmulski, J. Colloid Interface Sci. 135 (1990) 590. 126. F. A. Rodrigues, P. J. M. Monteiro and G. Sposito, J. Colloid Interface Sci. 211 (1999) 408.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

409

Chapter 10

Principles of Reactivity from Studies of Organic Reactions on Model Oxide Surfaces A. B. Sherrill and M. A. Barteau Center for Catalytic Science and Technology, Department of Chemical Engineering University of Delaware, Newark, Delaware, 19716 USA

1. INTRODUCTION AND SCOPE The surface science of metal oxides has come of age over the past decade, stimulated by a variety of technological challenges. Examples include the needs to understand metal-metal oxide interfaces in semiconductor devices, sensors, and catalysts. Consequently, new techniques for surface analysis and modeling have been exploited to probe the complex surface-adsorbate interactions that govern chemical processes on oxide surface. Several recent reviews have explored relationships between the surface and bulk characteristics of metal oxide materials, and the ways in which these characteristics subsequently influence the surface chemical pathways [1-5]. Instead of simply adding a new edition to that list, two "case-studies" are presented, each highlighting the critical roles of surface anion-cation pairs and the redox properties of the surface in determining the product slate. The reactions of small organic molecules on titanium dioxide surfaces provide the central focus of these examples. Comparisons with reactions on single crystal surfaces of other oxides illustrate some of the key structural and electronic issues in oxide surface reactivity and the importance of local interactions between surface sites and adsorbed molecules. 2. TITANIUM DIOXIDE AS A MODEL FOR TRANSITION METAL OXIDES Transition metal oxides exhibit a number of properties that are conducive to catalytic applications, including thermal and mechanical stability needed to survive severe reaction conditions. More importantly, transition metal cations can typically exist in several different valence states. Titanium dioxide has a bulk band gap energy of about 3.2 eV, but electrons can be placed in (3d) gap states

410

about 0.7 eV below the Fermi level created by vacuum annealing. In contrast, for main group metal oxides such as MgO (bulk band gap ca. 7.7 eV), the d orbitals are not accessible [4]. Further, the lattice oxygen anions electronically isolate cations, thus surface-adsorbate bonding on oxides is highly localized. Consequently, transition metal oxides are often compared to transition metal complexes in solution. This so-called "cluster-surface analogy" has been used to advance molecular approaches to heterogeneous catalysis [6]. The surface science literature of titanium dioxide is replete with studies of the thermodynamic stability of different surface structures, as well as studies of the adsorption of metals, light gases, light and condensable organic molecules to the those surfaces. The full arsenal of surface analytical techniques has been applied to study Ti02 surfaces, including electron and vibrational spectroscopies, scanning probe microscopies, desorption spectroscopies, and computational models of varying complexity [1,3-5]. Experimental studies of single crystal Ti02 surfaces have utilized almost exclusively the rutile crystal structure, which is more stable than the anatase or brookite structures. Most of the work reviewed here was carried out on the (110) or (001) crystal planes, so a short introduction to the structures and coordination environments of those planes is presented first. The Ti02(l 10) crystal face (figure 1) is the most stable face of rutile titanium dioxide as well as the most thoroughly characterized. TiO2(110) has been imaged with low-energy electron diffraction (LEED), scanning tunneling microscopy (STM) and atomic force microscopy (AFM), and the electronic states have been characterized with X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS). Polished TiO2(110) exhibits a (1x1) LEED pattern, although the surface can reconstruct into a (1 x 2) pattern during high temperature annealing in vacuum. The TiO2(110) surface is characterized by alternating parallel rows of bridging oxygen anions (and corresponding six-fold coordinate titanium cations beneath) and five-fold coordinate titanium cations [4,7-15]. The TiO2(001) surface (figure 2) is comparatively less stable, as all of the surface cations are four-fold coordinate in the ideal structure [4]; because of this, the TiO2(001) surface facets upon annealing to high temperatures. Upon annealing to 750 K, the surface facets into a {011} reconstruction with all of the cations in a five-fold coordinate arrangement. Further annealing to 900 K creates {114} faceting of the surface [4]. This structure ideally produces equal populations of four, five, and six-fold coordinate cations (thus retaining an average coordination number of five). A third reconstruction into a (7V2 x V2) R45° arrangement has recently been reported for (OOl)-oriented crystals annealed to 1473 K and rapidly quenched. The TiO2(001) surface has been imaged with AFM and STM, and has been characterized using Auger electron

411

Fig. 1. Structure of the unrelaxed Ti02(l 10) surface. Black and grey spheres are Ti and O atoms, respectively. H. Idriss and M.A. Barteau, Advances in Catalysis, v. 45 (2000). Reproduced by permission of Academic Press.

Fig. 2. Structure of the unreconstructed TiO2(001) surface. Black and grey spheres are Ti and O atoms, respectively. H. Idriss and M.A. Barteau, Advances in Catalysis, v. 45 (2000). Reproduced by permission of Academic Press.

412

spectroscopy (AES), XPS, LEED, UPS, and near-edge X-ray absorption fine structure (NEXAFS) [4,16-22]. 3. CASE STUDY I: FORMIC ACID AS A PROBE OF SURFACE PROPERTIES Formic acid is a popular molecule for probing the catalytic properties of metal oxides [23-28]. The selectivity of its decomposition has frequently been used as a measure of the acid-base properties of oxides. This is a tempting generalization to make; oxides that produce dehydration products (H2O and CO) are described as acidic oxides, while their basic counterparts produce dehydrogenation products (H2 + CO2). It has been shown that in many cases the product selectivity is better connected to the surface redox behavior of the oxide [29]. Thus, more reducible surfaces produce higher yields of CO2. Consequently, particular attention has been paid in surface science studies to the interaction between adsorbed formate ions (the primary reaction intermediate) and surface metal cations, as well as to the participation of lattice oxygen anions in the surface reaction mechanism. 3.1. Formic Acid Decomposition on Metal Oxides As discussed in detail below, several reaction pathways are available to formic acid in addition to the "simple" dehydration and dehydrogenation reactions noted. Interactions between neighboring molecules and between the adsorbed molecules and the surface both influence the product selectivity. Formic acid may adsorb molecularly on oxide surfaces as shown in reaction (1). Formic acid also adsorbs by proton transfer to surface oxygen anions (reaction 2). The resulting intermediate is a surface formate. Details about the structure of this intermediate on different crystal planes will be presented below. The formate intermediate may decompose by dehydration, producing carbon monoxide, or by dehydrogenation, producing CO2. Gas phase molecules are labeled with "(g)" (e.g., HCOOH(g)) Adsorbed molecules are marked with asterisks (e.g., HCOOH*), surface lattice oxygen atoms are denoted by Os, and hydroxyl groups formed by protonation of them are designated with OsH. (1) (2) (3) (4)

HCOOH(g) ^ HCOOH* HCOOH(g) + O s ^ HCOO* + OsH HCOO* -^ CO(g) + Os + H* HCOO* -> C02(g) + H*

(dehydration, e.g. MgO [26]) (dehydrogenation, e.g. ZnO [25])

Formic acid decomposition has been examined on a number of metal oxides using a wide assortment of surface analytical techniques to probe the decomposition pathways. One might expect the most facile adsorption to occur

413

on the most basic surfaces. Indeed, formic acid adsorbed on thin films of MgO (grown on Mg single crystals), and on stoichiometric and defective single crystals, easily dissociates into surface formate, even at 150 K and below [26,3033]. The formate intermediate dehydrates on MgO(lOO) surfaces exclusively at 520 K. The amount of formic acid adsorbed on MgO does not appear to be structure sensitive. The MgO(lll) and (100) surfaces adsorbed the same amount of formic acid as measured by XPS [30]. On both surfaces thermal decomposition produced exclusively CO, the nominal dehydration product, demonstrating that dehydration is not an indication of surface acidity [26,30,32]. Magnesium oxide is widely considered to be a strong base, but the formation of surface formates also occurs readily on bare surfaces of oxides that have been typically considered to be less basic (e.g. ZnO) or even acidic (e.g. Ti02). Such observations point out the limited utility of these labels. The zinc-terminated ZnO(OOOl) surface also readily dissociates formic acid to form surface formates; the subsequent decomposition of these yields both dehydration and dehydrogenation products at 575 K [25,34-39]. The dehydrogenation reaction results in the reduction of surface zinc cations as evidenced by desorption of Zn metal atoms coincident with CO2 desorption [39]. Pretreating the ZnO(0001)-Zn surface with chlorine atoms prior to formic acid exposure blocks Zn^"^ sites (each Zn cation on the surface has a single coordination vacancy, and roughly three zinc cations are blocked by each chlorine atom) and enhances the selectivity of the dehydrogenation reaction to form CO2. Coincidentally, chlorine was observed to promote the abstraction of lattice oxygen anions by the surface hydrogen produced by surface formate decomposition [37]. Earlier work on the ZnO(0001)-Zn surface demonstrated a correspondence between the extent of surface reduction produced by the adsorption and reaction of different molecules, e.g., formic acid, formaldehyde, and methanol, from which formates were synthesized. The greater the extent of surface reduction produced by the formate generation reaction, the greater was the selectivity to CO observed in formate decomposition [25]. The effect of chlorine adsorption on formate chemistry is consistent with this trend. Chlorine is electron withdrawing; consequently, chlorine adsorption causes the energy of the valence band maximum to shift away from the vacuum level and create a depletion layer. Thus, the chlorine adsorption may be thought of as increasing the extent of surface oxidation [4,37]. The change in surface electronic character and the corresponding increase in the formate dehydrogenation selectivity following chlorine adsorption are consistent with the selectivity trends for formate reactions on the clean ZnO (0001) surface: the more highly oxidized the surface, the higher the dehydrogenation selectivity. Formic acid decomposition has been studied on the (110), (001), and (100) surfaces of Ti02 [23-25,40-51]. The degree to which surface reducibility influences the reaction paths (e.g., dehydrogenation vs. dehydration; unimolecular reactions vs. bimolecular ones) will be explored in more detail

414

below. Here it is sufficient to understand that the primary reaction of formic acid in temperature programmed desorption experiments on titanium dioxide single crystals is via dehydration at 550 K [43]. The TiOiCllO) surface facilitates both molecular and dissociative adsorption of formic acid. The dissociative adsorption of formic acid (to form surface formates) is mediated by surface oxygen anions; low-energy electron diffraction studies have shown that formate is ordered into ( 2 x 1 ) domains with formates bridging the surface titanium cations. This ordered layer is disrupted on heating; formate may recombine with surface hydroxyl groups to desorb formic acid (reverse of reaction 2) or it may decompose to form the dehydration products CO and H2O, as well as small amounts of CO2 and H2 [43]. Net dehydration reactions dominate the surface decomposition chemistry of formic acid on {011}-TiO2(001) faceted surfaces. Temperature programmed desorption experiments on these surfaces revealed that formic acid desorption resulted from both recombinative (390 K) and decomposition (560 K) processes. X-ray photoelectron spectroscopy measurements demonstrated that formic acid adsorbed on the {011}-TiO2(001) surface at 165 K both as formate and formic acid. This reconstruction of the TiO2(001) surface exposes titanium cations in a five-fold coordinate manner only, leaving one free coordination site per cation for interaction with adsorbates. Formic acid adsorbed on TiO2(110) at 180 K also formed a layer containing both molecularly and dissociatively adsorbed species; heating to 270 K left only adsorbed formate ions in a (2 x 1) pattern. This ordered arrangement vanished on heating to 350 K, coincident with the desorption of water and formic acid; further heating decomposed the remaining formate species, forming CO2 and CO as the principal carbon-containing products [43,52]. In both cases the reaction products are generated from exclusively unimolecular decomposition [45]. Surfaces with four-fold coordinate Ti"*"^ cations are capable of bimolecular coupling of surface formate to form formaldehyde; these sites can be created by more severe thermal faceting [1,3,23]. Reaction (5) illustrates this, where R may be a hydrogen atom or an alkyl group, and where Os denotes surface lattice oxygen. R

(5)

R

co,^ 0/ -Ti— / \

^

o R-c-R

+ CO.2(g)

This bimolecular coupling is a structure-sensitive reaction, and it illustrates a key characteristic of metal oxides: multiple coordinative unsaturation of surface metal cations may facilitate coupling of ligands in a manner similar to that for unsaturated metal complexes in solution. Examples of other coupling reactions

415

observed on TiO2(001) surfaces include alkyne cyclotrimerization (on reduced surfaces) and alkoxide etherification reactions (on oxidized surfaces). Alkynes are readily cyclized into aromatics on reduced surfaces of TiO2(001) exposing Ti^^ cations on the surface; this chemistry is analogous to catalysis of alkyne cyclization by low-valent mononuclear transition metal complexes in solution. In both cases, the metal center must be able to complete a two-electron oxidation and reduction cycle in order to form a metallacycle intermediate and eliminate the aromatic product [53,54]. A reaction more closely related to the bimolecular ketonization reaction described above is alcohol etherification via coupling of surface alkoxides. It was observed that {114}-faceted TiO2(001) surfaces containing adsorbed methoxide species produced dimethyl ether, while the {011}-faceted TiO2(001) surface produced no detectable ether products [55]. Titania surfaces with oxygen vacancies reduce oxygen-containing intermediates such as formates and methoxides in the process of the filling the oxygen vacancy [41]. Vacancy healing is also important in the reductive coupling of aldehydes and ketones to form symmetric olefins on reduced TiO2(001) surfaces. This reaction is analogous to the well known McMurry reaction, wherein slurries of low-valent titanium carry out the coupling process [20]. Thus, not only do coordinatively unsaturated metal cations affect the local electronic structure of the oxide surface, but they also serve as active sites for a variety of bimolecular coupling reactions with potential synthetic utility. The formation of formaldehyde from formic acid is not expected to occur on stoichiometric TiO2(110) because the surface cations are exclusively fivecoordinate, and formaldehyde formation in the absence of reduced surface sites has been shown to be dependent on the presence of low-coordinate metal cations on titanium dioxide surfaces. Iwasawa et al. noted the absence of formaldehyde from the product slate of formic acid reaction products in their studies on the TiO2(110) surface [43]. 3.2. The Participation of Oxygen Vacancies in Formic Acid Decomposition The extent of lattice oxygen incorporation into the decomposition products is linked to the surface redox properties of the oxide. Recent SSIMS and FTRAIRS studies have provided initial evidence for the transient formation of oxygen vacancies on the surface of TiO2(110). Henderson has proposed that surface oxygen vacancies may explain the formation of trace amounts of formaldehyde from formic acid on TiO2(110) (figure 3) [41]. Both reactions (6) and (7) are proposed to occur below 500 K; the water produced from formic acid exposure desorbed at 475 K, before the onset of formate decomposition to form CO. Formaldehyde and CO2 were produced in minor quantities relative to the production of CO [41]. Formaldehyde was formed from formic acid on reduced

416

Oxygen vacancy formation on TiO 2(110) during bridging hydroxyl combination water desorption

bridging ^'''''P^

\ > J % L

A v X

^ - ^ vacancy newly formed five-coordinate Tl'^+ cation sites (one cation hidden from view)

Fig. 3. Schematic model for the creation of oxygen vacancies on Ti02 (110) by combination of bridging hydroxyl groups to form water. M.A. Henderson, Journal of Physical Chemistry B, V. 101 (1997). Reproduced by permission of the American Chemical Society.

surfaces of TiO2(001), and the amount of oxygen deposited on the surface was consistent with that required to convert the available Ti^"^ sites to Ti"^^ [23]. Consequently, the low yield of formaldehyde and CO2 likely represents the small surface population of oxygen vacancies on the crystal surface in the work by Henderson [41]. Experiments with ^^0-enriched surfaces did demonstrate incorporation of lattice oxygen into the CO product of formic acid decomposition, illustrating the relative ease of lattice oxygen exchange and abstraction [40,41]. (6) (7) (8)

HCOOH* + Os ^ HCOO* + OsH 2 OsH -> H20(g) + Os + vacancy 2 HCOO* + vac. -> Os + H2C0(g) + C02(g)

The formation of formaldehyde at reduced titanium surface cation sites can also occur, but via formate reduction rather than via bimolecular formate coupling. (9)

HCOO* + H* -> H2CO + Os

417

Idriss et al noted that the decomposition temperature of surface formate is nearly constant across a wide range of different TiO2(001) faceted and reduced surfaces, suggesting that this reaction occurs via a conmion intermediate, whether the products are those of dehydration or reduction. These workers proposed scission of the formate C-H bond as the rate-determining step [23]. Catalytic experiments on single crystals of TiO2(110) suggested that transportlimited reaction conditions favor unimolecular dehydration, and that lower surface temperatures favor bimolecular dehydrogenation [43]. Thus, it is important to realize that interactions with neighboring molecules may influence the surface chemistry as much as surface-adsorbate interactions. The case for oxygen vacancies is supported by recent work with Fourier Transform Reflection-Absorption Infrared Spectroscopy (FT-RAIRS) by Hayden and coworkers that indicated the presence of two types of surface formate on the TiO2(110) surface [49]. Adsorption of formic acid at 300 K yielded a surface coverage of about 0.6 ML (from XPS) [50,51]. One type of surface formate (accounting for about 2/3 of the surface coverage) has the OCO plane aligned in the <001> direction. The less populous type of surface formate is oriented orthogonally to the first, in the <110> direction (figure 4). FT-RAIRS indicated that this second variety of formate interacts with oxygen vacancies on the surface created during the formation of formate along the <001> direction in the manner

6.49 Ang. 0^.oxygan(toprow)^ I formate

titanium

hydrogen

Fig. 4. Schematic showing the possible adsorption sites for formate species on TiOa (110). B.E. Hayden, A. King, and M.A. Newton, Journal of Physical Chemistry v. 103 (1999). Reproduced with permission of the American Chemical Society.

418

of reactions (6) and (7) above. The population of the second variety of formate was increased when formic acid was adsorbed at 450 K, where vacancy creation by hydroxyl recombination would be more energetically favorable [49]. Impingement on the surface of low-density electron beams can also be used to form Ti^"^ sites on the crystal surface. Recent work has shown that additional surface capacity for bridging formate created by these new surface sites corresponds to the amount of additional formic acid uptake by the surface [50,51]. The adsorbed formate did not heal the oxygen vacancy until the formate decomposed, lending additional credence to a reaction step other than defect healing controlling the formation of formaldehyde at Ti^"^ sites. Sputter reduction of the TiO2(110) surface greatly increased the surface capacity (to ca. 0.92 ML), and subsequent exposure of a formate-saturated, sputtered surface to O2 produced almost no defect healing [49]. The formation of formaldehyde and the attendant formation of CO2 reveals a likely interaction with oxygen vacancies, and when contrasted with decomposition chemistry on the MgO(lOO) surface, highlights the importance of surface redox properties in carboxylic acid chemistry on metal oxide surfaces. 3.3. Tracking Formic Acid Decomposition with Scanning Probe Microscopy Appealing as the traditional probes for formic acid chemistry may be, recent work with temperature programmed scanning tunneling microscopy (STM) and non-contact atomic force microscopy (NCAFM) has generated new interest in oxide surface science by dynamically illustrating surface diffusion and reaction of formates [8,24,56-60]. The strong interaction of formate with the metal oxide surface helps make the surface reaction visible to STM at elevated temperatures; in contrast, surface diffusion of adsorbates on metal surfaces above cryogenic temperatures is often fast on the time scale of the microscope, making identification of important reaction intermediates difficult [61]. Iwasawa and coworkers have published several accounts of STM and NCAFM investigations of formic acid decomposition on the stoichiometric surface of TiO2(110), and have imaged surface titanium cations, oxygen anions, and oxygen vacancies [8,24,52,56-59,61]. Formate ions on the surface were observed in ordered (2x1) domains with STM, in agreement with earlier LEED and FT-RAIRS work. Surface diffusion and reaction pathways were probed with STM as well. The migration of formate ions was observed by using the microscope tip to clean off a portion of the scanned area and monitoring the back-filling of the bare patch from the surrounding domains. The bare area was initially roughly square; subsequent images indicated that surface migration was anisotropic, favoring transport along the rows of titanium cations between the oxygen ridges. When viewed under catalytic conditions, Iwasawa et al.

419 Step

/

(001]

Ti203 rowj Upper terrace

Step Lower terrace

Fig. 5. A post-reaction STM image of the Ti02 (110) surface heated to 450 K and exposed to 1.3 X 10'^ Pa HCOOD gas for 10 min. One product particle is marked by the white circle. One ( 2 x 1 ) unit of the formate monolayer is shown with the solid rectangle. Y. Iwasawa, H. Onishi, K. Fukui, S. Suzuki, and T. Sasaki, Faraday Discussions, v. 114 (1999). Reproduced by permission of the Royal Society of Chemistry.

observed the formation of bright particles across the surface in addition to the (2x1) array of formate (figure 5). The bright features were located atop oxygen anion positions and were about the same size as formate ions. Based on the size of the bright particles and the preference for dehydrogenation on surfaces with high formate coverage, surface carbonate ions were proposed. Such species may be formed as intermediates in the production of CO2, e,g. via reactions such as the bimolecular ketonization of acetate [46,47]. Thus, surface formate and hydroxyl species may not be the only participants in formic acid decomposition [8,24,56-59]. Now consider the fate of formic acid adsorbed on two (1x2) reconstructions of TiO2(110). When annealed in vacuum at temperatures near 900 K, doublerow strands of Ti203 were formed on the surface of the crystal. Formic acid dosed onto these surfaces did not make formate [8,25]. The inability of the acid to undergo dissociation is reminiscent of the oxygen-polar ZnO(OOoi) surface; on that surface, the overlayer of oxygen anions obscured the zinc cations. Thus, by extension, there must not be Ti cations accessible to the adsorbed formic acid. Recent work by Bowker on TiO2(110) surfaces annealed to higher temperatures (ca. 1200 K) has revealed the cross-linked structures on the surface [62]. The surface included added rows of Ti02 running in the <001> direction.

420

Temperature programmed STM experiments under oxygen led to new growth of (1 X 1)-Ti02 on the crystal surface. Formic acid dissociated on the cross-linked surface to make formate ion. As the temperature ramp progressed, the population of formate diminished while small islands of (1 x 1) termination formed on the terraces of the crystal. Bowker and coworkers proposed that the decomposition of formate ions supplies oxygen atoms to interact with interstitial titanium cations and generate the Ti02 islands [60]. The nucleation of new (1x1) titanium dioxide is not based on the healing of oxygen vacancies, but it does highlight the role that the surface redox environment plays in formate decomposition. The change in reactivity of formate on (1 x 2) reconstructed TiO2(110) shows that although the coordination environment surrounding the cation occupies a central role in determining product selectivity, the surface electronic structure should not be ignored. 3.4. Catalytic Reactions of Formic Acid on Titanium Dioxide (110) Iwasawa and coworkers conducted molecular beam experiments to investigate the catalytic decomposition of formic acid on single crystal TiO2(110) surfaces [43]. These experiments revealed competing reaction channels (figure 6) for formate dehydration (rO and dehydrogenation (r2) on the single crystal surface under catalytic conditions that could not be accounted for by the chemistry observed in TPD experiments alone. At surface temperatures below 650 K, the predominant catalytic pathway was concluded to be dehydrogenation of formic acid through a bimolecular reaction between an adsorbed formate and a molecule of formic acid. An activation-energy of 15 ± 10 kJ/mol and a reaction order of 0.9 ±0.1 with respect to the gas phase formic acid pressure (at 650 K) were estimated from catalytic rate studies. The surface coverage of adsorbed formate at saturation was estimated at 0.5 ML based on C (Is) and O (Is) XPS measurements; this value was consistent with the (2x1) LEED pattern observed when formic acid was adsorbed on the clean surface at 300 K. The adsorbed formate ions were assumed to align themselves between top-layer oxygen anion rows in a bidentate fashion. Thus, each bridging formate is coordinated to two five-fold coordinate Ti"^^ cations. UHV TPD data indicated that surface formate decomposes at 570 K. The first-order rate dependence mentioned above was also observed at 600 K, suggesting that formic acid must interact with formate ions on the near-saturated surface [43]. A mechanism for catalytic dehydrogenation was proposed to explain these observations. The researchers used deuterium-labeled formic acid to discriminate against background water and hydrogen. In the scheme below, Os represents surface lattice oxygen, while oxygen originating from formic acid is denoted by Of*. Adsorbed species are labeled with (*).

421

H0.5

0

Fig. 6. Reaction rates simulated as a function of reaction temperature with a collision frequency of 0.5 s'^ site'\ ri: dehydration; T2: dehydrogenation; 0: the coverage of formates. H. Onishi, T. Aruga and Y. Iwasawa, Journal of Catalysis, v. 146 (1994). Reproduced by permission of Academic Press.

(10)

(11) (12)

-OsD DCOOD(g) + Os ^ D C O O * + ( D C O O * -> CO(g) + OfD D C O O * + DCOOD(g) -> C02(g) + D2(g) + D C O O *

The break-point temperature in dehydration (above which the rate was temperature insensitive) matched the maximum temperature for dehydrogenation, suggesting that a conmion intermediate exists for each reaction, and that the product selectivity is determined by interactions with other molecules and the surface. Above 650 K, the catalytic dehydration channel dominates, but the rate-determining step changes above 700 K. Below 700 K, the reaction rate is nearly independent of the partial pressure of formic acid (ca. 0.2 order). Above 700 K, the rate of the reaction is essentially independent of temperature, implying that reaction is limited by formic acid adsorption and dissociation; thus, above 700 K, the rate becomes first-order with respect to the partial pressure of formic acid. Higher pressures of formic acid over the crystal surface should therefore increase the transition temperature - this behavior was observed by Iwasawa and coworkers, and the turnover frequency for catalytic dehydration approached the collision frequency of formic acid at high

422

temperatures. Consequently, a mechanism was proposed to account for the catalytic dehydration of formic acid on the titania surface. Reaction (13) is the decomposition reaction of formate ions on titania, and is rate-determining below 700 K. Reaction (14) is a proton transfer reaction that forms water and formate ions, and reaction (15) highlights the scrambling reaction possible between lattice oxygen anions and formate-derived oxygen atoms [43]. (13) DCOO*-^ CO(g) + OfD (14) DCOOD(g) + OfD -> D20(g) + DCOO* (15) OfD + Os-^ OsD + Of* It should not be surprising that, except in those cases where the coordination environment around the cation is suitable for bimolecular coupling reactions, UHV decomposition reactions tend toward unimolecular processes. The bimolecular dehydrogenation process that occurs during steady-state reaction on the single crystal TiO2(001) surface should serve as a reminder that steady-state processes involve interaction between the surface, surface species, and vapor phase components, and that it is not appropriate to attribute product selectivity exclusively to oxide characteristics. 3.5. Extension to Higher Carboxylic Acids The decomposition of formate on titanium dioxide is a model for the decomposition of higher carboxylic acids. Adsorption and reaction of acetic acid on TiO2(110) demonstrated behavior similar to formic acid [63,64]. Experiments with LEED showed that acetate ion on the (110) surface formed ( 2 x 1 ) overlayers that vanished on heating to 450 K. The return of the (1 x 1) LEED pattern agreed with the disappearance of long-range order as acetic acid and water desorbed from the surface. Williams et al. argued, based on LEED data, for a shortening of the Ti-Ti bond distance because of the formation of bridging acetate [63]. Heating to 450 K did not change the angular distribution of H"^ ions desorbing from the surface during ESDIAD experiments, which implied that acetate ions retained a bridging configuration [63,64]. Non-contact AFM of (2 X 1) acetate overlayers on TiO2(110)-(l x 1) has illustrated electronic interactions of acetate species with neighboring acetate ions and hydroxyl groups [65]. Scanning tunneling microscopy has also been used to study acetate orientation on the (110) surface. Temperature programmed STM demonstrated the formation of (2 x 1) ordered domains of acetate and subsequent disappearance of acetate from the surface on heating to 540 K. From this, Iwasawa and coworkers estimated a first-order rate constant of about 4 x 10"^ s'^ at 540 K, in reasonable agreement with prior TPD experiments [61]. ESDIAD, STM and LEED investigations of the orientation of benzoic acid on the (110) surface again focused on the formation of the adsorbed carboxylate

423

ion. ESDIAD confirmed that the molecule stood upright on the surface of the crystal, agreeing with earlier studies with formate and acetate. Benzoate also formed a (2 x 1) LEED pattern at saturation. STM of benzoate on the (110) surface detected (2 x 2) patterns; this discrepancy was attributed to intermolecular interactions between phenyl rings of adjacent intermediates forming dimers on the surface [64]. Thermal reaction of larger carboxylic acids on titanium dioxide surfaces traced reaction pathways similar to those of formic acid. The {110}-faceted surface of TiOiCOOl) selectively decomposed acetate and propionate via unimolecular dehydration to form ketene and methyl ketene. The dehydration yields from both reactants were quite similar, as were the desorption temperatures of the products [46]. Acetic acid decomposed on the {114}-faceted of the TiOa (001) surface to produce ketene as well as acetone [44]. The acetone generated arose from bimolecular coupling of pairs of surface acetates at four-fold coordinate cations; this is analogous to the production of formaldehyde from surface formate on identically prepared surfaces. The reaction of propionic acid corresponded directly to the reaction of acetic acid, producing methyl ketene and 3-pentanone [46]. Adding bifunctional character to the molecule adds another dimension to the overall reaction on the surface. For example, acrylic acid (which contains both carboxyl and vinyl groups) adsorbed on {110}-faceted TiO2(001) to form surface acrylates. The acrylate ion not only decomposed via unimolecular dehydration to form ethylene and CO, but was also reduced to form acrolein. Coupling of acrylates to form divinylketone was not observed on this surface, but was observed on the {114}-faceted surface, where divinylketone was formed at the expense of acrolein formation. Two mechanisms are at work here. On annealing the crystal to higher temperatures, the surface becomes slightly more oxidized, decreasing the acrylate reduction. Simultaneously, surface restructuring to form {114}-TiO2(001) exposes four-fold coordinate titanium cations, providing sites for the bimolecular coupling reaction [44]. Reactions of maleic anhydride on {Oil }-faceted TiO2(001) produced ketene, ethylene, acetylene, vinylacetylene, benzene, butene, CO and CO2 [66]. The reaction was sensitive to the population of surface defects; increased surface reduction (via sputtering) produced a corresponding increase in production of hydrocarbon coupling products (benzene, vinylacetylene, and butene). Reaction of maleic anydride was proposed to occur at five-coordinate Ti"^"*" cations via a ring opening process leading to a six-membered metallacycle incorporating a lattice oxygen anion. This intermediate was proposed to form acetylene by decarboxylation; subsequent cyclotrimerization of acetylene produced benzene [66].

424

The most complicated carboxylic acid examined to date is likely biisonicotinic acid (2,2'-bipyridine-4,4'-dicarboxylic acid), adsorbed on the TiOiCllO) surface by Patthey et al. [67] Their investigation with NEXAFS experiments and INDO calculations revealed that both carboxylate groups were bound to the surface in a bridging fashion [67]. DeSegovia and coworkers examined glycine (comprising both carboxyl and amino substituents) on the (1 X l)-TiO2(110) surface with photoelectron spectroscopy (PES). Their experiments suggested that multilayers of glycine on the (110) surface have a zwitterionic form [68]. Glycine adsorbed on the (1 x 2) reconstructed surface at submonolayer coverage predominantly formed adsorbed carboxylates and quenched Ti(3d) features in the PES spectra [69]. The reaction chemistry of carboxylates depends on many factors, including the surface coordination environment, as demonstrated by the bimolecular coupling of formate ions to form formaldehyde or acetate ions to form acetone on the {114}-reconstruction of TiO2(001). The oxidation state of the cation center is critical as well, as the reduced-surface chemistry of formic acid (forming formaldehyde) attests. Steady-state catalytic experiments with formic acid demonstrated that gas phase interactions with adsorbates lead to an unexpected dehydrogenation pathway, and that it is important to consider vapor phase interactions. In UHV experiments, the product selectivity from formic acid decomposition across a wide range of oxides is most often attributed to the ease with which the oxide surface is reduced. Considering the important role of oxygen vacancies in the decomposition process, it is unsurprising that surface redox properties affect reaction selectivity. Consequently, the reactions of carboxylic acids (even multifunctional ones) on metal oxide catalysts are dominated by carboxylate-surface interaction. 4. CASE STUDY II: THERMAL AND PHOTOREACTIONS OF METHANOL ON TiOi Like formic acid, methanol decomposition has also been used to probe the acidbase properties of metal oxides [70]. However, methoxide decomposition is dependent on surface structure in much the same way as formate decomposition. For example, methanol undergoes parallel dehydration and dehydrogenation reactions on the same crystal surface of zinc oxide [25]. Once again, product selectivity ratios may not necessarily serve as a diagnostic of acid-base properties alone. Alcohol decomposition does provide additional insight into the interaction of adsorbates with metal oxide surfaces. The reactions of alkoxides on titanium dioxide have been used to probe thermal and photo-reactivity of powder and single crystal samples.

425

4.1. Thermal Reactions of Methanol on TiOi Single Crystals Methanol decomposes on titanium dioxide surfaces by mechanisms that are similar to those by which formic acid decomposes. Methanol can reversibly adsorb on single crystal surfaces of titania (reaction 16) in a molecular state, or it may dissociatively adsorb by interaction with surface lattice oxygen anions, forming a surface methoxide (reaction 17). Reaction (18) represents the disproportionation reaction of hydroxyl groups on the surface of the metal oxide. (16) (17) (18)

CHsOHcg) ^ CH3OH* CH30H(g) + Os-> CH3O* + OsH OsH + OsH -^ H20(g) + Os

It is important to note that the production of a molecule of water via reactions (17) and (18) creates an oxygen vacancy on the surface. This vacancy can be filled by decomposition of a methoxide leaving a methyl group bound to the surface. This methyl group can decompose to leave carbon and hydrogen on the surface, or it can incorporate hydrogen (from the decomposition of other methyl groups or from the dissociation of methanol to form methoxide) from the surface to produce methane. Methoxide may also combine with surface hydrogen to form methanol. This step is similar to the production of formic acid at higher temperatures from adsorbed formate. (19) (20) (21) (22)

CH3O* ~> CH3* + Os CH3* -> C* + 3H* CH3* + H * - > CH4(g) CH3O* -^ CH20(g) + H*

Methanol desorbs from the (110) surface in low and high temperature states, as recently verified in separate studies by Henderson et al. [71,72] and Vohs et al [73] The low temperature desorption state has been assigned to desorption of molecularly adsorbed methanol. The higher temperature states have been assigned as recombinative desorption of methanol with surface hydroxyl groups based on SSIMS and HREELS identifications of surface species [52,71-73]. The TiO2(110) surface can be modified to alter the surface chemistry of adsorbed methoxides. Recently Vohs et al. [73] have reported temperature programmed desorption studies of methanol on TiO2(110)-supported V2O5. Deposited monolayer films of vanadia converted some of the adsorbed methanol to formaldehyde and water, while multilayer films of vanadia on the TiO2(110) surface were found to be inactive for methanol oxidation. Furthermore, adsorption studies of formaldehyde indicated that formaldehyde production from

426

methanol is reaction-limited and that there is little further decomposition of formaldehyde to CO or CO2. Consequently, the authors proposed that methoxide is the key intermediate in the oxidation of formaldehyde on V2O5 monolayers [73], although spectroscopic confirmation is lacking to date. Henderson et al. [71,72] have recently used SSIMS and HREELS to study methanol adsorbed on hydroxylated, clean, and oxygen-precovered vacuumannealed surfaces of TiOiCllO). The vacuum-annealed surface contains oxygen vacancies at a surface concentration of between five and ten percent. Coveragedependent TPD of methanol on these defective surfaces demonstrated the presence of both molecularly and dissociatively adsorbed methanol molecules on the surface of the oxide; these give rise to methanol desorption between 165 and 480 K. HREELS experiments suggested that both molecularly and dissociatively adsorbed methanol molecules were present on adsorption at 135 K. SSIMS experiments showed that the low-temperature desorption event at 290 K involved molecularly adsorbed species, while the high temperature feature at 480 K involved recombination of surface methoxide and hydroxyl groups. The absence of surface hydrogen atoms may explain why methane is not detected among the products desorbing from the (110) surface; the methoxides present on the surface may be isolated, with other surface species unavailable for reaction. Methanol adsorbed on an ^^0-enriched surface of TiOiCHO) did not produce CHs^^OH; thus it was proposed that methoxy groups bound at vacancy sites were immobile [71,72]. Methanol co-adsorbed with water displaced most of the water on the surface; methanol co-adsorbed with oxygen formed surface methoxides stable to 625 K. Oxygen pretreatment of the surface did lead to the formation of a species assigned as formaldehyde, which Henderson proposed to be formed via a disproportionation reaction between two methoxides. Spectroscopic probes of the controlling intermediate were inconclusive [71,72]. Methanol decomposition has also been studied on sputtered, {Oil}-, and {114}-faceted crystal faces of TiO2(001). Chemisorbed methanol formed surface methoxide species; about half of the methoxides formed recombined with surface hydroxyl species to form methanol at 365 K. The remainder of the methoxide species decomposed at higher temperatures to form some combination of methanol, methane, dimethyl ether, formaldehyde, and CO depending on the preparation of the crystal surface [74]. The sputtered (001) surface produced methanol as the primary hightemperature desorption product - more than half of the methoxy species present on the surface at 400 K recombined to form methanol at 580 K. The balance of surface methoxides decomposed into methane and CO in a 2:1 ratio between 590 and 600 K. X-ray photoelectron spectroscopy measurements for methanol adsorbed on the sputtered (001) surface at 300 K demonstrated that two types of carbon-containing species were present on the surface; subsequent flashing to

427

450 K desorbed molecularly bound methanol and shifted the binding energy to 286.8 eV, characteristic of surface methoxy groups [74]. The {011}-faceted (001) surface generated methanol, methane, and formaldehyde between 610 and 640 K. Again, methanol accounted for about half of the high temperature products, with the balance of products identified as formaldehyde and methane in a roughly 1:5 ratio. Deuterium labeling of the hydroxyl group illustrated the formation of surface methoxy groups; CH3OD desorbed at 370 K, while CH3OH desorbed at 620 K. Further confirmation of surface methoxide was provided by XPS. Heating the methanol-dosed surface to 300 K desorbed the molecular state, leaving only one type of carbon-containing intermediate (FWHM = 1.9 eV) at 286.8 eV, as observed on sputtered (001) surfaces. [74] The identification of surface methoxide as a precursor for decomposition products is in agreement with methanol adsorption experiments on ZnO[25], Ti02 (110) [71,72], Zr02 [75] and UO2 [76]. Formaldehyde production was associated with a net dehydrogenation of surface methoxides [74]. Methanol decomposition on the {114}-faceted surface of TiO2(001) produced methanol, methane, and formaldehyde, as well as a product not demonstrated on the other crystal faces, dimethyl ether. Methanol desorption on the {011}-faceted and sputtered crystal surfaces typically accounted for about half of the high-temperature products, and the {114}-faceted surface gave rise to a similar result. Formaldehyde and methane were produced in a 1:1 ratio on the {114}-faceted surface; while formaldehyde production increased slightly on the more oxidized surface, production of methane dropped to less than half of the production on the {011}-faceted surface. Generation of dimethyl ether, the product of a bimolecular reaction, occurred at 400 K and accounted for about eight percent of the total carbon yield. No CO was produced on the {114}faceted surface. X-ray photoelectron spectroscopy experiments revealed that the total methoxide coverage was about sixty-five percent of the methoxide coverage on the {011}-faceted surface. This corresponds to the population decrease of coordinatively unsaturated cations on the surface. One-third of the five-fold coordinate titanium cations on the {011} reconstructed surface are converted to six-fold coordinate sites after reconstruction to the {114}-faceted surface (another third are converted into four-fold coordinate sites, preserving the fivefold surface average). It is likely that dissociative adsorption depends on coordinatively unsaturated titanium cations. In fact, the relative coverage change matches the relative population change in available unsaturated centers (down by about a third in both cases), demonstrating the importance of cation coordination vacancies in surface reactions on oxide surfaces. Water desorption occurs concurrently with methanol desorption on the oxidized surfaces, albeit in low yields (ca. 2-5%). Production of water requires dissociation of two methanol molecules; consequently, there is a small net reduction in the population of

428

surface oxygen anions that ultimately affects subsequent thermal reactions of methoxide. Methane selectivity increased as the sputtered surface was oxidized to form the {Oil}-faceted surface, and then decreased upon further oxidation to the {114}-faceted surface [74]. The ideal {011}-faceted surface contains exclusively five-coordinate titanium cations, so a bimolecular disproportionation reaction (see reaction (23)) to produce the ether product is precluded. Although the sputtered surface contains low-coordinate titanium cations and ample oxygen vacancies, the formation of dimethyl ether is exclusive to the {114}-faceted surface. In fact, the defect density drives all of the methoxides to fill those oxygen vacancies, and the resulting carbon is deposited on the surface, only to be burned off as CO, as illustrated in reaction (24) [74]. (23) (24)

2 CH3O* -> HsCOCHscg) + Os C* + Os ^ CO(g)

4.2. Thermal Reactions of Methanol on TiOi Powders As evidenced by the chemistry on the (001) surface, the activity and selectivity of methanol decomposition is dependent on defect and coordination vacancy populations; thus, application of single crystal studies to polycrystalline powders requires some caution. In fact, single crystal surface chemistry has been scaled up to powders before. For example, aldol condensation of acetaldehyde to crotonaldehyde on the {114}-faceted Ti02 (001) crystal face, while not a structure sensitive reaction, has been shown to be similar to acetaldehyde aldolization on rutile powders [77]. The presence of four-, five-, and six-coordinate titanium cations on the {114}-faceted surface has been suggested to provide a good model for polycrystalline materials; this conclusion is based on experimental studies of a variety of small oxygen-containing molecules that illustrate the utility of carefully selected model surfaces for understanding more complex, high surface area materials [55,78]. Methanol decomposition has been studied on the anatase and rutile polymorphs of titanium dioxide [55,78]. Methanol uptake experiments on anatase TiOi demonstrated a molecular coverage of about 3.2 molecules-nm"^ [78]. Both reversibly adsorbed methanol and irreversibly adsorbed methoxides were present on the surface at low temperatures. Infrared spectroscopy showed that surface methoxides were responsible for desorption of the parent alcohol and the decomposition products. Surface methoxy groups decomposed into both dehydration and dehydrogenation products at 650 K. Temperature programmed desorption experiments demonstrated a net weight loss for the powdered sample at high temperatures; the starting weight could be recovered by oxidizing the sample at 400 K. The weight loss was directly related to the product distribution since oxidative dehydrogenation to form formaldehyde and water involved the

429

abstraction of lattice oxygen. Major reaction products at 650 K included formaldehyde, dimethyl ether, methane, CO, and H2O. Small amounts of H2 and methanol also desorbed at 650 K [78]. Methanol reacted on rutile Ti02 to form a similar product distribution. Methanol uptake on rutile was about 1.7 times higher per unit area than on anatase Ti02. Earlier work has suggested that the estimated ratio of exposed titanium cations on rutile and anatase polycrystalline surfaces is about 1.5, in good agreement with the methanol uptake ratio [55]. The selectivity to formaldehyde was approximately the same on both anatase and rutile, although the yield was higher on rutile, reflecting greater methanol uptake. The selectivity ratio for dehydration products (methane and dimethyl ether) was also about 1:1 [55]. 4.3. Thermal Reactions of Methanol on SnOi - Another Rutile Oxide Because titania and tin oxide have the same (rutile) bulk structure, comparison of their surface chemistry permits other influences on reactivity to be examined. One important difference is that the metal-oxygen bond energy is lower for Sn02 than Ti02. Titania is often compared to Sn02, although differences in the electronic structure of Sn02 result in weaker bonding of lattice oxygen anions to the surface. Consequently, it is considerably easier to remove bridging oxygen anions from the (110) surface of Sn02 by annealing or sputtering than from Ti02. Thus, one might expect that surface methoxides would form more oxidized products on tin oxide than on titania [4]. Consistent with the connection between the ease of reduction and product selectivity of surface oxygenates, recent work on SnO2(110) [79] has demonstrated that methanol decomposition produced formaldehyde exclusively. The formation of methoxide was determined to be dependent on the availability of bridging oxygen vacancies on the surface; stoichiometric surfaces converted only about ten percent of adsorbed methanol. It was demonstrated that it was also possible to over-reduce the surface as well. Surfaces characterized by fourcoordinate Sn^"^ cations created by vacuum annealing at 700 K converted as much as half of the adsorbed methanol, while ion-bombarded surfaces annealed to 950 K converted only five percent of methanol to formaldehyde. This disparity in conversion was attributed to in-plane oxygen vacancies created by the more severe annealing (figure 7). These in-plane defects were responsible for a new, higher temperature desorption state for formaldehyde that did not exist for surfaces containing only bridging oxygen vacancy sites. The presence of this additional high-temperature state implies greater stabilization of methoxide on the surfaces annealed to high temperature, but the decreased yield of formaldehyde from these surfaces illustrates the difficulty of dissociatively adsorbing methanol at in-plane sites [79].

430

(a) [110]

(b)

^Bridging vacancy

(c)

•in-plane vacancy

Fig. 7. Ball model illustrations of: (a) the ideal, stoichiometric surface of SnOi (110), (b) the surface missing all bridging oxygen anions and (c) the surface with in-plane oxygen vacancies. The small solid circles represent Sn cations while the large open circles represent O ^" anions. V.A. Gercher, D.F. Cox, and J.-M. Themlin, Surface Science, v. 306 (1994). Reproduced by permission of Elsevier Publications.

Methanol decomposition on {114}-faceted TiO2(001) was accompanied by the formation of dimethyl ether attributed to the presence of four-coordinate Ti"^"^ surface cations, yet no ether was formed on the vacuum annealed SnO2(110) surface, which also contains four-coordinate metal cations. However, the tin cations present on the surface are present as Sn^"*^; SnO has a bulk fourcoordinate structure, hence there is only one coordination vacancy, again precluding bimolecular coupling reactions [79]. 4.4. CH3OH Decomposition on Fluorite Metal Oxides Methanol decomposition has also been studied on fluorite metal oxides. In this bulk structure each metal cation is octahedrally coordinated to oxygen anions, and each anion is tetrahedrally coordinated to metal cations. Methanol decomposition on Zr02 has been the subject of considerable surface analysis largely because the decomposition pathway has demonstrated structure sensitivity [75]. The ZrO2(100) crystal face is characterized by square arrays of zirconia bonded to four subsurface oxygen anions. Methanol adsorbed on ZrO2(100) at 125 K decomposed into methoxide and hydroxyl groups. Seventy-five percent of the adsorbed methanol desorbed as the parent molecule at 230 K through a recombinative pathway. The balance of surface methoxide reacted between

431

ZrOa (100)

•

:Zr*'^ in 1st layer

® . iZr'"'^ in 2nd layer

ZrOa (110)

Q .

-O'^ •" Ist layer lO'"^ in 2nd layer

Fig. 8. Models of the (100) and (110) surfaces of cubic zirconia. Both top and side views are presented. P.A. Dilara and J.M. Vohs, Surface Science, v. 321 (1994). Reproduced by permission of Elsevier Publications.

610 K and 630 K to liberate CO and methane. Comparison between lightly sputtered and stoichiometric Zr02 surfaces demonstrated that oxygen vacancies were not essential for the formation of methoxide [75]. The ZrO2(110) surface, characterized by rectangular arrays of zirconium cations and oxygen anions in the same plane, exhibited substantially different surface chemistry in the high temperature desorption state. Figure 8 illustrates both the (110) and (100) basal planes of zirconia. Although nearly three-quarters of the adsorbed methanol desorbed at 240 K via recombinative generation of the parent molecule, the remaining methoxides primarily formed formaldehyde and methane in a 2.5:1 ratio. Approximately equal amounts of methane and CO were formed on the Zr02 (110) surface. Temperature programmed HREELS recorded the dissociative adsorption of methanol to form methoxide. However, HREELS also indicated the formation of a new intermediate after heating to 300 K, identified as dioxymethylene (CH2O2), which could decompose to produce formaldehyde [75]: (25) (26)

CH3O* + Os ^ CH2O2* + H* CH202*^CH20(g)+Os

432

The production of dioxymethylene from methoxides competed with dehydrogenation to form CO and methane [75]: (27) (28)

CH3O* ^ CO(g) + 3 H* CH3O* + H* -^ CH4(g) + Os

Interaction of methoxide with the in-plane oxygen anions on the (110) surface is necessary to form the dioxymethylene intermediate [75]. CH 3

A1

O—Zr-0

H3C^

p I O—Zr-0

CH,

•

o

o ^ "^ *—Zr-O

However, the methoxide would be expected to have a much more difficult time forming the dioxymethylene intermediate on the (100) surface, as the oxygen anions are below the surface plane (see figure 8). Consequently, this surface produced methane instead. The reaction to produce methane is also driven by the coordinative unsaturation around the zirconium cations on the ZrO2(100) surface. The Zr"^"^ cations are four-fold coordinate on ZrO2(100) as compared to six-fold coordinate on ZrO2(110), giving the ZrO2(100) surface a greater affinity for oxygen and therefore those site may be better able to cleave the carbon-oxygen bond in the methoxide [75]. Methanol reactions have also been studied on polycrystalline wafers of UO2 [76]. Two parent desorption states existed for methanol adsorbed at 90 K. Molecularly adsorbed methanol desorbed at 110 K, and methanol generated by surface recombination of methoxides and protons desorbed at 180 K. Carbon (Is) XPS demonstrated that methanol dissociatively adsorbed on the urania surface and that methoxide was the only surface intermediate present above 150 K. Primary reaction products were methane and carbon monoxide at 480 K. Oxygen atoms not removed from the surface as CO were incorporated into the oxide surface; isotopically labeled U^^02 surfaces did not exchange oxygen with methoxide to produce C^^O [76]. 4.5. Decomposition of Higher Alcohols on Titanium Dioxide Single Crystal Surfaces Decomposition reactions of larger aliphatic alcohols have been examined in detail on the {Oil}-faceted TiO2(001) surface [80]. Ethanol adsorbed at 300 K exhibited a low temperature desorption peaks for ethanol and water at 365 K and a high temperature desorption state for decomposition products at 588 - 595 K. Half of the ethanol adsorbed on the surface desorbed as ethanol at 365 K. Half of the remaining surface ethoxide groups desorbed as ethanol at 588 K. The

433

remaining twenty-five percent of carbon-containing products desorbed as ethylene and acetaldehyde in a 5:1 molar ratio between 588 - 595 K. Temperature programmed XPS measurements for ethanol adsorbed at 300 K revealed no intermediates distinguishable from surface ethoxide (e.g., no carboxylates), and measured peak area changes corresponded well to quantitative changes in TPD desorption events. High-temperature production of olefins from ethoxide is consistent with hydride elimination from the p-position observed on polycrystalline anatase powders (reaction (29)), while the concomitant production of ethanol must occur via reaction of from the released hydrogen with other ethoxides (reaction (30)). Acetaldehyde is also produced on the (001) surface (reaction (31)), the result of ethoxide dehydrogenation via hydride elimination from the a-position [80]. (29) (30) (31)

CH3CH2O* -> H2C=CH2(g) + Os + H* CH3CH2O* + H* -^ CH3CH20H(g) CH3CH2O* -^ CH3CH0(g) + H*

Qualitatively similar results were obtained for reaction and desorption of normal and iso-propanol on the {Oil}-faceted TiO2(001) surface. In the case of normal propanol, almost half of the molecules initially adsorbed desorbed as the parent molecule at 370 K, while half of the remaining surface species reacted to form propanol at 580 K. The ratio of propene to propionaldehyde generated at 580 K was 10:1. Desorption of isopropanol quantitatively mirrored the desorption of normal propanol in two desorption states at 365 and 512 K. Isopropanol did not generate any dehydrogenation products (e.g., acetone), and the surface did not generate any bimolecular coupling products for any of the probe alcohol molecules. The absence of ether formation on the {011}-faceted surface is consistent with the need for double-coordination vacancies to facilitate that reaction, and the absence of such sites on this surface of titanium dioxide [80]. Campbell et al. have recently studied the reactions of surface ethoxy groups on the TiO2(110) surface [81]. Temperature programmed desorption studies of isotopically labeled ethanol were used to probe the surface intermediates in ethanol decomposition. As in the other work discussed here, low temperature desorption features (< 250 K) were associated with the desorption of molecularly adsorbed ethanol. Between 250 and 400 K, ethanol desorbed as a result of the recombination of surface ethoxy groups with surface hydroxy] groups. Seventyfive percent of the surface ethoxy groups produced by saturation exposure of ethanol to the (110) surface generated ethanol below 400 K. The remaining onequarter of the initial ethoxy population desorbed at 650 K as ethanol and ethylene. The high temperature desorption channel was thought to involve

434

hydrogen elimination from the p-position (earUer proposed on {Oil }-TiO2(001)) as a result of further isotopic labeling experiments. The CD3CH20H-dosed TiO2(110) surface produced primarily CD2CH2 [81]. Had the reaction proceeded through hydrogen elimination from the aposition and subsequent hydride shift, the expected product would have been CD2CHD. A kinetic isotope effect of about 2 kJ/mol was detected for ethylene production in TPD of CD3CH2OH versus CH3CH2OD, consistent with C-D vs. C-H scission as the rate determining step in ethoxide decomposition. Furthermore, TPD analysis suggested that the dehydrogenation reaction was not concerted with ethylene desorption. To explain these observations, Campbell proposed that initial ethanol exposure to the TiO2(110) surface generates surface hydrogen bound to bridging oxygen anions and surface ethoxides bound to titanium cation centers [81]: (32)

CH3CH2OH* -^ Ti-OCH2CH3 + Obridging-H

Seventy-five percent of the ethoxy groups recombine with the surface hydrogen to reform ethanol as the reverse of reaction (32). The remaining twenty-five percent are stable to 500 K. In fact, these ethoxides were irreversibly formed; when exposed to water at low temperature, they did not

Ti02 (110) Surface

(^J) Bridging Oxygen ((V) I Surface Oxygen (QJ

#

Ti''

V^r = Bridging Oxygen Vacancy

Bulk Oxygen (O^)

Fig. 9. Representation of the TiOi (110) surface. Binding of-OR groups (to represent either OH or OEt) is shown at both Ti ^'"*" sites and bridging oxygen vacancies. L. Gamble, L.S. Jung, and C.T. Campbell, Surface Science, v. 348 (1996). Reproduced by permission of Elsevier Publications.

435

react to form any ethanol. This suggests that the surface ethoxides converted to a different surface binding site. Examination of the CH3CH2OD TPD revealed coverage-dependent production of HDO and D2O in addition to backgroundadsorbed water. Consequendy, Campbell proposed a water-generation step (reaction (33)) stoichiometrically analogous to reaction (18) that involved the creation of surface oxygen vacancies. Figure 9 illustrates the two binding sites proposed [81]. (33)

2 Obridging-H - > H20(g) + Obridging + ^ bridging

(34)

T i - O C H 2 C H 3 + Vbrldging - ^ Obridging-CH2CH3

Reaction (34) represents surface diffusion of the ethoxide to fill in oxygen vacancy. The resulting Obridging-CH2CH3 remains on the surface until the Phydrogen is eliminated, forming ethylene [81]. 4.6. Higher Alcohols: Extension to Poly crystalline TiOi and Other Materials Ethanol decomposition has been studied on anatase and rutile powders of titanium dioxide. Much like methanol decomposition, it was demonstrated that the bulk form of titania did not influence the surface chemistry significantly [55]. Furthermore, TPD studies of ethanol decomposition on clean and hydroxylated titania powders have also demonstrated the insensitivity of high-temperature reaction channels to the initial presence of surface hydroxyl groups, while lowtemperature desorption modes of ethanol and water are affected by pre-exposure of the oxide surface to water, where it acts as a site-blocker [82]. These results were in good agreement with similar experiments on the (110) single crystal face conducted by Campbell et al, and together they suggest that there are two separate adsorption modes for ethanol on the surface of titanium dioxide [81]. Ethanol TPD on anatase powders revealed the existence of two desorption states (390 and 590-620 K). Approximately sixty percent of the adsorbed ethanol desorbed as the parent molecule via the low-temperature state in powder TPD experiments. The remainder decomposed into acetaldehyde, ethylene, diethyl ether, and water. Small amounts of ethanol and butene were produced as well. Reaction and decomposition of n-propanol and isopropanol on anatase titanium dioxide powders were also characterized by two desorption states; much like ethanol and methanol, the low temperature state was characterized by desorption of the parent molecule, while the high temperature state contained decomposition products. Major decomposition products from normal propanol reaction included propylene, di-n-propyl ether, propionaldehyde, and water, while isopropanol decomposition produced mostly propylene and water. It was suggested that steric effects inhibited the bimolecular coupling of isopropoxides to form the corresponding ether [78].

436

Recent work by Idriss and Seebauer has focussed on using thermal reactions on ethanol on many oxide powders to correlate material properties with oxide reactivity [83]. In steady-state catalytic experiments at low conversion over Ti02, ethanol product selectivity heavily favored the production of acetaldehyde (selectivity ca. 93%) and its derivatives, ethyl acetate and acetone. Acetaldehyde is produced via oxidative dehydrogenation of surface ethoxides, the formation of which Idriss and Seebauer determined was related to the degree of coordinative unsaturation of surface cations. Furthermore, the dehydrogenation reaction requires that the surface be reducible and reoxidizable due to the participation of lattice oxygen in the surface reactions. Acetaldehyde reacted further to produce ethyl acetate via a proposed Tischenko reaction involving hydrogen transfer between two aldehydes and subsequent coupling of the resulting aldehyde/alkoxide pair [83]. (35)

H3C(CH)=0* + H3C(CH)=0* ^

H3C(CO)-0-CH2CH3(g)

Acetone was suggested to be formed by oxidation of acetaldehyde followed by coupling of the resulting acetate intermediates [83]: (36) (37)

CH3CHO* + Os ^ CH3COO* + H* 2 CH3COO* -> C02(g) + (CH3)2C=0(g) + Os

Two conditions must be satisfied for acetone formation from acetaldehyde on the metal oxide surface. The cation must be reducible, as lattice oxygen must be abstracted to form the surface acetate group. Si02, which will dehydrogenate ethanol to form acetaldehyde, is not capable of this, as the oxygen-silicon bond is too strong. The metal cation must also possess two coordination vacancies. For example, MgO will also produce acetaldehyde but not acetone in UHV. Magnesia possesses a rocksalt structure with six-coordinate occupation around the cation, and cleavage of the (100) surface simply truncates the bulk lattice, leaving a single coordination vacancy at each magnesium cation. Single crystal work with MgO(lOO) has demonstrated that acetates on this surface will not couple, but instead undergo a Cannizzaro-like disproportionation reaction [83]. There is a rich array of chemistry centered on alkoxides and carboxylates such as the reaction of acetic acid to form ketene on titania [45,46] and silica [84], and aldol condensation of acetaldehyde to form on titania [77]. In fact, both of these reactions represent cases where single crystal work ultimately provided the basis for understanding more complex powder-catalyzed chemistry.

437

4.7. Photoreactions of Alcohols on TiOi Photoreactions on titanium dioxide have been the focus of considerable interest for some time. Titania offers the opportunity to oxidize organic compounds in polluted environments, and has also been exploited to generate titania-supported nanoparticles of metals (e.g., silver) via photoreduction reactions [85]. While there is not enough room here to thoroughly treat photocatalytic processes, a brief introduction to the subject is presented below. Readers seeking detailed treatments of this subject are referred to a recent review by Yates et al. on titania-facilitated photocatalysis [86]. The utility of titania for photocatalytic applications is based on the band gap of the semiconductor. By definition, there are no states in the band gap to facilitate the recombination of electron-hole pairs; thus, excitation across the bandgap may allow charge transfer to an adsorbate and therefore its activation. Studies have concluded that the lifetime of excited states across the bandgap is on the order of nanoseconds, which is long enough to facilitate charge transfer. Two competing processes determine the fate of the electron-hole pair. The first is simple electron-hole recombination, leading to no reaction and a return to the initial state. The second route allows the semiconductor to donate the electron to an electron-accepting adsorbate. The hole may then combine with an electron from an oxidizable adsorbate. The efficiency of this process is conventionally referred to as the quantum yield, which is defined as the number of reaction events per photon adsorbed. If all incident light is assumed to be absorbed, it is more effectively defined as the ratio of the rate of charge transfer to adsorbates to the sum of all electron-hole consuming events. On titanium dioxide, the concentration of electrons tends to exceed that of holes, because the transfer of electrons to adsorbed oxygen is relatively slow [86]. Early studies on titania powders showed that methanol generated methyl formate as the principle photooxidation product. Molybdena- and vanadiamodified Ti02 catalysts demonstrated at least an eighty percent drop in activity relative to pure titania, although selectivity to dimethoxymethane (and thus suppression of further oxidation products) was almost total [87]. Recently, Lin et al. have studied methanol photochemistry on Ti02 powders with IR in order to spectroscopically identify the important surface intermediates [88]. Investigation of methanol reactivity in the absence of oxygen indicated the presence of surface methoxides which decomposed to form methanol, formaldehyde, water and hydrogen. No additional IR features were detected during UV exposure reactions, although the surface concentration of methoxide decayed on consumption during the reaction. In the presence of oxygen, UV exposure caused features in the IR spectrum associated with surface formate to grow at the expense of methoxide. The rate of methoxide decomposition was initially slower in the oxygen atmosphere than without oxygen, although methoxide concentration monotonically decreased throughout the experiment.

438

The surface methoxide concentration approached a stable population when oxygen was absent from the system (figure 10) [88]. Electron spin resonance studies of aqueous methanol reactions with colloidal titania have suggested the appearance of methoxy radicals [89]; the participation of similar radicals was suggested for methanol decomposition in the absence of vapor phase oxygen. Lin suggested that methoxy groups absorb holes and generate methoxy radicals on the surface. (38) (39)

Ti^^-0-Ti^^-0CH3 + h^ -> Ti^^-0-Ti^^-0CH2- + [Hi* Ti^^-0-Ti^^ + CH2O,'(g) Ti^^-0-Ti'^-0CH2

The rapid decrease in rate shown in figure 10 was likely due to electron/hole recombination at the surface as the concentration of surface electrons increased with continued irradiation [88]. The apparently slower rate of methoxide decomposition in the presence of oxygen is likewise dependent on the photoexcitation process. As stated above,

(A) -oin the absence of oxygen

2000 4000 6000 8000 Photo-irradiation Time (sec)

10000

Fig. 10. Relative concentration of CH30(a) as a function of UV irradiation time in the absence of (A) O2 and in the presence of (B) 10 Torr of O2. C.-C. Chuang, C.-C. Chen, and J.-L. Lin, Journal of Physical Chemistry B, v. 103 (1999). Reproduced by permission of the American Chemical Society.

439

charge transfer to adsorbed oxygen is a relatively slow process. Lin postulated that the methoxy radicals incorporated photoexcited oxygen in the form of peroxide radicals [88]: (40) (41) (42)

O2* + H* + e- -^ HOO* CH3O* + hv + h"^ + 02(g) -^ *0CH200» HOO* + *0CH200* -> [OCH2OOOOH]* -^ HCO2* + H2O + O2*

Thus, the formation of the organoperoxide transition state is dependent on the availability of the peroxide radical formed by photoexcitation of adsorbed oxygen, the rate limiting step of the process [88]. Methyl formate photodecomposition has also been studied on titanium dioxide powders. In the presence of oxygen, methyl formate decomposed to form CO, CO2, formaldehyde, and water [90]. 5. SUMMARY Selectivity on metal oxide catalysts is ultimately determined by complex intermolecular and surface-adsorbate interactions. Competing reaction channels are facilitated or hindered by the coordination geometry around metal cations, the ease of reduction of the surface, and the resulting stabilization of surface intermediates. The decomposition of relatively simple organic molecules like methanol and formic acid can be surprisingly complex, but attention to a few concepts may help to understand the reaction processes: 1) The population of surface defects and coordination vacancies drives alkoxide and carboxylate formation and decomposition. When cations have at least two coordination vacancies, bimolecular reactions are possible (e.g., acetone from acetate ions, dimethylether from methoxy groups). 2) The ease of reduction of the metal oxide is often coupled to the product selectivity due to the involvement of oxygen vacancies in the reaction. For example, MgO(lOO) exclusively dehydrates formic acid to form CO, while the more reducible ZnO(0001)-Zn surface readily produces CO2 by dehydrogenation of formate. 3) Reduced cation oxidation states can drive reduction reactions of oxygenated adsorbates. Sputter-reduced surfaces of TiO2(001) readily converted formic acid into formaldehyde and methanol into methane and CO. 4) Interactions with other molecules can affect the ultimate reaction selectivity of the process. For example, the catalytic dehydration of formic acid on TiO2(001) occurred as a unimolecular process at high temperatures and low formate coverage, while a bimolecular dehydrogenation process dominated at near-saturation coverage of the titania crystal.

440

Even though many of these principles appear to be applicable to for multifunctional carboxylates and alkoxides, it is important to recognize that more complex molecules may be more or less influenced by differences in surface conditions than others. Small, monofunctional molecules ably serve to highlight site requirements for some reactions, but there is no reason to expect that a large, multifunctional molecule will necessarily interact with a metal oxide catalyst as a mere combination of its functionalities. Consequently, it is important to characterize the adsorption and synthesis of larger molecules where possible, to determine the limitations of the principles explored here, and to develop an understanding of adsorption and reaction characteristics that will lead to more selective catalysts. REFERENCES [I] [2] [3] [4] [5] [6]

[7] [8] [9] [10] [II] [12] [13] [14] [15] [16] [17] [18] [19] [20]

H. Idriss and M.A. Barteau, Advances in Catalysis, 45 (2000) 261. M.A. Barteau and J.M. Vohs in Handbook of Heterogeneous Catalysis, G. Ertl and H. Knozinger, VCG, V^einheim, (1997) Vol. 2, pp. 873. M.A. Barteau, Chemical Reviews, 96 (1996) 1413. V.E. Henrich and P.A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, Cambridge (1994). H.-J. Freund, Faraday Discussions, 114 (2000) 1. J.-M. Basset, B.C. Gates, J.-P. Candy, A. Choplin, M. Leconte, F. Quignard and C. Santini, Eds., Surface Organometallic Chemistry: Molecular Approaches to Surface Catalysis, Kluwer Academic Publishers, Dordrecht (1986) Vol. 231. M.A. Henderson, Surface Science, 419 (1999) 174. H. Onishi, K. Fukui and Y. Iwasawa, Bulletin of the Chemical Society of Japan, 68 (1995) 2447. M. Li, W. Hebenstreit, L. Gross, U. Diebold, M.A. Henderson, D.R. Jennison, P.A. Schultz and M.P. Sears, Surface Science, 437 (1999) 173. M. Ashino, T. Uchihashi, K. Yokoyama, Y. Sugawara, S. Morita and M. Ishikawa, Applied Surface Science, 157 (2000) 212. S. Petigny, H. Mostefa-Sba, B. Domenichini, E. Lesniewska, A. Steinbrumm and S. Bourgeois, Surface Science, 410 (1998) 250. M. Brause, S. Skordas and V. Kempter, Surface Science, 445 (2000) 224. S. Tanaka, K. Mase, M. Nagasono, S. Nagaoka and M. Kamada, Surface Science, 451 (2000) 182. M. Li, W. Hebenstreit, U. Diebold, A.M. Tyryshkin, M.K. Bowman, G.G. Dunham and M.A. Henderson, Journal of Physical Chemistry B, 104 (2000) 4944. M. Li, W. Hebenstreit and U. Diebold, Surface Science, 414 (1998) 1951. V.S. Lusvardi, M.A. Barteau, J.G. Chen, J. J.J. Eng, B. Fruhberger and A. Teplyakov, Surface Science, 397 (1998) 237. B.A. Watson and M.A. Barteau, Chemistry of Materials, 6 (1994) 771. G.S. Herman, Y. Gao, T.T. Tran and J. Osterwalder, Surface Science, 447 (2000) 201. H. Norenberg, F. Dinelli and G.A.D. Briggs, Surface Science, 446 (2000) 183. H. Idriss and M.A. Barteau, Catalysis Letters, 26 (1994) 123.

441 [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44]

[45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59]

J. Muscat and N.M. Harrison, Surface Science, 446 (2000) 119. H. Norenberg, F. Dinelli and G.A.D. Briggs, Surface Science, 436 (1999) L635. H. Idriss, V.S. Lusvardi and M.A. Barteau, Surface Science, 348 (1996) 39. Y. Iwasawa, H. Onishi and K. Fukui, Catalysis Letters, ((in press)). J.M. Vohs and M.A. Barteau, Surface Science, 176 (1986) 91. X.D. Peng and M.A. Barteau, Catalysis Letters, 7 (1990) 395. P. A. Dilara and J.M. Vohs, Journal of Physical Chemistry, 97 (1993) 12919. L. Wang, K.F. Ferris, G.S. Herman and M.H. Engelhard, Journal of Vacuum Science and Technology A, 4 (2000) 1893. R.N. Spitz, J.E. Barton, M.A. Barteau, R.H. Staley and A.W. Sleight, Journal of Physical Chemistry, 90 (1986) 4067. H. Onishi, C. Egawa, T. Aruga and Y. Iwasawa, Surface Science, 191 (1987) 479. T. Shido, K. Asakura and Y. Iwasawa, Journal of Catalysis, 122 (1990) 55. X.D. Peng and M.A. Barteau, Surface Science, 224 (1989) 327. X.D. Peng and M.A. Barteau, Langmuir, 7 (1991) 1426. T. Shido and Y. Iwasawa, Journal of Catalysis, 129 (1991) 343. S. Crook, H. Dhariwal and G. Thornton, Surface Science, 382 (1997) 19. W.T. Petrie and J.M. Vohs, Surface Science, 245 (1991) 315. A.W. Grant, A. Jamieson and C.T. Campbell, Surface Science, 458 (2000) 71. R. Davis, J.F. Walsh, C.A. Muryn, G. Thornton, V.R. Dhanak and K.C, Prince, Surface Science Letters, 298 (1993) 1196. J.M. Vohs and M.A. Barteau, Surface Science, 197 (1988) 109. M.A. Henderson, Journal of Physical Chemistry, 99 (1995) 15253. M.A. Henderson, Journal of Physical Chemistry B, 101 (1997) 221. G. Munuera, Journal of Catalysis, 18 (1970) 19. H. Onishi, T. Aruga and Y. Iwasawa, Journal of Catalysis, 146 (1994) 557. K.S. Kim and M.A. Barteau in Structure-Activity and Selectivity Relationships in Heterogeneous Catalysis, R.K. Grasselli and A.W. Sleight, Elsevier Science Publishers, Amsterdam, (1991) pp. 327. K.S. Kim and M.A. Barteau, Langmuir, 6 (1990). K.S. Kim and M.A. Barteau, Journal of Catalysis, 125 (1990) 353. K.S. Kim and M.A. Barteau, Langmuir, 4 (1988) 945. S. Thevuthasan, G.S, Herman, Y.J. Kim, S.A. Chambers, C.H.F. Peden, Z. Wang, R.X. Ynzunza, E.D. Tober, J. Morals and C.S. Fadley, Surface Science, 401 (1998) 261. B.E. Hayden, A. King and M.A. Newton, Journal of Physical Chemistry B, 103 (1999) 203. L. Wang, K.F. Ferris, D.R. Baer and M.H. Engelhard, Surface Science, 380 (1997) 352. L. Wang, A.N. Shultz, D.R. Baer and M.H. Engelhard, Journal of Vacuum Science and Technology A, 14 (1996) 1532. H. Onishi, T. Aruga, C. Egawa and Y. Iwasawa, Surface Science, 193 (1988) 33. K.G. Pierce and M.A. Barteau, Journal of Physical Chemistry, 98 (1994) 3882. J.P. Collman, Inorganic Chemistry, 7 (1968) 1298. V.S. Lusvardi, M.A. Barteau and W.E. Fameth, Journal of Catalysis, 153 (1995) 41. Y. Iwasawa, Surface Science, 402-404 (1998) 8. Y. Iwasawa, H. Onishi, K. Fukui, S. Suzuki and T. Sasaki, Faraday Discussions, 114 (2000) 259. K. Fukui, H. Onishi and Y. Iwasawa, Applied Surface Science, 140 (1999) 259. K. Fukui, H. Onishi and Y. Iwasawa, Chemical Physics Letters, 280 (1997) 296.

442

[60] [61] [62] [63] [64] [65] [66] [67]

[68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90]

R.A. Bennett, P. Stone, R.D. Smith and M. Bowker, Surface Science, 454-456 (2000) 390. H. Onishi, Y. Yamaguchi, K. Fukui and Y. Iwasawa, Journal of Physical Chemistry, 100 (1996) 9582. P. Stone, R.A. Bennett and M. Bowker, New Journal of Physics, 1 (1999). Q. Guo, I. Cocks and E.M. Williams, Journal of Physical Chemistry, 106 (1997) 2924. Q. Guo and E.M. Williams, Surface Science, 433-435 (1999) 322. K. Fukui and Y. Iwasawa, Surface Science Letters, 464 (2000) L719. J.N. Wilson, D.J. Titheridge, L. Kieu and H. Idriss, Journal of Vacuum Science and Technology A, 4 (2000) 1887. L. Patthey, H. Rensmo, P. Persson, K. Westermark, L. Vayssieres, A. Stashans, A. Petersson, P. Bruhwiler, H. Siegbahn, S. Lunell and N. Martensson, Journal of Chemical Physics, 110 (1999) 5913. E. Soria, E. Roman, E.M. Williams and J.L. deSegovia, Surface Science, 433-435 (1999)543. E. Soria, I. Colera, E. Roman, E.M. Williams and J.L. deSegovia, Surface Science, 451 (2000) 188. J.M. Tatibouet, Applied Catalysis A, 148 (1997) 213. M.A. Henderson, S. Otero-Tapia and M.E. Castro, Faraday Discussions, 114 (2000) 313. M.A. Henderson, S. Otero-Tapia and M.E. Castro, Surface Science, 412/413 (1998) 252. G.S. Wong, D.D. Kragten and J.M. Vohs, Surface Science, 452 (2000) L293. K.S. Kim and M.A. Barteau, Surface Science, 223 (1989) 13. P.A. Dilara and J.M. Vohs, Surface Science, 321 (1994) 8. J.A. Lloyd, W.L. Manner and M.T. Paffett, Surface Science, 423 (1999) 265. J.E. Rekoske and M.A. Barteau, Langmuir, 15 (1999) 2061. K.S. Kim, M.A. Barteau and W.E. Fameth, Langmuir, 4 (1988) 533. V.A. Gercher, D.F. Cox and J.-M. Themlin, Surface Science, 306 (1994) 279. K.S. Kim and M.A. Barteau, Journal of Molecular Catalysis, 63 (1990) 103. L. Gamble, L.S. Jung and C.T. Campbell, Surface Science, 348 (1996) 1. V.S. Lusvardi and M.A. Barteau, Journal of Physical Chemistry, 100 (1996) 18183. H. Idriss and E.G. Seebauer, Journal of Molecular Catalysis A, 152 (2000) 201. R. Martinez, M.C. Huff and M.A. Barteau, Applied Catalysis A, 200 (2000) 79. W.E. Farneth, R.S. McLean, J.D. Bolt, E. Dokou and M.A. Barteau, Langmuir, 15 (1999) 8569. A.L. Linsebigler, G. Lu and J. J.T. Yates, Chemical Reviews, 95 (1995) 735. T. Carlson and G.L. Griffin, Journal of Physical Chemistry, 90 (1986) 5896. C.-C. Chuang, C.-C. Chen and J.-L. Lin, Journal of Physical Chemistry B, 103 (1999) 2439. O.I. Micic, Y. Zhang, K.R. Cromack, A.D. Trifunac and M.C. Thurnauer, Journal of Physical Chemistry, 97 (1993) 13284. C. Chuang, W. Wu, M. Huang, I. Huang and J. Lin, Journal of Catalysis, 185 (1999) 423.

Oxide Surfaces D.P. Woodruff, editor © 200J Elsevier Science B. V. All rights reserved.

443

Chapter 11

The Structure of TiOi surfaces Ulrike Diebold Department of Physics, Tulane University, New Orleans, LA 70118, U.S.A. 1. INTRODUCTION Unraveling the relationship between the atomic surface structure and other physical and chemical properties is probably one of the most important achievements of surface science. Because of the mixed ionic and covalent bonding in metal oxide systems, the surface structure has an even stronger influence on local surface chemistry as compared to metals or elemental semiconductors [1]. A vast amount of work has been performed on Ti02 over the years, and this is certainly the best-understood surface of all the metal oxide systems. This chapter starts with a brief description of the bulk structure of titanium dioxide crystals, and their stable crystal planes. Because bulk nonstoichiometries influence the surface properties of Ti02 in a variety of ways, we include a short discussion of bulk defects. The longest part of the chapter is devoted to the rutile (110) surface. The (bulk-truncated) (1x1) surface is known with a very high accuracy from experimental as well as theoretical studies. Even more importantly, there has been given a reason on why theory and experiment disagree in some aspects [2]. Surface defects are categorized as step edges, oxygen vacancies, line defects (closely related to the (1x2) reconstruction), common impurities, and the manifestation of crystallographic shear planes on surfaces. The long-standing argument of the structure of the (1x2) phase seems to be settled, as discussed in section 3.2.2. Scanning Tunneling Microscopy and, more recently. Atomic Force Microscopy, studies have revealed the complexity of the seemingly simple rutile (110) surface, hence the section on Ti02(l 10) commences with a recommendation on how to best to prepare this surface. The two other low-index planes, rutile (100) and (001) are described in sections 4 and 5, respectively. Until fairly recently the (1x3) reconstruction of the rutile (100) seemed well understood, but inconsistencies in theoretical calculations as well as new interpretations of xray diffraction data show that closer look on the structure of this phase may be

444

needed (4.2.2.). Very recent developments on structural investigations of anatase surfaces are included at the end.

2. BULK STRUCTURES Titanium dioxide crystallizes in three different structures; rutile (tetragonal; D4hi4-P42/mnm, a = b = 4.584 A, c = 2.953 A, ), anatase (tetragonal; D4hi9I4i/amd, a = b = 3.733 A, c = 9.37A), and brookite (rhombohedrical, D2h^^Pbca, a = 5.436 A, b = 9.166 A, c = 5.135 A) [3]. Only rutile and anatase play any role in the applications of Ti02 and have been studied with surface science techniques. Their unit cells are shown in Fig. 1. In both structures, the basic building block consists of a titanium atom surrounded by six oxygen atoms in a more or less distorted octahedral configuration. In each structure, the two bonds between titanium and the oxygen atoms at the aspices of the octahedron are slightly longer. A sizable deviation from a 90° bond angle is observed in anatase. In rutile, neighboring octahedra share one corner along <110> type directions, and are stacked with their long axis alternating by 90°. This results in three-fold coordinated oxygen atoms. Rutile Ti02 single crystals are widely available. They can be bought in cut and polished form from companies such as Commercial Crystal Laboratories, U.S.A.; Kelpin Kristallhandel, Germany; Goodfellow Ltd., U.K.; or Earth Jewelry, Japan; and many others. A very small roughness is achieved by grinding the sample, and then polishing the surface for many hours with a chemo-mechanical treatment. This is also referred to as epitaxial polish. Practical aspects of surface preparation and handling are discussed in ref [4].

Anatase

Rutile 1.946 A

titanium

[010] ygen

1.983

[001] [010] [100]

1.937 A

Fig. 1. Unit cells of rutile and anatase. The tetragonal bulk unit cells (rutile a = b = 4.584 A, c = 2.953 A; anatase a = b = 3.733 A, c = 9.37A) are indicated. In both structures, slightlydistorted octahedra are the basic building units. The the bond lenghts and angles of the octahedrally-coordinated Ti atoms are indicated.

445

Ramamoorthy et al. [5] calculated the total energy of periodic Ti02 slabs using a self-consistent ab initio method. The (110) surface has the lowest surface energy, and the (001) surface the highest. This is expected from the considerations of surface stability, based on electrostatic and dangling-bonds arguments discussed in section 3.1.1. below. The thermodynamic stability of the (100) surface was also considered, and was found to be stable with respect to forming (110) facets. The experimental results on the three lowindex rutile surfaces discussed below fit rather well with the stability expected from these calculations. For rutile, the (110), (001), and (100) surfaces have been studied, with (110) being the most stable one. These three surfaces are discussed in this chapter. The two simple approaches that are commonly used to estimate the stability and structure of oxide surfaces are exemplified in detail for the rutile (110) surface. For anatase, the (101) and the (100)7(010) surface planes are found in powder materials, together with some (001). The (101) surface was calculated to have the lowest surface energy, even lower than the rutile (110) surface [6]. The first experimental results on this surface appear to confirm this theoretical prediction [7,8]. 2.1. Bulk defects The titanium-oxygen phase diagram is very rich with many stable phases with a variety of crystal structures, see Fig. 2 (ref [3]). Consequently, Ti02 can be reduced easily. Bulk reduction and the resulting color centers are reflected in a pronounced color change in Ti02 single crystals from initially transparent to light and, eventually, dark blue [9]. These intrinsic defects result in n-type doping and a high conductivity. This high conductivity makes Ti02 single crystals such a convenient oxide system for experimentalists. As has been pointed out recently [10], bulk defects play a major role in a variety of surface phenomena where annealing to high temperature is necessary , e.g. during the encapsulation of Pt [11-13], in bulk-assisted re-oxidation [14,15], in restructuring and reconstruction processes [9,16], and in gas adsorption [17]. The relationship between crystal color, conductivity, bulk defects as characterized by EPR measurements, and surface structure of rutile (110) has been investigated systematically by Li et al. [10]. The bulk structure of reduced Ti02-x crystals is quite complex with a various types of defects such as doubly charged oxygen vacancies, Ti3+ and Ti4+ interstitials, and planar defects such as crystallographic shear (CS) planes. The defect structure varies with oxygen deficiency which depends on temperature, gas pressure, impurities, etc. Despite years of research, it is still subject to debate which type of defect is dominant in which region of oxygen deficiency [18,19]. It was shown that the dominant type are Ti interstitials in the region from TiOi.9996 to TiOi.9999 (from 3.7x10^8 to 1.3x1919 missing O atoms per cm^) [19]. CS planes precipitate on cooling crystals across the Ti02-x (0 < X < 0.0035) phase boundary [20]. They show a very strong

446

1200

0.8

1.2

1.6

O/Ti ratio

Fig. 2. Phase diagram of the Ti-0 system re-drawn from ref. [3]. The region Ti203-Ti02 contains Ti203, Ti305, seven discrete phases of the homologous series Tin02n-l (Magneli phases), and Ti02. See ref [3] for a more detailed description.

dependence on the cooling history and are absent in quenched specimen. The formation mechanism was reviewed by Smith et al. [20-22]. Such CS planes may extend all the way to the surface [23-27] and their appearance is discussed in section 3.1.4 below. The diffusion mechanism for the various types of defects is quite different; oxygen migrates via a site exchange (vacancy diffusion) mechanism, while excess Ti diffuses through the crystal as interstitial atoms. The interstitial diffusion happens especially fast through the open channels along the (001) direction (the crystallographic c-axis) [28,29]; see Fig. 3a. A Ti interstitial located in these channels is in an octahedral configuration, similar to the regular Ti sites. Consequently, the diffusing species in oxidation reactions of reduced TiaOb surfaces (where a > b/2 but probably less than b) produced by sputtering and/or Ti deposition is the Ti atom and not the O vacancy, as has been shown in a series of elegant experiments with isotopically labeled '^O and 46Ti by Henderson [14,15].

3. THE STRUCTURE OF THE RUTILE TiO2(110) SURFACE The rutile (110) surface is the most stable crystal face and simple guidelines can be used to essentially predict the structure and estimate the stability of the TiO2(110)(lxl) surface. Because these concepts are very useful for the other crystal faces of Ti02 as well other oxide materials, they are exemplified for this surface. The relaxations from the bulk-terminated coordinates are reviewed as

447

well as the types and shapes of defects. Although the TiO2(110) surface is very stable, it nevertheless reconstructs and restructures at high temperatures under both oxidizing and reducing conditions. 3.1. The rutile TiO2(110)(lxl) surface 3.1.1. Bulk truncation Two concepts have been introduced to predict the stability of oxide structures. Tasker [30] discussed the stability of ionic surfaces based on purely electrostatic considerations. The second concept, autocompensation, was originally developed for surfaces of compound semiconductors and applied to metal oxide surfaces by LaFemina [31]. The most stable surfaces are predicted to be those which are autocompensated, which means that excess charge from cation-derived dangling bonds compensates anion-derived dangling bonds. The net result is that the cation- (anion-) derived dangling bonds are completely empty (full) on stable surfaces. This model allows for the partially covalent character found in many metal oxides, including Ti02. Both concepts are used in a complementary way, and represent a necessary (but not sufficient) condition for stable surface terminations. With very few exceptions, stable metal oxide surfaces for which the structure is known are non-polar [30]a^(i fulfill the autocompensation criterion [31]. Tasker's and LaFemina's approaches are exemplified in creating a stable (110) surface (Fig. 3). In Tasker's concept, the dipole moment of a repeat unit perpendicular to the surface must be zero in order for the surface energy to converge. He introduced three categories for ionic (or partially ionic) structures. Type 1 (neutral, with equal numbers of cations and anions on each plane parallel to the surface) is stable. Type 2 (charged planes, but no dipole moment because of the symmetrical stacking sequence) is stable as well. Type 3 surfaces (charged and a dipole moment in the repeat unit perpendicular to the surface) will generally be unstable. Consider, for example, the rutile structure as being composed of (110)oriented planes such as drawn in Fig. 3 a. The top plane in Fig. 3a consists of the same number of Ti and O atoms. In a purely ionic picture, the titanium and oxygen atoms have formal charges of 4+ and 2-, respectively. Hence, the top layer has a net positive charge. The next two layers consist of oxygen atoms, hence both of them have a net negative charge. A type-2 repeat unit is outlined by the dashed lines A and B in Fig. 3a. It consists of a mixed Ti, O layer, sandwiched between two layers of oxygen atoms. The total unit does not have a dipole moment. A crystal, cut or cleaved^ to expose a (110) surface, will naturally terminate with the surface created by cutting along line A (or B) in Fig. 3a. In Fig. 3b, the top of the model is shifted along the (110) direction (cutting* the crystal in a 'Gedankenexperiment'). The resulting surface is very

Unfortunately, Ti02 fractures and does not cleave well.

448

corrugated because one 'layer' of oxygen atoms is left behind. As shown in below, there is overwhelming evidence that the (1x1) surface of TiO2(110) closely resembles the 'bulk-terminated' structure depicted in Fig. 3b. The same surface structure is also predicted using the rules of autocompensation. In Fig. 3b, the same number of oxygen-to-titanium bonds are broken as titanium-to-oxygen. Transferring electrons from the dangling bonds on the Ti cations will just compensate the missing charge in the dangling bonds on the O anions. Hence, the surface is autocompensated [31]. Note that only the longer (and, consequently, weaker) bonds are broken when the crystal is sliced in this way. This is yet another indication that this surface termination is a likely one. The rutile (110)(lxl) surface in Fig. 3b contains two kinds of oxygen atoms. Along the [001] direction, rows of 6-fold coordinated Ti atoms (as in the bulk) alternate with 5-fold coordinated Ti atoms with one dangling bond perpendicular to the surface. Two kinds of oxygen atoms are created as well. Within the main surface plane, they are three-fold coordinated as in the bulk. The so-called bridging oxygen atoms miss one bond to the Ti atom in the removed layer and are two-fold coordinated. These bridging oxygen atoms are

;S[v W ^M y^

w'Jk ^^^ ^ ^ ^

[110]^

ii [001] 1]

^^T^^'r

^l^^^^iH^^/^^l^^^^il^^^^^ii

3.

B

X^=(?>^m, F^^^^vJinl

Y^^

^

uO ^" j l

r^ ^^^^^jX

^rp\^ ~^

^i7!iS

l)Lny

TZ)

/rS

Y\_^^^^^^^^fej J l [no]

Fig. 3(a) Ball-and-stick model of the rutile crystal structure. It is composed of slightly distorted octahedra, two of which are indicated. Along the [110] direction these octahedra are stacked with their long axes alternating by 90°. Open channels are visible along the (001) direction. The dashed lines A and B enclose a charge-neutral repeat unit without a dipole moment perpendicular to the [110] - direction (a 'type 2' crystal plane [30]). (b) The crystal is 'cut' along line A. The same number of Ti -> O and O -> Ti bonds are broken, and the surface is autocompensated [31]. Note that only the longer bonds are broken. The resulting (110) surface plane is stable and overhelming evidence exists for such (1 x 1 )-terminated TiO2(110) surfaces.

449

subject to much debate. Because of their coordinative undersaturation, atoms from these rows are thought to be removed relatively easily by thermal annealing. The resulting point defects (section 3.1.4.) affect the overall chemistry of the surface, even in a macroscopic way [32]. A (1x1) LEED pattern is generally observed upon sputtering and annealing in UHV. To my knowledge no quantitative LEED study has been reported, probably because of the difficulty of creating defects when the sample is bombarded with electrons (see section 3.1.4). A medium-energy electron diffraction study (MEED) study of TiO2(110) employed an ESDIAD optics with a channelplate; this setup is more sensitive than a conventional LEED apparatus, and allows for very small electron currents to be used. The results of this study are basically consistent with the (1x1) structure depicted in Fig. 3b. X-ray photoelectron diffraction (XPD) spectra also fit the expected (1x1) termination [33], as do STM and AFM results (which are discussed in section 3.1.3 below). 3.1.2. Relaxations Every surface relaxes to some extent. In recent years, the geometry of the Ti02(l 10)(lxl) surface has been studied in some detail both experimentally and theoretically. The results of a surface x-ray diffraction experiment (SXRD) [34] and of several total-energy calculations are listed in Table 1. The experimentally-determined directions of atoms in the first layers are sketched in Fig. 4. As is expected from symmetry, the main relaxations occur perpendicular to the surface. Only the in-plane oxgyens ('4', '5' in Fig. 4) move laterally towards the 5-fold coordinated Ti atoms. The bridging oxygen atoms (labeled '3' in Fig. 4) are measured to relax downwards considerably, and the 6-fold coordinated Ti ('1') atoms upwards. The 5-fold coordinated Ti atoms ('2') move downwards and the neighboring 3-fold coordinated oxygen atoms ('4', '5') upwards, causing a rumpled appearance of the surface. The relaxations in the second Ti02 layer are roughly a factor of two smaller. The most striking feature is the large relaxation of the bridging oxygen atoms by -0.27 A. The measured geometry would indicate a very small bond length between the six-fold coordinated Ti atom ('1') and the bridging oxygens ('3') of only 1.71± 0.07A instead of the 1.95 A expected from the bulk structure. The relaxation results in vertical distances of 0.89 ±0.13A and 1.16 ±0.05A from the 6-fold ('1') and 5-fold coordinated ('2') Ti atoms, respectively. This is in agreement with ion scattering measurements, where vertical distances of 0.87 A and 1.05 ±0.05A were found [35,36]. (Another ion scattering study found the height of the bridging oxygen atoms comparative to that of the bulk structure but the interlayer distance largely relaxed with about -18 ± 4% [37].) Photoelectron diffraction results [38] are also in agreement with relaxations from the x-ray diffraction work given in Table 1.

450

A [110]

[110]

[001]

Fig. 4. Model of the TiO2(110) surface. The relaxations of surface atoms, determined with SRXD are indicated [34]. The labels refer to the relaxations listed in Table 1. Charlton

Harrison Harrison FPLAPW 7 layers

SXRD experiment

0.12*0,05 -0.16±0.O6

Tl(1)(6-fold) Ti(2)(5-fold) p(3)bridging) 0(4,5) [110] [1 1 0] 0(6)

ppiiii; 0{9) TK13) _ TK14)

,

0(15) O(16,17)[110] [110] 0(18) 0(19)

-

0,08 -0.23

0,23 /-0.17

-0.27±0.08 0.05±0.05 ±0.16±0.08 0.03±0.08

-0.16 0.09 ±0.06 -0.09

0.07*0.04 •<),09i0.04

0,07 . 0.14 -0i3 .. -0.10

0.00±0.08 0.02±0.06 ±0.07±0.06 -0.09±0.08

O(10,11)[110] [110] 0(12)

LCAO 7 layers

•

"

'

.

'

.

'

;

'

-

-

"- - ; _

-0.12±0.07

-

-0.05 -0.04 ±0.03 -0.04

/ 0.02 \, -0.08 -0.07 -0.03 ±0.02 -0.02 0.02

-0.02 0.03 ±0.05 0.02

0.00 0.03 ±0.03 -0.01 0.05 -0,06 0.01 0.01 ±0.02 0.01 -0.01

Bates RamaLindan moorthv PW-PPPWPW-PPLDA GGA GGA 5 layers 5 layers 3 layers 0.13 0.09 0.23 -0,11 0.12 ; -0.17 -0.06 -0.09 -0.02 0.12 0.11 0.18 ±0.04 ±0.05 ±0.05 -0.07 -0.05 0.03 0.06 0,12 -0.08 -0.06 : - . 0.02 0.03 -0.03 0.00 ±0.05 ±0.02 -0.01 0.03 .

•

*

-

•

:

,

-

.

•

'

Vogtenhuber FPLAPW 3 layers -0.05 '-v'.-^.1$ -0.16 -0.12 ±0.07

-

Reinhardt HF-LCAO 3 layers 0.09 / -0.15 -0.14 0.07 ±0.08 -0.07

" ,•'<> ,; '' ' ->4, -

:

-0.02

-

-

'- -\ -, . ' -' .. :.,- ' ' ' - " 1 '" - 1 '

-

-

-

,

•

•

.

- -=

-

"r

-

Table 1 Displacements (in A) determined experimentally and theoretically by several groups using different computational techniques. The atomic labels and the directions of the experimentally-determined relaxations are given in Fig. 4. The results are grouped by first authors; Charlton [34], Harrison [2], Ramamoorthy [5], Bates [155], Lindan [41], Vogtenhuber [39], and Reinhard [40]. Acronyms are used for surface x-ray diffraction (SXRD), full-potential linear augmented plane wave (FP-LAPW), linear combination of atomic orbitals (LCAO), Hartree-Fock (HF), plane-wave pseudopotential (PW-PP), local density approximation (LDA), and generalized gradient approximation (GGA). The number of Ti02 repeat units used for the various calculations are indicated.

451

The results of total-energy calculations by several groups [2,5,39-42] are compared to the measured relaxations in Table 1. Two complementary approaches were used, the linear combination of atomic orbitals (LCAO) and plane-wave techniques. Either periodic or free-standing supercells with different numbers of layers (in the sense of Tasker's non-polar repeat units in Fig. 3a) were used. For example, the configuration drawn in Fig. 4 represents part of the upper half of the 7-layer slab used by Harrison et al. [2]. Because of the localized nature of the Ti3d electrons in the Ti02 structure, plane-wave expansions are challenging. A rather high-energy cutoff needs to be used for convergence, and the functional for the LDA or GGA based calculations may also influence the results [43]. In addition, the thickness of the slab may play a role in the accuracy of the calculated geometry. In (almost) all the theoretical papers the directions of the calculated relaxations agree with the experimentally determined coordinates. The quantitative agreement is not as good as one could expect from state-of-the art ab initio calculations, however. As Harrison et al. pointed out [2], the extensive experience of calculations on bulk oxides which has been built up in recent years leads one to expect that DFT and HF calculations will reproduce experimental bond lengths to somewhat better than 0.1 A. For example, the bulk structural parameters of Ti02 rutile agree better than 0.06A using softcore ab initio pseudopotentials constructed within the LDA, and a plane-wave basis [44]. In particular, all the calculations find a much smaller relaxation for the position of the bridging oxygen atom. A possible reason for this disagreement was given by Harrison, Wang, Muscat, and Scheffler [2]. All the theoretical results listed in Table 1 are strictly valid only at zero temperature. It is conceivable that strong anharmonic thermal vibrations at the Ti02(l 10) surface cause the discrepancy in the experimental and theoretical results. Based on these theoretical results, the finite temperature has to be taken into account for a proper evaluating diffraction results. Hopefully, future experiments will show whether a better agreement with theoretically predicted relaxations can be achieved. When considering surface reactions, one also needs to depart from a static picture of this and other oxide surfaces, and has to keep in mind the substantial distortions and bond length changes that take place during such large-amplitude vibrations. 3.1.3. Appearance in STM and AFM Naturally, scanning probe techniques are extremely useful tools for studying atomic-scale structures Ti02 and other metal oxide surfaces, where local changes in stoichiometry or structure can severely affect surface reactivity. On TiO2(110), scanning tunneling microscopy and, more recently, non-contact atomic force microscopy, have been used by many different groups. They have provided valuable and very detailed insight into local

452

surface structure. However, the interpretation of STM images of oxides is somewhat challenging because of strong variations in local electronic structure, and because tips can easily 'snatch' a surface oxygen atom, causing a change in tip state and the resulting 'artifacts' in STM images. The question on what is 'really' observed with STM on this surface is still not resolved, although different groups are converging to a common consensus. My personal view of the ongoing discussion is summarized in the following. The first question is the dominant tunneling site on Ti02(l 10) surfaces. A priori it is not clear as to whether the image contrast is governed by geometric or electronic-structure effects. For Ti02, atomic-resolution STM is only successful when imaging unoccupied states (positive sample bias) on reduced (n-type) samples. In reduced Ti02 crystals, the Fermi level is close to the conduction band minimum (CBM) in the 3 eV gap, so that the electrons are much more mobile than the holes. Under a typical bias of+2V, electrons can thus tunnel from the tip into states within ~ 2 eV above the CBM, and be conducted away from the surface. On the one hand, these CBM states have primarily cation d character (the valence band having primarily O 2p character, see Chapter 14 in this book) so that one might expect to image the metal atoms as the "white" features in STM topographs. On the other hand, the bridging oxygen atoms protrude above the main surface plane and dominate the physical topography. Hence it seems equally plausible that geometrical considerations might dominate the contrast in STM images. Figure 5 shows an STM image of a stoichiometric (1x1) surface. Bright and dark rows run along the [001] direction. The distance between the rows is

Fig. 5. STM image (200 A x 200 A) of a Ti02(l 10)(lxl) surface. The surface was sputtered and annealed in UHV to 1100 K for 10 min. The contrast in STM is normally electronic rather than topographic, and the bright lines in the image correspond to the position of the Ti atoms rather than the bridging oxgyen atoms. The image shows point defects. Features labelled with 'A' have been assigned as oxgyen vacancies. From ref. [53], with permission.

453

Fig. 6 Contour plots of [001]-averaged charge densities associated with electron states within 2 eV of the conduction band minimum for (a) the relaxed stoichiometric (1x1) surface, and (b) the relaxed missing-row type (1x2) surface. Contour levels correspond to a geometric progression of charge density, with a factor of 0.56 separating neighboring contours. To first approximation, the STM tip will follow one of the equal-density contours several angstroms above the surface. After ref. [47,53]^ with permission.

6.3±0.25 A, in agreement with the unit cell dimension of 6.5A along [110]. At neighboring terraces they are staggered by half a unit cell. The "bridging oxygen" rows protrude from the surface plane on a relaxed Ti02(l 10) surface (see Table 1), so if STIU were dominated by topographical effects, they would appear as rows with high contrast in Fig. 5. There is strong evidence that, normally, this is not the case, and that the Ti sites are imaged bright in this and similar images. Onishi et al. have observed formate ions (which are expected to adsorb to Ti sites) on top of the bright rows [45]. This is now confirmed for many other adsorbates, e.g., chlorine [46] and sulfur [17] appear as bright spots on top of bright or dark rows when adsorbed on Ti sites or oxygen sites, respectively. A theoretical approach to determine the image contrast in STM is shown in Fig. 6. Pseudopotential calculations were used to analyze the local density of states in the vacuum region above the surface [47]. In rough correspondence with the experimental bias conditions, the charge density of conduction-band states were summed up from 0 to 2 eV above the conduction-band minimum. This quantity was then averaged over the [001] direction and plotted as a function of the other two coordinates as shown in Fig. 6. Under constantcurrent tunneling conditions, the STIVI tip is roughly expected to follow one of the equal-density contours several angstroms above the surface. The plot in Fig. 6(a) clearly shows that the charge-density contours extend higher above the 5-fold coordinated Ti atoms, in spite of the physical protrusion of the

454

bridging oxygen atoms. This confirms that the STM is imaging the surface Ti atoms, i.e., that the apparent corrugation is reversed from naive expectations by electronic-structure effects. The slab in Fig. 6b has a (1x2) symmetry with every other bridging oxygen row missing. This configuration has been proposed originally to account for the (1x2) structure observed in LEED [48]. More recent experimental evidence, resulting predominantly from STM measurements, have shown that this is not a likely structure (see section 3.2.1). However, the charge contours in Fig. 6b indicate that single vacancies in the bridging oxygen rows are expected to appear as bright features on the dark oxygen rows. A different and computationally less expensive computational approach has been taken by Giilseren and coworkers [49]. They have used a firstprinciples atomic-orbital base scheme with limited self-consistency. From analyzing the radial distributions of O and Ti wave functions, it was concluded that the STM tip should sample electrons from different surface atoms, depending on the height. For close distance (<4A) the contributions from the oxygen atoms should dominate, for larger separations, the Ti atom. Hence a 'reversal' of the tunneling site should be possible. STM images, taken with high tunneling current and relatively low bias voltages (Itunnel = 2.0 nA, Vsample ^ +0.75 V) seem to confirm this conclusion [50]. Such images show an enhanced resolution, and alternating rows of individual dots and white rows. A high resolution was also reported in reference [51] and has been explained as tunneling centered at the 5-fold coordinate Ti4+ ions and at the bridging oxygen ions. After 'functionalizing' the STM tip by scanning over a Si surface with +10V sample bias and 20 nA, enhanced resolution has been observed by another group [52]. These images were interpreted as tunneling into the sixfold coordinated Ti atoms underneath the bridging oxgyens. The tunneling process was interpreted as being influenced by a strong chemical interaction and formation of a partial chemical bond between Si and surface oxygen. However, as pointed out in ref [53], tip changes occur often on Ti02, and can make a definite interpretation of 'unusual' images (with seemingly very high resolution or reversal of image contrast) rather difficult. For the sake of studying surface structure and adsorbates, it is maybe better (and often sufficient) to focus on results obtained with the "normal" tip state that can be obtained reproducibly by 'cleaning' with high voltage/high current pulses. Recently, non-contact AFM has been introduced as a complementary technique to study Ti02 surfaces with atomic resolution. An image of the TiO2(110)(lxl) surface, obtained by Fukui et al. [54], is reproduced in Fig. 7. The frequency modulation technique described in ref [55] was used as the feedback signal. While the contrast formation of atomically-resolved AFM images using this technique is also somewhat controversial [56], one would assume that the physical geometry should dominate in AFM. The registry of (1x2) strands (see section 3.1.4) in AFM images was consistent with the bright

455

4i

^ - / * t^ r # f ^ fW ,^f

-?

"

# #

Fig 7. Non-contact AFM image of a Ti02(l 10)(lxl) surface. Reprinted with permission fromref [54].

rows in Fig. 7 being the protruding bridging oxygens. Consequently, the black spots in Fig. 7 have been assigned as oxygen vacancies. 3.1.4. Surface defects The ability to control the amount of defects is one of the main attractions of Ti02 as a 'well-characterized' model system. Imperfections such as vacancies introduce changes in the electronic structure (as discussed in detail in Chapter 14 of this book). These defects have been investigated with spectroscopic techniques for years. From recent with scanning probe techniques much has been learned about the structure of defects. In the following are discussed steps; vacancies produced by thermal annealing, sputtering, and electron bombardment; line defects (strands which are related to the (1x2) reconstruction); common impurities such as Ca and FI; and the manifestation of crystallographic shear planes on the surface of Ti02(l 10).

Step Edges Sputtering and annealing in UHV at (not too high) temperatures renders flat (1x1) surfaces. As is expected for annealing of sputter-damaged surfaces, the terrace size increases with annealing temperatures. This has been shown nicely in an STM work by Fischer et al. [57]. The correlation length in SPALEED measurements (which corresponds to the average terrace size) increased with a ti/4 dependence at temperatures above 800 K [58]. Interestingly, Ar implanted during the sputtering process at relatively moderate ion energies (1000 eV) and fluences (typically 1 |iA and 30 min) does not completely leave the near-surface region during annealing up to 1000 K and is still visible in XPS andAES[17]. An example for a typical terrace-step structure is shown in Fig. 8. Step edges on annealed surfaces run predominantly parallel to <001> and < l T l >

456

Fig. 8. STM image of a clean stoichiometric TiOaCl 10)(lxl) surface after sputtering and annealing to 1100 K in UHV. The step structure is dominated by step edges running parallel to <1 1 1> and <001> directions. A kink site at a <1 1 1> step edge is marked with 'K'. Smooth ('UR') and rugged ('R'econstructed) <001>-type step edges appear with roughly equal probability and are marked with arrows. The lower part of this figure shows a balland-stick model of two terraces of Ti02(l 10). The step edge FG runs parallel to <111> type directions. The smooth and rugged step edges along <001> in the STM image are attributed to step edges AB and HI, respectively. After ref [53], with permission.

type directions [53,57,59]. These steps are measured to be 3.2 A high, in agreement with the value expected from the rutile structure [60]. In the STM image in Fig. 8, a kink site at the point where a type step edge turns into a type step edge is labeled with K. Such kink sites are located at the end of dark rows (the bridging oxygens). Two kind of <001>-type steps are pointed out by arrows in Fig. 8. One kind appears

457

smooth (marked as 'UR'), the other one rugged (marked as 'R' for reconstructed) with a high number of kinks. Both types of step edges appear with roughly equal probability in images. It is relatively straightforward to construct models for step edges following the rules of autocompensation [31]. For example, Fig 8 shows a ball-and-stick model of two layers of TiO2(110) that contains several step edges [53]. As outlined above (section 3.1.1.) the same number of O -> Ti bonds and Ti -> O bonds need to be broken when cutting a Ti02 crystal, for example by forming a step edge by removing part of the upper terrace. A step edge (parallel to the diagonal of the surface unit cell) runs between the corners labeled 'F' and 'G'. A change in step orientation from the to the direction occurs always at the bridging oxygen rows ('K' in Fig. 8). The local environment of the atoms at such a kink site is not different from the rest of the step edge. Similarly-constructed models of step edges oriented along [1T2] and [iTs] directions are given in ref [61]. There are several possibilities to form step edges parallel to the <001> direction. One can either cut next to the Ti atoms underneath the bridging oxygens (parallel to the arrow labeled '1' on the upper terrace in Fig. 8) or between the in-plane oxygen and titanium atoms (parallel to arrow '2'). If one cuts at position 1, an autocompensated step edge is formed (step AB in Fig. 8). If one cuts a terrace parallel to the <001> direction at position 2, the step edge that is formed is not autocompensated. A 'reconstruction' was proposed [53] by removing three of four Ti atoms as well as the neighboring in-plane O atoms (step edge HI in Fig. 8). This detailed insight into step geometries and coordination number of atoms at step edges and kink sites is important, because a decrease in coordination number of surface atoms often correlates with an enhancement in chemical reactivity. For example, pyridine molecules have been found to be more strongly adsorbed at the 4-fold coordinated Ti atoms at step sites than at the five-fold coordinated Ti atoms on the terraces [61]. A different effect has been observed by Iwasawa et al. [62]. Adsorption of formic acid at 400 - 450 K resulted in particles with a strongly suppressed presence in the vicinity of step edges. Possibly, electrostatics plays a role. To my knowledge, virtually no theoretical work has been done to determine step geometry with the same detail as flat surfaces. Because step edges break the symmetry, first-principles calculations would require huge unit cells. With the advent of ever faster computers and more powerful programs such calculations may soon become viable. Oxygen vacancies created by annealing There is overwhelming spectroscopic and chemical evidence for the presence of point defects on samples sputtered and annealed in UHV. These are attributed to vacancies in the bridging oxygen rows. Their concentration is typically reported as several percent [63,64]. (The defect concentration may

458

increase somewhat when crystals become more reduced, but no systematic study on the correlation between defect concentration and bulk properties has been performed.) These defects are of high importance for the surface properties of TiO2(110). Thermally-induced point defects are visible as distributed black points on the bright oxygen rows in the ATM image in Fig. 7. As already pointed out, STM results are harder to interpret because of strong electronic effects. STM images of titanium dioxide surfaces that have been annealed in UHV often exhibit two kind of atomic-scale features (Fig. 5). These appear as short bright features centered on dark rows (labeled 'A') and connecting the neighboring bright rows, and as dark spots on the bright rows (labeled 'B'). The features labeled A have a density of 7 ± 3% per surface unit cell, consistent with O vacancy density estimates from spectroscopic measurements. The size of the bright spots is of the order of one single atom (FWHM 5 A in line scans along the [001] direction) when taking into account convolution with the tip. They always appear as isolated spots with no apparent short-range ordering but a slight tendency to be staggered perpendicular to the rows. It is well-known that oxygen vacancies are healed by dosing a reduced Ti02 sample with oxygen [63,65,66]. When the sample in Fig. 5 was stepwise exposed to oxygen, the density of the bright spots (A) decreased. This, and the contrast expected from electronic structure calculations (missing bridging oxygens give rise to a hump in empty-states charge-density contours, see Fig. 6b) led to the assumption of the bright spots being missing oxygen vacancies [47,53]. As was reported in ref. [53], their appearance in STM is strongly tip-dependent. The dark features ('B' in Fig. 5 are less common (~ 1 - 2% per surface unit cell) than the oxygen vacancies. They can extend over several unit cells. Upon adsorption of oxygen, the number of the dark spots stays constant within the statistical error. They were tentatively assigned as sub-surface oxygen vacancies [46], however, first-principles total-energy calculations show that such a configuration is highly unlikely [67]. Their nature is unclear at this point. In a recent STM/ESD work, Fukui et al. report that oxygen vacancies are hydroxylated, and that the bright spots between bright rows should be hydrogen atoms on bridging oxgyens rather than vacancies [68]. The presence of H impurity atoms was concluded from a combined STM/ESD study [68]. STM images show bright spots on dark rows with a coverage of 0.25 - 0.05 ML, depending on sample history. Their density decreases upon irradiation with 20-eV electrons (with an energy below the O desorption threshold in ESD, see below) and increases upon exposure to atomic hydrogen. A residual species that was not desorbed upon electron irradiation was interpreted as H atoms bound to oxygen vacancies. For the sake of image contrast, this does not change the arguments, as hydroxylated bridging oxygen atoms would allow to assign the position of Ti and bridging O rows equally well as O vacancies. From an electronic structure point of view, the STM contrast of hydrogen

459

atoms adsorbed on top of bridging oxgyens is expected to be similar to the one calculated for oxygen vacancies [67]. Oxygen vacancies created by other means Oxygen vacancies can also be created by bombardment with electrons. Ti02 is the classic example for a maximum-valency compound material where electron-stimulated desorption occurs via the Knotek-Feibelman process [69]. Bombardment with energetic electrons creates a core hole in the T\3p level. With a certain probability, this hole is filled through an inter-atomic Auger process from a neighboring O atom. If two (instead of the usual one) valence electrons are emitted during the Auger decay, the oxygen anion becomes positively charged. The previously attractive Madelung potential changes into a repulsive one, and an O"^ ion is emitted [70]. The whole process has a threshold energy that correlates with the Tiip edge as discussed in detail in ref. [71]. Such electron-stimulated defects behave somewhat different that thermally-created ones [72]. It is generally assumed that electron bombardment results in the ejection of bridging oxygen atoms, but direct evidence from STM studies points towards more complicated structures [73]. The high current and high field provided by an STM tip has been used to create humps and craterlike hollow structures [74]. Defects can also be created by irradiation with UV light, but, again nothing is known about their structure [75]. Sputtering with rare-gas ions reduces the surface oxygen content. Usually, the long-range order of the surface is lost, and the LEED pattern disappears. Spectroscopic measurements as well as adsorption experiments indicate the defects are more complex, involve more than one atoms, and are partially sub-surface [72]. There are indications that sputtering does not completely randomize the surface but creates some short-range order [76]. Irradiation with high-energy electrons (300 keV) induced a TiO as well as an intermediate Ti02-II phase [77]. Generally, sputter-induced damage can be removed easily by annealing in UHV [15]. Line defects Scanning Tunneling Microscopy images of UHV-annealed surfaces (which exhibit a (1x1) LEED pattern) often show dispersed bright stands, typically several tens of Angstroms long. They are distributed across terraces and have a tendency to grow out of step edges onto the lower terrace (see Fig. 9). The strands are centered on top of bright rows of the lower terrace (on top of the 5-fold coordinated Ti atoms). STM shows a bright spot at the end. A double-strand structure is often resolved in high-resolution images (Fig. 9). As shown in Fig. 9 these strands are precursors for the (1x2) reconstruction. Conflicting geometric models have been proposed for this reconstruction. These are discussed in section 3.2.1. below.

460

Fig. 9. (a) STM image of strings growing out of the upper terrace. These strings are centered on the 5-fold coordinated Ti rows of the lower terrace. They are precursors of the (1x2) reconstruction. The surface has been annealed at 1020 K for 1 min., Sample bias: +0.7 V, tunneling current: 0.3 nA. (b) Variable current scan of the (lx2)-ordered strings on the surface annealed at 1150 K, 70 x 70 nm^, sample bias: +2.0 V, tunneling current: 0.3 nA. The vertical axis in both images is parallel to the [001] direction. Reproduced from ref. [60].

The presence of such dispersed strands is sample-dependent. Li et al. have heated to different reduction states samples cut from the same specimen. After sputtering and annealing at 973K for 10 min, strands were present on dark blue samples, but not on the lighter ones [9]. The structure of the strands may also depend on the reduction state of the crystal [16]. Investigations with numerous Ti02 samples in my own laboratory have shown that small amount of bulk impurities (well below the detection limit of commonly-used surface analytical techniques) can also cause strands on the surface. This is in addition to the bright strings caused by Ca segregation discussed in the next section. Impurities Commercial Ti02 single crystals are generally quite clean. The most common impurity is calcium, which tends to segregate to the surface upon high-temperature annealing [78-80]. Typically, Ca can be depleted from the near-surface region with a few sputter/annealing cycles. For high coverages, it forms a well-ordered overlayer which can clearly be observed in LEED [7880].

Zhang et al reported a

structure in LEED and [001]-oriented,

bright strands in STM. The formation of a CaTiOs-like surface compound was tentatively proposed. This structure was questioned by Norenberg et

461

aL[80], who reported a p(3xl) reconstruction and antiphase boundaries which could give rise to the LEED pattern observed in [78]. Model calculations suggest that Ca substitutes five-fold coordinated surface Ti atoms. Two vacancies form in the three-fold coordinated O atoms located next to the Ca atom. Ordering is energetically favorable, and the relaxed structure is predicted to be quite buckled [80]. The Ca-segregated surface is very flat which points to a rather low surface free energy. Such a flattening takes place also in other segregated oxide systems [81]. Ca segregation also appears to affect the formation of CS planes (see next section). In addition to Ca, some Mg segregation is sometimes observed. Persistent Al impurities were detected with static SIMS [82]. This technique is very sensitive to certain elements. The Al probably resulted from the polishing procedure. All these impurities can be removed to a large extent by sputtering/annealing cycles. It appears that H atoms, adsorbed at bridging oxygen atoms or at oxygen vacancies [68], are almost always present on TiO2(110)(lxl) surfaces. They can stem from dissociative adsorption of water on oxygen vacancies [64] (which is a precursor-mediated process and can hardly be avoided even in the best vacuum conditions). Possibly, H impurities may also come from the bulk of the crystal [68]. Instead of the usual sputter/annealing preparation of new crystals, oxygen plasma treatment appears to be a quite efficient treatment for producing clean, (1x1) terminated terraces [83]. Plasma treatment removed Ca (that often segregates to the surface initially) quickly. At relatively low temperature it produced larger terraces (1500 A in width) and straight step edges than typically found on sputtered/annealed surfaces.

Crystallographic Shear Planes As pointed out above, the oxygen loss through thermal annealing that occurs in the bulk of Ti02 crystals can lead to crystallographic shear (CS) planes. An overview of CS planes and their relation to surface structure is given by Bennett et al. [26]. The CSP's are formed by shifting the normally edge-sharing octahedra (see Fig. 3a) to a face-sharing arrangement. The crystal can be thought as consisting of small, undisturbed volumes of the regular rutile lattice, separated by CS planes (see Fig. 10).. As their concentration increases, they can order into a regular array known as Magneli phases (see Fig. 2). Two homologous series of Magneli phases are known to exist with shear planes along {121} and {132} directions (see Table 1 in ref [26]). The CS planes intersect the (110) surface with adjacent sections of the crystal being displaced by 1.6 A. A good example of CS planes intersecting with a (1x2) -reconstructed surface is shown in Fig. 10. One of the first STM studies of reduced Ti02 showed various periodic structures which were interpreted as CSP defects [23,84]. Ti02 crystals that

462

<335>

JSlormal step. 1/2 step CSP step edges

Interstitial site

CSP centre slab

Fig. 10. (top) STM image (400 A x 400 A, 0.2 nA, 1 V, 773 K) showing a CSP pair running in the <335> direction across (1 x 2)- reconstructed terraces, (bottom) Schematic showing a <001> projected view of a pair of CSPs terminating at the [110] surface. Oxygen ion distorted octahedra centred on the Ti ^'^ ions (dark dots) are indicated by the diamond shapes. The interstitial Ti populates the <001> channels in the bulk lattice as indicated in the bottom left while a normal step edge is shown top left. The distorted octahedra forming the CSPs are shown in grey and have not been relaxed from their bulk positions in the normal and displaced lattices. Where the CSPs terminate at the surface 1/2 height steps are formed which have differing structures. From ref. [86], reproduced by permission of the Royal Society of Chemistry.

were subjected to heat-treatment only (without sputtering) showed Ca segregation and step edges along [110], [ 1 1 1 ] and [111] directions [24]. These were interpreted as CSP's belonging to a close packed family of {112} planes [24]. The substantial Ca segregation may have influenced this structure. The surface structure of the most oxygen deficient Magneli phase, Ti407, was investigated by the same group [85]. After repeatedly sputtering and heating (1223 K) a TiO2(110) crystal (which diminished the initially observed Ca segregation), Bennett et al. report STIVI and LEED results of CS planes on a crystal with dark blue/black color. The crystal showed a rippled texture that was visible with the naked eye [26].

463

The Ti(390 eV)/O(510 eV) AES ratio was 1.22 for the heavily reduced surface (compared to a ratio of 1.14 for the (1x1) terminated surface). On different parts of the crystal, streaks along the [111] and [112] directions, or a superposition of the two, were observed with LEED. STM images show a large concentration of steps running along < 112> and < 1 1 2 > with the expected step height of 1.6 A, as well as a strong variation in mesoscopic morphology [26]. The CS planes also act as nucleation sites for re-growth of new Ti02 layers during high-temperature oxidation (see section 3.2.2) [86]. 3.2. Reconstructions 3.2.1. Reconstruction under reducing conditions: The structure (s) of the (1x2) phase The most common reconstruction observed on TiO2(110) surfaces has a (1x2) symmetry with a doubling of the periodicity along the [110] direction. Various models have been suggested for this reconstruction and are depicted in Fig. 11 [50]. A (1 X 2) LEED pattern was originally observed after high-temperature annealing of a reduced Ti02(l 10) sample in ultrahigh vacuum (UHV). Based on Ti:0 AES ratios it was interpreted as alternate rows of bridging oxygen missing from the regular (1 x 1) surface ("missing-row moder'[48], Fig. 11a). One of the first atomically-resolved STM results on this surface was also interpreted as missing bridging oxygen rows [51,87]; however, the Ti atoms underneath the missing oxygen atoms had to be shifted by half a unit cell in [001] direction to account for the observed image contrast. A structure with a (3x2) symmetry was reported by one group [88]. A model for the this reconstruction was proposed where this symmetry is achieved by removing 1/3 or 2/3 of the oxygen in the bridging oxygen rows. However, such a reconstruction has not been reported by other groups. The simple missing row model for the (1x2) structure in Fig. H a has been discounted on the basis of more recent results. In STM the (1x2) reconstruction appears as a series of bright strings along the [001] direction [50,51,60,87,89-91]. At low coverage, the strings grow preferentially out of the upper terrace onto the lower one (Fig. 9 a) [60]. At first, they are scattered across the terraces with a minimum distance of 13A along the [001] direction. They consist of bright double strings (although the double-ridge structure is often not resolved), with a bright dot at the end (see Fig. 9 ). Antiphase boundaries are observed in high-resolution images of a fullydeveloped (1x2) reconstructed surface [50]. Higher periodicities, i.e., a local (1x3) reconstruction, have also been observed [90,92,93]. In STM images the (1x2) strands generally have an apparent height smaller than a regular Ti02 step edge of 3.2 A, and are in registry with the bright rows of the (1x1) substrate. Because most researchers report empty-state images and because these are dominated by the tunneling into mostly Ti 5(i-derived states bright

464 (a) missing row

(b) added TizOa' row

iMfiM (a) added row

Fig. 11. Models for the Ti02(l 10)(lx2) surface. Small white balls, Ti, large black balls, O. (a) The 'missing row' model, obtained by removal of one row of bridging oxygens, was originally proposed by Moller et al. [48]. This model is inconsistent with more recent STM images and first-principles calculations, (b) The 'added row model' has Ti203 stoichiometry as proposed by Onishi and Iwasawa [59]. (c) The 'missing unit' model was proposed by Pang et al. [92] . Recent evidence suggests that the structures in b) and c) might both be present at Ti02(l 10)(lx2) under different conditions. From ref [50], with permission.

strands centered on top of bright substrate rows imply that the (1x2) strands are at the position of 5-fold coordinated Ti atoms and not at the bridging oxygen atoms. STM images of a simple missing row structure are expected to show a bright feature above the missing bridging oxygen row (provided the STM tip is a reasonable distance from the surface [49]), inconsistent with the registry observed experimentally. The rows can be removed by tunneling under 'extreme conditions' (Vsampie = +1.5 V, I = 10 nA [50]). First-principles total energy calculations show that the missing row structure is energetically equivalent to a (2x1) structure (where every other bridging oxygen is removed) [41]. For all these reasons, the missing row reconstruction is no longer considered a viable model. Early on, Onishi and Iwasawa [59,60] suggested a quite different model. It consists of double rows of Ti cations in a distorted tetrahedral configuration

465

(Fig. 11). The structure has a Ti203 stoichiometry, and the model is often called 'added Ti203 rows'. However, it needs to be emphasized that the structure does not resemble the one found in corundum Ti203 crystals. Selfconsistent total-energy and electronic structure calculation found that this added 'Ti203' row structure has a lower surface free energy than the missing row structure and that it is consistent with the expected contrast in STM [94]. It is also supported by ESDIAD [95], high-resolution STM [50], and ion scattering [93] measurements. Based on the fact that (1x2) rows extend out of step edges, a modified model of the missing row structure has been proposed by Murray et al. [27] which involves narrow rows with missing bridging oxygens that are effectively part of the upper terrace. Lateral relaxations were also included [27]. This model was shown to be consistent with calculated surface charge densities [49]. Based on the same scheme, an 'added row model' was proposed by Pang et al. [92] for the fully-reconstructed surface. This consist of narrow, long regions of the regular Ti02 structure with all the atoms in bulk-like positions and with all bridging oxygen atoms missing (Fig. 11). The black grooves between the bright rows are then due to the missing Ti02 units separating the rows. (Consequently this model has also been called 'missing unit' structure [50].) The stoichiometry of the added rows is Ti305. The model was based on STM images with unusually high resolution and the observation of a (1x3) phase consisting of thicker rows. Charge-density calculations by Pang et al. supported the added row model, but were inconsistent with either the added Ti203 or the simple missing row model (Figs. 1 la and b, respectively). The off-normal lobes in ESDIAD images were supposed to stem from the O atoms at either side of the added Ti02 'rows', adjacent to the missing units. Pang's model has been questioned by Tanner et al. [50,96,97]. The 1x2 strands that are part of 'restructured' surfaces after low-temperature oxidation (see the next section) are consistent with the added Ti203 rather than the added row model [98]. Ion scattering measurements are also consistent with the added 'Ti203' model [93] although (somewhat surprisingly) extra oxygen atoms at the position of the 5-fold coordinated Ti atoms were postulated according to these measurements. To my knowledge, no total-energy calculations have been made that would support the added row structure. Such calculations could help to decide which of the added row models has a lower energy. As is discussed in the next section, the two added row models appear not to be mutually exclusive, and the formation of one or the other structure may just depend on the preparation and crystal parameters. 3.2.2. Restructuring under oxidizing conditions The first atomic-level investigation of the dynamic processes that occur when reduced Ti02 crystals are exposed to oxygen was reported by Onishi and Iwasawa [99]. These authors used a blue crystal with a resistivity of 2 Qm. They acquired STM images while the sample was kept at a temperature of 800

466

K and under an O2 background pressure of IxlO-^ Pa. Added rows (interpreted as Ti203 rows, Fig. lib) and 'hill-like features', appeared while imaging the surface, and disappeared when the oxygen was pumped out from the chamber. This effect was not tip-induced; the same structures were observed on areas of the sample that were not imaged during the hightemperature oxygen exposure. It is now established [9,16,100-103] that such an oxygen-induced surface restructuring effect is attributed to the re-oxidation of the reduced crystal, as already suggested by Onishi and Iwasawa [99]. As mentioned above, the bulk defects in sub-stoichiometric Ti02-x consist partially of Ti interstitials which show a high diffusitivity at elevated temperatures. When these interstitials appear at the surface, they can react with gaseous oxygen and form additional Ti02 (or TiaOb) structures. At high oxygen partial pressures, a complete reoxidation of the whole crystal can be achieved [9]. The reoxidation process has pronounced effects on the surface structure. The kinetics of the oxygen-induced restructuring process as well as the resulting surface phases depend on sample temperature, annealing time, gas pressure, and reduction state (i.e., 'age' or color) of the crystal. These parameters have been investigated in detail by Li et al. [9,98,100-102]. For example. Fig. 12 shows the effect of annealing in 1x10-^ mbar O2 at various

Fig. 12. STM images (500 A x 500 A) of a Ti02 (110) surface taken at room temperature. The surface was exposed to I8O2 (1 x 10-6 ^ibar) at (a) 500 K, (b) 520 K, (c) 550 K, (d) 660 K for 10 min, (e) 710 K for 15 min, and (f) 830 K for 20 min. Before each gas exposure the sample was sputtered and annealed in UHV at 880 K for 30 min which renders flat, (1x1) terminated surfaces. Fromref [102].

467

temperatures [102]. Before each gas exposure, a flat (1x1) terminated surface was prepared by sputtering and annealing in UHV. The morphology of the sample is very temperature dependent. For medium temperatures, surfaces are relatively rough with many small-scale features. Isotopically labeled 18O2 gas was used for the annealing excursions. In SSIMS and low-energy He+ ion scattering measurements, the signal from i^O atoms can clearly be separated from the (naturally much more common) i^O isotope in the crystal. All the surfaces in Fig. 12 showed an enrichment with i^O, with a maximum at ^^O surface concentration around 660 K. The structures that form for intermediate annealing temperatures (520 K - 660 K) are better seen in the small-scale image in Fig. 13a. They consist of small, (1x1) terminated islands and irregular networks of connected 'rosettes', i.e., six bright spots in a pseudohexagonal arrangement, as well as small strands. A model for the rosette network is shown in Fig. 13b. It consists of atoms in bulk-like positions with some atoms missing from the regular (1x1) structure. LDAbased first-principles calculations [102] have shown that such a rosette structure is stable. The same calculations also predict sizable relaxations. The rosette structure can be explained simply by the formation of (partially incomplete) Ti02 layers through a growth process where the Ti atoms come from the reduced bulk and the i^O from the gas phase. The kinetics of the growth determines the relative concentration of the incomplete structures (the rosettes and strands) and the (1x1) islands on the surface. The surface is flat and (1x1) terminated when the growth is slow in comparison with surface diffusion processes. This can be achieved in various ways, either during annealing at high temperatures. Fig. 12f, or when the flux of one of the constituents is small (in very light samples with a small concentration of interstitials [9]) or at lower O2 background pressures. Conversely, on very dark crystals, or at intermediate temperatures, a substantial part of the surface can be covered with rosette networks. Note that 60% of the Ti atoms in the rosettes are 4-fold coordinated, whereas the Ti atoms exposed on the (1x1) surface are 5-fold coordinated. One should take this into account when preparing rutile (110) surfaces for surface chemistry experiments. The surface structure in Fig. 12e, obtained after annealing at 710 K in 1x10-6 mbar 1^02, shows the presence of (1x2) strands on otherwise flat, (1x1) terminated surfaces. This is in agreement with Onishi and Iwasawa's results described above [99]. From atomically-resolved images of strands connected to rosettes as well as UHV annealing experiments of restructured surfaces it has been shown that these strands also have the Ti203 structure depicted in Fig. 11 [98].

468

(b) Fig. 13. (a) An atomically resolved STM image (150 A x 150 A) of a surface prepared as in Fig. 1 Ic. Small (1x1) terminated islands and patches of connected pseudohexagonal rosettes are seen, (b) Atomic model (top and side view) for the oxygen-induced structure observed in (a). A bulk-terminated (1x1) island is shown on the right side and the unit cell is indicated. Small white balls are Ti atoms. Shadowed large balls represent oxygen atoms; darker shading indicates higher z-positions. The rectangle indicates the unit cell of the (1x1) structure. The network patch ('R') on the left side consists of an incomplete Ti02(l 10)(lxl) layer and contains only atoms at bulk position. The strands probably have a structure similar to the added Ti203 model in Fig. 1 Ic. From ref. [102] .

The dependence of the restructuring effect (as well as the type of (1x2) reconstruction) on the reduction state of the bulk was shown by Bennett et al. [16]. High-temperature STM studies were performed on two different Ti02 samples. On a dark blue/black crystal (that showed already evidence for CS plane formation) two different structures were observed (see Fig. 14.) The dark and bright strings in Fig. 14 were attributed to added Ti203 rows and added rows of a bulk-terminated Ti02 layer, but with bridging oxgyens at the center of the strands, respectively. The latter structure also appears crosslinked with partial 'rosettes' (Fig. 13b). Both, the added row structure and the rosettes are just incomplete Ti02 structures that form during the growth of additional Ti02(l 10)(lxl) layers. The authors have published impressive webbased STM 'movies' [104] (which can be viewed at http://www.njp.org/) of the growth process that results from the cyclic completion and new formation of the cross-linked added-row structures. In contrast, the dark rows (the Ti203 added rows) appeared relatively unreactive for additional growth.

469

148

^3

iLine Profile A-B

148

Line Profile C-D

A

D) 2

W^^ ^ 0

A

l'\

\d

0 10 20 30 40 60 60

0

20 40 60 80 100

Line Length (A)

Fig. 14. High temperature STM image of oxygen-induced features on a rather dark, nonstoichiometric TiO2(110) crystaL The crystal was exposed to oxygen (5.5 x 10"^ mbar, 833 K), stopped midreaction by removal of the oxygen overpressure, and imaged at the same temperature. The two types of strands with different apparent height (dark strands, DS and bright strands, BS) are attributed to the presence of different (1x2) structures. The added Ti203 rows account for the dark strings. The bright strands are interpreted as a slightly modified (addition of bridging oxgyens at the strands) added row structure (Fig. l i e ) . Additionally, bright rows and vacancies are visible on the (1 x 1 ) surface (marked BR and V, respectively). Line profiles are taken in the fast scan direction to minimize thermal drift and tip change problems. Reproduced from ref. [16], with permission...

3.3 Recommendations for surface preparation Although the Ti02(l 10)1x1 surface is considered the 'bestcharacterized', prototypical metal oxide surface, the above summary clearly shows that its atomic-level structure is significantly more complex than originally assumed. The recent STM results clearly indicate that both the oxidation conditions and the history of the TiO2(110) sample have significant influence on the morphology of the surface, the presence of strands, rosettes.

470

or CSP's. The variations in the surface structure with O2 pressure, crystal temperature and bulk defect density are so vast that one could suspect chemistry of the TiO2(110) surface could also be significantly variant for samples oxidized under different conditions. For example, the issue of whether water is molecularly or dissociatively adsorbed on TiO2(110) [63,64,72,105,106], may be significantly clouded in the literature because of studies in which the morphology of the surface was unknowingly disordered by the presence of the rosettes and/or strands observed recently by STM. This level of ambiguity may also permeate many other adsorption studies on TiO2(110). Guidelines of surface preparation of TiO2(110) can be extracted from this recent work [9,16,83,102,103]. If solely (1x1) terminated surfaces are desired, light blue crystals should be used. Annealing in oxygen will then result in stoichiometric, (1x1) terminated surfaces. On the other hand, more complex morphologies with a range of different coordination sites can be formed if dark crystals are annealed in oxygen. Oxygen plasma-treatment might be a good alternative to the more commonly used sputter/annealing cycles [83], but is not accessible in many standard surface science set-ups. The rich array of surface structures that can be produced on TiO2(110) may provide a playground for surface science experiments where the influence of different adsorption sites can be tested.

4. THE STRUCTURE OF THE RUTILE (100) SURFACE 4.1. The TiO2(100)(lxl) surface The rutile (100) surface has received considerably less attention than the (110) crystal face. The rules of autocompensation and creation of non-polar surfaces discussed above (3.1.1 allows a straightforward prediction of the stable surface termination (Fig. 15a). Again, the same number of Ti -> O as O -> Ti need to be broken, as indicated by the line in Fig. 15 b. This results in a strongly corrugated surface, with rows of bridging oxygen atoms at the outermost, [001]-oriented ridges (Fig. 15 a). Indeed a (1x1) terminated LEED pattern is observed on this surface after sputtering and annealing, and STM and noncontact AFM images are consistent with this model [107,108]. Several theoretical calculations have determined likely relaxations of the (1x1) surface [5,40,109,110]. In ref [109] different theoretical approaches and basis sets were tested. All these calculations agree in the general motions of the atoms, although the amount of relaxations differ somewhat. As expected from symmetry, no relaxations occur along [001]. In the [100] direction only the 5-fold coordinated Ti atoms show appreciable relaxations (downwards) [5]. Substantial relaxations occur along the [010] direction with the two-fold coordinated and the three-fold coordinated oxygen atoms moving in opposite direction of the five-fold and six-fold coordinated Ti atoms.

471 2-fold coordinated oxgyen s.foid coordinated titanium 3-fold coordinated ^^ _ , ^ oxgyen

A

(110) ^°S

f^ micro-

interstitial site: A

trigonal prismatic coordination: B

Fig. 15. (a) Geometry of the unreconstructed TiO2(100)(lxl) surface. This surface results when the same number of Ti -> O as O -> Ti bonds are broken in a bulk crystal, see dashed line in (b). The formation of the (1x3) microfaceted surface, originally proposed by Zschak et al. [118] is displayed in (b). Removal of the volume labeled y produces a stochiometric surface. In order to reconcile the reduced character of the (1x3) surface (observed in photoemission), the outermost bridging oxygen atoms (Y) are thought to be missing as well [119]. The transition from the (1x1) surface to the (1x3) microfaceted surface was proposed to occur via breaking the bonds a which leads to the relaxation indicated by the small arrow and the creation of an intermediate (1x3) phase [107]. (c) A surface x-ray structure analysis by Zajonz et al. has proposed a modified, heavily relaxed model (From ref. [124], with permission), (d) The glancing angle x-ray diffraction data that have lead to the original microfacet model have been re-evaluated by Landree et al. [125] and a new structure has been proposed. The octahedra model schematic representations in (d) show of the microfacet model (top) and the new model (bottom) (From ref [125], with permission).

In Fig. 15a, O atoms would move to the right and Ti atoms to the left. The net effect of these displacement is to increase the effective coordination of the 5-fold coordinated Ti atoms [109]. To my knowledge, no experimental data on relaxations of the TiO2(100)(lxl) surface exist. X-ray photoelectron and Auger electron diffraction were performed but are insensitive to the details of the surface structure [111].

472

4.2. Reconstructions 4.2.1. The microfacet model of the rutile TiO2(100)(lx3) surface In addition to the (1x1) terminated surface, a (1x3) reconstructed surface forms relatively easily upon annealing to high temperatures in UHV. It is partially reduced as is clearly evident from photoemission experiments [112115]. Initially, the observed reconstruction was interpreted as removal of every third row of the outermost, 'bridging' oxygen atoms [115]. Such a proposal was very much in line with the initially proposed [116] (and now largely abandoned) 'missing row' model for the (1x2) structure of the TiO2(110) surface (see Fig. 11a). The step structure on this surface may lead to a misinterpretation as additional (1x5) and (1x7) reconstructions as pointed out by Muryn et al. [115]. The first STM work on a reconstructed TiO2(100) surface was reported by Clark and Kesmodel [117]. A glancing angle X-ray diffraction and low energy electron diffraction study [118] suggested a 'microfaceted' model, see Fig. 15 b. Removal of the volume assigned with y creates facets of the lowest-energy, (110) plane. Note that, again, the same number of O -> Ti as Ti -> O bonds are broken. This results in a stoichiometric surface, with 2-fold coordinated bridging oxygen atoms at the outermost ridges, as is the case for the TiO2(110) surface. The unit cell in [010] direction is 3 times wider. Atomically-resolved STM images showed bright ridges with a geometry consistent with the microfaceted model [119,120]. The apparent height between top and bottom of the reconstruction was measured as 3 A with STM, instead of the expected 5A. This has been associated with tip effects [119,120]. Scanning tunneling spectroscopy (STS) measurements showed considerable difference between dl/dV spectra taken on and in between the bright ridges, respectively [119,121].. This was interpreted as missing oxygen atoms on the outermost ridges, (i.e., removal of the oxygen atoms labeled y in Fig. 15 c). This results in a surface termination with 3-fold coordinated Ti atoms in the outermost plane, consistent with the observation of a reduced surface in photoemission [115]. A photoelectron diffraction study of the Tiip level was performed by Hardman et al [114]. The PED curves of the (curve fitted) Ti3+ feature were evaluated. These Ti^"^ features are supposed to be come from Ti atoms located next to oxygen vacancies, and diffraction effects should give information about their surface geometry. Three different positions for vacancies on the microfaceted (1x3) surface were tested. The missing oxygen atoms at the outermost ridge of the (110) facets was determined as the most likely configuration [114]. In this context it is also interesting to note that the reduction state of the substrate plays a role in the formation of the (1x3) reconstruction. Almost stoichiometric, Nb-doped TiO2(100) films are thermally much more stable than reduced Ti02 substrates and do not reconstruct at temperatures where reduced Ti02 substrates already show a clear (1x3) structure [122].

473

A model involving 'discrete bond breaking' was proposed for the formation of the (1x3) surface [107,108]. Surfaces were prepared that exhibited both, the (1x1) termination and the ridges typical for the (1x3) surface. STM and non-contact AFM images showed an intermediate phase, which had a (1x3) symmetry, but did not possess the characteristics of the micro faceted structure. Raza et al. [107] suggested that the bonds labeled a in Fig. 15 are broken, which allows the Ti atoms to relax towards the other rows of bridging oxygen atoms. 4.2.2. Is the simple microfacet model valid? Despite the experimental results discussed in the previous section, it is presently not clear if the microfacet model for the (1x3) surface (Fig. 15b) is valid. In fact, recent evidence indicates that it may be a oversimplification. Theoretical calculations are not in agreement with the microfacet model. A tight-binding calculation derives a higher surface energy for the relaxed, micro faceted surface as compared to the (1x1) terminated one [123]. DFT ab initio calculation [110] also showed a considerably higher surface free energy. This implies that there would be no driving force for the reconstruction. These surprising results were attributed to the fact that stoichiometric surfaces has been considered in the calculations (i.e., the oxygen atoms y. Fig. 15c, were not removed). Experimental spectroscopic data show that the (1x3) surface is partially reduced. As was pointed out by Lindan et al., the Ti3+ state associated with the reduction should correctly be treated with spin-polarized DFT calculations [110]. A recent grazing incidence x-ray diffraction (GIXD) analysis shows strong lateral and vertical relaxations (1 A and more!) of the titanium and oxygen atoms in the top layer [124]. The coordination of the surface Ti atoms differ considerably from the simple microfacet model, see Fig. 15c. Threefold coordinated Ti atoms were found at the facet ridges, in agreement with previous work. In addition, various Ti atoms were found in different configurations, e.g., Ti in an oxygen bridge site, titanium B in a trigonal prismatic coordination, as well as an interstitial Ti site A. Obviously, this model deviates strongly from the simple microfaceted structure. The original GIXD data, which were the basis for the microfaceted model, were re-evaluated recently by Landree, Marks, Zschack, and Gilmore [125]. In ref [118] the data were interpreted using Patterson functions which show only interatomic vectors, not the 'true' atomic positions. The reevaluation was based on a technique known as 'direct method', in essence a Fourier transform of the measured data in connection with a search for possible (unknown) phases [126]. It was found that the microfacet model gave very poor agreement which the data as compared to a model that contained four Ti and six to eight O atoms in the surface unit cell. In this model, the Ti atoms reside in edge- and corner-sharing octahedral units, as opposed to the normal rutile structure which is composed of comer-sharing octahedra only

474

[001]

- • [010] - • [010] [010]

Fig.l6. Surface termination of the rutile TiO2(001) surface. Only one possibility exists to cut a Ti02 crystal in this direction, see the side view at the left side. Surface Ti atoms are 4fold coordinated and surface O atoms 2-fold coordinated.

(see Fig. 15d). The reconstructed structure is rationalized as a standard configuration for non-stoichiometric defects such as CS planes, see Fig. 10. The reduced states observed with spectroscopic measurements would then be accommodated similar as in the bulk. The surface unit cell of the TiO2(100)(lx3) structure is quite large with many atoms, hence it is no surprise that it is hard to determine conclusively the exact surface geometry. While the electron diffraction [114], STM and AFM results [107,108,119-121], together with ESDIAD measurements [127] are consistent with the microfacet model of the (1x3) surface, neither of these techniques gives direct evidence of surface geometries. The (1x1) and (1x3) reconstructions provide a convenient system for site-sensitive surface chemistry experiments, because one can reversibly cycle a TiO2(100) crystal between the (1x1) surface and the reconstructed one [128,129]. Hence, it would be quite important to have additional experimental as well as theoretical support for one of the models currently suggested and depicted in Fig. 15.

5. RUTILE (001) There is only one way to cut a rutile crystal in (001) direction (Fig. 16.) Although this creates a non-polar, autocompensated surface, it does not represent a low-energy configuration. This becomes clear immediately when reviewing the coordination of the surface atoms. All the Ti atoms are 4-fold coordinated, and all the O atoms 2-fold coordinated. Hence the number of broken bonds on this surface is higher than on the other low-index rutile surfaces discussed so far. Consequently, the (001) surface has a high surface

475

Fig. 17. A TiO2(001) surface after annealing at a very high temperature (1300°C, 1 h in UHV). The image has been taken with an optical microscope. It shows lines due to slip along certain crystallographic directions. From ref [133], with permission.

energy and tends to facet. Based on LEED studies, (Oil) and (114) facets have been identified by Tait and Kasowski [130] and Firment [131]. Several additional crystal planes were identified in an STM/LEED study by Porier et al. [132]. . The image in Fig. 17 shows the drastic changes that can occur upon very high-temperature annealing. The micrograph has been taken with an optical microscope after heating a TiO2(001) crystal for Ih at 1300°C in UHV [133]. The lines are caused by slip. A non-equilibrium structure was observed after rapidly quenching a TiO2(001) crystal from a similar high-temperature anneal [134]. 6. VICINAL SURFACES Vicinal surfaces of TiOi have not been studied extensively. A study of Na adsorption on a stepped (441) surface was reported by Onishi et al [135]. Unpublished experiments in my own laboratory with a similarly cut crystal showed macroscopic faceting upon annealing. The most detailed investigation was performed recently on a TiO2(210) surface [136]. In a formal sense, TiO2(210) lies midways between (110) and (100), and is the most simple vicinal surface. Atomistic simulations, based on Coulombic interaction between ions and a short-range repulsive interaction, predicted an asymmetric sawtooth-like structure of the surface, consisting of {110} nanofacets. The width of each nanofacet is 1.5 times the width of the surface unit cell of the (110)(lxl) structure (i.e., 3a/V2). The nanofacets terminate with a row of Ti

476

atoms carrying bridging oxygen atoms. The surface energy of this structure is predicted to be 2.07 J/m^. (This is to be compared by a surface energy of 1.78 J/m2 for (110) derived using a similar calculation [136].) STM images showed a (1x1) terminated surface that could be consistent with this structure, although the interpretation was again made difficult by balancing electronic effects with the very strong corrugations of this surface.

7. ANATASE SURFACES Commercial titania powder catalysts are a mixture of rutile and anatase (e.g., the most often-used Degussa P-25 contains approx. 80 to 90% anatase and 10 to 15% rutile [137]), and, for reasons that are not completely understood, anatase is the photocatalytically more active form. It also shows enhanced activity in (non-photoinduced) catalysis, and behaves different than rutile in gas sensing devices. High-purity titania powder catalysts are typically made in a flame process from titanium tetrachloride [137]. Many additional synthetic techniques processes have been applied [138-142]. The shapes of the crystallites vary with preparation techniques and procedures. Typically, the (101) and the (100)/(010) surface planes are found, together with some (001) [6]. Several theoretical studies have predicted the stability of the different low-index anatase surfaces [6,143,144]. The (101) face is the thermodynamically most stable surface. Calculated surface energies are 0.52 and 0.81 J/m^ for the (relaxed) anatase (101) and (001) surfaces. Compared with the (also GGA-calculated) value for rutile (110) surface [145] it appears that the anatase (101) surface is even the most stable surface of all the Ti02 polymorphs! Experimental investigations on single-crystalline anatase, are just starting. Meaningful surface science investigations necessitate singlecrystalline samples. While rutile crystals are readily available, sufficiently large and pure anatase crystals are more difficult to obtain. Because anatase is a metastable phase, it transforms into rutile at relatively low temperatures [139], with the transition temperature dependent on impurities, crystal size, sample history, etc. However the recent progress in synthesizing singlecrystalline anatase samples with high purity [138] as well as the successful growth of epitaxial thin-films on appropriate substrates [7,146,147] has rendered the first surface science investigations of this material. 7.1. Anatase (101) Two reports on the structure of anatase (101) surfaces have appeared very recently [7,8]. The (101) surface on an anatase sample grown by chemical transport [138] showed a (1x1) surface after mild sputtering and annealing [8]. A mineral sample was used in an STM study by Hebenstreit et al. [7]. In order to avoid the contaminations in this natural single crystal, a 700 A thick, epitaxial film was grown on the surface. Sputtering and annealing again

477

film was grown on the surface. Sputtering and annealing again produced a (1x1) termination in LEED. The surface has only pm symmetry, giving rise to a preferential orientation of step edges (Fig. 18a). Based on the rules of autocompensation for step edges a detailed model for steps is given in Fig. 18 [7]. Titanium atoms at the terraces they have five-fold and six-fold coordination, and titanium atoms at the step edges are four-fold coordinated. These have a higher reactivity against gas adsorption [7]. Two-fold coordinated oxygen atoms are located at the ridges of the saw-tooth like structure. According to ref. [6], these atoms relax. The tunneling site in the atomic-resolution images probably extends across both, the two-fold coordinated oxygen atoms and the five-fold coordinated Ti atoms [7]. In an image taken with a higher tunneling current (12 nA), where the tip was probably closer to the surface, the 2-fold coordinated oxygen atoms are distinguished as single features [7]. In correspondence to the rutile (110) surface, one might expect that the twofold coordinated oxygen atoms are removed easily upon annealing in UHV, giving rise to point defects. Several imperfections with atomic dimensions are identified in the atomic-resolution images [7]. At this point it is not clear which one of these features, if any, are indicative of oxygen vacancies. Their number density is very small, certainly smaller than the usually quoted 5 - 10 % [53,64] on rutile (110). This would be in agreement with the (calculated) low surface energy of the (101) surface [6]. Calculations of the electrostatic potential by Woning and van Santen also predict that the rutile (110) surface can be reduced easier than anatase [148]. 2-fold^

5-fold Ti

r

Fig. 18 (a) STM results of an anatase (101) single crystal. The monoatomic terraces terminate with step edges that run predominantly in certain preferred orientations, (b) Atomic models (side and top view) of the anatase (101) surface. From ref [7], with permission..

478

7.2. Anatase (001) As shown in Fig. 19, the stable, autocompensated anatase (001) surface exhibits exclusively 5-fold coordinated Ti atoms, as well as two-fold and threefold coordinated oxygen atoms. Calculations show that the corrugation increases somewhat upon relaxation, from 0.82 A to 0.92 A [6]. The most detailed structural investigations on this surface so far have been taken on thin films, grown epitaxially on the SrTiO3(001) substrate [146,147]. The SrTiO3(001) surface shows a very good lattice match with the anatase (001) surface, but a poor one with the rutile phase. Their anionic sublattices bear substantial resemblance despite their overall crystallographic dissimilarities [149] As Ti02 is deposited, heteroepitaxial growth can be regarded as a continuous formation and extension of the oxygen atom network from the substrate into the film. Within this oxygen sublattice the (relatively small) Ti cations become arranged in their appropriate sites. Formation of interfaces where the oxygen sublattice continues and the metal cation sublattice changes abruptly is often exploited for thin film heteroepitaxy of metal oxides [150]. On SrTiO3(001) epitaxial, stoichiometric anatase thin films of high crystalline quality have been grown by several techniques [149,151]. Recent, unpublished results show that MOCVD-grown films are stable up to 1000°C [152]. During the growth, the anatase (001) films are (1x1) terminated. The (1x1) termination has been confirmed ex-situ with LEED and XPD by Herman et al. [146]. The surface was only outgassed at 100°C and not sputtered. Hence, a carbon coverage of 0.6 monolayers was estimated based on XPS measurements. The surface was well ordered and the structure in XPD was fully consistent with the bulk-like termination depicted in Fig. 19. A twodomain (1x4) reconstruction formed on a similar sample after sputtering and annealing the (1x1) surface in UHV [147]. Based on angle-resolved mass5C-Ti

2C-0

(001) ,[010] • [100]

^ vv.^., V . - • /

I

^ ^^ QQj

anatase (001) (1 x4), microfacet model

Fig. 19 The surface structure of teh anatase (001)(lxl) surface and the proposed model for the (1x4) reconstruction found after annealing a clean anatase (001) surface in UHV (after ref [147], with permission).

479

spectroscopy of recoiled ions (AR-MRSI) a 'microfaceted' model was proposed. In this model (103) facets are exposed which contain two-fold oxygen and both four-fold and five-fold coordinated Ti atoms, see Fig. 19. Such a model resembles in many ways the microfacet model for the rutile (100) (1x3) surface discussed in section 4.2. It remains to be seen if this model will be confirmed by additional experimental and theoretical work. There is no reason to assume that epitaxial anatase films, provided they are thick enough, would have a different surface structure than an single crystals. In fact, LEED studies on single-crystalline anatase samples confirm the results by Herman et al. A (1x1) termination was observed after sputtering and annealing in oxygen at 400°C an iron contaminated mineral anatase sample [153 ]. A single crystal grown with chemical transport [138] showed a (1x1) termination initially and a (1x4) reconstruction after sputtering and annealing to 900K [154]. 8. CONCLUSION Much has been learned about the surface structure of the titanium dioxide system in recent years. Two somewhat contradictory lessons can be drawn from all this work: simple approaches work well for obtaining a first guess of surface structures and oxide surfaces are even more complicated than anticipated. It is comforting that the elemental rules for predicting surface terminations outlined in section 3.1.1. work so well for predicting the structure of the (1x1) terraces and the step edges of all the orientations of both rutile and anatase. The extensive theoretical work has helped to refine the understanding of surface relaxations, and the level of detail on the atomic geometry of the TiO2(110) surface is certainly comparable to that of certain elemental semiconductors or metals. On the other hand, scanning probes techniques have unraveled a very rich picture of surface structures. One interesting theme is the interplay between surface structure and bulk defects; the reduction state of the crystal is quite important for the presence of different structural features under exactly the same preparation conditions. It remains to be seen if such a behavior is also present on other metal oxides. Total-energy calculations have helped enormously to confirm models for surface reconstructions, and have acted as a warning sign when models derived from experimental results were too naive. It appears that, again, bulk structures should act as a guideline for devising models for new surface reconstructions. The expanding data base has made rutile the model system for metal oxides. Nevertheless, there are still many open questions concerning the crystal structure of rutile surfaces as pointed out throughout this Chapter. One interesting aspect is the advent of surface studies on anatase. From the recent

480

progress in synthesizing single-crystalline anatase samples with high purity as well as the successful growth of epitaxial thin-films on appropriate substrates, I would expect a rapid increase in the interest in anatase in the near future.

REFERENCES 1. M. A. Barteau, J. Vac. Sci. Technolog. A 11 (1993) 2162 . 2. N. M. Harrison, X. G. Wang, J. Muscat and M. Scheffler, Faraday Discuss 114 (1999) 305. 3. G. V. Samsonov, The Oxide Handbook (IFI/Plenum, New York, 1982). 4. U. Diebold, in: Specimen Handling, Treatments, Beam Effects and Depth Profiling, edited by C. A. Czanderna, C. J. Powell and T. E. Madey 1999) 5. M. Ramamoorthy and D. Vanderbilt, Phys. Rev. B 49 (1994) 16721. 6. A. Vittadini, A. Selloni, F. P. Rotzinger and M. Gratzel, Phys. Rev. Lett. 81 (1998) 2954 . 7. W. Hebenstreit, N. Ruzycki, G. S. Herman, Y. Gao and U. Diebold, Phys. Rev. B (2000, submitted) 8. R. Hengerer, B. Bolliger, M. Erbudak and M. Gratzel, Surf. Sci. 460 (2000) 162. 9. M. Li, W. Hebenstreit, U. Diebold, A. M. Tyryshkin, M. K. Bowman, Glen G. Dunham and M. A. Henderson, J. Phys. Chem. B 104 (2000) 4944. 10. U. Diebold, M. Li, O. Dulub, E. L. D. Hebenstreit and W. Hebenstreit, Surf. Rev. Lett, in press (2000) 11. F. Pesty, H. P. Steinruck and T. E. Madey, Surf Sci. 339 (1995) 83. 12. K. D. Schierbaum, S. Fischer, M. C. Torquemada, J. L. d. Segovia, E. Roman and J. A. Martin-Gato, Surf Sci. 345 (1996) 261. 13. O. Dulub, W. Hebenstreit and U. Diebold, Phys. Rev. Lett. 84 (2000) 3646. 14. M. A. Henderson, Surf Sci. 343 (1995) LI 156. 15. M. A. Henderson, Surf Sci. 419 (1999) 174. 16. R. A. Bennett, P. Stone, N. J. Price and M. Bowker, Phys. Rev. Lett. 82 (1999) 3831. 17. E. L. D. Hebenstreit, W. Hebenstreit and U. Diebold, Surf Sci. in press (2000) 18. P. Kofstad, J. Less-Common Metals 13 (1967) 635. 19. E. Yagi, R. Hasiguti and M. Aono, Phys. Rev. B 54 (1996) 7945. 20. D. J. Smith, L. A. Bursill and M. G. Blanchin, Philisophical Magazine A 50 (1984) 473. 21. L. A. Bursill and D. J. Smith, Nature 309 (1984) 319. 22. L. A. Bursill, M. G. Blanchin and D. J. Smith, Proc. R. Soc. London,[Ser.] A 391 (1984) 351 and 373. 23. G. S. Rohrer, V. E. Henrich and D. A. Bonnell, Science 250 (1990) 1239. 24. H. Norenberg, R. E. Tanner, K. D. Schierbaum, S. Fischer and G. A. D. Briggs, Surf Sci. 396(1998)52. 25. H. Norenberg and G. A. D. Briggs, Surf Sci. 404 (1998) 738. 26. R. A. Bennett, S. Poulston, P. Stone and M. Bowker, Phys. Rev. B 59 (1999) 10341. 27. P. W. Murray, N. G. Condon and G. Thornton, Phys. Rev. B 51 (1995) 10989. 28. J. Sasaki, N. L. Peterson and K. Hoshino, J. Phys. Chem. Solids 46 (1985) 1267. 29. H. B. Huntington and G. A. Sullivan, Phys. Rev. Lett. 14 (1965) 177. 30. P. W. Tasker, J. Phys. C: Solid State Phys. 12 (1979) 4977. 31. J. P. LaFemina, Crit. Rev.Surf Chem. 3 (1994) 297.

481 32. R. Wang, K. Hashimoto, A. Fujishima, M. Chikuni, E. Kojima, A. Kitamura, M. Shimohigoshi and T. Watanabe, Nature 388 (1997) 431 . 33. Y. Gao and S. A. Chambers, Mater. Res. Soc. Symp. Proc. 401 (1996) 85. 34. G. Charlton, P. B. Hoowes, C. L. Nicklin, P. Steadman, J. S. G. Taylor, C. A. Muryn, S. P. Harte, J. Mercer, R. McGrath, D. Norman, T. S. Turner and G. Thornton, Phys. Rev. Lett. 78(1997)495. 35. B. Hird and R. A. Armstrong, Surf. Sci. 420 (1999) L131. 36. B. Hird and R. A. Armstrong, Surf. Sci. 385 (1997) L1023. 37. E. Asari, T. Suzuki and R. Souda, Phys. Rev. 61 (2000) 5679. 38. A. Verdini, M. Sambi, F. Bruno, D. Cvetko, M. D. Negra, R. Gotter, L. Floreano, A. Morgante, G. A. Rizzi and G. Granozzi, Surf Rev. Lett. 6 (1999) 2101. 39. D. Vogtenhuber, R. Podloucky, A. Neckel, S. G. Steinemann and A. J. Freeman, Phys. Rev. B 49 (1994) 2099. 40. P. Reinhardt and B. A. HeB, Phys. Rev. B 50 (1994) 12015 . 41. P. J. D. Lindan, N. M. Harrison, M. J. Gillan and J. A. White, Phys. Rev. B 55 (1997) 15919. 42. S. P. Bates, G. Kresse and M. J. Gillan, Surf Sci. 409 (1998) 336 . 43. D. R. Hamann, Phys. Rev. B 56 (1997) 14979 . 44. K. M. Glassford and J. R. Chelikowsky, Phys. Rev. B 46 (1992) 1284. 45. H. Onishi and Y. Iwasawa, Chem. Phys. Lett. 226 (1994) 111. 46. U. Diebold, W. Hebenstreit, G. Leonardelli, M. Schmid and P. Varga, Phys. Rev. Lett. 81 (1998)405. 47. U. Diebold, J. F. Anderson, K. O. Ng and D. Vanderbih, Phys. Rev. Lett. 77 (1996) 1322. 48. P. J. Moller and M.-C. Wu, Surf Sci. 224 (1989) 265 . 49. O. Gtilseren, R. James and D. W. Bullett, Surf Sci. 377-379 (1997) 150. 50. R. E. Tanner, M. R. Castell and G. A. D. Briggs, Surf Sci. 412 (1998) 672. 51. A. Szabo and T. Engel, Surf Sci. 329 (1995) 241. 52. Q. Guo, I. Cocks and E. M. Williams, J. Phys. D: Appl. Phys. 31 (1998) 2231. 53. U. Diebold, J. Lehman, T. Mahmoud, M. Kuhn, G. Leonardelli, W. Hebenstreit, M. Schmid and P. Varga, Surf Sci. 411 (1998) 137. 54. K.-i. Fukui, H. Onishi and Y. Iwasawa, Phys. Rev. Lett. 79 (1997) 4202. 55. T. R. Albrecht, P. Grutter, D. Home and D. Rugar, J. Appl. Phys. 69 (1991) 668. 56. M. Reichling and C. Barth, Phys. Rev. Lett. 83 (1999) 768. 57. S. Fischer, A. W. Munz, K. D. Schierbaum and W. Goepel, Surf Sci. 337 (1995) 17. 58. B. Grossmann and P. Piercy, Phys. Rev. Lett. 74 (1995) 4487. 59. H. Onishi, K.-c. Fukui and Y. Iwasawa, Bull. Chem. Soc. Jpn. 68 (1995) 2447. 60. H. Onishi and Y. Iwasawa, Surf Sci. 313 (1994) L783. 61. S. Suzuki, Y. Yamaguchi, H. Onishi, K.-i. Fukui, T. Sasaki and Y. Iwasawa, Catal. Lett. 50(1998)117. 62. Y. Iwasawa, H. Onishi, K. Fukui, S. Suzuki and T. Sasaki, Faraday Discuss. 114 (1999) 259. 63. J. M. Pan, B. L. Maschhoff, U. Diebold and T. E. Madey, J. Vac. Sci. Technol., A 10 (1992) 2470. 64. M. A. Henderson, Surf Sci. 355 (1996) 151. 65. W. Gopel, G. Rocker and R. Feierabend, Phys. Rev. B 28 (1983) 3427. 66. L. Q. Wang, D. R. Baer and M. H. Engelhard, Surf Sci. 320 (1994) 295. 67. D. Vogtenhuber, (private communication)

482 68. S. Suzuki, K. Fukui, H. Onishi and Y. Iwasawa, Phys. Rev. Lett. 84 (2000) 2156 . 69. M. L. Knotek and P. J. Feibelman, Surf. Sci. 90 (1979) 78. 70. U. Diebold and T. E. Madey, Phys. Rev. Lett. 72 (1994) 1116. 71. E. Bertel, R. Stockbauer and T. E. Madey, Surf. Sci. 141 (1984) 355. 72. L. Q. Wang, D. R. Baer, M. H. Engelhard and A. N. Shultz, Surf Sci. 344 (1995) 237. 73. S. A. Joyce, (private communication) 74. A. Berko and E. Krivan, J. Vac. Sci. Technol., B 15 (1997) 25. 75. A. N. Shultz, W. Jang, W. M. I. Hetherington, D. R. Baer, L. Q. Wang and M. H. Engelhard, Surf Sci. 339 (1995) 114. 76. T. E. Madey and B. L. Maschhoff, (unpubUshed results) 77. M. R. McCartney and D. J. Smith, Surf. Sci. 250 (1991) 169. 78. L. P. Zhang, M. Li and U. Diebold, Surf Sci. 412 (1998) 242. 79. H. Norenberg and J. H. Harding, Appl. Surf Sci. 142 (1999) 174. 80. H. Norenberg and J. H. Harding, Phys. Rev. B 59 (1999) 9842. 81. J. F. Anderson, M. Kuhn, U. Diebold, K. Shaw, P. Stoyanov and D. Lind, Phys. Rev. B 56(1997)9902. 82. M. A. Henderson, S. Otero-Tapia and M. E. Castro, Surf. Sci. 412 (1998) 252. 83. S. Gan, Y. Liang and D. R. baer. Surf Sci. 459 (2000) L498 . 84. G. S. Rohrer, V. E. Henrich and D. A. Bonnell, Surf Sci. 278 (1992) 146. 85. H. Norenberg and G. A. D. Briggs, Surf Sci. 404 (1998) 738. 86. R. A. Bennett, Phys. Chem. Comm. 3 (2000) (web. 87. M. Sander and T. Engel, Surf Sci. 302 (1994) L263. 88. M. Wagner, O. Kienzle, D. A. Bonnell and M. Ruble, J. Vac. Sci. Technol., A 16 (1998) 1078. 89. P. W. Murray, N. G. Condon and G. Thornton, Surf Sci. 323 (1995) L281. 90. A. Berko and F. Solymosi, Langmuir 12 (1996) 1257. 91. C. Xu, X. Lai, G. W. Zajac and D. W. Goodman, Phys. Rev. B 56 (1997) 13464. 92. C. L. Pang, S. A. Haycock, H. Raza, P. W. Murray, G. Thornton, O. Gulseren, R. James and D. W. Bullett, Phys. Rev. B 58 (1998) 1586. 93. E. Asari and R. Souda, Phys. Rev. B 60 (1999) 10719. 94. K. O. Ng and D. Vanderbilt, Phys. Rev. B 56 (1997) 10544. 95. Q. Guo, I. Cocks and E. M. Williams, Phys. Rev. Lett. 77 (1996) 3851. 96. R. E. Tanner, M. R. Castell and G. A. D. Briggs, Surf Sci. 437 (1999) 263. 97. C. L. Pang, S. A. Haycock, H. Raza, G. Thornton, O. Gulseren, R. James and D. W. Bullet, Surf Sci. 437 (1999) 261. 98. M. Li, W. Hebenstreit and U. Diebold, Phys. Rev. B 61 (2000) 4926. 99. H. Onishi and Y. Iwasawa, Phys. Rev. Lett. 76 (1996) 791. 100. M. Li, W. Hebenstreit and U. Diebold, Surf Sci. 414 (1998) L951. 101. M. Li, W. Hebenstreit, U. Diebold, M. A. Henderson and D. R. Jennison, Faraday Discuss 114 (1999) 245. 102. M. Li, W. Hebenstreit, L. Gross, U. Diebold, M. A. Henderson, D. R. Jennison, P. A. Schultz and M. P. Sears, Surf Sci. 437 (1999) 173. 103. R. A. Bennett, P. Stone and M. Bowker, Faraday Discuss 114 (1999) 267. 104. P. Stone, R. A. Bennett and M. Bowker, New J. Phys. 1 (1999) 8. 105. M. A. Henderson, Langmuir 12 (1996) 5093. 106. M. B. Hugenschmidt, L. Gamble and C. T. Campbell, Surf Sci. 302 (1994) 329. 107. H. Raza, C. L. Pang, S. A. Haycock and G. Thornton, Phys. Rev. Lett. 82 (1999) 5265 . 108. H. Raza, C. L. Pang, S. A. Haycock and G. Thornton, Appl. Surf Sci. 140 (1999) 271 .

483

109. J. Muscat, N. M. Harrison and G. Thorton, Phys. Rev. B 59 (1999) 2320. 110. P. J. D. Lindan, N. M. Harrison, J. M. Holender, M. J. Gillan and M. C. Payne, Surf. Sci. 364(1996)431. 111. P. J. Hardman and P. L. Wincott, Phys. Rev. B 60 (1999) 11700. 112. W. J. Lo, Y. W. Chung and G. A. Somorjai, Surf. Sci. 71 (1978) 199. 113. Y. W. Chung, W. J. Lo and G. A. Somorjai, Surf. Sci. 64 (1977) 588. 114. P. J. Hardman, N. S. Prakash, C. A. Muryn, G. N. Raikar, A. G. Thomas, R. J. Blake and C. A. Muryn, Phys. Rev. B 47 (1993) 16056. 115. Muryn, (1991) 116. M. C. Wu and P. J. Moeller, Surf. Sci. 224 (1989) 250. 117. G. W. Clark and L. L. Kesmodel, Ultramicroscopy 41 (1992) 77. 118. P. Zschack, J. B. Cohen and Y. W. Chung, Surf Sci. 262 (1992) 395. 119. P. W. Murray, F. M. Leibsle, H. J. Fisher, C. F. J. Flipse, C. A. Muryn and G. Thornton, Phys.Rev. Lett. 46 (1992) 12877. 120. P. W. Murray, F. M. Leibsle, C. A. Muryn, H. J. Fisher, C. F. J. Flipse and G. Thornton, Phys. Rev. Lett. 72 (1994) 689. 121. P. W. Murray, F. M. Leibsle, C. A. Muryn, C. F. J. Flipse and G. Thornton, Surf. Sci. 321(1994)217. 122. Y. Gao, Y. Liang and S. A. Chambers, Surf. Sci. 365 (1996) 638. 123. P. M. Oliver, S. C. Parker, J. Purton and D. W. Bullett, Surf. Sci. (1994) 124. H. Zajonz, H. L. Meyerheim, T. Gloege, W. Moritz and D. Wolf, Surf. Sci. 398 (1998) 369. 125. E. Landree, L. D. Marks, P. Zschack and C. J. Gilmore, Surf. Sci. 408 (1998) 300. 126. L. D. Marks, R. Plass and D. L. Dorset, Surf. Rev. Lett. 4 (1997) 1. 127. Q. Guo, I. Cocks and E. M. Williams, Surf. Sci. 366 (1996) 99. 128. M. A. Henderson, The journal of physical chemistry 99 (1995) 15253. 129. M. A. Henderson, Surf. Sci. 319 (1994) 315. 130. R. H. Tait and R. V. Kasowski, Phys. Rev. B 20 (1979) 5178. 131. L. E. Firment, Surf. Sci. Phys. Rev. B (1982) 5178. 132. G. E. Poirier, B. K. Hance and M. White, J. Vac. Sci. Technol. B 10 (1992) 6 . 133. H. Norenberg, F. Dinelli and G. A. D. Briggs, Surf. Sci. 446 (2000) L83. 134. H. Norenberg, F. DineUi and G. A. D. Briggs, Surf. Sci. 436 (1999) L635. 135. H. Onishi, T. Aruga, C. Egawa and Y. Iwasawa, Surf. Sci. 199 (1988) 54. 136. A. Howard, C. E. J. Mittchell, D. Morris, R. G. Egdell and S. C. Parker, Surf. Sci. 448 (2000)131 . 137. C. N. Satterfield, Heterogeneous Catalysis in Industrial Practice (McGraw-Hill, Inc., New York, 1991). 138. L. Kavan, M. Gratzel, S. E. Gilbert, C. Klemenz and H. J. Scheel, J. Am. Chem. Soc. 118(1996)6716. 139. J. M. G. Amores, V. S. Escribano and G. Busca, J. Mater. Chem. 5 (1995) 1245 . 140. C. Byun, J. W. Jang, I. T. Kim, K. S. Hong and B.-W. Lee, Mat. Res. Bulletin 32 (1997) 431. 141. P. Arnal, R. J. P. Corriu, D. Leclercq, P. H. Mutin and A. Vioux, J. Mater. Chem. 6 (1996) 1925. 142. T. Oyama, Y. limura, K. Takeuchi and T. Ishii, J. Mater. Sci. Lett. 15 (1996) 594 . 143. A. Fahmi and C. Minot, Surf. Sci. 304 (1994) 343 . 144. T. Bredow and K. Jug, Surf. Sci. 327 (1995) 398 .

484 145. J. Goniakowski, J. M. Holender, L. N. Kantorovich and M. J. Gillan, Phys. Rev. B 53 (1996)957. 146. G. S. Herman, Y. Gao, T. T. Iran and J. Osterwalder, Surf. Sci. 447 (1999) 201. 147. G. S. Herman, M. R. Sievers and Y. Gao, Phys. Rev. Lett. 84 (2000) 3354. 148. J. Woning and R. A. V. Santen, Chem. Phys. Lett. 101 (1983) 541. 149. S. Chen, M. G. Mason, H. J. Gysling, G. R. Paz-Pajult, T. N. Blanton, K. M. Chen, C. P. Fictorie, W. L. Gladfelter, A. Franciosi, P. I. Cohen and J. F. Evans, J. Vac. Sci. Technolog.All (1993)2419. 150. D. S. Lind, S. D. Berry, G. Chern, H. Mathias and L. R. Testardi, Phys. Rev. B 45 (1992) 1838. 151. W. Sugimura, A. Yamazaki, H. Shigetani, J. Tanaka and T. Mitsuhashi, Jpn. J. Appl. Phys. 36 (1997) 7358 . 152. Y. Gao and G. S. Herman, (private communication) 153. G. Durinck, H. Poelman, P. Clauws, L. Fierman, J. Vennick and G. Dalmai, Solid State Commun. 80 (1991) 579 . 154. R. Hengerer, L. Kavan and M. Gratzel, J. Electrochem. Soc. 147 (2000) 1467. 155. S. P. Bates, G. Kresse and M. J. Gillan, Surf. Sci. 385 (1997) 386.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

485

Chapter 12

The Anisotropy of Metal Oxide Surface Properties G.S. Rohrer Department of Materials Science and Engineering, Carnegie Mellon University, Pittsburgh PA, 15213-3890, U.S.A. 1. INTRODUCTION The surface properties of metal oxides influence the rates of heterogeneous chemical reactions, the growth of heteroepitaxial films, and the sintering of particles during the consolidation of ceramics. Just as certain bulk properties of crystalline materials are anisotropic, surface properties also depend on orientation. The dependence of surface properties on orientation can be rationalized by recognizing that the atoms on crystallographically distinct facets have different coordination environments, as illustrated schematically in Fig. 1. This same figure also illustrates that in binary and more complex materials, surfaces with identical orientations can be terminated by different atomic layers with distinct compositions. Throughout this chapter, we shall take the "character" of a surface to be defined by its orientation, {hkl}, and its atomic termination layer. The relationship between surface properties and surface character allows the macroscopic properties of surface active materials to be manipulated by controlling the distribution of the exposed surfaces. For particulate materials, this distribution is determined by the crystal habit. The character of surfaces exposed by a dense polycrystal can be controlled by the introduction of some crystallographic texture and thin film materials can be oriented through heteroepitaxial seeding. While techniques for structural control are generally available to the material scientist, what is less frequently known is how the physiochemical properties of a material vary with surface character. In this chapter, efforts to measure the surface properties of oxides as a function of character will be described. Section two of this chapter reviews some specific examples of the relationship between surface character and properties that have been derived from studies of particulate oxides, single crystals, and thin films. In section three, the factors that determine surface morphology and the range of achievable orientations are described. The factors that influence surface stoichiometry and the composition of the termination layer are described in

486

• • • •

o •

o •

o

o •

o •

o •

o

o •

o •

o •

o •

o •

o •

o •

o •

o

o

•

o

o

•

•

o

o

o o • o• o • • o • o • o o • o• 0 • • o • o • o o • o• o • • o • o • 0 o • o• o • • o • o • o •

->-/7

•

o#o#o«o« o • o• o •

o • o o• • o o• • o o • •

M O

Fig. 1. Schematic illustration of a two-dimensional metal (M) oxide (O) crystal. Note that atoms on the {10} type faces have different numbers of nearest neighbors than the atoms on the {11} faces. Furthermore, the (11) surface is terminated by M atoms while the (11) surface is terminated by O atoms.

section four. Two specific examples illustrating anisotropic reactivity are reviewed in sections five and six, which deal with partial oxidation reactions on M0O3 and photochemical reactions on Ti02, respectively. 2. SURFACE CHARACTER.PROPERTY RELATIONSHIPS 2.1 Particulate oxidation catalysts One clear piece of evidence for surface character-property relationships lies in the observation that particulate catalysts with the same composition, but different crystal habit, have different reactivities. This phenomenon, sometimes referred to as a catalytic anisotropy or a structure-sensitive catalytic reaction, has been observed in a variety of Mo, V, and W containing materials [1]. Because the catalytic anisotropy of a-Mo03 has been studied in relatively greater depth, this exemplary case is described below [1-10]. The orthorhombic (a) phase of M0O3 has the highly anisotropic layered structure illustrated in Fig. 2 [11], Because of the weak, secondary bonding between the sheets of corner and edge-sharing MoO^ octahedra perpendicular to [010], the crystals are micaceous. The most prominent facet on the growth form is {010}. The crystals are bounded laterally by {001} and {hkO} facets, where the h/k ratio is about 2. In some case, the {001} and {hkO} facets may be replaced by {101} and {100} facets, respectively. The Mo atoms on the (010) surface have their full complement of six oxygen nearest neighbors. However, termination of the bulk structure in other

487

[010]

Fig. 2. The structure of a-MoOs crystals, (a) The structure is built of double octahedral layers stacked along [010]. The spacing between adjacent double layers is 6.9 A. (b) A projection on the (010) plane shows two superimposed rectangular nets of comer-sharing octahedra. The shaded net is below the unshaded one. (c) A typical growth form of aM0O3. All indexing is with respect to space group Fbnm.

directions leaves the surface Mo atoms on lateral facets with fewer than six neighboring O. Investigations of molybdenum trioxide's catalytic anisotropics have been based on the hypothesis that this difference in coordination number will lead to differences in the catalytic properties of the different facets. This hypothesis has been experimentally tested in the following way. Chemically identical particulate samples were prepared in different ways so that they exhibited different characteristic particle habits and aspect ratios. Under the assumption that all of the particles in a given sample have approximately the same shapes, the fractional contribution that each type of facet makes to the total surface area of each specimen was determined by microscopic inspection. The role of each facet in a given reaction can then be inferred by comparing the properties of specimens with different distributions of external surfaces. Based on such studies, it has been concluded that the lateral facets are active for the oxidation of methane to formaldahyde [10] and the oxidative anmionalysis of toluene [7], while the (010) facet is active for the conversion of methanol to formalahyde [1]. Studies of the oxidation of propene to acrolein illustrate that it is not always easy to relate overall activities or selectivities to the presence of a single face [3, 5, 8, 9]. Since the overall reaction is composed of several elementary steps, it is possible that different steps occur on different facets. For example, it has been proposed that the mechanism for the oxidation of propene to acrolein begins with the activation to an allyl intermediate on a lateral facet and ends with the addition of O on a basal facet [5]. The (210) facet, which is thought to consist of terraces with (010) character and ledges with (100) character, should be able to perform both elementary steps. This explanation has been used to rationalize the observation that the (210) surface is especially active for the conversion of propene of acrolein [9]. Using similar

488

experimental techniques, the functionality of specific facets in other layered materials, including vanadium pentoxide [12] and vanadyl pyrophosphate [13], has been identified. While the knowledge gained from these studies is valuable, a number of limitations should be mentioned. First, the conclusions rest mainly on the observed macroscopic shape of the crystals and little is known about the morphology and structure of the individual facets on the particles. Furthermore, the surface morphology of the catalyst is expected to change during reactions, as stoichiometry compensating defects are continuously created and annihilated. Finally, it should be noted that conclusions from studies of particulate materials are not always consistent with observations from other experiments. For example, by comparing the properties of different particulate samples, Tatibouet and Germain [2] identified the (010) facet of M0O3 as being responsible for the conversion of methanol to formaldehyde while Farneth et al. [14] concluded, on the basis of temperature programmed desorption studies, that this facet was largely inactive for the partial oxidation of methanol. We shall see in section 5 that such discrepancies might be explained by morphological changes that occur on the surface of the oxide while in service. 2.2 Single crystal surfaces Perhaps the most convincing demonstrations of structure sensitivity come from highly controlled single crystal experiments in ultrahigh vacuum [15-18]. For example, Vohs and Barteau [15] compared the ability of the ZnO(OOOl) and ZnO(OOOT) to abstract protons from alkynes. While the alkynes (C2H2, CH3CCH, C6H5CCH) decomposed on the (0001) Zn-terminated surface, the (000T) O-terminated surface was observed to be inactive for dehydrogenation. Barteau's [16-18] group conducted parallel studies on Ti02 surfaces in UHV. Flat TiO2(001) orientated crystals can be induced to facet into nonplanar configurations terminated by {101} or {114} facets, depending on the thermal treatment. This phenomenon was exploited to show that when either methanol [16] or acetic acid [17] was reacted with Ti02, the distribution of products was governed by whether the crystal was terminated by {001}, {101}, and {114} surfaces. While the effect of orientation was convoluted with the level of surface oxidation, it was still possible to draw some conclusions about the property-character relationships for these surfaces. For example, the fourcoordinate Ti"^"^ cations found on the {114} faceted surface are capable of coupling carbon containing species to form higher order products (for example, dimethyl ether from methanol and acetone from acetic acid) while the fivecoordinate cations on the {101} faceted surface initiate only dehydration and dehydrogenation reactions. The rationale for this difference is that the lower coordinate cations on the {114} surface are able to adsorb both of the necessary

489

reactants to a single site while the cations on the {101} surface can adsorb only one [18]. 2.3 Thin films As noted briefly in the introductory section of this chapter, specifying only the orientation of a surface can sometimes lead to an ambiguity with respect to the chemical composition of the termination layer. Consider, for example, a crystal with the perovskite structure in the (100) orientation (see Fig. 3). The structure can be terminated either on the AO plane or on the BO2 plane. The composition of the termination layer has been studied most extensively for SrTiOg, which is commonly used as a substrate for the growth of superconducting cuprate phases such as YBa2Cu307. SrTiO3(100) substrates annealed at about 1000 °C exhibit two characteristic types of terraces. Terraces with smoothly curved edges are associated with SrO terminated surface; Ti02 terminated terraces have polygonized edges made of <010> and <001> step segments [19]. It has been demonstrated that it is possible to prepare surfaces that have a uniform termination layer. Ti02-terminated (100) surfaces can be prepared by chemical etching [20, 21] and SrO surfaces can be prepared by depositing a monolayer thick film of SrO [22, 23] or Sr2Ti04 [24]. It has been observed that the SrO and Ti02 surfaces stabilize different polytypes of SrCu02 during thin film deposition [24] and that YBa2Cu307 films are less susceptible to decomposition when deposited on SrO terminated substrates [23]. While SrTi03 is the most well-studied example, the influence of termination layer on film growth has been demonstrated to occur in other systems as well. For example, [0001] oriented sapphire can be terminated by three distinct planes. One is oxygen terminated and the other two are Al terminated. Bench et al. [25] have observed that sapphire (0001) surfaces heated

AO layer

BO2 layer (001)

B Fig. 3. Schematic illustration of the perovskite structure. Note that surfaces perpendicular to (001) can be terminated by and AO or a BO2 layer.

490

^ A

#

»

111 #

». •

*

*

Fig. 4. Scanning electron microscope image of Cu.O deposits on the Al.O.lOOOl) surface. The different particle sizes and morphologies correlate to different surface compositions [25].

for 6 h in air at 1400 °C exhibit a combination of these possible terminations on a sample with a constant average orientation. When CU2O islands were deposited on these surfaces, two distinct island sizes and morphologies were observed that correlated to differences in the terrace termination (see Fig. 4). When Ti02 was deposited on these surfaces, one of the terrace types promoted the growth of an unanticipated (and unidentified) phase that was suspected to be a reduced form of titania or an Al-Ti-0 ternary phase. The results cited above were selected to clearly illustrate the influence that the character of a surface has on its properties. To be more specific, the ability of an oxide surface to catalyze specific reactions or nucleate specific phases is known to depend on both the surface orientation and the termination layer. We now turn to the question of what determines the surface character and what characters are reasonably attainable. 3. FACTORS DETERMINING SURFACE MORPHOLOGY In equilibrium, the morphology of a surface is determined by the anisotropy of the surface energy. While the surface morphologies that we observe in most practical situations are not in global equilibrium, local equilibrium can be achieved at the intersection of two facets or at the point were a grain boundary intersects the surface. Observations of surfaces in local equilibrium give us the

491

opportunity to determine the anisotropy of the surface energy. Conversely, knowledge of the surface energy anisotropy allows us to specify possible morphologies. The two important conclusions from this section are that because of anisotropy, it is usually not possible to realize all surface orientations and that experimental observations can be used to determine the range of allowable orientations. The equilibrium shape of a free particle is determined by the surface energy, which for crystalline solids is usually anisotropic. The microscopic origin of this anisotropy can be understood if one considers that the work required to create a new surface should (ignoring relaxations or reconstruction) increase with the number of pair-wise interatomic bonds that must be broken. Those surfaces with a higher density of broken bonds will have a higher relative energy. Therefore, we can assume the surface energy is a function of the surface normal, Y(n), and for any element of the crystal surface, dA, the surface energy is y(h)dA, where fi is normal to the element dA. For a single crystal with fixed volume in thermodynamic equilibrium, the shape is that which minimizes the total surface energy, E [26]: E = JY(n)dA

(1)

For an isotropic material, the equilibrium shape is a sphere. For an anisotropic material, the equilibrium shape will depart from sphericity so that crystallographic orientations with relatively low values of y will be exposed. Based on Eq. 1, it has been demonstrated that for a crystal in equilibrium, the ratio of the energies of two orientations (y^ and Y2) is equal to the ratio of the perpendicular distances from the center of the crystal to the facet (Z^ and l^) [27]:

h

yi

In other words, the lower the surface energy of the facet, the closer the facet is to the center of the crystal and the greater the surface area of the facet. This is the basis of the Wulff [28] construction, which specifies the relationship between the surface energy of a crystal and its equilibrium shape. To apply the Wulff construction, we begin by imagining that a normal vector represents each distinct crystallographic orientation and the magnitude of each normal vector is equal to the surface energy of the orientation that it represents. Further, we assume that the vectors share a common origin. If a perpendicular plane is situated at the end of each vector, then the inner envelope of the planes defines the equilibrium crystal shape.

492

The Wulff construction is important because it allows the relative energies of different observed surfaces can be determined by direct microscopic measurement of crystals (or voids within crystals) with an equilibrium shape. By measuring the distances from the center of a crystal shape to the various facets (//), the relative surface energies (7/) can be computed using Eq. 2. While this method has been applied to many crystals, applications to oxides have been less frequent [29-31], Most recently, two groups reported results for AI2O3 (see Table 1) [32, 33]. Underpinning such studies is the assumption that the particle (or small cavity) is in equilibrium. To reach this state, it is necessary to have sufficient mass transport and surface attachment kinetics and, in the case of faceted particles, step producing defects that reduce the barrier for the nucleation of new layers [34]. Table 1. Experimentally measured surface energies for AI2O3 Plane (0001) (1011) (1012) (1120) (1123)

1600 °C Ref. 32

Ref. 33

1800 °C Ref. 33

1.0

1.0

1.0

1.07±0.02 1.05±0.02 1.22±0.05 1.06±0.02

0.947±0.016 0.855±0.017 0.947±0.026 0.957±0.026

1.042±0.019 0.950±0.030 1.080±0.017 1.029±0.016

That the data in Table 1 are not in agreement illustrates the difficulty of making such measurements. The constraint of global equilibrium is particularly challenging. An alternative method for experimentally assessing the anisotropy of the surface energy has been developed based on the assumption of local equilibrium between free surfaces and grain boundaries at thermal grooves formed by surface diffusion [35]. The results of this measurement are illustrated in Fig. 5. While the observed anisotropy is not large, the effects on surface morphology are significant. Furthermore, it should be pointed out that there are geometric limits to observable anisotropics for nonperpendicular surfaces in equilibrium experiments. For example, if the energy of the (110) surface of a cubic crystal were more than V2 times the energy of the (100) surface, then the surface would lower its energy by faceting into (100) and (010) surfaces that would meet along lines in the [001] direction and form ridges. Therefore, the energy of the (110) orientation has an upper limit of V2YJQQ. Similarly, if the energy of the (111) surface were more than V3 times the energy of the (100) surface, it would lower its energy by faceting into trigonal pyramids bound by (100), (010), and (001) facets. In fact, at lower temperatures where the anisotropy of the surface energy is expected to be higher, the MgO(lll) surface has been observed to break up into (100) facets [36]. Calculations for MgO

493

1.08 T 1.06 4o o

>=^ 1.04 +

(100)

3 ^

90

120

angle, Fig. 5. Plot of the relative surface energy of MgO in air at 1400 °C, around the perimeter of the unit triangle, from (100) to (111), then to (110), and back to (100).

suggest an anisotropy so high that only {100} facets should be observed [37-40]. While this is consistent with relatively low temperature vacuum experiments, it is not consistent with higher temperature observations in air [35, 41]. In the present context, where our interest is in controlling surface character, it is important to note that surface energy anisotropy frequently leads to missing crystallographic orientations. For example, consider the two-

i (a)

«2 (b)

Fig. 6. (a) A hypothetical two-dimensional equilibrium crystal shape. A continuous range of orientations exists between hi and n^, the complex facet. Between he and %, there is a range of missing orientations (n^ is an example), (b) The surface of a crystal whose orientation is macroscopically constrained to a missing orientation will break up into facets with orientations n^ and n2.

494

dimensional Wulff shape illustrated in Fig. 6. The surfaces of constrained crystals with orientations between fi^ and n^ will be flat, as this will minimize the total surface energy. However, surfaces with orientations between n^ and ^2 will be unstable with respect to faceting and in equilibrium, the surface will be covered with tent-shaped ridges bounded by surfaces with orientations fi^ and ^2- The relative proportion of the two surfaces will depend on how far the orientation is from either of the stable orientations. If it is very close to ^2' then the surface will be composed of large terraces with small steps whose orientations are n^. On the other hand, if the orientation is half way between the two stable surfaces, the surface will be composed of equal parts of the two orientations. The appearance of facets on surfaces with unstable orientations is one conmion manifestation of anisotropy. Surfaces with macroscopic orientations that are missing from the equilibrium shape can be prepared at low temperature by polishing. When heated, such surfaces form microscopic facets while maintainihg their global orientation. The ranges of stable and unstable orientations can be mapped out by examining the surfaces of individual grains in a polycrystalline ceramic that has been polished and then annealed at high temperature. Data for the orientation stability of MgO at 1400 °C in air are shown in Fig. 7. The orientations of more than 100 grain surfaces were determined by electron backscattered diffraction. AFM images of the same surfaces were used to discriminate flat from faceted surfaces (see Fig. 7).

(100)

(110)

Fig. 7. An orientation stability map for magnesia at 1400 °C. Each point corresponds to the orientation of an observed grain. Faceted orientations are marked with an x and smooth orientations are marked with an empty square. An AFM image of a typical faceted surface is shown in the upper left hand part of the figure. The facets are approximately 500 A high.

495

Because of the relatively high energy of the polar (111) surface, this orientation and orientations within about 20 ° decompose into lower energy complex facets [35]. The relative stability of the different orientations is conveniently represented on a stereographic projection, as proposed by Cahn and Handwerker [42]. These orientation stability diagrams are composed according to an analogy with isothermal sections of ternary phase diagrams. Ranges of orientations that are flat (or form part of a continuously curved surface on a particle) are represented as a shaded region and are analogous to a solid solution. Orientations that break up into two separate surfaces are placed on tie lines that connect the two stable facets, in analogy with two-phase regions; orientations that break up into three facets make up empty polygons whose vertices are stable

Fig. 8. (a) Orientation stability diagram for SrTiOg at 1200 °C, in air. There are a range of stable orientations near {100}, The {110} (black circles) and {111} (white triangles) orientations are also stable. Typical AFM images of surfaces from (b) a stable region, (c) a two facet region, and (d) a three facet region are illustrated below. Each shows a 5|xm x 5|im topographic image.

496

orientations. As an example, Fig. 8 illustrates an orientation stability diagram for SrTiOg annealed at 1200 °C in air for 6 h. The observations indicate that the {100}, {110}, and {111} surfaces are stable under these conditions. In addition, there is an extended range of stable surfaces vicinal to {100}. We can therefore say that the stable shape of SrTiOg (after heating for 6 h in air at 1200 °C) has continuously curved surfaces at each of the cube faces which meet flat {110} and {111} surfaces at sharp edges. In practice, it is much easier to derive information on the equilibrium crystal shape from observations of orientation stability than from small crystals or cavities. This is because orientation stability measurements require only local equilibrium at the points where facets meet, whereas global equilibrium is required for equilibrium crystal shape measurements. However, because only a portion of the equilibrium form is observed at any time, the relative energies of different continuously curved regions of the equilibrium crystal shape can not be determined. It is, however, possible to estimate the relative energies of any curved and flat regions of the equilibrium form that intersect [43]. Consider the example discussed above where and orientation h^ breaks up into to other orientations, fi2 and h^. Under the assumption of local equilibrium at the intersection between h2 and n^, then the Herring [44] equation gives the following relation: ^ = cos0-—^sin0 Yc

Yc de

(3)

If the anisotropy is small enough that the differential term can be ignored, then the ratio of the surface energies is simply related to the cosine of the angle at which they meet. Note that one criterion for the application of Eq. 3 is that the facets are in local equilibrium. For this to be true, there must be no net growth or evaporation. If this is true, then thermal grooves at surface-grain boundary intersections should maintain a quasistatic profile and increase in width with the one quarter power of time [45]. 4. FACTORS INFLUENCING SURFACE STOICHIOMETRY The factors influencing the surface composition of pure materials are described in this section. While the factors influencing surface morphology and surface composition are treated in different sections, it should be emphasized that they are closely related. We begin by considering the solid-vapor equilibrium and how this influences surface and bulk stoichiometry. The equilibrium defect structure, like the equilibrium morphological structure, might not always be obtained. Regardless, it remains useful to know the destination state of a surface whose composition is evolving at elevated temperature. The section concludes

497

with a brief consideration of the composition of oxide surfaces produced at low temperature by cleavage. 4.1 The solid-vapor equilibrium We begin by noting that the bulk stoichiometry of an oxide is a function of the temperature and composition of the surrounding atmosphere [46]. For a hypothetical metal oxide (MO), we can write the appropriate defect chemical reactions in Kroger-Vink notation and equilibrium constants for congruent evaporation (k^o)? the production of vacancies by evaporation (k^ and kg), and Shottkey defect formation (kg): MM + Oo = I/2O2 ^,, + M f,,

PM PO2'^ = kMo

(4)

MM = M^^^ + VM*

PM[VM*] = kM

(5)

Oo = 1/20, ,,^ + Vo*

Po2'"[Vo*] = ko

(6)

0 = VM* + VO*

[VM*][Vo*] = ks

(7)

For the purpose of simplicity, the vacancies are taken to be neutral and we do not consider the Frenkel or anti-Frenkel processes. In both cases, extending the analysis does not alter the important parts of the results. From Eqs. 5-7, it can be shown that: [Vo*]=koPo2-^^'

(8)

[VM*] = {^,IK)Po2"

(9)

The equations are represented schematically in Fig. 9. There are several important conclusions to draw from this simple analysis. First, the defect concentration is a function both of temperature (through the equilibrium constants) and the partial pressure of oxygen. Second, there is a single composition of the gas phase where the compound is stoichiometric. Furthermore, the range of accessible compositions is limited by either the appearance of new phases or by experimental limitations. For example, the ranges of stoichiometry available to Fe304 are limited at high and low partial pressures of oxygen by the formation of Fe203 and FeO, respectively. For Ti02, the appearance of the Magneli phases at low partial pressures of O2 limit the maximum concentration of oxygen vacancies at 1000°C to x = 0.008 in Ti02.x [47]; the upper limit of oxygen pressures is determined only by the strength of the vessel that contains the experiment.

498

O

[Vnn*]

X''' [Vo*]

O

o o

o >

lnpo2 Fig. 9. Schematic graph illustrating how the composition of the surrounding atmosphere influences the stoichiometry of a metal oxide.

If we take the total pressure to be p^ + PQ2, we can use Eqs. 4-6 to show that the total pressure is: Piot ~ P02 "•" ^MO Po2

(10)

By minimizing Eq. 10 with respect to Po2' it can be shown that there is a minimum partial pressure of oxygen at which the compound can be in equilibrium. In this case: P02,

,= l/2(2kMo^)"^

(11)

As long as the partial pressure of O2 is greater than PQ2, min' the stoichiometry of the compound will adjust to equilibrate with its surroundings. Below this minimum, some of the crystal will have to sublime to establish the minimum vapor pressure. In a dynamic vacuum, where subliming material is constantly pumped away, this is clearly not possible. Under such conditions, the stoichiometry of the compound is determined by kinetic factors and it might not be possible to conduct experiments with reproducible stoichiometrics. While the preceding statements apply to the bulk of the crystal, the surface of the crystal is subject to the same constraints with some additional complexity. Specifically, because different point defects are likely to have different free energies of formation, there is an enhanced concentration of one of the two defects in the near surface region and a compensating layer of charge on the

499

surface [48, 49]. Suppose, for example, that it is easier to create O vacancies than M vacancies. The enhanced concentration of O vacancies that would result from this situation would lead to a net positive charge in the near surface region that would be have to be compensated by negative charge on the surface. It has been pointed out that the amount of charge that it is possible to store in the subsurface region will be limited by the capacity of the surface to accommodate the compensating charge [50]. As with the bulk situation, equilibrium surface defect concentrations will be established only at a fixed and sufficiently high partial pressure of oxygen. One of the more well-documented cases that illustrates the effect of the annealing atmosphere on the surface composition concerns the Al2O3(0001) surface [51]. After heating in air, 1 X 1 LEED patterns are observed. High temperature annealing (1350 °C) in UHV leads to the loss of O and the formation of an Al terminated surface that has a (VSl X V31)R±9° LEED pattern. The model for this reconstruction is two pure and nearly perfect Al planes on an O terminated Al2O3(0001) surface. The same structure can be produced by depositing Al metal from the vapor phase onto an Al2O3(0001) surface with the 1 X 1 structure [52, 53]. Therefore, and alternate way to view this reconstruction is that the Al2O3(0001) surface is decomposing during high temperature annealing in a low partial pressure of oxygen atmosphere to form Al, which adopts a low energy epitaxial structure. Similarly, Fe2O3(0001) surfaces are found to be enriched in metal and transform to Fe304(lll) and FeO(lll)-like reconstructions after being ion bombarded, annealed in UHV, and then heated in 1 x 10"^ mbar of O2 [54, 55]. As in the case of the Al2O3(0001) surface, the observed restructuring of the surface is consistent with O loss in a low PQ2 environment. One certain implication of the solid-vapor equilibrium is that when an oxide is ion bombarded or heated in UHV, the resulting bulk stoichiometry is neither well defined nor likely to be reproducible. This is a likely explanation for the diverse range of observations that have been reported for the TiO2(110) surface which include not only a (1 X 1) structure [56-62], but also structures related to Magneli phases [63,64], and (1x2) [56,57,59,61], (1x3) [61], (2x2) [59], c(2xl) [56], and (2x3) [65] structures. Despite the difficulty of producing oxides with reproducible bulk stoichiometrics in vacuum, it must be emphasized that kinetically stable surfaces with reproducible structures have been prepared in the UHV environment. The creation of the TiO2(110) (1x1) surface in UHV is an excellent example [62]. Not surprisingly, STM images of the surface illustrate that it has a remarkably high concentration of oxygen vacancies. Surfaces prepared by cleavage at ambient temperature, if not annealed or ion bombarded in vacuum, are likely to have relatively well defined and reproducible stoichiometrics. While not generally in equilibrium, they should be kinetically stable at room temperature. However, the composition of such a

500

surface is not always easy to specify. In very simple situations, cleavage can produce two identical surfaces and there is no ambiguity about the surface termination. In other cases, cleavage leads to two distinct surfaces. It is also possible that there are multiple possible cleavage planes. Consider, for example, the V^Ojg structure shown in Fig. 10 [66]. Cleavage at the points labeled A, B, or C lead to distinct terminations. The analysis of STM and AFM images of the surface and comparison to calculated electron density maps are consistent with the conclusion that cleavage occurs at C. Based on studies of tungstates, molybdates, and vanadates, it is possible to make the general statement that oxides cleave at the longest, weakest set of bonds perpendicular to the cleavage plane [66-70].

(a)

(b)

Fig. 10. (a) Idealized polyhedral projection of the V6O13 structure, showing the three possible planes of cleavage, (b) An STM image of the plane, identified as C.

501

5. REACTION ANISOTROPIES ON M0O3 SURFACES In section 2 of this chapter, surface character-property relationships for the catalytic properties of M0O3 were described. These relationships were deduced largely from experiments on particulate materials. In this section, AFM experiments are described that were designed to discriminate the functions of the basal and lateral facets by direct observation. Experiments were conducted on single crystals surfaces with a clearly defined orientation and composition that were treated in a reactor system capable of reproducing the range of temperatures and environmental conditions relevant to the oxidation of alcohols [71-75]. Under these conditions, both the morphology and the stoichiometry of the surface evolved in a complex way. Room temperature AFM imaging at the conclusion of the treatment was used to determine how the surface structure changed. The AFM images in Fig. 11 show a variety MoOgCOlO) surfaces before and after reactions with methanol [73, 74]. The images in Fig. 11a and b show the surface structure before the reaction. Cleavage surfaces, such as those illustrated in Fig. 11a, are characterized by atomically flat terraces separated by steps that have heights that an integer multiple of about 7 A. This characteristic height is half the unit cell repeat length along [010] and corresponds to the distance between the van der Waals gaps which separate the adjacent layers of the structure (see Fig. 2). The steps on cleavage surfaces are most frequently aligned along [001]. The configuration of steps on a growth surface (see Fig. lib) is determined by the positions of the screw dislocations. Defects such as the one near the center of the image in Fig. l i b provide a continuous source of steps that facilitates crystal growth. In the reactivity studies described below, there was no detectable difference in properties of the growth surfaces and the cleavage surfaces. When MoO3(010) surfaces are reacted with air-N^-MeOH mixtures at 300 °C, loops of steps surrounding shallow pits nucleate on the previously flat terraces (see Fig. llc-e). The pits in these images are always bounded by half unit steps and are therefore only 7 A deep. While the pitting process was observed in all of the air-N2-MeOH mixtures tested (airiN^ ratios between 100:0 and 02:98), the shapes of the pits and their evolution depended on the composition of the mixture. Under oxygen rich conditions, the pits had a rectangular or ovular shape, and were always elongated along <001>. As the air concentration is reduced, the pits elongate along <100> and assume a triangular habit (see Fig. lie). The pits are only formed when some methanol is included in the feed. When air is completely eliminated from the feed, H^MoG^ precipitates (the white contrast in Fig. llf). These precipitates are formed when the hydrogen liberated during the chemisorption of the alcohol reacts with the M0O3 to form the molybenum bronze, rather than reacting with lattice O to form

502

water [73]. The same reaction occurs when MeOH is replaced by higher order alcohols.

Fig. 11. AFM images of the MoO3(010) surface. The arrows show the direction of the [001] axis. A cleavage (a) and growth (b) surface before the reaction, (c) After 2 h at 300 °C in air saturated with methanol, (d) After 2 h at 300 °C in a 50:50 air : N2 mixture saturated with methanol, (e) After 2 h at 300 °C in a 30:70 air : N2 mixture saturated with methanol, (f) After 6 min at 330 °C in N2 saturated with methanol.

503

The most plausible explanation for the formation of the pits during the reactions with the methanol-air mixtures is that the surface methoxide species that form during the initial dissociative chemisorption of methanol are capable of desorbing as molybdenum-oxide-methoxide molecules, which are then carried away in the flowing gas. The observation that crystalline Mo^O^COCH^)^ forms at the reactor exhaust supports this mechanism for evaportation [74]. These observations, made far from equilibrium, illustrate that the morphological structure of the M0O3 surface in service is continuously evolving and depends on the composition of the gas feed. Of particular interest is the generation of step edges on the surface. As discussed in section 2, the step edges provide undercoordinated Mo sites (similar to those found on the lateral facets of the

Fig. 12. AFM images of defect related pits that form on the MoO3(010) surface. In each case, the arrow indicates the [001] direction, (a) After heating for 3.5 h in air at 400 °C, shallow pits form at surface-dislocation intersections, (b) After 8 min at 400 °C in a 90:10 N2:H2 mixture containing 20 ppm water, (c) After 5 min at 400 °C in a 90:10 N2:H2 mixture containing 20 ppm water, (d) After 8 min at 400 °C in a 90:10 N2:H2 mixture containing 20 ppm water, followed by a 3.5 h anneal in air at the same temperature.

504

crystal) that are thought to have a different function than the basal sites. It has been observed that the elastic strain associated with extended crystal defects also promotes the formation of pits in the (010) surface when heated to approximately 400 °C in the presence of water vapor. It is therefore possible to generate controlled distributions of undercoordinated Mo to test their relative importance in the reaction with methanol. There are two primary crystal defects that create enough strain to nucleate a pit during reactions with water vapor: dislocations and crystallographic shear planes. Several examples of pits that can be formed at the sites where defects intersect the surface are illustrated in Fig. 12. Water vapor is known to enhance the sublimation of M0O3 through the formation of volatile molybdenum oxyhydroxides such as Mo02(OH)2 and is thought to play a similar role in the formation of pits near the points where defects intersect the surface [75]. For crystals heated to 400 °C in oxidizing atmospheres containing water vapor, pits form only at the points where screw dislocations intersect the surface. The image in Fig. 12a shows decending spiral steps into the etch pit formed around a dislcoation with a Burgers vector of approximately 14 A, which is equal to the lattice spacing along [010]. By examing dislocations with different Burgers vectors, it was concluded that the amount of material removed in a fixed time period was proportional to the stored elastic strain around the defect. Because the average dislocation density was only 1.5 x lOVcm^, such pits are rather rare. However, in reducing atmospheres, the wide spread formation of crystallographic shear planes (defects that acconmiodate oxygen loss) leads to pits densities up to four orders of magnitude greater [71]. A surface with a high density of small pits (black contrast) is shown in Fig. 12b. Higher resolution images show that 1.5 A high steps that are characteristic of surface/CS plane intersections invariably intesect the pits (see Fig. 12c). When a surface such as the one shown in Fig. 12b is reheated in an oxidizing atmosphere, the CS planes are eventually annihilated and and the remaining pits reorient to form acicular trenches along the <001> direction, as illustrated in Fig. 12d. Because pit formation is activated by elastic strain from extended defects (which can be deliberately introduced), we can exert some degree of control over the size and density of the pits and, therefore, the available population of undercoordinated Mo. To test the influence that these sites have on the chemisorption of alcohols, a series of model surfaces were created with different densities of undercoordinated Mo associated with pits and then reacted with methanol saturated N2 at 330 °C. These conditions lead to the formation of the H^Mo03 phase, which was used as a quantitative and local indicator of the dissociative chemisorption reaction. The images in Fig. 13 illustrate selected results from these experiments. In Fig. 13a and b, it is clear that the acicular H^MoOg precipitates are growing from the lateral walls of the surface pits [75].

505

The images in Fig. 13c and d show that when the density of the pits increases and the distance between adjacent pits is small compared to the size of the precipitates, the preferential nucleation is not obvious. However, the total amount of the hydrogen bronze phase produced still correlates with the total area of the pits. Both crystals in Fig. 13 c and d were reduced for 5 min at 400 °C in forming gas to form the pits. One was then oxidized in ambient air at 410 °C for 1 h so that the pits could grow, while the other was not. As a results, the pits shown in Fig. 13c are all less than 200 A deep while those on the surface that received the oxidation treatment are all greater than 500 A deep (see Fig. 13d).

mi.% ~*l-% -*4,^i

Fig. 13. AFM topographs of MoOsCOlO) surfaces, (a) A crystal that was reduced for 10 min at 400°C in forming gas and then oxidized for 2 hr at 400°C in the ambient prior to being reacted for 20 min at 330 °C in N2 saturated with MeOH. The black-to-white contrast is 100 A. (b) A surface that was reduced in forming gas at 400°C for 10 min, oxidized for 3 hr at 410°C in air, and then exposed to N2 saturated with MeOH for 15 min at 330 °C. The black-to-white contrast in the image is 100 A. (c & d) comparison of precipitation on surfaces with different pit sizes (see text). Each was reacted for 10 min at 330 °C with N2 saturated with MeOH. The black-to-white contrast in (c) is 200 A, while that in (d) is 400 A.

506

Both surfaces were then reacted for 10 min at 330 °C in N^ saturated with MeOH at room temperature; the precipitates on the surface with the larger pits (Fig. 13d) are larger and more numerous than those on the surface with more shallow pits (Fig. 13c). When the total precipitate volume on each surface is estimated, the results indicate that the surface with the larger pits produced four times as much of the H^MoOg phase. It should be pointed out that the stoichiometry of these two surfaces was not the same. The one that was heated in air was more oxidized than the other and, therefore, had fewer surface oxygen vacancies. Since surface oxygen vacancies are also sources of undercoordinated Mo, one must consider the possibility that they compete with the lateral facets as chemisorption sites. However, when the upper limit of the oxygen vacancy concentration (lO'"^) is compared to an estimate of the number of undercoordinated Mo associated with the lateral walls of pits, one finds that the surface oxygen vacancy concentration is not a significant factor [75]. This is consistent with the observations that the sample with fewer oxygen vacancies (but deeper pits) formed more of the hydrogen bronze phase. In summary, the results presented in this section indicate that at the temperatures and pressures relevant for partial oxidation reactions, it is the {hOk} surfaces that are active for the dissociative chemisorption of alcohols. Furthermore, the morphological structure of the catalyst particles is continuously evolving during the reactions, introducing a higher than anticipated density of undercoordinated Mo on {hOk} surfaces while the catalyst is in service. While proposed mechanisms for partial oxidation usually concentrate on the exchange of O between the catalyst and the gas phase, the current results suggest that the exchange of H and Mo may also be important. 6. THE ORIENTATION DEPENDENCE OF THE PHOTOCHEMICAL REACTIVITY OF TiOj The photochemical properties of titania surfaces are of interest for several reasons. They determine the stability of pigmented paint systems [76], the rate at which pollutants can be degraded in systems designed to purify air and water [77], and are the root cause of poorly understood phenomena such as water photolysis [78] and "super hydrophilicity" [79]. Using thin rutile epilayers with five low index orientations, it has been shown that the relative rates of photochemical reactions catalyzed by titania depend on the surface orientation [80]. In this chapter, experiments used to map the complete orientation dependence of the relative photochemical reactivity of TiO^ are described [8183]. In this case, the relevant reactions are carried out at room temperature and this gives us the opportunity to fix both the surface morphological structure and stoichiometry. For the reactions described here, all of the surfaces were

507

annealed in air for several hours at 1200 °C to define the stoichiometry and morphological structure. The challenge associated with measuring the photochemical properties of titania as a function of orientation lies in the number of required measurements. Each possible surface orientation is defined by two independent variables. If each orientation parameter can be measured to within ±3°, then for a tetragonal material such as rutile, there are 436 distinct surface orientations. It is clear that the preparation and analysis of oriented single crystal surfaces would be prohibitively tedious. This challenge can be overcome by using polycrystalline specimens. Assuming a 50 micron grain size, a 5 mm^ area of a polished section through a random microstructure reveals approximately 2000 grain surfaces. Single grain orientations were determined based on electron backscattered diffraction patterns (EBSPs) obtained in a scanning electron microscope (SEM). For unfaceted grains, the index of the surface plane can be determined directly from the EBSP. Faceted grains, however, are terminated by more than one surface plane as descibrd in section 3. In such cases, atomic force microscopy (AFM) was used to measure the inclination of each facet with respect to the surface normal; when these data are combined with knowledge of the grain orientation from the EBSPs, the facet planes can be indexed. To make a local measurement of the photochemical activity of each grain, we used a well established probe reaction (the reduction of aqueous Ag+ to Ag^) that deposits metallic silver on the surface as a reaction product [84-85]. The amount of silver deposited on each grain's surface during a given reaction, which is determined from atomic force microscopy (AFM) images, is taken to be a quantitative indicator of the grain's relative photochemical reactivity. The reactivity can then be correlated to surface orientation and/or the relative area of each facet on the surface. The AFM images in Fig. 14 illustrate how this technique is used to determine the anisotropy of the reactivity [81]. The rutile poly crystal used in this experiment was polished and then annealed in air at 1473 K for four hours to heal polishing damage. Note that grain boundaries and pores appear in the images as black contrast. Figure 14a shows several contiguous grains in the specimen. Fig. Ic shows a triple grain junction, and Fig. 14e shows the step structure on the surface of a single grain. The sample was immersed in aqueous AgNOs, exposed to ultraviolet (UV) light, and the same areas were located and imaged again. The results are shown in Fig. 14 (b,d,f). Note that after the reaction, heterogeneous white contrast (corresponding to elevated regions) is apparent in the images. On the basis of energy dispersive X-ray spectroscopy conducted in an SEM, there is a one-to-one correlation between the elevated regions of white contrast and high concentrations of Ag. Fig. lb illustrates that some surfaces appear to be almost completely covered with silver, while others have very few deposits. On the basis of these

508

observations, three types of grains (illustrated in Fig. 14d) are distinguished: very reactive (upper left), moderately reactive (lower left), and relatively unreactive (right). The very reactive grains are uniformly covered with silver so that the underlying surface structure can no longer be distinguished. This occurs when there are more than 4 Ag islands per jLim^. The moderately reactive grains

Fig. 14. AFM images of polycrystalline rutile surfaces before (a,c,e) and after (b,d,f) the photochemical deposition of silver (white contrast).

509

have easily distinguishable silver islands with a density between 2 and 4 per |im2, while the morphology of the TiO^ surface remains visible in the background. The relatively unreactive grains have an inhomogeneous distribution of Ag and there are less than 2 silver islands per ^im^. While parts of unreactive grains in the vicinity of grain boundaries, residual polishing scratches, pores, and other defects do show deposited silver, the flat regions of the surface are indistinguishable before and after the reaction. It is the inhomogeneous distribution of silver that distinguishes the relatively unreactive grains from the moderately reactive grains. Based on knowledge of the grain orientation and surface inclinations measured by AFM, it is possible to index all of the grain surfaces. The dependence of the reactivity on the grain and surface orientation is illustrated simultaneously in Fig. 15, which shows an inverse pole figure that combines photochemical reactivity and surface orientation stability data. As discussed in section 3, the areas with the constant gray shading are regions where the surface orientation is stable and the grain has a flat surface. In such areas, the surface orientation and the grain orientation are the same. The areas filled by tie lines indicate orientations that break up into two facets. In this case, the two stable surface facets are found at the ends of the tie lines. The white areas show orientations that break up into three stable surfaces whose orientations are found at the vertices of the surrounding triangle. Based on the results in Fig. 15, we can make the following statements. First, the most reactive grains have orientations near {101}. Second, with the exception of a few outliers, the grains that have high reactivity have surface {001

•

100}

> 4 deposits/cm^

X > 2 deposits/cm^ O < 2 deposits/cm^ 110} Fig. 15 Inverse pole figure summarizing reactivity and orientation stability data for titania.

510

orientations that are unstable and break up into two or three separate facets. Furthermore, one of the stable facets is always {101}. Further analysis indicated that the relative amount of silver on surfaces containing {101} facets correlates with the fractional {101} area. Based on this result, we conclude that the {101} surface is the most reactive and this is the key microstructural feature for the design of high reactivity materials. This conclusion allows microstructures optimized for high reactivity to be specified. Planar structures should be oriented such that the surface normal breaks up into complementary {101} facets (the surface area of a faceted structure is greater than a single {101} type facet). Particulate catalysts should have a pseudo-octahedral habit with {101} planes bounding the crystal. The observation that the most photochemically reactive surface orientations all contain {101} facets suggests that the enhanced reactivity is not associated specificially with the bulk crystal orientation, but is a property of this particular surface plane. It is likely that there are special atomic configurations on this plane that either alter the efficiency with which photogenerated carriers are trapped at the surface or the rate at which they are transferred across the solid-liquid interface. Based on bulk geometry alone, there is nothing that sets the {101} plane apart from less reactive surfaces. On this surface, Ti cations are coordinated by five O, as they are on the (100) surface (which is inert). In the absence of higher resolution microscopy results, it is not possible to say if special molecular configurations are created by reconstruction or defect formation. 7. CONCLUSION The surface properties of metal oxides are anisotropic and because of this, there is a potential to control the properties of surface active materials by manipulating the distribution of exposed surface characters. Under equilibrium conditions, the distribution of surface characters is determined by the anisotropy of the surface energy and the solid-vapor equilibrium. While one is not necessarily limited to equilibrium configurations, materials at sufficiently high temperatures will evolve in this direction and significant morphological and stoichiometric changes can occur. By systematically evaluating the properties of surfaces with different characters, it is possible to define optimized morphological structures. ACKNOWLEDGEMENT The author acknowledges the support of the US National Science Foundation under grant DMR 0072151. The work was supported in part by the MRSEC program of the National Science Foundation under award number DMR0079996. Contributions to the research described in this chapter from Richard

511

L. Smith, Jennifer B. Lowekamp, David M. Saylor, Jennifer L. Giocondi, P.A. Morris Hotsenpiller, J.D. Bolt, and W.E. Farneth are also gratefully acknowledged. The author thanks Prof. C.B. Carter for the permission to use Fig. 4. REFERENCES [I] J.E. Germain in: M. Che, G.C. Bond, (Eds.), Adsorption and Catalysis on Oxide Surfaces, Elsevier Science Publishers, B.V., Ansterdam, 1985, p. 355. [2] J.M. Tatibouet, J.E. Germain, J. Catal., 72 (1981) 375. [3] J.C. Volta, J.M. Tatibouet, J. Catal., 93 (1985) 467. [4] J.M. Tatibouet, Ch. Phichitkul, J.E. Germain, J. Catal., 99 (1986) 231. [5] K. Briickman, R. Grabowski, J. Haber, A. Mazurkiewicz, J. Sloczynski, T. Wiltowski, J. Catal., 104(1987)71. [6] K. Bruckman, J. Haber, T. Wiltowski, J. Catal., 106 (1987) 188. [7] A. Anderssen, S. Hansen, J. Catal., 114 (1988) 332. [8] B. Mingot, N. Floquet, O. Bertrand, M. Treilleux, J.J. Heizmann, J. Massardier, M. Abon, J. Catal. 118(1989)424. [9] M. Abon, J. Massardier, B. Mingot, J.C. Volta, N. Floquet, O. Bertrand, J. Catal, 134 (1992) 542. [10] M.R. Smith, U.S. Ozkan, J. Catal., 141 (1993) 124. [II] L. Kihlborg, Arkiv Kemi, 21 (1963) 471. [12] M. Gasior, T. Machej, J. Catal., 83 (1983) 472. [13] E. Bordes, Catal. Today, 6 (1992) 21. [14] W.E. Farneth, F. Ohuchi, R.H. Staley, U. Chowdhry, A.W. Sleight, J. Phys. Chem., 89 (1985) 2493. [15] J.M. Vohs, M.A. Barteau, J. Phys. Chem., 91 (1987) 4766. [16] K.S. Kim, M.A. Barteau, Surf. Sci., 223 (1989) 13. [17] K.S. Kim, M.A. Barteau, J. Catal., 125 (1990) 353. [18] M.A. Barteau, J. Vac. Sci. Technol. A, 11 (1993) 2162. [19] J. Fompeyrine, R. Berger, H.P. Lang, J. Perret, E. Machler, Ch. Gerber, J.-P. Locquet, Appl. Phys. Lett., 72 (1998) 1697. [20] M. Kawasaki, K. Takahashi, T. Maeda, R. Tsuchiya, M. Shinohara, O. Ishiyama, T. Yonezawa, M. Yoshimoto, H. Koinuma, Science, 226 (1994) 1540. [21] G. Koster, B.L. Kropman, G.J.H.M. Rijnders, D.H.A. Blank, H. Rogalla, Appl. Phys. Lett., 73 (1998) 2920. [22] H. Tanaka, H. Tabata, T. Kawai, Thin Sohd Films, 342 (1999) 4. [23] M. Kawasaki, O. Ohtomo, R. Tsuchiya, J. Nishino, K. Koinuma, Mat. Res. Soc. Symp. Proc, 474 (1997) 303. [24] P.A. Salvador, B. Mercey, O. Perez, A.M. Haghiri-Gosnet, T.-D. Doan, B. Raveau, Mat. Res. Soc. Symp. Proc, 587 (2000). [25] M.W. Bench, P.G. Kotula. C.B. Carter, Surf. Sci., 391 (1997) 183.

512 [26] C. Herring, in: R. Gomer C.S. Smith. (Eds.), Structure and Properties of Solid Surfaces, The University of Chicago Press, Chicago, 1952, p. 5. [27] P. Curie, Bull. Soc. Min. de France, 8 (1885) 145. [28] G. Wulff, Z. Krist., 34 (1901) 449. [29] R.S. Nelson, D.J. Mazey, R.S. Barnes, Philos. Mag., 11 (1965) 91. [30] J.C. Heyraud, J.J. Metois, Acta Met., 28 (1980) 1789. [31] Z.Y. Wang, M.P. Harmer, Y.T. Chou, J. Am. Ceram. Soc, 69 (1986) 735. [32] J.-H. Choi, D.-Y. Kim, B.J. Hockey, S.M. Wiederhom, C.A. Handwerker, J.E. Blendell, W.C. Carter, A.R. Roosen, J. Amer. Ceram. Soc, 80 (1997) 62. [33] M. Kitayama, A.M. Glaeser, Key Engineering Materials, 159-160 (1999) 193. [34] W.W. Mullins, G.S. Rohrer, J. Amer. Ceram. Soc, 83 (2000) 214. [35] D.M. Saylor, D.E. Mason, G.S. Rohrer, J. Amer. Ceram. Soc, 83 (2000) 1226. [36] V.E. Henrich, Surface Science, 57 (1976) 385. [37] P.W. Tasker, D.M Duffy, Surf. Sci., 137 (1984) 91. [38] M. Causa, R. Dovesi, C. Pisani, C. Roetti, Surf. Sci., 175 (1986) 551. [39] W.C. Mackrodt, Phys. Chem. Minerals, 15 (1988) 228. [40] J. Goniakowski, C. Noguera, Surf. Sci., 319 (1994) 68. [41] M. Gajdardziska-Josifovska, P.A. Crozier, J.M. Cowley, Surf. Sci. Lett., 248 (1991) L259. [42] J.W. Cahn, C.A. Handwerker, Materials Science and Engineering A,162 (1993) 83. [43] A.J.W. Moore, Acta Met., 6 (1958) 293. [44] C. Herring in: W.E. Kingston (Editor), The Physics of Powder Metallurgy, McGraw Hill, New York, 1951, p. 79. [45] W.W. MulHns, J. Appl. Phys., 28 (1957) 333. [46] W. van Gool, Principles of Defect Chemistry of Crystalline Solids (Academic Press, New York, 1966) p. 44. [47] P. Kofstad, Nonstoichiometry, Diffusion, and Electrical Conductivity in Binary Metal Oxides (Robert E. Krieger Publishing Company, Malabar, PL, 1983) p. 139. [48] J. Frenkel, Kinetic Theory of Liquids (Oxford University Press, New York, 1946) p. 36. [49] K.L. Khewer, J.S. Koehler, Phys. Rev., 140 (1965) A1226. [50] R.B. Poeppel, J.M. Blakely, Surf. Sci., 15 (1969) 507. [51] M. Gautier, G. Renaud, L.P. Van, B. Villette, M. Pollak, N. Thormat, F. Jollet, J.-P. Duraud, J. Amer. Ceram. Soc, 77 (1994) 323. [52] T.M. French, G.A. Somorjai, J. Phys. Chem., 74 (1970) 2489. [53] M. Vermeersch, R. Sporken, Ph. Lambin, R Caudana, Surf. Sci., 235 (1990) 5. [54] N.G. Condon, P.W. Murray, F.M. Leibsle, G. Thornton, A.R. Lennie, D.J. Vaughan, Surf. Sci., 310 (1994) L609. [55] N.G. Condon, F.M. Leibsle, A.R. Lennie, P.W. Murray, D.J. Vaughan, G. Thornton, Phys. Rev. Lett., 75 (1995) 1961. [56] M. Sander, T. Engel, Surf. Sci. Lett., 302 (1994) L263. [57] H. Onishi, Y Iwasawa, Surf. Sci. Lett., 313 (1994) L783. [58] A. Szabo, T. Engel, Surf. Sci., 329 (1995) 241.

513 [59] P.W. Murray, N.G. Condon, G. Thornton, Phys. Rev. B, 51 (1995) 10989. [60] S. Fischer, A.W. Munz, K-D. Schierbaum, W. Gopel, Surf. Sci., 337 (1995) 17. [61] A. Berko, F. Solymosi, Langmuir, 12 (1996) 1257. [62] U. Diebold, J.F. Anderson, K.-O. Ng, D. Vanderbilt, Phys. Rev. Lett., 77 (1996) 1322. [63] G.S. Rohrer, V.E. Henrich, D.A. Bonnell, Science, 250 (1990) 1239. [64] G.S. Rohrer, V.E. Henrich, D.A. Bonnell, Surf. Sci., 278 (1992) 146. [65] M. Wagner, D. A. Bonnell, M. Ruble, Appl. Phys. A, 66 (1998) 1165-70. [66] R.L. Smith, G. S. Rohrer, K.S. Lee, D.-K. Seo, M.-H. Whangbo, Surf. Sci., 367 (1996) 87. [67] R.L. Smith, W. Lu, G.S. Rohrer, Surf. Sci., 322 (1995) 293. [68] G.S. Rohrer, W. Lu, R.L. Smith, A. Hutchinson, Surf. Sci., 292 (1993) 261. [69] W. Lu, N. Nevins, M.L. Norton, G.S. Rohrer, Surf. Sci., 291 (1993) 395. [70] R.L. Smith, G.S. Rohrer, J. Solid State Chem., 124 (1996) 104. [71] R.L. Smith, G.S. Rohrer, J. Catalysis, 163 (1996) 12. [72] R.L. Smith, G.S. Rohrer in: K.S.Ramesh, M. Misono, and P.L. Gai (Eds.), Catalyst Materials for High-Temperature Processes, American Ceramic Society, Westerville, OH, 1997. p. 139. [73] R.L. Smith, G.S. Rohrer, J. Catalysis, 173 (1998) 219. [74] R.L. Smith, G.S. Rohrer, J. Catalysis, 180 (1998) 270. [75] R.L. Smith, G.S. Rohrer, J. Catalysis, 184 (1999) 49. [76] H.G. Volz, G. Kaempf, H.G. Fitzky, A. Klaeren in: S.P. Pappas, F.H. Winslow, (Eds.), Photodegradation and Photostabilization of Coatings, American Chemical Society, WashingtionD.C, 1981. p. 163. [77] A. Wold, Chem. Mater., 5 (1993) 280. [78] A. Fujishima, K. Honda, Nature, 238 (1972) 37. [79] R. Wang, K. Hashimoto, A. Fujishima, M. Chikuni, E. Kojima, A. Kitamura, M. Shimohigoshi, T. Watanabe, Nature, 388 (1997) 431. [80] P.A. Morris Hotsenpiller, J.D. Bolt, W.E. Fameth, J.B. Lowekamp, G.S. Rohrer, J. Phys. Chem. B, 102 (1998) 3216. [81] J.B. Lowekamp, G.S. Rohrer, P.A. Morris Hotsenpiller, J.D. Bolt, W.E. Fameth, J. Phys. Chem. B, 102 (1998) 7323. [82] J.B. Lowekamp, G.S. Rohrer, P.A. Morris Hotsenpiller, J.D. Bolt, W.E. Fameth, to be published. [83] J.B. Lowekamp, Ph.D. Thesis, Camegie Mellon University, 1999. [84] W.C. Clark, A.G. Vondjidis, J. Catalysis, 4 (1965) 691. [85] J.-M. Herrmann, P. Pichat, J. Catalysis, 113 (1988) 72.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

514

Chapter 13

Vibrational Spectroscopy at Oxide Surfaces Brian E. Hayden Department of Chemistry, University of Southampton, Highfield, Southampton, S017 IBJ, United Kingdom.

1.

INTRODUCTION

Vibrational spectroscopy of surfaces has proved itself a powerful tool for the study of both clean and adsorbate covered surfaces. The characterisation of the vibrational modes on clean surfaces provides information concerning interatomic potentials that can be used in the interpretation of structural and dynamical properties. Vibrational spectroscopy of adsorbates provides one of the most important sources of information concerning the chemical and physical nature of the adsorbed species. This chapter concerns the development and application of vibrational spectroscopy on well characterised metal oxide surfaces to the measurement of adsorbate and localised surface phonon modes of the substrate. The development of the techniques applied to metal oxides has been proceeded by their more extensive application to studies on metal and semi-conductor surfaces, for which there is a comprehensive review literature [1-8]. The most important and commonly applied techniques include various forms of infrared spectroscopy, such as Reflection Absorption InfraRed Spectroscopy (RAIRS), Surface Electromagnetic Wave Spectroscopy (SEWS), Emission Spectroscopy (ES) and Multiple Internal Reflection Spectroscopy (MIRS). High Resolution Electron Energy Loss Spectroscopy (HREELS) and Inelastic Atom Beam Scattering (lABS) have also found wide application to surface vibrational spectroscopy, and have the advantage of an intrinsic surface sensitivity in the scattering event. ES and SEWS are the least commonly applied of the infrared techniques, and no attempt has been made to apply them to metal oxide surfaces. MIRS, which has been extensively applied to semi-conductor surfaces, has not so far been applied to metal oxide surfaces. This is probably because absorption in the substrate resulting from more extensive and strong dipole active phonon modes in metal oxides would make internal reflection experiments difficult. The spectral limitations of lABS to <400cm"^ confines it

515

more to the study of surface phonons, and in the case of oxides studies the technique has been restricted to the characterisation of non-localised modes. The main techniques applied so far to metal oxide surfaces have been HREELS andRAIRS. At the time of a recent review [9], there remained very few examples of vibrational studies of adsorbate, or localised substrate modes, at metal oxide surfaces. By far the majority of studies concerned the characterisation by HREELS of phonon modes (such as Fuchs-Kliewer modes) pertaining to the properties of the bulk structure, rather than the surface, or to electronic transitions. Such studies have been excluded from this review in order to concentrate on the vibrational spectroscopy of surface vibrations on wellcharacterised metal oxide surfaces such as single crystals or epitaxially grown oxide films, for which there is now a substantial literature. Nevertheless, it is important to briefly describe the electronic and phonon properties of oxides in order to understand the constraints and difficulties in carrying out RAIRS and HREELS with sufficient sensitivity to observe adsorbate vibrations, and more localised substrate vibrational modes.

2.

THEORETICAL AND EXPERIMENTAL CONSIDERATIONS

The intrinsic band gap of metal oxides is sufficiently large that the electronic contribution to the substrate dielectric is real. The absence of screening electrons in the vibrational energy regime result in low IR absorption and the consequences of this for RAIRS measurements are outlined below. The substrate phonon modes range from essentially zero meV for long-wavelength acoustical modes (there are always three branches), to several hundred meV for optical modes. For a primitive unit cell with N atoms, there are (3N-3) optical branches: in all oxides there are at least two atoms per unit cell so that there are always optical phonons present. In the case of strongly ionic metal oxides, strong dynamic dipole fields result from the relative motion of the ions in the optical modes, and these can couple strongly to externally applied fields. The result is a strong interaction of optical phonons with not only IR radiation (hence the name optical phonons), but also with charged particles such as electrons. The polarisation of the optical phonon modes, and hence their dipole activity as a fimction of scattering geometry, may be determined from the crystal structure and symmetry by group theory arguments [10]. For crystals of high symmetry, some "optical" phonon branches correspond to net ionic motions that do not result in electric fields in the long-wavelength (dipole) limit. The dipole active phonon modes of some important crystal structures are summarised elsewhere [9]. In addition to the bulk modes, there are also surface phonon modes, which result from the termination of the bulk structure. The

516

associated wavefunctions correspond to oscillatory, travelling waves in directions parallel to the surface, but have an exponentially damped amplitude normal to the surface into the bulk of the crystal. They may be observed experimentally using inelastic helium scattering or using HREELS at low incident energy EQ. They are localised in a layer of about one phonon wavelength deep at the surface, and the same criteria for the number of surface phonon modes of each type apply as for bulk modes. Since optical phonon frequencies lie in the range 300-800cm"\ RAIRS measurements carried out outside this frequency range are characterised by a uniform dielectric characteristic of the bulk electronic structure. So far there are no examples of RAIRS measurements of adsorbates carried out at energies where optical phonons will result in strong changes in IR reflectivity. HREELS measurements reveal, however, that electron coupling to the surface optical phonon modes is strong, and the probability of energy loss can account for up to one half of the elastic scattering probability. The result is that there is also a significant probability that successive loss events can take place during the scattering event, and strong multiple loss peaks are observed. The origin of these is quite different from multi-phonon absorption in IR spectra, which is intrinsically weak since it results from anharmonic effects [10]. It is the multiple loss peaks in the HREELS spectrum that makes vibrational spectroscopy of adsorbates difficult, since these extend into the spectral range of intra-molecular vibrational modes. 2.1.

High Resolution Electron Energy Loss Spectroscopy (HREELS) HREELS spectra of oxide surfaces are dominated by the so-called surface optical phonons, or "Fuchs-Kliewer" modes [11]. The strongest loss peaks in HREELS from solids result from the long-range electrostatic fields associated with dipole-active excitations [2]. For a beam of electrons incident on a surface with a normal velocity component Vn, energy loss may take place corresponding to excitation of a mode with frequency co. A change in parallel momentum may also take place during the scattering process. Assuming that the total momentum is conserved during scattering, the change in momentum can be related to k, the parallel wave vector of the excitation. Describing the substrate using an isotropic dielectric function s(co), the probability of such a loss event is given by: P(a),k)=

4Q\ nW{k'H(d/yrf}'

Im{-l/e((B)+l}

(1)

P(co, k) peaks sharply at low k under normal experimental conditions i.e. long wavelength phonons are predominantly excited because the electric field associated with such phonons extends outside of the solid further than short

517

wavelength phonons, and the interaction with electrons is correspondingly stronger. For typical spectrometer solid angles of collection, one can assume that an effective angular integration of the loss spectrum occurs [12]. Performing the appropriate integration of P(co, k) over k yields: P(co)=

le

(2)

Im{-l/8(co)+l}

hVnCO

This function is a maximum at frequencies satisfying the condition Re{s(co) = 1} for surface optical phonon (Fuchs-Kliewer) excitation. Such modes have an energy slightly less than the corresponding bulk longitudinal optical phonon purely as a result of electrostatic effects resulting from different boundary conditions. Fuchs-Kliewer modes are localised at the surface, but because of their long wavelength under normal excitation conditions, their exponential decay into the bulk is very long (ca. 10 nm) and they are therefore insensitive to changes in bonding and lattice dynamics at the surface. Note from Eqn. 2 that P(co) a Eo'^''^, the primary electron energy. This has been demonstrated experimentally during measurements of the probability of exciting the fimdamental mode of the surface optical phonon in MgO [13], and the result is shown in Fig. 1. This dependence highlights the first tool available in the experimentalist arsenal for reducing the influence of the phonon modes when attempting to achieve the sensitivity required for studying adsorbate modes. Unlike investigations on metal surfaces where Eo is kept at low values in order to maximise spectral resolution, Eo is generally held at significantly higher values on oxide surfaces in order to optimise the scattering probability in favour of adsorbate vibrations over the surface phonons.

100

Fig. 1. The scattering probability P measured experimentally [13] corresponding to the excitation of the fundamental mode of the MgO surface optical phonon as a function of Eo. The fitted curve represents the theoretical expectation that P((D) a Eo"^^^ (Eqn.2).

518

Because the coupling to the surface optical phonon modes is strong in HREELS, there is also a significant probability that successive loss events can take place during the scattering event, and strong multiple loss peaks are observed. It has been demonstrated both experimentally and theoretically [2, 14] that the total probability PN for an electron to excite N quanta, losing the energy Nhco during the scattering process, is given by the Poisson distribution (Eqn. 3). (3)

P N = (P^/N!)exp(-P)

P is the single (fundamental) phonon loss probability. This results in an even greater enhancement in the reduction of the intensity of the multiple loss features by the increase in energy predicted above. Eo = 46.2eV

MgO/CHjOH

8.5 off specular X46.7

1000

2000

3000

4000

waven umber / cm" Fig. 2. HREELS spectra taken from the clean (bottom panel) and methanol covered (upper panel) surface of epitaxially grown MgO(lOO) [13]. The two spectra in the lower panel are taken at different primary energies Eo demonstrate the drastic reduction in the intensity of the multiple loss features by increasing Eo. These spectra were acquired in the specular direction with the beam incident at 60° with respect to the surface normal. The spectra of the adsorbate covered surface in the upper panel were collected at Eo = 46eV, and using off specular (8.5°) detection: a) 180L CH3OH at 90K; b) flash to 160K; c) flash to 273K; d) flash to 380K; e) flash to 490K. Flashing to 160K (spectrum b) desorbs methanol multilayers. The weak features associated with the adsorbed methanol monolayer in the range 1000-4000cm'^ can be sensitively detected by off specular measurement at high Eo.

519

This is exemplified in the results obtained on MgO [13] where it was shown that the excitation of the multiple phonon modes was indeed predicted by the Poisson distribution. The overall effect of increasing Eo is best presented in the measurements of the spectra at two different energies (Fig. 2). At Eo = 46.2eV, all losses due to the excitation of the multiple quanta of the MgO surface phonon virtually disappear at frequencies above the third multiple loss. Indeed, P3 has decreased more than two orders of magnitude by increasing Eo from 3.2eVto46.6eV. The adherence to the Poisson distribution of the multiple loss peak scattering probabilities has lead to a generalised technique for the calculation of the multi-phonon structure in HREELS using a Fourier transform algorithm [15]. This involves the calculation of the single phonon loss spectrum P((o) using Eqn.2, and obtaining its Fourier transform (p). The complete multiphonon expectation is then obtained as the Fourier transform of exp(p). By inverting this procedure, a method is available for the removal of the multiphonon components of the HREELS spectrum. The effectiveness of this approach for the removal of the phonon contributions was demonstrated for NiO(lOO) [15], and this technique of phonon overtone deconvolution was extended to other studies of adsorption on single crystal Fe2O3(012) [16], ZnO(OOOl) andZnO(lOlO) [17], Cr2O3(001) [18], and TiO2(110) [19-23] where both phonon-phonon, and phonon-adsorbate combination losses where removed. The main criticism of this approach to the removal of the multiphonon components in HREELS spectra is that it is essentially an "indirect" technique. Of particular concern are the spurious features that can be introduced in the spectrum by the Fourier deconvolution procedure itself A further improvement in the direct technique of reducing the multiple phonon contribution to HREELS spectra by increasing Eo is to make measurements at off specular angles [13, 24], in the "impact" scattering regime. This reduces the coupling to the Fuchs-Kliewer optical surface phonon which, as pointed out earlier, is characterised by a small wave vector component parallel to the surface, and a long penetration depth. In the case of MgO [13] it has been demonstrated that the scattering probability Po decreases by over an order of magnitude at 18° off specular over the specular measurement (Fig. 3). It can be seen from Fig. 3 that this benefit decreases, however, for increasing multiple phonon loss (N). In addition, the benefit of working at off specular angles is reduced by the fact that off specular measurements are characterised by count rates which are unrealistically low for resolving adsorbate peaks. Nevertheless, off specular HREELS measurements carried out at high incident energy have been shown to be sufficiently sensitive for obtaining vibrational spectra of adsorbates at monolayer coverages on oxides. A spectrum is shown in Fig. 2 for multilayer and monolayer coverages of CH3OH measured at an off

520

6

I . I I . I 8 10 12 14 16 18 oflf specular angle / degrees

Fig.3. The intensity of the phonon loss and its multiples as a function of ofF-specular angle in HREELS on epitaxial MgO(lOO), Eo = 3.2eV [13].

specular angle of 8°. Clearly adsorbate modes are observable in monolayer coverages in the range 1000-4000cm ^ It should also be noted, however, that the surface acoustical Rayleigh mode, which are dipole inactive and hence do not couple strongly to the incident electrons and hence contribute in specular HREELS, can appear in off specular measurements through impact scattering [25]. It is much less easy to establish the advantages of thin film substrates over single crystal oxide surfaces in carrying out HREELS on adsorbates at the metal oxide surface. One potential advantage of the thin oxide film is in reducing the amount of charging which may result in electron spectroscopies of insulating oxides. A second may be any reduction in the intensity of the surface optical phonon mode, which propagates more deeply in the bulk material (ca. lOnm) than the thickness of many epitaxially grown thin film oxide surfaces. It is also clear that the metal substrate on which such oxide films are grown will screen parallel components of dynamic dipoles, but this aspect of the differences between single crystal oxide measurements and thin film oxide measurements has manifested itself only in RAIRS measurements. Nevertheless it appears that a similar number of HREELS measurements of adsorbates have been made to date on thin film [13, 26] [27] [13, 28-30] and single crystal oxide substrates [16-23].

521

2.2.

Reflection Absorption Infrared Spectroscopy (RAIRS) RAIRS has proved to be a powerful vibrational spectroscopic technique for the study of adsorbates on metal surfaces, allowing not only the identification of the surface species, but also information concerning molecular geometry and chemical environment. The application of RAIRS to measurements on single crystal metal surfaces has been the subject of a number of reviews [1, 3-8, 31, 32], and both the theoretical and experimental aspects have been discussed extensively for measurements on metals. The extension of RAIRS to oxide surfaces requires foremost a consideration of the difference in optical response of the substrate. This aspect had already been examined in attempts to extend RAIRS to measurements on semi-conducting surfaces [7, 3235], which have a similar optical response to many metal oxide surfaces through much of the IR. In the case of a metal substrate, the dielectric response of the substrate in the IR results in high reflectivity, and the screening of the parallel components of the net incident field and the dynamic dipole of the adsorbate. The result is that in addition to the dipole selection rule, absorption of radiation is governed also by the so-called surface selection rule, which requires a change in dipole perpendicular to the metal surface. In the case of a semi-conducting surface that is non-absorbing, reflectivity is low but parallel components are not screened. In order to consider these extreme regimes for RAIRS, and the possible intermediate regimes for metal oxide surfaces, it is instructive to first consider the isotropic three layer mode [36, 37].

i

Z

Pt

^^^—

>^

Ss Fig. 4. Reflection of IR radiation at a surface, showing incident vectors for P- and Spolarised radiation, and the components of the P-polarised radiation at the surface.

522

Consider an IR beam incident on the surface at an angle (p with respect to the surface normal (Fig. 4). The incident radiation can be resolved into components parallel (S-polarised) and normal (P-polarised) to the incident plane. The S-polarised radiation only has a component (S) parallel to the surface (in the y direction). However the p-polarised radiation has components parallel or tangential (Pt) to the surface, and perpendicular (?„) to the surface. Each layer (vacuum (e =1), adsorbate (e) and substrate (cs)) is characterised by an isotropic complex dielectric constant (e) which is defined as 8 = (n + ik/, where n in the refractive index, and k is the absorption coefficient. The change in reflectivity (AR) resulting from the adsorbate layer of thickness d for S- and Ppolarised radiation is usually expressed as a ratio to the reflectivity (AR/R): ARs =87id coscp Im fe - ER) Rs I (l-8s)

(4)

ARp =87cd C0S9 Im (e - eg) 0 - (\/ZER)(E + ER) sin^9) Rp X (l-8~s) (1 - (l/8s)(l + 8s) sin^G)

(5)

Given that, for a metal |8s| » |8| ~ 1, and d/X « 1 for IR frequencies of ~10"^m and d « 0.1 nm, it can be seen that the reflectivity change of S-polarised light is very small. In order to consider the effect of change in the optical response of the substrate between the two extremes of a metal (ks=30), and a non-absorbing (ks=0) or slightly absorbing (ks=0.05) semi-conductor or insulator. Fig. 5 shows the values of AR/R as a function of (p calculated using equations 1 and 2 for the various values of ks , with 8s = (3+ ksi)^. A moderately absorbing adsorbate layer has been assumed, with 8 = (1.3+O.li)^. AR is defined as R-Rads, where R and Rads are the reflectivities of the clean and adsorbate covered surfaces respectively: A positive value of AR corresponds to an adsorbate induced absorption band. Such a calculation shows why for a metal substrate (ks = 30), there is no coupling of the dipole to the parallel S component of the radiation, and the optimum incident angle for RAIRS using P polarised radiation is cp = 80-90° [36, 38, 39]. Under such conditions, one can expect an absorption band (AR/R is positive) at the adsorbate resonant frequency. The results in Fig. 5 show that for RAIRS on an oxide surface (ks » 0), a complex behaviour of AR/R is exhibited. The most significant influence of the absorbing overlayer on the P-polarised radiation is observed on either side of the Brewster angle, cpe . The absorbing overlayer will induce a reduction in reflectivity (absorption band) below cpe, and an increase in reflectivity (transmission band) above cps . S-polarised radiation also interacts with the absorbing overlayer, all be it weakly, with an increase in reflectivity induced at all (p. For P-polarised radiation, maximum sensitivity in RAIRS would therefore be expected just above or below cpB with ks=0. A small amount of absorption by

523

the substrate (ks = 0.05) results in a retention of the sensitivity for cp > cpB, but a reduction for (p > (ps. Typical values [40] of the optical constants for metal oxides in the range 1000-3000 cm'^ are ns=1.5-3.5, and ks=0-0.07, similar to silicon, with stronger absorption at energies < 1000cm"\ 0.003 1

0.002

ks=30

0.00000 -0.00002 H

f^
-0.00004

-0.00006 -0.00008 i -0.00010

Fig. 5. AR/R calculated as a fiinction of incident angle cp for S- and P-polarised radiation, and for various values of the substrate absorption coefficient ks, with ns = 3 using Equations 1 and 2. The optical constants of the adsorbate layer are n = 1.3 and k = 0.1, with d = 0.5nm and X = 5000nm. AR is defined as R-Rads: A positive AR corresponds to an absorption band, and a negative AR a transmission band.

524

The absolute reflectivity R of a semiconductor or oxide, however, is smaller (and more (p dependent) than a metal, and it has been suggested [35] that one should consider the value of AR rather than R in order to assess the optimum angle of maximum sensitivity in RAIRS on semi-conductors. This is because for low reflectivity, and hence small signal levels, random noise may limit the signal to noise ratio rather than total intensity noise [4, 41]. The effect is to push the optimum angle for P-polarised RAIRS away from either side of (PB- In the case of Si, it has been suggested that the optimum angle for RAIRS is in the range 80°< cp < 87° for P-polarised radiation [34]. In experimental measurements using external reflection [35] on Si, an angle of (p = 85° was used, and on GaAs(lOO), cp = 80°. External reflection has also been used for the study of spontaneously adsorbed layers from solution (ethyl xanthate) on a semiconductor (cuprous sulphide) at various incident angles (p [42]. S-polarised radiation will clearly interact most weakly at grazing incidence (Fig. 5), and parallel components will be best probed at more normal incidences by reflection, of transmission through the sample. At an energy where a substrate is not absorbing e.g. > 900cm"^ on Si, it has been demonstrated that internal reflection methods produce higher sensitivity [34], and the majority of IR measurements on silicon, for example, are carried out using such an approach [33, 34, 43-45]. It is clear, however, that the presence of any absorption in the substrate will restrict measurements to external reflection. The extension of phonon absorption to higher energies on many metal oxide surfaces makes the application of internal reflection measurements less attractive on such surfaces. An alternative on semiconductor surfaces has been to use metal ion implanted substrates which exhibit dielectric propertied in the IR more similar to metal substrates [46-50]. Under such conditions, the conventional RAIRS geometry for metal surfaces of P-polarised radiation at grazing incidence is optimal, with strong screening of parallel components and absorption bands at the resonant frequency. An important consequence of carrying out RAIRS on a substrate exhibiting no absorption, or only weak absorption, is that coupling to both parallel and perpendicular components of the surface dynamic dipole can take place. This is evident clearly in the case of S-polarised radiation (Fig. 5) which predicts a small transmission band for all values of (p. For P-polarised radiation, the isotropic three layer model (Fig. 5) predicts a transmission band at the resonant frequency for (p > Brewster angle, and absorption bands for cp < (pB. This simple model, however, is insufficient to separate the coupling of the normal Pn and tangential Pt components of the IR field to the isotropic adsorbing overlayer (Fig.4). It is also ineffective in treating the anisotropic and microscopic characteristics of molecular adsorption, such as molecular alignment or dipole coupling.

525

The microscopic characteristics of a real adsorbate layer have been considered [33, 34] by separating the x, y and z components of the electric field at the interface (Fig.4), and applying the Lorentz oscillator model to microscopically represent the adsorbate in the three-layer model. For the case of external reflection at a vacuum/semi-conducting, where Ss is real (no absorption) and isotropic, w e can write: ARS = Rs

271V ly (8"ydy) coscp

(6)

ARp = .iTiv. {Ix (8"xdx) + Iz _ s s ' _ ( s " z d z ) } Rp

coscp

(7)

8'z^ + S"z^

Where Ix,y,z are the corresponding components of the field intensities: ly = 4cos^(p

(8)

1-8S

Iz = 4cos^(p l-8s

l/8s sin^cp {((l+8s)/8s) sin's-1}

(9)

Ix = 4cos'(p 1-Bs

WER sin'cp-n {((l+8s)/ss)sin'(p-l}

(10)

While the three layer "macroscopic" model (Eqn.4,5 and Fig.5) allows a calculation of AR/R for substrates in which 8s is complex, there are no simple explicit expressions corresponding to Eqn. 6,7 for complex 8s, and in practice normal reflection or transmission has been used for which the equations are greatly simplified [51, 52]. Nevertheless, Eqn. 6-10 can be considered a reasonable approximation for non-absorbing and weakly absorbing substrates. The separation of the field components, and the expression of anisotropic adsorbate dielectrics ( 8"x, 8"y , 8"z ) becomes usefiil when 8" is expressed in terms of meaningfiil microscopic quantities (eg. The dynamic charge of the adsorbate mode), and the Lorentzian oscillator [53] parameterisation of an adsorbate layer has proved usefiil [34] for this purpose. For any particular vibrational mode, the effective dielectric fimction 8(co) is given by: 8"(co) = 8oo + cop' coo^ - 0)' - iyco

(11)

526

(p = 45"

^ 4.00E-04

20|50

2070

2090 \

wavenumber/cm*

/2110 \[^

2130

2150

S

-8.00E-04 l.OOE-03

9 = 80" 20|50

2070

2090 1

2^0

wavenumber / cm'

Fig. 6 Model calculations (employing the Lorentzian ocillator approximation and a three layer optical model) of AR/R for RAIRS on semiconducting or insulating (isotropic and nonabsorbing) substrates (8s=3) for P- and S-polarised radiation. Calculations are shown for two values of incident angle ((|)), below and above cpe The adsorbate layer is assumed isotropic (8"x=8"y=8"z) with e*/e=0.5, v=2100cm"\ Y=5cm'\ The convention AR=R-Rads is used, so that positive resonances correspond to absorption bands, and negative values transmission bands.

Where Coo is the electronic part of the dielectric function, cop is the plasma frequency associated with the adsorbate layer, and y is the natural line width of the oscillator. Fig. 6 shows the resonance associated with the coupling of the Spolarised radiation to the y component of an oscillating dipole (S), and the coupling of P-polarised radiation to the x and z components through Pt and Pn

527

respectively, calculated using Eqn. 6-11. For the purpose of the calculation it has been assumed that the adsorbate layer is isotropic with a dynamic charge e*/e = 0.5 to represent a mode with a high extinction coefficient such as v(C-O). Calculations have been made at two incident angles, one below (cp = 45°) and one above (cp = 80°) (pB. The advantage of such an approach is that it starts to take account of the orientation of the adsorbate dynamic dipole in addition to the dynamical interactions resulting from dipole coupling in a thin layer or monolayer. A spectral shift from the singleton frequency can be expected in coupling to the z (normal) oscillator component as a result of the net dipole coupling of a two dimensional array of parallel oscillators [54]. This effect shows up as a thickness (d) dependence in the optical response in the z direction of the adsorbate layer (Eqn.7), within the modified three layer model. For a strongly adsorbing monolayer such as CO, this shift is calculated (Fig. 6) to be -18cm'\ in agreement with Chabal [33, 34]. The result is that for an isotropic adsorbed thin film on an oxide surface where the parallel mode may be observed, the parallel and normal components will be observed in RAIRS at different frequencies, and will be manifested as either absorption or transmission bands depending on (p (Fig.6). Measurement with S-polarised radiation at (p = 45° or (p = 80° (either side of (pe) will result in transmission bands (Fig. 6), as predicted also by the simpler model (Fig. 5). For P-polarised radiation at (p = 80°, the coupling of Pn and Pt to the adsorbate oscillator results in a transmission and an absorption band respectively. This situation is expected to be reversed at cp = 45°, below (p B. Owing to the relatively low reflectivity of semiconductor surfaces, and hence the low detected signal expected for the external reflection experiments, very few RAIRS studies of adsorbates on semiconductors have been performed, and most experiments have been carried out using internal reflection (MIRS). However, studies using P polarised radiation on the hydrided Si(lOO) surface at (p=85'' [35], and unpolarised IR of trimethylgallium deposited on GaAs(lOO) [55] confirm the existence of transmission bands predicted theoretically on semiconductors. For metal oxide surfaces, where the substrate is isotropic and nonabsorbing or weakly absorbing, the microscopic model predictions concerning the type of band and shifts for the coupling of the various polarised components (Fig. 6) will be valid. One may therefore expect (with cp > CPB) that for a thin film of an isotropic physisorbed layer, coupling of S or Pt to an adsorbate mode with a high extinction coefficient will result in transmission and absorption bands respectively at a frequency similar to the solid state value, as a result of the "infinite" dimensions of the film parallel to the surface. The coupling of the component Pn will result in a transmission band at a higher frequency (Fig. 6). Fig. 7 shows an FT-RAIRS spectrum taken [56] of a isotropic physisorbed multilayer of Rh2C04Cl2 on Ti02(l 10) with cp = 83°.

528

T 1 r""T 1—I I 2030 2120 2100 2080 2060 2040 2020 wavenumber / cm'^ Fig.7 FT-RAIRS spectrum of Rh2C04Cl2 on TiO2(110) [56]. Measurements are carried out at (p=83°, and can be compared directly with the predictions of the model used for an isotropic film on an isotropic, non-absorbing substrate (Fig.6). The lines indicate the expectation values for the modes in the condensed phase.

As predicted by the calculation, the absorption features of the P-polarised spectrum (coupling to Pt) correspond to the transmission features of the S polarised spectrum, and also correspond to the expectation values for the assigned modes of the bulk crystalline material [56]. The transmission features of the P-polarised spectrum (coupling to Pn) correspond to the blue shifted bands associated with a net dipole coupling of the normal components of the dynamic dipoles in a thin film. The microscopic model, however, cannot take into account net coupling of dynamic dipoles oriented parallel to the surface for the thin (microscopic) film. Such coupling in adsorbed monolayers has been shown [57] by probing an otherwise disallowed transition on a metal through a combination band, to result in a red shift from the singleton frequency. This effect of parallel and normal dipole components can be best exemplified by comparison of the RAIRS spectrum of an isotropic physisorbed molecule with a very strong dipole oscillator, v(C-O) in Mo(CO)6, with the gas phase value (singleton frequency) [58]

529

7 / / / /"Z

ttttftttttttt ttttttttttttt Fig. 8 FT-RAIRS spectrum of a physisorbed layer of Mo(CO)6 on TiO2(110) [58]. The net dipole coupling in the thin film shifts the singleton (gas phase) frequency coo either up or down in energy for normal and parallel components respectively.

This is shown in Fig. 8. The coupling of P polarised radiation to the normal and parallel components of the oscillator (the Tiu v(C-O) mode) results in the expected transmission and absorption bands respectively. The former is blue shifted, and the latter is red shifted. This can be understood simply by considering the dynamic dipole coupling lattice sum expected for aligned oscillators perpendicular and parallel to the surface [59], represented schematically in the inset of Fig. 8. The difficulty in carrying out RAIRS, with respect to metals, on metal oxides and semiconductors is that reflectivity and sensitivity are low. The advantages of RAIRS on metal oxides results from the fact that coupling of Ppolarised radiation to can take place to both normal and parallel components of the adsorbate dynamic dipole, not only in isotropic physisorbed layers, but more importantly in oriented chemisorbed monolayers. In addition the modes can be separated as a consequence of the net optical response (transmission or absorption bands), and the adsorbate geometry becomes accessible through the characteristic of both components. It will also become evident that since the parallel component of dynamic dipoles are IR active, the azimuthal dependence of the RAIRS spectrum can be used to obtain the alignment of adsorbed species. The number of RAIRS experiments carried out to date on single crystal oxide surfaces is relatively small. An alternative approach to studies on oxides is to use metal oxide substrates epitaxially grown on metal substrates. RAIRS on such systems is characterised by the dielectric response of the metal since the oxide films are thin with respect to IR wavelengths. The resuh is that only the

530

normal components of the dynamic dipole couple to the radiation, and all bands are absorption bands. The advantage of this approach is that sensitivity is concomitantly higher. In principle, the technique of enhancing RAIRS signals by metal ion implantation, a technique used for semiconductors [46-50], can be adapted to single crystal metal oxide systems. Indeed this approach has been applied to the study of SiOi supported CuO model catalysts on a metal implanted Si substrate [60]. The three-layer theory can, in principle, be extended to the more complex, but perhaps more interesting case with respect to modelling heterogeneous catalysis, of RAIRS on single crystal oxide supported metal particles or thin films. While no explicit calculations are available to date, it has been shown experimentally that RAIRS of adsorbates on small particles supported on oxide surfaces result in the dominance of the metal oxide dielectric response. As particle size (or film thickness) increases, the metallic response eventually dominates resulting in strong RAIRS absorption bands. The particle size or film thickness at which such a transition takes place is discussed below. RAIRS does provide a tool for differentiating between adsorption at the oxide itself, and on small and large particles supported on the oxide.

3.

SURFACE VIBRATIONAL MODES OF THE SUBSTRATE

The "surface" Fuchs-Kliewer modes, like the Rayleigh modes, should be regarded as "macroscopic" vibrations, and may be predicted from the bulk elastic or dielectric properties of the solid with the imposition of a surface boundary condition. Their projection deep into the bulk makes them insensitive to changes in local surface structure, or the adsorption of molecules at the surface. True localised surface modes are those which depend on details of the lattice dynamics of near surface ions which may be modified by surface reconstruction, relaxation or adsorbate bonding at the surface. Relatively little has been reported on the measurement of such phonon modes, although they have been the subject of lattice dynamical calculations [61-67]. Localised phonon excitations are in principle best studied by neutral-atom scattering, or off specular HREELS, in order to reduce the strong dipole excitation of Fuchs-Kliewer modes. Two off specular HREELS measurements on MgO(lOO) have been reported [25, 68], however there is some disagreement concerning the energy and assignment of the substrate derived loss peaks. Since the microscopic surface modes are expected to be sensitive to the surface structure, it has been suggested [9] that the differences may be associated with differences in surface preparation. Perhaps the most clear cut evidence for the observation of a localised surface phonon comes from a recent HREELS study a clean and adsorbate

531

covered CriOaCOOOl) surface [67]. Fig. 9 shows the specular HREELS spectrum of CriOsCOOOl) measured with Eo=7.5eV. The CriOsCOOOl) film was grown as a 40A film produced through the oxidation of Cr(l 10). Spectra are shown for the clean Crz^^OsCOOOl) and Cr2^^O3(0001) surface. The spectrum of the clean Cr2^^O3(0001) surface is characterised by losses at 21.4, 51.7, 78.6, 85.0 and 88.5meV, and combinations of these losses at higher energies. The latter four are identified as Fuchs-Kliewer phonon modes, and the intensity and energy of these modes are found to be uninfluenced by the adsorption of CO or O2 at 90K.

84.6

88.5

Cr2 0 3

^lOcr/^03 I

0

"T" 50

^

I

100 150 energy / meV

r^ 200

Fig. 9. The HREELS spectrum of clean Cr2^^O3(0001) and Cr2^^O3(0001) [67]. The loss at 21.4meV in the spectrum of Cr2^^O3(110) disappears on exposure to CO or O2, and is ascribed to a localised surface mode. Its isotopic shift of 2.4% is predicted theoretically, and is significantly smaller than that observed for the Fuchs-Kliewer modes of 4.5-6%.

532

Similarly the combination of two of these modes at 139.8meV (88.5+51.7meV) does not change on CO adsorption. In contrast, the loss at 21.4meV is attenuated with increasing CO and oxygen adsorption, and it is this attenuation which is given as evidence of a localised surface mode of the substrate. A weak loss peak observed at ca 40meV appearing at high CO coverages was associated with either the substrate-CO mode, or the CO modified Cr2O3(0001) counterpart of the clean surface localised surface phonon. The rather weak band at ca. llOmeV was attributed to a combination of the SS.SmeV Fuchs-Kliewer mode and the 21.4meV localised surface phonon, and consequently also disappears on CO adsorption. The small loss at llSmeV was assigned to oxygen atoms adsorbed (following diffusion from the bulk) at terminating chromium ions. Fig.9 shows that localised phonon mode at 21.4meV was also observed to undergo the smaller (theoretically predicted) isotopic shift than the Fuchs-Kliewer modes.

4.

VIBRATIONAL SPECTROSCOPY OF ADSORBATES

Both HREELS and RAIRS have been applied extensively of the study of adsorbates on metal surfaces. The extension of the techniques to semiconducting or insulating oxide surfaces is hampered by a number of problems. The result is that until even relatively recently [9] there were only a couple of examples of RAIRS studies on oxides, and these were confined to polycrystalline systems. Most early HREELS studies were concerned with the characterisation of the intrinsic phonon modes of metal oxide surfaces. This contrasts strongly with the extensive literature concerning the vibrational characterisation of adsorbates and intermediates on powdered oxide surfaces that have been obtained by transmission or diffuse reflection IR techniques. Both techniques will be influenced by the appearance of strong intrinsic phonon bands. This has been less of a problem for RAIRS measurements when confined to energies higher than the fundamental optical phonons. The intrinsic resolution of RAIRS also ensures that there is little contribution of absorption due to the phonon modes away from their oscillator frequency. In contrast the strong coupling of the scattered electron in HREELS to the phonons results in the excitation of not only the fundamental modes, but also their overtones, and these lie close in energy to even intramolecular adsorbate vibrational modes. This, coupled with lower spectral resolution, results a requirement to subtract the underlying phonon signal from the adsorbate spectrum in order to identify adsorbate bands. There are a number of methods that have been developed to deconvolute phonon overtones from HREELS spectra so that adsorbate losses can be observed (Section 2.1).

533

Fig. 10. FT-RAIRS spectrum of rhodium dicarbonyl Rh(C0)2 on TiO2(110) [56, 69, 70] following dissociative adsorption of Rh2C04Cl2 (a). The molecule is reformed after desorption of the CO at 450K following exposure to lOOL of CO (b). Reaction of Rh(C0)2 with hydrogen forms the monocarbonyl Rh(H)CO [71].

Because of the additional problem of sensitivity in RAIRS measurements made on semiconducting or insulating oxide surfaces, it is perhaps not surprising that in parallel to the history of RAIRS measurements on metal surfaces, the first measurements on a single crystal metal oxide surface involved a molecule with a carbonyl stretch. The dissociative adsorption of the MOCVD precursor Rh(CO)4Cl2 on TiO2(110) results in the chemisorption of the well known rhodium geminal dicarbonyl species Rh(C0)2 [56, 69, 70]. This species had been formed through the same route or through the reaction of CO with supported Rh^ particles, and characterised extensively, on high area powder oxides by IR because of its importance in heterogeneous catalysis. Two nondegenerate v(C-O) modes, the symmetric and anti-symmetric stretches, are expected for the oxide supported species at Vsym(C-O) » 2110cm"^ and Vasym(CO) « 2030 cm"^ The Rh(C0)2 species was formed exclusively on reaction of Rh(CO)4Cl2 with TiO2(110) at 300K to form 1/3 of a monolayer, and

534

characterised by a combination of XPS, TPD and FT-RAIRS [56, 69, 70]. The FT-RAIRS spectrum is shown in Fig. 10. The spectrum of Rh(C0)2 on TiO2(110) exhibits the two expected bands for the mono-dispersed geminal dicarbonyl, with Vsym(C-O) = 2112cm"^ observed as a transmission band, and Vasym(C-O) =2028cm"^ observed as an absorption band. Since the incident angle cp = 83° and the incident radiation was P polarised, the result is consistent Rh(C0)2 bound to TiO2(110) with C2v symmetry. The analysis of the theoretically expected changes in reflectivity due to adsorbates on a semiconducting or insulating substrate is given in Section 2.2. There is a coupling of Vsym(C-O) with Pn (AR is expected to be -ve, i.e. an increase in reflectivity following adsorption), and a coupling of Vasym(C-O) with Pt (AR is expected to be +ve, i.e. a decrease in reflectivity following adsorption). The advantages of RAIRS on the metal oxide surface are evident from this measurement in that modes polarised both normal and parallel to the surface can be observed, and can be differentiated by appearing as transmission or absorption bands in the RAIRS spectrum. A combination of FT-RAIRS, XPS and TPD allowed a study of the stability and reactivity of Rh(CO)2 on TiO2(110) [56]. Fig. 10 shows that following the thermal decomposition of Rh(CO)2 (a), the molecule can be reformed through carbonylation at 300K (b). The reaction with hydrogen produces a monocarbonyl Rh(H)CO in which v(CO) = 2065cm"^ is polarised normal to the surface, and couples to Pn resulting in a transmission band [71]. Since the parallel components of the dynamic dipole are active in RAIRS, it is possible to use the azimuthal dependence to obtain the orientation of the adsorbate at the surface. A similar technique has been applied to adsorbates on metals in HREELS measurements made off specular in order to observe parallel modes through impact or resonant scattering processes. This was first demonstrated for the Rh(CO)2 molecule on anisotropic TiO2(110) surface [72]. The results of this study also allow a test of the three layer model theory (Fig.5,6) as applied to S-polarised radiation. Fig. 11 shows the FT-RAIRS spectrum for 1/3 monolayer of Rh(C0)2 on TiO2(110) measured with P and S polarised radiation. Measurements have been made as a function of azimuthal angle, where d=0°corresponds to the incident radiation being incident along the <110> azimuth. Note that this differs from the definition used in [72], but is chosen for consistency with the convention used elsewhere for FT-RAIRS measurements of formate on TiO2(110) [73] shown below. The angle of incidence cp = 83°. The mode polarised normal to the surface Vsym(C-O) only couples to the Ppolarised radiation through Pn (transmission band). The mode polarised parallel to the surface Vasym(C-O) couples to the P polarised radiation through Pt (absorption band), and also to the S-polarised radiation. The latter gives rise to a transmission band, as predicted by the three layer model (Fig. 5,6).

535

titanium

—I

1

2160

1

1

2100

1

1

2800

1

r

2400

2000

1960

wavenumber / cm

<110>

Fig. 11. Azimuthal dependence of FT-RAIRS spectra for TiO2(110)-Rh(CO)2 [72]. The azimuthal angle (j) is defined as 0° when the incident radiation is aligned in a plane parallel to the <110> direction. The Vsym(C-O) dynamic dipole is aligned normal to the surface and couples to Pn (transmission band), and Vasym(C-O) is aligned parallel to the surface in the <110> direction, and couples to Pt (absorption band). Two possible adsorption geometries consistent with the FT-RAIRS azimuthal dependence are shown for the gem-dicarbonyl.

The azimuthal dependence of the intensity of Vasym(C-O) in the Ppolarised radiation shows a maximum at d=90° indicating an ahgnment of the Rh(CO)2 in the <110> direction. Since the S and Pt fields are orthogonal, using S-polarised radiation at S = 0° Vasym(C-O) is not observed, but is observed at S = 90°. The two most likely adsorption geometries of the adsorbed gem-dicarbonyl are shown in Fig. 11, both with the C-O bonds in a plane aligned in the <110> direction. The Rh(C0)2 molecule has also been observed in FT-RAIRS studies of AI2O3 supported Rh particles [74]. This provides an interesting comparison with the single crystal TiOiCllO) measurements, since the alumina substrate was grown a thin film on NiAl(l 10). The film was sufficiently thin to ensure that the optical response in the IR at the oxide surface was dominated by the metallic substrate. Fig. 12 shows the FT-RAIRS spectra taken for CO adsorbed on metal particles of increasing size supported on the alumina film. For the smallest particle sizes, a sharp band at about 2120cm'^ is observed, and is associated with Vsym(C-O) of dispersed Rh(CO)2. The corresponding Vasym(C-O) band is not observed because of the screening of the parallel modes.

536

v^™(C-0) 9 13 20 30 55

100 200 450 720

I I

2000

I

I I I I I I j I I I I I I

1900 1800 1700 wavenumber/ cm"*

Fig. 12. A series of RAIRS spectra taken following deposition of increasing amounts of Rh on an alumina film grown on NiAl(llO) and subsequent CO saturation at 90K[74]. The average number of Rh atoms per particle is indicated. Vsym(C-O) of the gem-dicarbonyl is observed (indicated) as an absorption band at low Rh coverages, while Vasym(C-O) is screened by the underlying metal substrate. The large broader band observed at higher Rh coverages (particle sizes) results from v(C-O) of CO adsorbed on Rh° particles.

The broad band observed around 2080cm"\ predominating at higher particle sizes, is due to CO adsorbed on Rh^ particles. The RAIRS spectra obtained on oxide supported metal particles (single crystal oxides and thin film oxides) is discussed in detail in Section 5. A number of HREELS measurements of adsorbates have also been carried out on TiO2(110) single crystal surfaces [19-22]. One of these, the study of adsorbed formic acid [20], provides a useful comparison of HREELS and FTRAIRS on single crystal oxides. Fig. 13 shows the HREELS spectra at increasing exposure for formic acid adsorbed at llOK on TiO2(110). The spectra presented (including those of the clean surface) have been Fourier deconvoluted [23] in order to remove the contributions of multiple phonon losses. The positive and negative features in the background spectrum signify a less that ideal deconvolution of the data, probably a result of less than ideal spectrometer tuning and/or sample positioning [20]. It is such features, and the fundamental phonon features, that distort the adsorbate spectra, particularly in the spectral range <2300cm"^ where many molecular vibrations are present.

537

multiple phonon remnant V3y,(0C0)

^'

1

1 1

1

v(C-H) ->^,,

1 li

1 1

I

X600

^^"^

c

:i

1 '/

^-

b

\

y[

J L/^

V\

V

x25

.'—^

N_/

1 a

H""" 1 1 1 1 1 1 1 1 1 1 T'f'l I I I I T i i i i | i i i i i i i i i | i i i i i 111 ¥ 1 1111 11

0

1000

2000

3000

1 1

4000

wavenumber / cm"^ Fig. 13. HREELS spectra of the formate species adsorbed on Ti02(l 10) taken from [20]. The clean surface (a) was exposed to (b) 1.6x10^"* and (c) 2.4x10^"^ molecules cm"^ of HCOOH at 11 OK. The Vsym(OCO) and v(CH) of the adsorbed formate species are indicated. At higher exposures, the modes associated with the physisorbed parent molecule were observed (not shown).

Nevertheless, bands associated with the adsorbed formate species, Vsym(OCO) = 1365cm"^ and v(CH) = 2920cm"^ where identified at intermediate exposures, and additional bands ascribed to formic acid observed at higher exposures. The residual signal from multiple phonon losses and/or limitations of spectral resolution (Fig. 13) prevented the observation of the corresponding Vasym(OCO) modc of the adsorbed formate species on TiO2(110) [20]. Since the measurement was carried out on a single crystal sample of the oxide, the absence of screening electrons should have made the mode dipole allowed even if the formate was adsorbed in a Civ geometry. Both Vasym(OCO) and Vsym(OCO) could be observed, however, in FT-RAIRS measurements, although limited sensitivity prevented the observation of v(CH) [73]. The FT-RAIRS spectra of the formate species on Ti02(l 10) is shown in Fig. 14 as a function of azimuthal angle ((])), with (|) = 0° corresponding to radiation incident in a plane aligned in the <110> direction.

538

^

klv,asymi

j In 'Vsym

::inHU

oxygen (top row) P ^ formate • titanium o hydrogen

<110>

In/v.asym 't /Vasvm vasym

k 't /Vsym

t!L_A —I—I—I—I—I—m—I—I—r

2000 1800 1600 1400 1200 1000

=

"

^

wavenumber / cm"^

Fig. 14. FT-RAIRS spectra measured as a function of azimuthal angle ^ of formate adsorbed on TiO2(110) [73]. The absorption band, corresponding to the coupling of Vasym(OCO) to Pn, is observed at ^= 90° through to ^= 0° because of the adsorption of two formate species (A and B) aligned orthogonal to each other; and the difference in bonding site is responsible for the spectral shift with ^. The transmission band corresponds to Vsym(OCO) coupling to Pt, and should remain independent of (j). However, while species A has C2v symmetry, species B has a lower symmetry, resulting in the coupling of Vsym(OCO) to Pt for species B at (t)=0°. The appearance of this additional (small) absorption band at ^=0°, and its disappearance at (|)=90° resuhs in the azimuthal changes in the transmission band Vsym(OCO) (schematic spectrum).

All spectra in the v(OCO) spectral region are characterised by two distinct vibrational features. The absorption band, shifting from 1535cm"^ to 1566cm"^ with (|), corresponds to Vasym(OCO) coupling to Pt. The transmission band, at 1380cm'\ corresponds to Vsym(OCO) coupling to ?„. The observation of Vasym(OCO) at ^= 90° through to ^= 0° lead to the conclusion that there were two types of formate species (A and B in Fig. 14) with orthogonal alignment on TiO2(110), a possibility predicted theoretically [75]. The spectral shift of Vasym(OCO) with (|) rcsults from the difference in chemical environment of the

539

two species: Pt couples to Vasym(OCO) of species A at (|)= 90°, and to Vasym(OCO) of species B at (|)= 0°. These two formate species could not be distinguished in HREELS (Fig. 13), but their co-existence provided an explanation [73] for the two decomposition channels for adsorbed formate on TiOiCl 10) [20]. One may have expected that the transmission band should remain invariant with (j), since ?„ couples to Vsym(OCO) of both species equally at all (j). In fact the doublet structure in the transmission band was ascribed to the presence of the two formate species. However, while species A has Civ symmetry, the symmetry of formate species B is lower because of the incorporation of one of the formate oxygen atoms in the TiOiCllO) top rows (Figl4). The result is that Vsym(OCO) for B also has a component parallel to the surface which can couple to Pt, resulting in an absorption band. This coupling is possible at (()= 90°, but not at (|)= 0°, resulting (because of its coincident energy) in the apparent ^ dependence in the Vsym(OCO) transmission band. On single crystal oxides, FT-RAIRS studies have been restricted to adsorbates with strong dynamic dipoles [56, 58, 69-73, 76-78]. A wider variety of adsorbates has been studied on thin films of Cr2O3(0001) [27, 28], NiO(lOO) [79], MgO(lll) [80, 81], ordered alumina [29], amorpous silica [80, 82] and ceria [83] because of the increased sensitivity. A good example of an FTRAIRS study of an adsorbate mode with a weaker dipole is that of v(O-H) of adsorbed hydroxl groups on ordered alumina films [29]. This study also provides an interesting comparison of FT-RAIRS with HREELS sensitivities. The v(O-H) band is observed with very similar signal to noise ratio with the two techniques, but at 3750cm'^ is far from the underlying signal in HREELS resulting from the alumina phonon modes.

5.

VIBRATIONAL SPECTROSCOPY OF ADSORBATES ON OXIDE SUPPORTED METALS.

The reactivity of oxide supported metals has received considerable attention because of the importance of such systems in heterogeneous catalysis. The morphology (structure and size) of the supported particle and its stability, the interaction of the particle with the support, and the crossover of adsorbed reactants, products and intermediates between the metal and oxide phases are all important in determining the overall activity and selectivity of the system. Because of the relative insensitivity of an optical technique such as IR to pressure above the catalysts, and the flexibility of transmission and diffuse reflection measurement techniques, vibrational spectroscopy has provides a considerable amount of information on high area (powder) oxide supported metal surfaces. Particularly remarkable was the pioneering work of Eichens and Pliskin [84] in which adsorbed CO was characterised by IR spectroscopy on

540

supported Ni, Pt and Pd catalysts, using metal carbonyl complexes to deduce adsorption configuration. The extensive vibrational spectroscopic literature now available on single crystal metal surfaces now provides a more powerful database for spectroscopic comparisons [3, 4, 8]. IR spectroscopy has thus enabled the identification of adsorbates and reaction intermediates on such surfaces, and through the use of "probe molecules" has allowed a characterisation of the metal particle itself The extension of such studies of structure and reactivity relationships to metals supported on epitaxially grown thin oxide films and single crystal oxide surfaces provides more exact control of what can be a complex and heterogeneous system. Vibrational spectroscopy, when combined with the increasingly large arsenal of techniques available to the surface scientist, continues to provide a powerful tool for the characterisation of such systems.

Linear

90K

(lll)Bridge

(edge) Bridge

11 (edge)Bridge I I 1 I I I I I I 2200 2000 1800

wavenumber / cm'*

2200

2000

1800

wavenumber / cm'*

Pd/AI,O,/NiAl(110)

Pd/Si/Mo(110)

Fig. 15. FT-RAIRS spectra of CO adsorbed at 90K on alumina supported palladium particles, as a function of mean particle size [85]: The two sets of spectra correspond to particles formed by depositing palladium at 90K or 300K. FT-RAIRS spectra are also shown for CO adsorbed at lOOK on silica supported Pd (pre-exposed to oxygen) following the annealing of the surface to various temperatures [86].

541

A good example of the progression of vibrational spectroscopy on oxide supported metals can be found in the supported palladium system. Since the transmission IR studies of adsorbed CO on powdered oxide supported Pd by Eichens and Pliskin [84, 87], RAIRS has been applied extensively to the study of CO on various single crystal faces of Pd, and has now been applied to the characterisation of Pd supported on thin silica [86, 88] and alumina [85, 89-91] films, and single crystal TiOiCllO) [59, 77] surfaces using adsorbed CO as a "probe molecule". Fig. 15 shows the FT-RAIRS spectra obtained for CO adsorbed at about lOOK on Pd particles supported on silica [86] and alumina [85] thin films grown on single crystal metal substrates. In both cases Pd was deposited by metal vapour deposition, and the Pd coverage and the sample temperature were used to control the mean particle sizes. The morphology of the alumina supported Pd was characterised by a combination of SPA-LEED and STM [92, 93], and the mean particle size is given for each spectrum. The FTRAIRS spectra of CO on silica supported palladium [86] have been re-plotted in Fig. 15 so that a downward band corresponds to absorption, and the energy scale is similar to the silica supported palladium spectra to enable easy comparison of the bands. The position of the CO band was used to identify the facets of the Pd aggregates on the surfaces by comparison with single crystal data in both cases, although there remains some differences in the interpretation in the case of alumina supported palladium [85, 89]. In the case of palladium on alumina, the aggregates formed by annealing to 300K were shown by STM and SP-LEED to be three dimensional particles (40-60A) with well ordered (111) facets, and deposition at 90K produced smaller particles (10-20 A) with less ordered facets which were more difficult to define. The v(C-O) bands were assigned to bridged or linear species associated with either edge sites or exclusively the (111) terrace, where previously one of the band energies (1970-2000cm"^) had been associated with Pd(lOO) facets [89]. The development of the spectra with particle size (Fig. 15) clearly provide detail of the particle morphology. In the case of the silica supported palladium [86] the spectra were used to identify particle morphologies and adsorption sites which may exhibit different activity for CO oxidation, and hence spectra were recorder as a function of mean particle size (annealing temperature) on oxygen covered supported palladium. In this case the band at 2131cm"^ was identified as being associated with CO adsorbed on oxygen covered Pd(lOO) facets [94], and these are conspicuously absent on the larger particles (formed at higher annealing temperature). In both sets of data, it is again important to note that all vibrational modes manifest themselves as absorption bands fi'om the smallest to largest metal particle sizes. In the case of FT-RAIRS spectra of molecules adsorbed on the thin oxide films grown on metal substrates, it is the latter which dominates the local dielectric response in the IR for molecules adsorbed on even very small metal particles.

542

L/L*

Bj

Bj

1.2ML 300K

15-30A 1.2ML 500K


lOOML TOOK

2200

2100

2000

1900

wavenumber/cm' Fig. 16. FT-RAIRS of CO adsorbed at saturation coverage at 300K on TiO2(110) supported Pd particles [59, 77]. The bands associated with bridge bonded CO (Bi and B2) were used as indicators of particle morphology. Linear bands (L) were associated with edge or terrace sites. Transition bands were associated with small Pd particles, and absorption bands with larger Pd particles.

This contrasts with the case of FT-RAIRS on metal supported on single crystal oxide surfaces where the local dielectric response, and the associated electric fields in IR, depend on the particle size or film thickness. It should also be mentioned that on thin film oxide substrates, because the underlying metal will always screen the parallel fields, molecules adsorbed on particle facets which result in e.g. v(C-O) being orientated parallel to the macroscopic surface will be screened and not be observed in the RAIRS spectra. FT-RAIRS spectra of adsorbed CO on palladium supported on TiOiCllO) has been combined with photoemission and LEED used to deduce particle morphology [59, 77]. A series of spectra have been selected (Fig. 16) from CO coverage dependent measurements made on Pd particles produced at various

543

coverages and annealing temperatures to emphasise the effect of particle size on particle morphology. At low palladium coverages (1.2ML), transmission bands dominate the FT-RAIRS spectrum because the particles are sufficiently small (15-30 A) that the local dielectric is dominated by the TiO2(110) substrate, and the band corresponds to a coupling of the P-polarised radiation to dipoles orientated primarily normal to the macroscopic plane. The dominant band at 1980cm"^ (Bi) was ascribed to adsorption of CO on (100)/(110) planes. The effect of annealing the surface was to produce similarly small particles but with a morphology which was dominated by CO linearly adsorbed at edge sites (L*), and a small contribution at 1940cm"^ (B2) assigned to CO adsorbed in bridge configuration on (111) planes. Intermediate (lOML) and high (lOOML) coverages of palladium, corresponding to larger mean particle sizes (50-100 A and >100 A respectively) indicate a similar conversion of the Bi to B2 band on annealing, again interpreted as a conversion of (100)7(110) type facets to (111) facets. The "as deposited (300K)" lOML palladium layer exhibits absorption as well as transmission features. This was ascribed to the onset of the influence of the metal partial dielectric on the local field as the particle size increased. Note that annealing the particles and re-adsorbing CO produced only transmission features indicating the reversion to the TiO2(110) local dielectric as a result of an increase in the dispersion of the particles. Large particles (>100 A) produced by deposition of lOOML always produced absorption bands for adsorbed CO because of the dominance of the metal particle dielectric. In the case of FTRAIRS of adsorbates on single crystal oxide supported metal surfaces, in addition to the changes of morphology the appearance of transmission or absorption bands can give an indication of the mean particle size. The point at which the local dielectric for the adsorbate changes from that of the oxide to the metal has not been treated theoretically, and may be complex because of the particulate and heterogeneous nature of the supported metal. Besides the example above for TiO2(110) supported Pd, where the Pd particle sizes were estimated on the basis of final state photoemission effects and LEED [56], experiments have been carried out on thin films of rhodium on TiO2(110) deposited by MVD as a fimction of film thickness for comparison with Rh particles produced by the agglomeration of atomically dispersed Rh in the form of the rhodium gem-dicarbonyl [56, 69, 76]. Fig. 17 shows a series of FTRAIRS and XPS spectra following the deposition of Rh on TiO2(110) using the MOCVD precursor Rh2C04Cl2, subsequent heating of the surface to 500K and re-exposure to CO. Initial adsorption of the precursor produces the spectrum of the rhodium gem-dicarbonyl species (Rh(C0)2) with the characteristic Vsym(C-O) and Vasym(C-O) modcs (Fig. 10). Heating the surface to 500K results in the desorption of the CO and the formation of metallic Rh^ aggregates of a range of sizes on the TiO2(110) surface.

544

Rh(CO),

Rh(CO),

V»ym(C-0)

Rh(CO), 500K 1.2X10'LCO

3.4x10* L CO 6.2x10'L CO

' I ' I ' I ' I' 2080

2000

1920

I—\—rn—I—r 306 310 314 318 binding energy / eV

wavenumber / cm' Fig. 17. FT-RAIRS and XPS spectra of the thermal decomposition and re-generation of Rh(C0)2 on TiO2(110) [56]. Small Rh^ particles are formed following CO desorption during heating to 500K. Re-exposure of the particles to CO at 300K results in the partial regeneration of Rh(C0)2, and CO adsorption on the remaining Rh° particles. The accompanying Rh(3d) XPS spectra show the change in Rh chemical environment on forming the Rh^ particles from Rh(C0)2 by heating, and show the co-existence of the two Rh chemical environments following re-generation.

The photoemission spectrum of the Rh(3d) levels shows that on desorption of the CO, the Rh is reduced from the oxidised state of the Rh(C0)2 to Rh , but that the broadness of the peaks indicate a distribution of Rh particle sizes. Re-exposure to CO results in the facile re-generation of some Rh(CO)2, and a slower adsorption of CO on the larger Rh° particles which have not undergone re-generation. The slowness of the adsorption results from the decoration of the Rh^ particles with oxygen (from the TiO2(110) substrate) which is reactively removed during CO exposure at 300K. The Rh(3d) XPS spectra reflect the duel chemical environment of the rhodium. Annealing the surface to higher temperature produces larger Rh^ particles none of which are able to undergo re-conversion to Rh(C0)2 on exposure to CO [56, 69, 76]. Note that the Rh(CO)2 species, which adsorbed directly on the TiO2(110) surface, are dominated by the dielectric response of the substrate, with a transmission band in FT-RAIRS corresponding to Vsym(C-O), and an absorption band associated with Vasym(C-O). The latter is not observed (Fig. 17) following regeneration by

545

CO exposure simply because of the measurement sensitivity, and is observed when a higher proportion of Rh(C0)2 is regenerated from smaller particles following annealing at 450K (Fig. 10). The CO adsorbed on Rh^ is characterised by a large absorption band associated with the linearly bound species, and hence indicates the dominance of the metallic dielectric response of the particles. A consequence of this is the disproportionate sensitivity of FT-RAIRS to the CO adsorbed on the metal particles. The actual concentration of adsorbed CO associated with the Rh particles (Fig. 17) was estimated from XPS as <0.04ML, less than one tenth of the CO associated with the regenerated Rh(C0)2.

CO exposure / L

IMLPd

I M ' I •I' 2120

2080

2040

•I M M M 2000

1960 wavenumber / cm'*

Fig. 18. FT-RAIRS spectra of CO adsorbed at 300K on IML and 20ML films of palladium deposited by metal vapour deposition on TiO2(110) [56]. The switch from a local titania dielectric (transmission band) to that of the metal (absorption band) takes place at about lOML of palladium. The singleton frequency and the coverage dependent dipole shift are similar for both palladium layers indicating little perturbation of the CO adsorption behaviour on the palladium by the Ti02(l 10) substrate.

546

In the same series of experiments [56], rhodium was deposited using MVD, and subsequently exposed to CO. Fig. 18 shows FT-RAIRS spectra for CO adsorbed on a IML rhodium layer, and a 20L rhodium layer, with both deposition and adsorption carried out at 300K. Under these deposition conditions it was concludes that initial growth was layer by layer with some three dimensional growth at higher coverages. Adsorption of CO is accompanied by the growth and shifting of a v(C-O) band of CO linearly bound to Rh^; it was concluded from FT-RAIRS that there was no difference in the chemisorption behaviour even in the case of a IML film when one may expect an influence of the TiOiCllO) substrate. The obvious difference between the spectra are that at IML, a transmission band is observed, and at 20ML and absorption band characterises the spectrum. This reflects the conversion from the dominance of the TiOiCllO) to the metallic dielectric. This conversion was shown to be complete at ca. lOML of rhodium, and between IML and lOML, a complex intermediate response was observed. FT-RAIRS measurements of CO have also been used to identify facets of oxide supported Cu particles [78, 82]. The low sensitivity of RAIRS on single crystal ZnO(OOOl) prevented the observation of adsorbed CO or CO2, despite their observation in NEXAFS [781, although the local metallic dielectric allowed CO to be observed on the Cu particles. There appear to be no examples of HREELS being used to carry out vibrational spectroscopy of adsorbates on oxide supported metal particles. A HREELS study of Ag on MgO(lOO) films [95] was used only to characterise the Ag induced attenuation in the substrate Fuchs-Kliewer phonons, and the appearance of the metal/oxide interfacial plasmon at higher energies. HREELS has also been used to characterise the oxide/oxide interface between NiO and thin film MgO(lOO) [96]. Similar measurements of substrate phonon attenuation were made in HREELS studies on Pt films grown on ZnO(OOOl) [97].

REFERENCES 1. 2. 3. 4. 5. 6.

J. Pritchard, in: M. W. Roberts and J. M. Thomas (Eds), Chemical Physics of Solids and Their Surfaces, Vol. 7, Chemical Society, London, 1979, p. 157. H. Ibach and D. L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations, Academic, New York, 1982. F. M. Hoffmann, Surf.Sci.Reports, 3 (1983) 107. B. E. Hayden, in Methods of Surface Characterisation, Vol. 4 (J. T. Yates and T. E. Madey, eds.). Plenum, 1987, p. 267. A. M. Bradshaw and E. Schweizer, in: R. E. Hester (Ed), Advances In Spectroscopy: Spectroscopy of Surfaces, Wiley, New York, 1988. M. A. Chesters, E. C. Hargreaves, M. Pearson, P. Hollins, D. A. Slater, J. M. Chalmers, B. Ruzicka, M. Surman, and M. J. Tobin, Nuovo Cimento Delia Societa

547

7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

28. 29. 30. 31. 32. 33. 34. 35.

36.

Italiana Di Fisica D-Condensed Matter Atomic Molecular and Chemical Physics Fluids Plasmas Biophysics 20 (1998) 439. A. Horn,, (R. J. H. Clark and R. E. Hester, eds.), John Wiley and Sons.Ltd., 1998, p. 273. P. Hollins, Surf. Sci. Reports, 16 (1992) 51. V. E. Henrich and P. A. Cox, The Surface Science of Oxides, Cambridge University Press, 1994. P. M. A. Sherwood, Vibrational Spectroscopy of Solids, Cambridge University Press, 1972. R. Fuchs and K. L. KUewer, Phys. Rev. A ,140 (1965) 2076. A. D. Baden, P. A. Cox, R. G. Egdell, A. F. Orchard, and R. J. D. Willmer, J.Phys.C: Solid State Phys., 14 (1981)L1081. M. C. Wu, C. A. Estrada, J. S. Corneille, and D. W. Goodman, J. Chem. Phys., 96 (1992) 3892. H. Ibach, Phys. Rev. Lett., 24 (1970) 1416. P. A. Cox, W. R. Flavell, A. A. Williams, and R. G Egdell, Surf. Sci. Reports, 152/153(1985)784. M. A. Henderson, S. A. Joyce, and J. R. Rustad, Surf. Sci., 417 (1998) 66. S. Crook, H. Dhariwal, and G. Thornton, Surf. Sci., 382 (1997) 19. M. A. Henderson and S. A. Chambers, Surf. Sci., 449 (2000) 135. M. A. Henderson, Surf. Sci., 400 (1998) 203. M. A. Henderson, J. Phys. Chem. B, 101 (1997) 221. M. A. Henderson, S. Otero-Tapia, and M. E. Castro, Faraday Discussions, 114 (1999) 313 S. A. Chambers, M. A. Henderson, Y. J. Kim, and S. Thevuthasan, Surf. Rev. Lett., 5 (1998)381. M. A. Henderson, Surf. Sci., 355 (1996) 151. M. C. Wu, C. A. Estrada, and D. W. Goodman, Phys. Rev. Lett., 67 (1991) 2910. C. Oshima, T. Aizawa, R. Souda, and Y. Ishizawa, Solid State Comm., 73 (1990) 731. C. M. Truong, M. C. Wu, and D. W. Goodman, J. Amer. Chem. Soc, 115 (1993) 3647. B. Dillmann, F. Rohr, O. Seiferth, G Klivenyi, M. Bender, K. Homann, L N. Yakovkin, D. Ehrlich, M. Baumer, H. Kuhlenbeck, and H. J. Freund, Faraday Discussions, 105 (1996) 295. L Hemmerich, F. Rohr, O. Seiferth, B. Dillmann, and H. J. Freund, Zeitschrift Fur Physikalische Chemie-Int. J. Research in Phys. Chem. & Chem. Phys., 202 (1997) 31. M. Heemeier, M. Frank, J. Libuda, K. Wolter, K. H., M. Baumer, and H.-J. Freund, Catal. Lett., 68 (2000) 19. M. C. Wu, C. M. Truong, and D. W. Goodman, J. Phys. Chem., 97 (1993) 9425. P. Dumas, Surf Interface Anal., 22 (1994) 561. P. Dumas, M. K. Weldon, Y. J. Chabal, and G. P. Williams, Surf. Rev.Lett., 6 (1999) 225. Y. J. Chabal, in: G. Le Lay, J. Derrien, and N. Boccara (Eds.), Semiconductor Interfaces: Formation and Properties, Vol. 22, Springer-Verlag, Berlin, 1987, p. 301. Y. Chabal, Sur.Sci.Reports, 8 (1988) 211. M. A. Chesters, A. B. Horn, E. J. C. Kellar, S. F. Parker, and R. Raval, in: D. J. ColeHamilton and J. O. Williams (Eds.), Proceedings of NATO workshop in 'Mechanisms of Reaction of Organometallic Compounds with Surfaces', St Andrews, Scotland, 2224 June, Plenum Press, 1989. R. G Greenler, J. Chem. Phys., 44 (1966) 301.

548 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.

J. D. E. Mclntyre and D. E. Aspnes, Surf. Sci., 24 (1971) 417. R. G. Greenler, J. Chem. Phys., 50 (1969) 1963. R. G. Greenler, J. Vac. Sci. Tech., 12 (1975) 1410. E. D. Palik, (Ed.), Handbook of Optical Constants of Solids, Academic, Vol.1 (1985), Vol.2 (1991). W. Suetaka, Surface Infrared and Raman Spectroscopy: Methods and Applications, Plenum Press, 1995. J. A. Mielczarski and R. H. Yoon, J. Phys. Chem., 93 (1989) 2034. M. K. Weldon, V. E. Marsico, Y. J. Chabal, D. R. Hamann, S. B. Christman, and E. E. Chaban, Surf. Sci., 368 (1996) 163. M. K. Weldon, B. B. Stefanov, K. Raghavachari, and Y. J. Chabal, Phys. Rev. Lett. 79(1997)2851. M. K. Weldon, J. Eng, B. E. Bent, Y. J. Chabal, and L. M. Struck, Abs. Amer. Chem. Soc. 213(1997)218. A. Borg, P. Apell, and O. Hunderi, Superlatt. Microstruct., 3 (1987) 103. S. L. Finke and G. L. Schrader, Spectrochim. Acta A, 46 (1990) 91. V. M. Bermudez and S. M. Prokes, Surf. Sci. Reports, 248 (1991) 201. W. Erley, R. Butz, and S. Mantl, Surf. Sci., 248 (1991) 193. Y. Kobayashi and T. Ogino, Surf. Sci., 368 (1996) 102. M. A. Olmstead, Surf. Sci. Reports, 6 (1986) 159. Chaiaradia, in: G. Laley and J. Derrien (Eds.), Semiconductor Interfaces: Formation and Properties, Vol. 22, Springer, Berlin, 1987, p. 290. F. Wooten, Optical Properties of Solids, Academic Press, 1972. G D. Mahan and A. A. Lucas, J. Chem. Phys., 68 (1978) 1344. Patel and Pemble, J. de Physique IV, 1 (1991) 167. J. Evans, B. E. Hayden, F. Mosselmans, and A. Murray, Surf Sci., 301 (1994) 61. B. E. Hayden, K. Prince, A. M. Bradshaw, and D. P. Woodruff, Phys. Rev. Lett., 51 (1983)475. J. Evans, B. E. Hayden, and G. Lu, J. Chem. Soc. Faraday Trans., 92 (1996) 4733. B. E. Hayden, in: R. M. Lambert and G. Pacchioni (Eds), NATO ASI Series E: Applied Sciences, Vol. 331, Academic, 1997, p. 215. K. Fukui, I. Oshima, H. Oosterbeek, and Y. Iwasawa, Chem. Phys. Lett., 299 (1999) 158. P. Masri and P. W. Tasker, Surf. Sci., 149 (1985) 209. P. Masri, P. W. Tasker, J. P. Hoare, and J. H. Harding, Surf. Sci., 173 (1986) 439. W. C. Mackrodt, P. W. Tasker, and E. A. Colbourn, Surf. Sci., 152/153 (1985) 940. G. Lakshmi and F. W. de Wette, Phys. Rev. B, 22 (1980) 5009. G. Lakshmi and F. W. de Wette, Phys. Rev. B, 23 (1981) 2035. W. Kress, F. W. de Wette, A. D. Kulkarni, and U. Schroder, Phys. Rev. B, 35 (1987) 5783. K. Wolter, D. Scarano, J. Fritsch, H. Kuhlenbeck, A. Zecchina, and H. J. Freund, Chem. Phys. Lett., 320 (2000) 206. P. A. Cox and A. A. Williams, J. Electron. Spec, 39 (1986) 45. J. Evans, B. E. Hayden, F. Mosselmans, and A. Murray, Surf Sci., 279 (1992) L159. J. Evans, B. E. Hayden, F. Mosselmans, and A. Murray, J. Am. Chem. Soc, 114 (1992) 6912. B. E. Hayden, A. King, and M. A. Newton, Surf. Sci., 397 (1998) 306. B. E. Hayden, A. King, and M. A. Newton, Chem. Phys. Lett., 269 (1997) 485. B. E. Hayden, A. King, and M. A. Newton, J. Phys. Chem. B 103 (1999) 203.

549 74. 75. 76.

77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97.

M. Frank, R. Kuhnemuth, M. Baumer, and H.-J. Freund, Surf. Sci., 427-428 (1999) 288. L.-Q. Wang, K. F. Ferris, A. N. Shultz, D. R. Baer, and M. H. Engelhard, Surf. Sci., 380(1997)352. J. Evans, B. E. Hayden, F. Mosselmans, and A. Murray, in: R. W. Joyner and R. A. Van Santen, Elementary Reaction Steps in Heterogeneous Catalysis, Vol. 398, Kluwer Academic Publishers, Dordrecht, 1993, p. 179. J. Evans, B. E. Hayden, and G. Lu, Surf. Sci., 360 (1996) 61. A. Gutierrez-Sosa, S. Crook, S. Haq, R. Lindsay, A. Ludviksson, S. Parker, C. T. Campbell, and G. Thornton, Faraday Discussions, 105 (1996) 355. J. Kubota, A. Bandara, A. Wada, K. Domen, and C. Hirose, Surf. Sci., 368 (1996) 361. S. K. Purnell, X. Xu, D. W. Goodman, and B. C. Gates, Langmuir, 10 (1994) 3057. S. K. Purnell, X. Xu, D. W. Goodman, and B. C. Gates, J. Phys. Chem., 98 (1994) 4076. X. P. Xu, S. M. Vesecky, and D. W. Goodman, Science, 258 (1992) 788. A. Siokou and R. M. Nix, J. Phys. Chem. B, 103 (1999) 6984. R. P. Eischens, S. A. Francis, and W. A. PHskin, J. Phys. Chem., 60 (1956) 194. K. Wolter, O. Seiferth, H. Kuhlenbeck, M. Baumer, and H. J. Freund, Surf. Sci., 399 (1998) 190. X. P. Xu and D. W. Goodman, J. Phys. Chem., 97 (1993) 7711. R. P. Eischens and W. A. Pliskin, Advances in Catalysis, 10 (1958) 1. X. P. Xu, J. Szanyi, Q. Xu, and D. W. Goodman, Catalysis Today, 21 (1994) 57. D. R. Rainer, M. C. Wu, D. I. Mahon, and D. W. Goodman, J. Vac. Sci. Tech.A, 14:(1996)1184. M. Frank, R. Kuhnemuth, M. Baumer, and H. J. Freund, Surf Sci., 454/456 (2000) 968. K. Wolter, O. Seiferth, J. Libuda, H. Kuhlenbeck, M. Baumer, and H. J. Freund, Surf. Sci., 402/404 (1998) 428. M. Baumer, J. Libuda, A. Sandell, H. J. Freund, G. Graw, T. Bertrams, and H. Naddermeyer, Ber. Bundsenges. Phys. Chem., 99 (1995) 1381. H. J. Freund, Angew. Chem. Int. Ed. Engl., 36 (1997) 452. E. M. Stuve, R. J. Madix, and R. J. Brundle, Surf. Sci., 146 (1984) 155. M. H. Schaffner, F. Patthey, and W. D. Schneider, Surf. Sci., 417 (1998) 159. Q. Guo, C. Xu, and D. W. Goodman, Langmuir, 14 (1998) 1371. W. T. Pitrie and J. M. Vohs, J. Chem. Phys. 101 (1994) 8098.

550

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

Chapter 14

Electronic structure of oxide surfaces R.G. Egdell Inorganic Chemistry Laboratory, South Parks Road, Oxford OXl 3QR, UK 1. INTRODUCTION Metal oxides display a wider and richer range of electronic properties than any other class of material. In terms of electronic conductivity at room temperature, oxides encompass some of the most robust wide-gap insulators such as AI2O3 as well as materials such as ReOa whose conductivity is comparable with that of Cu. The superconductor with the highest known transition temperatures HgBa2Ca2Cu308+x with Tc above 130K - is an oxide. The diverse range of behaviour cannot be understood solely in simple structural terms. Thus rocksalt oxides include insulators such as MgO, n- and p-type semiconductors typified by CdO and NiO respectively and metallic conductors such as TiO. Similarly oxides with the rutile structure include both insulators such as stoichiometric Sn02 and metallic conductors typified by RUO2. An important aspect of bulk oxide solid state science is that the electronic conductivity and optical properties are strongly influenced by non-stoichiometry and doping. Thus stoichiometric Ti02 is an optically transparent insulator with a bandgap of 3.06 eV, but nonstoichiometric Ti02.x assumes a deep blue colouration and is a semiconductor with an activation energy for conduction of only 0.02 eV for small values of x. Similarly WO3 is an insulator with a gap of 2.6eV at room temperature, but the sodium tungsten bronzes NaxWOs are metallic conductors for x values greater than 0.26. This chapter is concerned with experimental studies of the electronic structure of oxide surfaces. Obviously the complexities of surface electronic structure cannot be understood without first understanding the factors which determine bulk electronic structure: the next section outlines some of the key ideas that are needed to describe the bulk electronic structure of oxides. However, it is beyond our scope to dwell at length on experimental determination of bulk electronic structure: instead the main focus of the chapter is on aspects of electronic structure associated with surfaces that cannot be understood in terms of the bulk electronic structure alone. The publication of

551

"The Surface Science of Metal Oxides" by Henrich and Cox in 1994 [1] represented a milestone in the field and for the most part we concentrate on developments which have taken place since then. Reference to the earlier literature is not excluded but it is by no means comprehensive.

2. ELECTRONIC STRUCTURE OF OXIDES: GENERAL CONSIDERATIONS 2.1 Models of oxide electronic structure Following Cox [2,3], we introduce the ionic model as a useful starting point for the discussion of the electronic structure of oxides. Adoption of this approach does not presuppose that oxides are fully ionic and indeed the extent of covalency in oxides and how covalency might change at a surface are recurring themes in this chapter. Consider the schematic energy level diagram for MgO shown in figure 1. The atomic Mg 3 s level lies above the atomic O 2p level, but this ordering is reversed for the ions Mg^^ and O^". The free O^" ion is not stable in vacuum and the associated O 2p level is estimated to be just under 10 eV above the vacuum level. However the Madelung site potential in a fully ionic MgO lattice stabilises the O^" level and destabilises the Mg^^ level, thus restoring the original ordering of levels. The gap of 24 eV between the ionic O 2p and Mg 3 s levels is reduced when account is taken of polarisation, which lowers the energy of added-electron states and raises the energy of occupied O 02-/0-

Conduction bond (Mg 3s) Oh Bond gop

Energy (eV)

' ^"9 ^\|

k^

Mg*/Mg2*

-20 h

Free lOns (o)

Modeiung potent lol (b)

Polorisation

(c)

Valence bond (0 2p)

Overlop

(d)

Fig. 1. Schematic energy level diagram for MgO (a) free ion energies (b) ionic energy levels in point charge Madelung potential (c) ionic energy levels after allowance for polarisation (d) after allowing development of bandwidths due to interactions amongst ionic level. Adapted from ref. 1.

552

2p states, making it easier to remove electrons from these levels. Finally direct interactions between the ionic levels and indirect Mg-O-Mg and O-Mg-0 interactions give rise to appreciable O 2p and Mg 3s bandwidths, which finally lowers the bandgap to its experimental value of 7.8 eV. When dealing with transition metal oxides, this picture must be modified to take account of the splitting of the metal d levels and the occupancy of the resulting d bands. Where (as is often the case) the metal sites are of octahedral or approximately octahedral point symmetry, the d levels split into 3-fold degenerate t2g levels and 2-fold degenerate Cg levels. This splitting can be understood in terms of simple electrostatic interactions with the oxygen anions or in terms of the difference in the covalent interaction between jr-like tag levels and a-like Cg levels and the oxygen anion orbitals of the same local symmetry [4]. In an independent electron picture any transition metal oxide with partially occupied d bands will be metallic. However, this viewpoint ignores strong onsite repulsion between d electrons. When the on-site repulsion energy U exceeds the intrinsic d electron bandwidth B, electrons are unable to hop from site to site and when U » B the system is better described in terms of localised d"^ configurations [4]. In the intermediate situation, where U>B but the d electron bandwidth cannot be ignored, it is useful to think of an occupied lower Hubbard band of occupied d" states and an empty upper Hubbard band of states where there has been site to site transfer. When the lower Hubbard band is above the O 2p band the material is described as a Mott-Hubbard insulator. However there is growing evidence that for many insulating transition metal compounds, including oxides of later transition metals, the lower Hubbard band is located below the bottom of the O 2p band. The lowest energy excitations then involve charge transfer from O 2p states into the upper Hubbard band and such materials are described as charge transfer insulators [5]. 2.2 Covalency and band dispersion in oxides. The O 2p orbitals in oxides have opposite parity to the metal ns or nd orbitals upon which the conduction band states are based. For those many oxides where the cations occupy sites with effective inversion symmetry, this means that there can be no mixing between O 2p states and metal cation states at the zone centre. However, mixing is possible at the zone boundary. The effect of the covalent mixing is to lower the energy of the O 2p states, giving oxygen-metal bonding states at the zone boundary, and to raise the energy of the metal states whose acquisition of O 2p character makes them antibonding. Thus a general feature of computed bandstructures for many oxides (including those with rocksalt, rutile and perovskite-like structures) is that O 2p states disperse down in energy with increasing wavevector and metal based states disperse upward. Moreover the states at the bottom of the O 2p valence band have the most

553

pronounced metal character. These considerations suggest that bandgaps should be direct. Exceptions to this simple generalisation are found for structures with non-centrosymmetric coordination or where interactions between shallow core d states and O 2p states produce upward dispersion of O 2p states away from the T point, as for example in rocksalt CdO [6]. 2.3 Surface states in oxides In general terms surface states can arise even at unrelaxed, non-defective surfaces simply from the loss of translational periodicity which characterises the bulk. However, given the importance of the Madelung potential in determining the position of oxygen and metal levels for bulk oxides, it is informative to consider surface states in relation to changes in the site potential at the surface. The reduced coordination at surface sites might be expected to produce a decrease in the magnitude of the Madelung potential at both cation and anion surface sites, leading to new O 2p states above the valence band maximum and new cation states below the conduction band minimum i.e. a narrowing of the bandgap at the surface [7]. The reduced separation between O 2p and metal valence orbitals may in turn allow for enhanced metal-oxygen covalency at the surface. In addition the point group symmetry experienced by cations in surface sites will differ from that of bulk sites. This can produce first order changes in the pattern of crystal field splitting of cation atomic orbitals, which will be most obvious in transition metal oxides with localised d" configurations. In addition the change of symmetry may allow second order mixing between atomic orbitals at cation sites that is not allowed in the bulk. The first three case histories discussed in section 5 - namely MgO; transition metal rocksalt monoxides NiO and CoO; and Sn02 - are chosen to illustrate the influence of these three effects in producing surface states. Non-stoichiometry at oxide surfaces introduces further possibilities. The energy required to generate non-stoichiometric defects at surfaces will in general differ from the energy required in the bulk. Thus the equilibrium concentration of surface defects under well-defined conditions of temperature and oxygen partial pressure may well be greater than the bulk value. However, an additional complication is that it cannot always be assumed that equilibrium conditions have been attained. In particular the argon ion bombardment often used in sample cleaning leads to preferential sputtering of oxygen from most oxide surfaces and equilibrium surface stoichiometry may only be restored by very prolonged annealing in UHV or oxygen partial pressures. Finally the geometric arrangement of ions that accommodate non-stoichiometry at a surface may well be different to anything that is found in the bulk, leading to electronic states at energies different to those produced bulk non-stoichiometry. These ideas are developed in the context of non-stoichiometric Ti02 and WO3 surfaces.

554

3. EXPERIMENTAL TECHNIQUES FOR PROBING ELECTRONIC STRUCTURE OF OXIDES 3.1 Photoemission and inverse photoemission spectroscopies Photoemission spectroscopy involves measurement of the energy distribution of electrons emitted from a solid under irradiation with monoenergetic photons. In-house experiments are usually performed with He gas discharge lamps which generate vacuum UV photons at 21.2 eV (He l a radiation) or 40.8 eV (He Ila radiation ); or with Mg Ka (hv=1284.6 eV) or Al Ka (hv=1486.6eV) soft X-ray sources. UV photoemission is restricted to the study of valence and conduction band states, but XPS allows in addition the study of core levels. Alternatively photoemission experiments may be performed at national synchrotron radiation facilities. With suitable choice of monochromators it is possible to cover the complete photon energy range from about 5 eV upward to in excess of 1000 eV. The surface sensitivity of photoemission derives from the relatively short inelastic mean free path of electrons in solids, which reaches a minimum of about 5 A for electron energies ofthe order 50-100 eV. To a first approximation angle integrated photoemission measures the density of occupied electronic states, but with the caveat that the contribution of a given state to the spectrum must be weighted by appropriate ionisation cross sections. Comprehensive tabulations of ionisation cross sections calculated within an independent electron framework are available [8]. At X-ray energies cross sections for ionisation of second and third row transition metal d states are often very much greater than for ionisation of O 2p states, so that valence band X-ray photoemission spectra represent not so much the total density of states as the metal d partial density of states [9]. The independent electron picture of photoionisation breaks down completely at photon energies around the threshold for excitation of a core level [1]. Consider for example a first row transition metal compound with a 3d" configuration under irradiation with photons whose energies match that required to excite electrons from the shallow 3p core level. There is interference between the direct photoemission channel: (3p^3d") + hv

-^

(3p^3d"-^) + e

and a resonant channel involving 3p photoexcitation followed by super-CosterKronig decay: (3p^3d") + hv (3p^3d"^')*

~> -^

(3p'3d"^') (3p^3d"-') + e

555

The interference gives rise to a characteristic Fano-type lineshape in the ionisation cross section profile. Resonant photoemission, in which the photon energy is tuned to coincide with the maximum in the cross section profile, is of particular value in the study defect-induced metal d states. Measurement of constant initial state spectra (CIS), in which the photon energy and electron energy are varied in synchrony to keep fixed on the same initial state also provides a way of fingerprinting the atomic character of features in valence band photoemission in situations where the issue may be in doubt. For example, CIS spectra allowed identification of distinct Ti 3d and Fe 3d states when Fe was deposited on Ti02 [10]. Elsewhere it has been shown that bandgap states in Nbdoped TiOi are associated with electrons localised in Ti 3d orbitals rather than Nb 4d orbitals, whereas in V-doped Ti02 the analogous states are of V 3d character [11,12]. Angle resolved photoemission enables characterisation of the energy versus wavevector dispersion of electronic states [13,14]. The simplest experiment exploits momentum conservation parallel to the surface. For materials with layer-like structures in which there is little dispersion in electronic states normal to the surface, variations in binding energies with angle of off-take relative to the surface normal allows bands to be mapped out along directions parallel to the surface. Studies of this sort have proved to be particularly popular in the study of the dispersion of bulk states in layered oxides such as Lio.9Mo60i7 [15,16] and Nao.9Mo60i7 [17]. The technique would also be appropriate to measurement of the dispersion of 2D surface states on 3D materials. However for bulk states in 3D materials themselves interpretation of off-normal angle resolved spectra requires full knowledge of the bandstructure along a circular arc in k space. Direct band mapping in 3D materials can be achieved by the study of the dispersion of spectral features in normal photoemission spectra as a function of the energy of incident photons. This approach requires assumptions about the magnitude of the inner potential and the dispersion in the freeelectron-like final states. Variable photon energy normal photoemission experiments have now been carried out on a range of oxides. A general problem in these experiments is that the widths of the individual bands are quite large in relation to the separation between the individual bands within the bandstructure, so that individual spectral features are rarely well-resolved. In addition it is probable that the spectra contain a sizeable incoherent component arising fi-om breakdown of k conservation due to electron-phonon coupling [1]. Nonetheless the dispersion of O 2p states in materials such as ZnO [18,19] and SrTiOs [20] is in reasonable agreement with bandstructure calculations. Dispersion of 3d states in oxides such as CoO [21,22] and NiO [23] has also been observed, but it is questionable whether one-electron bandstructure calculations provide an appropriate framework for interpretation of these spectra.

556

Inverse photoemission involves a time reversed photoemission process in which electrons of varying energy incident upon a sample decay radiatively into empty electronic states. Photons of fixed frequency are counted and the energy dispersion of empty states may be investigated by varying the incident electron angle. Both photoemission and inverse photoemission require reasonable sample conductivity and their application to "hard" insulators such as MgO and AI2O3 is problematic. Both techniques also involve the complication that inelastic electron energy loss processes become convoluted with electron emission or decay. This may give rise to spectral features in regions where none are expected from the density of states [24,25] and care must always be taken to exclude these features before considering assignment to surface states. 3.2 Electron energy loss spectroscopy Two variants of the technique of electron energy loss spectroscopy (EELS) operated in the reflection mode have played a role in the characterisation of surface excitations on oxides. Moderate energy resolution of at best around 0.5 eV is achieved using non-monochromated electron beams whose energy width is determined by the thermal spread of energies in the cathode of the electron gun. The resolution may be further compromised if it necessary to apply a modulation in the analyser system in order to measure spectra in a differential dN(E)/dE mode. Nonetheless interband and other higher energy electronic excitations can be observed in N(E) or dN(E)/dE modes. The high resolution variant of the technique (HREELS) involves a monochromatised electron beam, allowing energy halfsvidths of 5 meV or better. Both phonon excitations and very low energy electronic excitations - such as the d to d excitations discussed in section 4.2 and low energy conduction plasmons in degenerately doped oxide semiconductors - can be observed in HREELS. At low beam energies inelastic scattering is usually dominated by a dipolar mechanism in which the electric fields associated with surface excitations couple over a long length range with the exciting electron in its trajectory toward and away from the solid [26]. The lobe of inelastically scattered electrons peaks strongly close to the specular direction with dominant k transfer of the order k=o}/v, where co is the frequency of the excitation and v is the electron beam velocity normal to the surface. The wavevector k in turn determines the decay length X of excitations into the bulk through X=l/k=v/co. Although the excitations do not propagate into the bulk, the decay lengths may be of the order of hundreds of Angstroms under typical experimental conditions, especially when dealing with surface phonons. Thus one is dealing with surface excitations in the sense that they arise from a dielectric discontinuity at the surface, but no information is conveyed about details of top layer structure. The

557

small angular spread of the inelastic dipole lobe when ^co«E allows analyser systems to collect all of the k integrated dipole loss intensity. Under these conditions the loss intensity I(co) as a function of frequency is related to the bulk dielectric function £(a>): I(co) =

2e^ ^ -1 Im hojvj^ £(co) -h 1

With increasing beam energy above about 100 eV, the mean free path of electrons into solids increases and the inelastic scattering is increasingly dominated by penetration of electrons into the bulk of the solid. The loss function is then dependent on Im(-l/£(co)), In the literature EELS data are often compared with bulk optical excitation data. However it should be remembered that bulk absorption spectroscopy measures Im(8(cD)) and that in general there are well-defined shifts between the transverse resonance frequencies found optically and the longitudinal frequencies for bulk losses in EELS. Surface loss energies lie between bulk transverse and longitudinal frequencies. The shifts are particularly obvious when dealing with vibrational losses, but the same principles apply to electronic excitations. 3.3 Scanning tunnelling spectroscopy Scanning tunnelling microscopy (STM) is a proximal probe technique in which an atomically sharp tip is held sufficiently close to a surface to allow significant overlap between the tails of the wavefunctions of electronic states associated with the tip and the surface respectively. Application of a small bias between the tip and the surface therefore allows electron tunnelling between the two. In the simplest models the magnitude of the tunnelling current varies as e'^, where d is the tip surface separation. The strong dependence on distance has allowed development of STM as a surface microscopy: as a tip is scanned across a surface the tunnel current will vary in a way that reflects the surface topography. However it must also be realised that STM is strongly influenced by electronic structure. Thus under conditions of positive sample bias, electrons tunnel into empty electronic states in the solid, whilst at negative bias electrons tunnel out of filled states in the solid into the tip. For an oxide with filled O 2p bands and empty metal-based bands it therefore follows that images taken at positive sample bias from a "flat" surface containing both O ions and metal ions in the same plane will be dominated by empty metal based states. At negative sample bias filled oxygen-localised states will be probed. In the TersoffHamman model [27] contours of constant tunnel current above a surface are simply contours that mirror the filled or empty state electron density in the relevant energy window. The variation of tunnelling current with sample bias is

558

intimately related to the density of states at the surface. Again assuming the simplest model, the density of states dN(E)/dE may be shown to be proportional to the normalised differential conductance i.e. dN{E) dI{E)ldV{E) dE °^ I(E)/V{E)

The measurement of tunnelling spectra in a scanning tunnelling microscope offers the potential of measuring the local density of states at spatially defined sites whose topography can be established at an atomic scale by STM. This information is only available however at the price of losing the information about the k-dependence of electronic states that is available in photoemission and inverse photoemission. In particular STS offers the prospect of measuring local densities of states at defect sites whose real space atomic structure can be established by STM. Given the potential of the technique there have been remarkably few studies of the electronic structure of defects by STS. One of the most satisfying early studies was of bandgap states at the (V5xV5)R26.6° surface of SrTiOsClOO) [28,29]. This surface was originally described as (2x2) [30], but is now believed to terminate in a TiOi surface plane with an ordered arrangement of in-plane oxygen vacancies. The vacancies are imaged as grey-scale maxima at negative sample bias in STM i.e. the image is dominated by filled electronic states just below the Fermi level. STS shows the vacancies to be associated with filled electronic state 1.35 eV below the Fermi energy. A similar state is observed in photoemission from oxygen deficient SrTiOsClOO) surfaces, but photoemission provides no guidance as to the spatial location of the state. The application of STS to defective TiOi and WO3 surfaces is discussed in section 4. 4. SELECTED CASE HISTORIES 4.1 The rocksalt MgO(lOO) surface*. MgO is a transparent insulator with a gap of about 7.8 eV separating a filled band of O 2p states from an empty band of Mg 3s states: Mg 3p states lie at higher energy. The lowest excitation in bulk MgO is not a simple band to band transition, but involves an excitonic process. "Conventional" excitons in oxides such as Li20 and SiOi are associated with localisation of the hole on an oxygen anions and the electron on an adjacent cation. The cation and anion involved undergo large asymmetric relaxations and there is therefore a substantial Stokes shift between the energies associated with the vertical excitonic absorption and * See the chapter by Freund et al for diagrams of the MgO(lOO) surface.

559

the excitonic luminescence from the self-trapped, relaxed configuration. In MgO by contrast it appears that the electron and the hole both reside on the same O site: the excitation may be viewed as an on-site O 2p->0 3s transition. The relaxation around the exciton is symmetrical and involves smaller energies than for the two-centre exciton. The Stokes shift is therefore small and the lower self trapping energy gives rise to much higher exciton mobility than is found for say Si02[31]. The (100) surface of rocksalt oxides typified by MgO represent perhaps the simplest of oxide surfaces. The ionic layer which terminates this surface is electrostatically neutral and contains equal number of Mg^"^ and O^" ions in interpenetrating square arrays. The (100) surface energy is lower than for any other surface of this material and bulk MgO crystals cleave easily to expose high quality (100) surfaces. LEED [32,33] and medium energy ion scattering demonstrate [34] that the rumpling of the Mg(lOO) surface is rather small, with inward relaxation of Mg ions and outward relaxation of O. In detail the surface structure as determined by surface X-ray diffraction involves [35] inward relaxation by 0.56% of the interlayer separation between MgO planes and 1.07% rumpling. These observations are reproduced by atomistic simulations [36], as well as periodic ab initio Hartree-Fock [37,38] and density functional calculations [39,40]. There is no particular indication of enhanced covalency at the surface as has been claimed on the basis of the low corrugation at the surface observed in He atom diffraction [41]. Experimental difficulties in the study of MgO surfaces by electron spectroscopic techniques may be anticipated because of the highly insulating nature of the material. Photoemission is indeed somewhat problematic, but techniques such as EELS and LEED, in which an electron beam impinges on the surface, prove to be viable even on bulk crystals. This is because the secondary electron yield coefficient (that is the number of secondary electrons emitted from the surface per incident electron) is greater than one above a threshold energy of around 20 eV and may reach values in excess of 10 at higher beam energy [42]. The sample therefore tends to charge positively and does not repel the incoming electron beam. HREELS experiments with beam energies below the 20 eV threshold are possible using a subsidiary high energy electron gun to promote an outward electron flux [43]. The alternative approach to dealing with sample charging is to prepare thin MgO overlayers on a metallic substrate. Goodman and co-workers [44-47] in particular have perfected recipes for growth of (100) oriented layers of MgO by evaporation of Mg onto an Mo(lOO) substrate in an oxygen ambient with pressures of the order 10'^ mbar. Two issues have been of interest particular in relation to the electronic structure of Mg(lOO) surfaces. The first is whether the electronic structure at the terrace surface differs in any significant and experimentally observable way

560

from the bulk. The second concerns the nature of electronic excitations associated with defects at (100) surfaces. Most features in angle resolved He(II) photoemission measurements on cleaved MgO(lOO) surfaces can be adequately rationalised in terms of a direct transition model based on the bulk bandstructure, although some features fall outside the range of bulk dispersion around the M point [48]. Owing to sample charging it is difficult to fix the position of the valence band edge relative to the vacuum level: the value of over 9 eV that was proposed by Sawatzky and coworkers [48] is probably too high. A more surface sensitive way of exciting valence band electron spectra is provided by metastable impact electron spectroscopy (MIES) [49-51]. In this technique a beam of He atoms in a ^S excited state associated with the configuration ls^2s^ impinges on the surface. At an insulating wide band gap surface such as that of MgO, the excited atoms decay via an intra-atomic Auger mechanism (sometimes called Penning deexcitation) in which an electron from the surface fills the Is hole and the 2s electron is emitted with the excess energy available. The MIES spectra from MgO(lOO) are much narrower than the photoemission spectra and essentially reflect the density of states associated the O 2pz orbitals which protrude out of the surface [49]. A more sophisticated approach to calculation of the MIES spectra involves projection of the surface density of states onto the Is orbital of the He* ion in its trajectory toward the surface [51]. These calculations provide closer agreement with experimental spectra, but reinforce the conclusion that the band of O 2pz surface states is narrower than the bulk valence band. The most direct evidence for differences between bulk and surface electronic structure comes from EELS. Using the technique in the differential mode Henrich et al found a loss peak at just over 6 eV on non-defective MgO(lOO) surfaces, an energy well below the onset of 7.5 eV for bulk excitations [52,53]. The relative intensity of this loss feature decreased dramatically on increasing the energy of the exciting beam from 100 eV to 2000 eV, as expected for a surface excitation [53]. Mg 2p and 2s core excitation spectra show a similarly shifted surface peak to low energy of the bulk core excitation threshold. HREEL spectra excited using a monochromatised electron beam from both bulk and thin film single crystal surfaces (fig. 2) illustrate the lowering of the excitation threshold even more dramatically [54,55]. The EELS data were originally interpreted in terms of simple bandgap narrowing at the surface, as expected from the reduced Madelung potential at surface sites [52,53,55,56]. However, Williams argued that the reduction in Madelung constant from 1.74765 in bulk sites to 1.68155 in surface sites is too small to produce the observed reduction in the excitation energy: in any case reduced valence and conduction bandwidths at the surface will oppose the Madelung term and the net effect might actually be to increase the gap [57]. It is informative therefore to consider quantitative

561 350

HREELS MgO(100)/Mo(100) 262

Ep=:48eV

N

X

t 175 (0

z

lU

87

1 2

3

4

5

6

LOSS ENERGY (eV)

7

8

9

^.2

Fig. 2. HREELS of MgO(lOO) thin film on Mo(lOO) excited at 48 eV beam energy. In moving upward the initially stoichiometric film is subject to annealing to the progressively higher temperatures indicated. Adapted from ref. 72.

attempts to model the electronic structure of MgO surfaces using bandstructure and cluster models. The first periodic calculation of energy levels for MgO(lOO) surfaces used a semiempirical LCAO model applied to an unrelaxed surface. Surfaces states were found w^hich fell outside the range of energies of the projected bulk band away from the T point. However the lowest energy direct surface excitations were found at 7.6 eV, very little below the calculated bulk bandgap of 7.8 eV [58]. Similarly small differences were found when relaxation was allowed using a tight binding total energy method [59]. On the other hand discrete variational cluster calculations using MgOe^^" to represent bulk MgO and MgOs^" to represent the (100) surface gave a clearcut indication that the bandgap should be reduced at the surface [60,61]. Essentially the bandgap narrowing arises from mixing between Mg 3s and 3p levels under the influence of the electric field gradient at the surface. This splits a hybrid 3s-3p state away from the bottom of the conduction, thus narrowing the bandgap. There is no significant lifting of

562

states above the top of the O 2p valence band. More recently ab initio calculations have been performed based on density functional theory applied to a fully relaxed MgO(lOO) surface represented by a periodic arrangement of slabs separated by vacuum gaps. As in the cluster calculations, a hybrid state localised mainly above the surface Mg ions in an 3s-3p hybrid state splits off from the bottom of the conduction band [40]. This result is independent of slab thickness. Despite these theoretical results, it is probably too naive to interpret the EELS results simply in terms of surface interband transitions, given that the threshold bulk excitation are excitonic in nature. In a survey of HREEL spectra of a wide range of compounds with the rocksalt structure it was found that a difference between surface and bulk excitation thresholds was only found for a limited range of materials including MgO, CaO, LiF but not NaCl, NaBr, KCl and KBr [54,57]. For compounds in the first group it was argued that the conduction band edge lies above the vacuum level giving a negative electron affinity - the high secondary electron yield coefficient for MgO discussed above has been attributed to this unusual arrangement of energy levels [62,63]. In this situation the lowest surface excitation is from the valence band edge to the vacuum level: binding of the electron at the vacuum level to the surface by its image potential gives rise a novel form of surface exciton where the electron sits outside the surface. This qualitative picture for the surface exciton receives support from calculations of Shluger et aL, who used a semiempirical intermediate neglect of differential overlap (INDO) model in which a Mg24024 cluster is embedded in a 5 layer periodic slab. The basis set included 12 floating Slater functions [31]. The surface exciton involved on site excitation of O 2p electrons, with outward displacement of the five near neighbour magnesium cations. The electron population of the floating orbitals located outside the surface plane was larger than those inside in the excited state. This model predicted a surface excitation energy of 6.4 eV, in excellent agreement with the experimental value. The surface exciton model of Cox and Williams was criticised by Saiki et al [64] who argued that the model does not account for the low energy metal core excitations observed for MgO and fluorides such as NaF. However there seems to be no reason why there should not be two distinct excitonic processes involved, with metal core level excitation into a surface metal sp hybrid and valence excitation into an on-site anion 3s state. Very recent calculations using an embedded cluster model give a value of 6.7 eV for the energy of the top of the O 2p valence band relative to the vacuum level [65]. This is in agreement with MEIS measurements on thin film MgO(lOO). Given the bulk bandgap of 7.8 eV, these calculations further reinforce the view that the conduction band minimum does indeed sit above the vacuum level. We turn next to consider defective MgO(lOO) surfaces. In contrast to the transition metal oxides to be discussed subsequently, argon ion bombardment

563

has little effect on UV photoemission spectra of MgO [48]. However Henrich et al [52,53] observed a loss feature in dN/dE energy loss spectra of UHV cleaved MgO(lOO) at about 2 eV that was obviously a surface excitation, as gauged by the decreasing intensity with increasing beam energy. These observations were confirmed by subsequent work using higher resolution [56,57,66]. The intensity of the feature increased upon electron irradiation of the crystal surface, but decreased on oxygen exposure. In most oxides electron beams induce oxygen deficiency, so the empirical observations support assignment of the loss feature to a surface F centre, that is a an oxygen vacancy with two associated electrons trapped in the vacancy site [52,53]. However, the excitation energy for a surface F centre was calculated to be 5 eV [67,68], leading to an alternative assignment [56,66] to a surface cation vacancy or V-centre whose excitation energy should be close to the value of 2.3 eV observed in non-differentiated EEL spectra [56]. This assignment receives some support from the observation that the loss feature is quenched by deposition of Cu [69], which would be expected to refill cation vacancy sites. On the other hand Tegenkamp et al found that loss intensity around 2 eV is enhanced by partial desorption of excess Mg from an MgO film surface: this experiment is hardly compatible with formation oi metal vacancies [70]. In a more detailed experimental study, Wu et al. [71] studied the effects of high temperature annealing in UHV on HREEL spectra of thin film singlecrystal MgO(lOO). The loss onset was just above 6 eV for as-prepared films, but UHV annealing at 1400 K led to growth of loss features at 1.15 eV, 3.58 eV and 5.33 eV. The first of these loss peaks is at lower energy than in other energy loss studies. However, it is difficult to compare HREEL data with data taken at lower energy resolution using non-monochromatised electron beams, especially when data are acquired in the differential mode. The 1.15 eV feature was assigned to surface F centre excitation, the loss at 3.58 eV to F centre aggregates and the loss at 5.33 eV to F centres in the "bulk" of the thin film. The third assignment here was based on two considerations. First the bulk F centre excitation in electron irradiated MgO crystals is at about 5 eV. Second it was observed that the relative intensity of the 5.33 eV loss decreased as the beam energy increased from 15 eV to 48 eV. The electron pathlength in the solid is undoubtedly greater at 15 eV than 48 eV, but the argument that the intensity variations establish the 5.33 eV loss as a bulk excitation is not valid if the dominant mode of inelastic scattering is by the surface dipole mechanism. The empirical observations relating to the low energy loss peak all seem to favour assignment to oxygen rather than magnesium vacancies. On the other hand more recent theoretical work has not led to a complete understanding of the experimental results. Periodic LDA calculations due to of Kantorvich et al establish that surface F centres have a formation energy of about 9.8 eV, which is significantly lower than the formation energy of 10.5 eV in the bulk. The electronic level of the relaxed surface F centre was only 2.3 eV above the

564

valence band maximum and therefore below the corresponding bulk F centre, which is 2.7 eV above the valence band [40]. This is in fair agreement with results from an embedded cluster calculation which places the surface F centre 3.25 eV above the valence band maximum and 3.43 eV below the vacuum level [65]. However these were both essentially ground state calculation that did not seek to calculate excitation energies explicitly. By contrast Pachioni and coworkers used cluster models to evaluate energies of excited states directly. Both neutral F centres and singly ionised F^ centres were considered. These calculations allowed for configuration interaction to take due account of correlation in both ground and excited states. The bulk singlet to singlet excitation energies for F and F^ centres were found to be at 6.00 eV and 5.75 eV [72], as compared with experimental values of 5.03 eV and 4.96 eV [73]. The bulk transitions correspond to dipole allowed Aig-^ Tiu and Aig-> T lu excitations. At the surface the threefold degeneracy of the Tiu states is removed. Surface F centre excitations corresponding to ^Ai->^Ai and ^Ai^^E processes are found at 3.24 eV and 4.22 eV, with corresponding doublet to doublet transitions for the surface F"^ centre at 3.47 eV and 4.52 eV. Application of a scaling factor of 0.84, which accounts for the error in the calculated bulk excitation energy, leads to a best estimate of 2.72 eV for the ^Ai^^Ai transition. This is of course higher than the experimental values of 2.3 eV and 1.15 eV discussed earlier. We note in passing that the ^Ai^^Ai transition is surface dipole allowed but the ^Ai—>^E excitations is forbidden and would not be observed in HREELS if a surface dipole scattering mechanism were predominant. There remain at least two further possibilities to account for the low energy surface excitations. The first is that the oxygen vacancies are found at 4- or 3-coordinate step or comer sites rather than 5-coordinate surface terrace sites. The lowest singlet to singlet excitation energies for step and comer F centres are calculated as 2.92 eV and 2.60 eV respectively [74], which scale down to 2.45 eV and 2.18 eV using the correction factor of 0.84. These energies are quite close to the values of 2.0-2.3 eV found in the earlier experimental work. Thermodynamically it is to be expected that oxygen vacancies will congregate at step and comer sites because the oxygen removal energies of 9.00 eV and 8.06 eV respectively are both lower than the value of 9.77 eV for terrace sites [40]. The second possibility is that low energy excitations are associated with F centre aggregates at the surface. Using a semiempirical Hartree-Fock approach Castanier and Noguera found that the energy required to create surface F centres decreased with increasing levels of oxygen deficiency, which can be interpreted in terms of an attractive interaction between oxygen vacancies [75,76]. However, later ab initio density functional calculations demonstrated that the interaction is repulsive [77]. The Hartree-Fock result is probably an artefact of the method, which requires the electron density associated with the F-

565

centre to be accommodated in neighbouring magnesium orbitals. This leads to a large quadrupole moment and attractive interactions between the quadrupoles. The more sophisticated density functional calculations suggest that the F centre electron density is trapped mainly in the oxygen vacancy site and that the localised electron densities repel each other. 4.2 Transition metal rocksalt monoxide (100) surfaces As discussed in section 2.1, the oxides MnO through to NiO are all insulators with localised d" configurations. In the simplest picture the bulk octahedral crystal field of the oxide ions splits the d levels into a lower tag set and an upper Cg set. This ordering is actually reversed in Hartree-Fock calculations on NiO, although the ground state configuration remains tigQg [78,79]. By contrast cations at the (100) surfaces sit in a site of C4V symmetry and the d levels split to give e
566

surface sensitivity. The importance of the surface site symmetry was neatly demonstrated by an adsorption experiment in which the NiO(lOO) film surface was saturated with NO. This molecule adsorbs on Ni^^ surface sites and restores the coordination number of the surface cations to 6. The surface induced splitting of the ^T states is predicted to be strongly attenuated, an effect which is clearly observed experimentally. Assignment of the surface and bulk excitation proposed by Freund and coworkers [83] has been confirmed by spin polarised HREELS experiments [86,87], which also allowed location of some of the singlet excited states. Very recent spin unrestricted Hartree-Fock calculations of bulk and surface d^^d excitations for NiO obtained from direct total energy differences [88] further confirm these assignments and allow additional assignment of a further surface loss at 2.18 eV [89,90] to the ^E component of the two electron excited state. It is worth noting in passing that even though the inversion symmetry is removed at the surface C4V sites, the d^^d excitations that

>* ^

Co'- (d^) Ci.

Oh

-h-I H •, + u-H- -ft-

+-

'>2 e

- ' - • ^ ' %

^A2g

2.(H ^29

'.%

/ ^T.

1.0• o - _ ^

"-J12 ^T„

0.0f

*A2 Co^*

Co'*

CoOf"

CoO|"

free ion

•PC

bulk

surface

CoO experiment bulk

Fig. 3. Schematic correlation diagram showing excitation energies for Co^^ in various environments. Left to right: free ion; ion in point charge crystal field; energies from cluster calculation in bulk site; energies from cluster calculations in surface site; experimental bulk excitation energies. The inset shows schematic one-electron 3d energy levels in bulk and surface sites. Adapted from ref 91.

567

have been observed in HREELS are all dipole forbidden: the ground state is ^Bi and the Ai normal dipole operator in C4V only allows transitions to other ^Bi states. These ideas may be extended to CoO where the results are even more clearcut [91]. The ground state configuration Xi^Q^ for bulk Co^^ ions gives an orbitally degenerate "^Tig ground state, with ^T2g and '^A2g states accessible by one electron excitation. At C4V surface sites, the ground state splits into a lower '*A2 component and a higher "^E component, separated in energy by about 50 meV. The '^T2g excited state splits into a strongly stabilised '*B2 state and a ^E state little shifted from its '^T2g parent. Finally the ^A2g in Oh state correlates with a strongly stabilised "^Bi state in C^v (fig. 3). The low energy 50 meV electronic excitation falls just below the strongly dipole allowed Fuchs-Kliewer surface phonon excitation and is clearly observed on both cleaved single crystal CoO(lOO) and lOA thick CoO(lOO) films (fig. 4). The improved sensitivity to true surface excitations in the latter is clearly apparent and arises from the fact

0.2 0.4 energy loss [eV]

^.06 0.0 mergy loss [eV]

0.06

0.12

Fig. 4. Left hand panel: HREEL spectra from CoO(lOO) up to 1.2 eV loss energy, (a) cleaved single crystal CoO(lOO) excited at 30 eV beam energy (b) clean CoO(lOO) film at 300K excited at 13 eV beam energy (c) CoO(lOO) film after adsorption of lOL of CO at 80K, with spectrum again excited at 13 eV beam energy. Right hand panel: Fuchs-Kliewer phonon region in HREELS of CoO(lOO) excited at 7 eV beam energy, (a) cleaved single crystal CoO(lOO) (b) clean CoO(lOO) film at 300K (c) clean CoO(lOO) film at 80K (d) CoO(lOO) film at 80K following adsorption of lOL of CO at 80K (e) clean NiO(lOO) film at 300K. Adapted from ref 91.

568

that the film thickness is much less than the effective decay length for FuchsKliewer surface excitations on bulk CoO under the experimental conditions employed. At higher energies, a broad feature at 0.45 eV was shown to correspond to the "^62 excitation, whilst structure between 0.7 eV and 1.1 eV contained contributions from both the "^Bi surface state and ^Aig bulk excited state. As with NO adsorption on NiO, CO adsorption on Co^^ surface cation sites at low temperatures essentially restores the octahedral coordination of the surface cations and shifts the surface excitation back toward their bulk values. The difference in the symmetry of the lowest empty states at CoO(lOO) and NiO(lOO) surfaces has interesting consequences in relation to STM imaging of these two materials. The low room temperature conductivity of the two oxides only allows STM imaging on bulk crystals at elevated temperatures. Subject to this difficulty, STM images of NiO(lOO) can be obtained in the constant current mode in the normal way at positive sample bias [92,93], whereas for CoO(lOO) only constant height images can be obtained [94,95]. From these the corrugation at Co(lOO) is estimated to be at least an order of magnitude lower than for Ni(lOO). Calculations within the local spin density approximations with on-site correlation added (LDSA+U) show that for NiO [96] the lowest empty state derives from empty A^z-^ orbitals of ai symmetry in C4V which protrude markedly above the surface. By contrast for CoO(lOO) the lowest empty state derives from 3dxy orbitals of b2 symmetry which lie mainly within the surface plane [94,95]. Computed contours of empty state charge density (fig. 5) make above the surfaces make clear why the bigger corrugation is found for NiO(lOO). The LSDA+U calculations clarify that there is contrast reversal on changing the sample bias in STM of NiO(lOO) because the highest lying levels derive from oxygen pz states, as expected for a charge transfer insulator. The calculations also confirm the assignment of defect features in STM as arising from Ni vacancies. Ni ions are coupled electronically most strongly not to nearest neighbour Ni ions, but to next nearest neighbours via the Ni-O-Ni superexchange pathway (fig. 6). Thus the electronic perturbation associated with an Ni vacancy extends over a (2x2) area [92,93]. The study of surface magnetism provides a subtle way of investigating surface electronic structure. Bulk NiO is antiferromagnetic with a Neel temperature TN=523.6 K. Spins within (111) sheets are aligned ferromagnetically, with spin vectors directed along the <112> direction: the coupling between sheets is mediated by antiferromagnetic superexchange. Paramagnetic NiO has a rigorously cubic structure (Fm3m) but below the Neel temperature the structure is rhombahedral with a slight contraction along one of the <111> axes of the cubic cell. The bulk phase transition is second order with a variation in the sublattice magnetisation proportional to (T-TN)^ where P=0.33. This exponent is

569

0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.0

I I I I I I I I I I I I I I I I I I I I I I I I I I I

CoO, empty states

Fig. 5. Contours of empty state charge density above NiO(lOO) and CoO(lOO) surfaces calculated by summing over a 1 eV interval within the conduction band. Distances are expressed in units of the respective lattice constants. Adapted from ref. 95.

not inconsistent w^ith the value expected for a Heisenberg system for w^hich P=0.345 [97]. Early studies of spin ordering on NiO(lOO) surfaces vv^ere first conducted by Palmberg et al [98-100] in the late 1960s. LEED patterns obtained below^ the Neel temperature contained half order spots consistent vsrith the (2x1) surface magnetic ordering expected from the bulk magnetic structure. The failure to observe superstructure peaks in He atom diffraction confirms that the superstructure spots cannot be attributed to periodic displacements in the ionic positions [101-103]. The variation in the intensity of the half order spots was studied in detail Namikaw^a and cow^orkers [104-106] w^ho took careful account of Debye-Waller factors in reducing the intensity of LEED beams at elevated temperatures. They concluded that the exponent P for the surface magnetic phase transition w^as 0.89, a value higher than expected for both Ising

570

Fig. 6. Left hand panel: empty state STM image (+1.2V sample bias, 1 nA tunnel current) of single point defect on NiO(lOO). The dimensions of the square primitive lattice is indicated by lines. The central defect is seen to produce an electronic perturbation on next nearest neighbour sites. Right hand panel: schematic representation of covalent O 2p - Ni 3d bonding at (100) surface of NiO. Because the two O 2p orbitals on the oxygen sites are orthogonal, the two square sublattices shaded dark and light are electronically decoupled. Thus an Ni vacancy produces an electronic perturbation on next nearest neighbour Ni sites, whereas an O vacancy site most strongly perturbs nearest neighbour O sites. These consideration suggest that the defects imaged in the left hand panel must correspond to Ni vacancies. Adapted from ref. 92.

(P=0.78-0.80) and Heisenberg (P=0.81) models. In none of the early v\^ork was there any indication of differences betw^een surface and bulk transition temperatures [98-100, 104-106]. More recently the surface magnetic ordering on NiO(lOO) has been studied by diffraction of velocity selected metastable ^S excited helium atoms (He*) [97,107-109]. As on MgO, de-excitation of He* at an NiO surface proceeds via the Auger decay mechanism. Where there is electron spin polarisation at a surface, the probability of survival of the He* in its excited state depends on the relative orientation of the spins associated w^ith the surface electrons and the Is electrons of the He*. This is because if the Is electron is "spin up", only "spin down" surface electrons may become involved in the de-excitation process. Thus if surface spins are arranged periodically, the de-excitation of an incoming He* beam will exhibit a periodic modulation. It follows that by studying the angular dependence of elastically scattered He* atoms which survive a deexcitation process at the surface, a diffraction pattern is obtained which reflects the surface magnetic ordering. Observation of half order diffraction peaks along

571 •

1

1

,

r—

1

— I

1

•

I"

1

•\

"5 200

•t

f

Bulk T„=523K

0

*

300

t

350

, 400

, 450

.

1 500

5S

T(K)

Fig. 7. Intensity of half order magnetic peak in He* diffraction from NiO(lOO) as a ftinction of temperature. The bulk Neel temperature of 523 K is indicated. Magnetic diffraction intensity persists above the bulk Neel temperature up to about 529 K. Adapted from ref 97.

<10> but not <11> directions confirmed [97] that the reconstruction is indeed (2x1), whilst the details of the surface magnetic phase transition were revealed by study of the intensity of the (1/2,0) peak as a function of temperature. The surface Neel temperature was found to be 529 K and therefore significantly higher than the bulk value of 523.6K (fig. 7). Superficially this could be taken to indicate that the surface next-nearest-neighbour antiferromagnetic superexchange interaction J2 is bigger than the bulk interaction. Such a reduction was predicted by a cluster model, essentially because of reduction in the charge transfer gap at the surface and enhanced Ni-O-Ni covalency [110]. However, using a model with a Ni20x cluster embedded in a Madelung potential, de Graaf et al concluded that the reduced coordination number at the surface actually leads to reduction of h by about 20% relative to it bulk value [111]. The elevation of the Neel temperature over its bulk value is consistent with this lowering of J2 at the surface provided proper account is taken of the anisotropy of the spin-orbit coupling at a surface site of C4V symmetry [97]. This anisotropy is not present in bulk sites of Oh symmetry and leads to enhancement of the effective surface exchange coupling strength. The He* diffraction data were consistent with a value of 0.125 for the critical exponent in the temperature regime between surface and bulk phase transitions, as expected in the Ising model. However the transition should be formally classified as an anisotropic extraordinary surface phase transition [36]. Transitions of this sort can arise when the effective surface superexchange coupling constant Js exceeds a critical value Js^ and the surface orders magnetically independent of the bulk. Ordinary

572

transitions are observed when Js<Js^. The possibiUty of determining surface magnon dispersion curves onNiO(lOO) has been considered theoretically [112]. Whilst the experiment appears to be feasible with current technology, it has yet to be executed. 4.3 Sn02 The oxide SnO^ adopts the tetragonal rutile structure with a=4.737 A and c=3.360 A. There are two formula units per cell and the Sn ions occupy 6coordinate sites with local Dah symmetry, approximating to Oh: the O ions are trigonally coordinate. The rutile structure is built up from ribbons of Ti06 octahedra sharing edges along the [001] direction and comers in orthogonal directions. The (110) surface of Sn02 has the lowest surface energy [113] and we confine our discussion to this surface. The surface structural chemistry of SnO2(110) is very much more complicated than that of the rocksalt(lOO) surfaces and needs to be understood in some detail before it is possible to discuss electronic structure. Viewed perpendicular to the (110) direction the rutile structure is built up of ionic planes with stoichiometry {0}-{Sn202}-{0}-{0}-{Sn202}- etc. carrying +

+

formal ionic charges {2"}-{4 }-{2-}-{2-}-{4 }. A stable quadrupolar sequence of charged layers [114] is therefore obtained if the outer {Sn202} layer is terminated by an outer {O} layer of oxygen ions. These oxygen ions complete the 6-fold co-ordination of half the surface Sn cations, leaving the remaining Sn cations 5-coordinate. This arrangement gives a (1x1) "stoichiometric" reconstruction. An alternative (1x1) reconstruction can be derived by removing all the bridging oxygen ions (fig. 8). This surface would be electrostatically unstable if the ions all retained their bulk charge, but is stabilised if the formal net charge on the 2xSn ions per surface unit cell is reduced from 8^ to 6^. Since the main oxidation states are 2^ and 4^ and Sn forms a bulk oxide SnO, the most appealing model for the reduction in charge would involve conversion of half the surface Sn"^^ ions to Sn^^. It was established by D.F. Cox et al that annealing SnO2(110) in a high partial pressure (1 torr) of oxygen for several minutes leads to a surface displaying a sharp (1x1) LEED pattern, with an intensity ratio between O and Sn peaks in ion scattering spectroscopy (ISS) characteristic of a surface terminated by bridging oxygen ions [115-117]. Heating to progressively higher temperatures up to lOOOK in UHV leads to a monotonic decrease in the intensity ratio between the O and Sn peaks, although the LEED pattern remains (1x1). This was attributed to loss of bridging oxygen ions to leave the bare Sn202 surface. In support of this hypothesis it was observed that it was possible to label the bridge sites with ^^O by annealing a "bare" UHV annealed surface in ^^02,

573

->[iio]

[001]

(a)

(b)

(c)

Fig. 8. Surface structures for SnO2(110). Large open circles are O: small filled circles are Sn. (a) Stoichiometric 1x1 reconstruction with rows of bridging oxygen ions. All the surface Sn cations are in the +4 oxidation state, (b) "Bare" 1x1 reconstruction in which all the bridging oxygen ions are removed to leave both 4- and 5-coordinate surface Sn cations. To maintain electrostatic neutrality half the surface Sn cations are reduced to the +2 state, (c) One possible model for a (1x2) reconstruction with alternately missing rows of bridging oxygen.

leading to a new^ peak in ISS [117]. An alternative method of surface preparation involves argon ion bombardment followed by annealing in UHV. At least 5 different reconstructions have been observed after bombardment and annealing [118], although the most reproducible sequence observed on progressively increasing the annealing temperature is (4xl)->(lx 1)^1x2)* [119-121]. The reverse change betw^een (1x2) and (1x1) can be induced by adsorption of formic acid followed by desorption at higher temperature [122]. The surface oxygen * We adopt the convention that (4x1) denotes quadrupling the periodicity along the [001] direction and (1x2) doubling along [110].

574

VBM

3-6 VBM

Annealing Temperature

Binding Energy (eV)

(

0 A 8 12 16 Energy relative to Ef/eV

o X

m

>

T3

V

Fig. 9. Top left hand panel: photoemission difference spectra excited at 21.2 eV photon energy for a fiilly stoichiometric SnO2(110) surface heated to the progressively higher temperatures indicated. Note the grow1:h of intensity close to the valence band maximum (VBM). Adapted from ref [115]. Lower central panel: schematic ionic energy level diagram show^ing how the Sn levels found in a centrosymmetric environment evolve on sw^itching on a perturbation V due to a noncentrosymmetric field w^hich allows 5s-5p mixing. Top right hand panel: inverse photoemission spectra of Sn02 as a ftmction of increasing exposure to 500 eV argon ions at 5 mA/cm^. (a) unexposed (b) Ihour exposure (b) 2 hour exposure (c) 3 hour exposure. In (d) the original spectrum is shown with dots emphasising the creation of new states centred 8 eV above the Fermi level. These are attributed to the empty 5s-5p hybrids shown m the central panel. Adapted from ref [214].

575

concentration seen in XPS or Auger spectroscopy increases as the structure evolves through these three reconstructions. It seems safe to assume that the intermediate (1x1) reconstruction corresponds to the bare Sn202 surface discussed above, although the nature of the (4x1) and (1x2) reconstructions remains a matter of debate. D.F. Cox et al [116] suggested that the (4x1) reconstruction involved an SnO(lOl) overlayer whose lattice parameter is 4/3 of that of SnO2(110) along [001]. Hov^ever atomically resolved STM measurements on SnOa support a model involving an ordered arrangement of oxygen vacancies within a "bare" Sn202 plane to give a stoichiometry Sn20i.5. This requires reduction of % of the surface layer Sn(IV) ions to Sn(II) [121]. Yet more recent STM observations suggest that the ordered in-plane vacancy reconstruction may co-exist with the SnO coincidence structure [123]. Likewise, the earlier suggestion [120] that the (1x2) reconstruction simply involved alternately missing rows of bridging oxygen ions must now be questioned in light of the controversies surrounding the structure of the (1x2) reconstruction on TiO2(110) to be discussed below. Despite these ambiguities of detail it is clear that all of the other reconstructions are oxygen deficient relative to the stoichiometric (1x1) and require some reduction of the Sn surface cations. Bulk Sn02 has a direct (but dipole forbidden) bandgap of just under 3.6 eV. Oxygen deficiency in Sn02-x gives n-type donor states about 0.15 eV below the conduction band minimum at the lowest x values, although the Fermi level moves upward with increasing x and moves into the conduction band for donor concentrations above about 10^^ cm"^ [124]. It thus follows that for a flat band surface, the bulk valence band onset in a photoemission experiment should be at about 3.6 eV relative to the Fermi energy. Several groups [115,116,118,125129] have now observed that at SnO2(110) surfaces reduced by UHV annealing and/or ion bombardment and annealing, well-defined photoemission intensity bandgap upon annealing a fully oxidised stoichiometric (1x1) SnO2(110) surface to progressively higher temperature. The appearance of the gap feature does not appear to be structure specific and very similar gap states have also been observed on reduced surfaces of ceramic [130,131] and thin film Sn02 samples [132-136]. Constant initial state photoemission spectra of the gap feature show pronounced resonant enhancement at around 35 eV photon energy, corresponding to the Sn 4d->Sn 5p excitation threshold [14]. Despite earlier claims to the contrary [137], it also appears that surface reduction produces a chemically shifted Sn^^ core level peak [131]. The reason why the Sn^^ defect levels lie close to the valence band edge has proved to be controversial. Based on parameters from the tight binding calculation of Robertson [138], Munnix and Schmeits [139-141] used a scattering method to calculate the surface band structure for the bare (i.e. no bridging oxygen) (1x1) reconstruction of Sn02(l 10). Projecting the surface

576

bands onto the bulk bandstructure, two surface states were found in the so-called stomach gap toward the bottom of the main O 2p valence band. In addition a number of weak resonances were found in the upper part of the valence band. In the conduction band region, three surface states were found, the lowest of which was only 0.1 eV below the conduction band minimum at the T point. Thus the bandgap region was found to be essentially free of surface states. More recent tight binding calculations by Godin and LaFemina [142] reached a similar conclusion. A way out of this dilemma is provided by the suggestion of P.A. Cox et al [130] that the surface state arises from mixing between 5s and 5p atomic orbitals at surface sites where the essentially centrosymmetric coordination of bulk sites is lost, as shown schematically in fig. 9. In general mixing between s and p levels under the influence of a potential V is described by a matrix element <5s|V| 5p> and will be non-zero only if V has a component belonging to the same irreducible representation as one of the 5p orbitals. This is never possible at a site with a centre of symmetry because the 5s and 5p orbitals are of opposite parity and cannot be mixed by centrosymmetric crystal fields. The condition for mixing is however satisfied by tin ions in the 4- and 5coordinate surface sites of the bare (1x1) reconstruction and will generally pertain at any surface cation site where the coordination is reduced below its bulk value of 6. The essential physics of this qualitative model has recently been confirmed [143,144] by first principles density functional calculations on SnOaCllO). In particular Mannasidis et al, [143] found a broad band of states in the bulk bandgap for the fully reduced (1x1) surface: on the half-reduced surface the band of gap states contracts to have a width of 1.5 eV and has maximum intensity about 1.1 eV above the valence band maximum. In the calculations the electron density associated with the gap states is mainly localised above bridging tin sites, with some weight on in-plane oxygens and elsewhere. This accords closely with the chemical picture of a directional 5s5p hybrid lone pair localised on Sn(II) ions. Gradient corrections within the local density approximation had little effect on the relaxed surface structure, although the surface energies are reduced by about 30% [145] The failure of tight-binding calculations to reproduce the observed gap states arises simply from the fact that the tight-binding Hamiltonian has no term which accounts for the 5s-5p mixing and includes no effects due to surface fields [143]. Aside from giving rise to the occupied 5s5p surface state, the hybridisation depicted in fig. 9 should also produce an empty surface state above the main band of empty Sn 5p states. Growth of intensity in this region has been observed in inverse photoemission spectra of ion bombarded Sn02 surfaces [146].

577

4.4Ti02*

4.4.1 Comparison ofTi02 andSn02 Ti02 is another oxide adopting the tetragonal rutile structure with a=4.594lA and c=2.9589A. Despite the fact that the metal oxygen bondlength is smaller in Ti02 than Sn02, metal-oxygen covalency is less pronounced in the former and the dynamic charge derived from analysis of IR reflection spectra is greater in TiOi (0.61 of the formal ionic charge) than in Sn02 (0.49 of the formal ionic charge). This has the consequence that the O 2p valence bandwidth and bandgap in Ti02 (6.3 eV and 3.062 eV respectively) are less than in Sn02 (where the corresponding values are 10.0 eV and 3.596 eV) [147,148]. These differences appear to derive from the fact that the conduction band states in Ti02 are based on relatively contracted Ti 3d atomic orbitals, whereas in Sn02 the conduction bands involve more diffuse 5 s and 5p orbitals. It should be emphasised that the covalency in Ti02 is not insignificant. In particular strong admixture of Ti 3d character into O 2p states toward the bottom of the valence band provokes strong resonance effects in photoemission at the Ti 3p->Ti 3d excitation threshold [149]. Resonance behaviour is also observed at the Ti 3p^'Ti 4s threshold [150]. Of course UV photoemission is intrinsically surface sensitive so that these experiments are in effect probing covalency at surfaces: using on and resonance data the total density of states can be decomposed into atomic components. Courths and co-workers [151,152] have raised the intriguing possibility of exploring differences between partial densities of states derived on the one hand from resonant photoemission and on the other from core and valence level photoelectron diffraction, which is not so strongly surface sensitive. However these ideas remain to be fully exploited. The bulk oxygen vacancy defect chemistry of Ti02 is very different to that of Sn02. In the tin-oxygen phase diagram, Sn02 tolerates only a narrow range of non-stoichiometry and there are no ordered defect structures connecting Sn02 and the lower oxide SnO. By contrast in the titanium-oxygen system, an essentially continuous range of compositions between Ti02 and TiOi.75 (i.e. Ti407) is possible. For very small values of x in Ti02-x, non-stoichiometry is accommodated by point defects, although there is some ambiguity as to whether these are oxygen vacancies or titanium interstitials. In the simplest ionic picture each oxygen vacancy is associated with reduction of two Ti"^^ ions to Ti^^. For x values greater than about 10"^ the defects cluster to give first Ti^^-Ti^^ pairs in face sharing octahedra, which then order into shear planes running along {132} crystallographic directions. The {132} shear planes order at yet higher deviation from ideal stoichiometry to produce members of the homologous series of Magneli phases Tin02n-i (15
578

decreases below 15, the shear planes swing round to the {121} direction, leading eventually to the oxide Ti407. Bulk transport measurements show that for very slightly non-stoichiometric Ti02.x with a carrier concentration in the range below 2.0x10^^ cm"^ the low-temperature activation energy for conduction is about 0.028eV, independent of x. For more strongly reduced material a so-called type b activation energy is found that increases with increasing degree of reduction. This unusual behaviour is attributed to an aggregation of point defects into shear planes with an enhanced trapping energy at shear plane sites [147]. However, even at the highest carrier concentration the activation energy is below 0.1 eV. Oxygen deficiency is associated with strong absorption in the visible/near IR region, peaking at about 1 eV. The optical absorption relates to the vertical excitation of trapped Ti 3d electrons, whereas the lower activation energy for conduction depends on adiabatic processes [148]. 4.4,2 TiO2(110) TiO2(110) has been more extensively studied by the contemporary techniques of surface science than any other oxide surface. The surface displays an even more complex range of behaviour than SnO2(110). A (1x1) reconstruction is obtained by cleavage in UHV [153] or more easily by a wellestablished recipe involving argon ion bombardment and subsequent annealing in modest pressures of oxygen, of the order of 10'^ mbar. The structure of the (1x1) reconstruction has been established in detail by surface X-ray diffraction [154] and is based on the "stoichiometric" termination discussed in section 4.3. There is some rumpling in the Ti202 layer, together with pronounced inward relaxation of the bridging oxygen ions by something of the order of 0.27 A. The earliest first principles calculations [155,156] failed to reproduce the large inward relaxation, and even in more recent work using the local density and generalised gradient approximations (LDA and GGA) to density functional theory the discrepancy remains [157-160]. This problem may possibly arise from a soft anisotropic and anharmonic surface vibrational mode that complicates analysis of the surface X-ray data [161]. In general the calculated surface related O 2p bands do not fall outside the range of energies spanned by the bulk O 2p bands [155,157,162]. This is in accord with the observation that the valence band onset in photoemission is at around 3 eV [153], as expected from the bulk bandgap in a flat band situation. Given the complexity of the bulk bandstructure it has not been possible to disentangle surface and bulk bands within the O 2p valence band in angle resolved normal photoemission experiments, which for TiO2(110) have been interpreted solely in terms of the occupied bulk bandstructure along the FZM line [163,164]. An interesting theoretical prediction that there should be a well-defined surface O 2s band associated with the bridging oxygen split off from the top of the bulk bands

579

[155,156] deserves further experimental attention. As regards empty states, some features in angle resolved inverse photoemission spectra from stoichiometric TiOiCllO) (1x1) surfaces [165] have been interpreted in terms of empty state surface resonances predicted by Munnix and Schmeits [166] on the basis of parameterised tight binding calculations. Terrace structure on TiO2(110) (1x1) has been observed with atomic or unit cell resolution by many groups in the past few years [167-180]. There is a general consensus that under conditions of positive sample bias the images are dominated by bright rows running along the [001] direction with a separation of V2a between the rows in the orthogonal [110] direction. The issue as to whether the rows correspond to bridging oxygen ions or 5-coordinate has provoked lively discussion. As demonstrated by XRD, the O ions sit about 1.5 A above the Ti ions in the Ti202 surface plane. However the empty electronic states that electrons tunnel into are predominantly localised on Ti. There is thus a delicate interplay between geometric and electronic effects. Calculations of contours of empty state density above the surface indicate unless the tip is very close to the surface, the Ti sites should appear as the grey scale maxima [178,181,182]. This interpretation is supported by the observation that formic acid [183] and chlorine [184] give rise to enhanced intensity on the bright rows. Chemically one expects both species to adsorb on bare Ti sites, thus supporting the view that electronic structure effects are dominant in imaging the bare surface. Under conditions of low sample bias and high tunnel current, when the tip must be closer to the surface than is normal, it is possible to image both bridging oxygen and in-plane Ti positions [185], in accord with the empty density of states contours [178,181]. Ti ions may also appear as greyscale maxima under abnormal tip conditions [186]. It has long been established that oxygen deficiency at TiO2(110) surfaces is induced by a number of different treatments including annealing in UHV at elevated temperatures [187-191], argon ion bombardment [149, 187-189, 191197] electron bombardment [192,198-202], laser irradiation [203,204] and deposition of submonolayer quantities of Ti [205,206]. At a microscopic level substoichiometric humps can be generated by operation of an STM under conditions of high sample bias and relatively low tunnel current. If high tunnel current is combined with high sample bias, crater-like defects with diameters of the order of 10 nm are generated [207]. The oxygen deficient surfaces are characterised by the appearance of a feature in photoemission within the bulk bandgap, typically with a binding energy of around 1 eV (fig. 10). In addition a new electronic excitation appears in EELS: the energy of this feature depends somewhat on the level of oxygen deficiency and on the energy of the electrons used to excite the spectra, but again energies are typically around 1 eV. Finally

580

N(E)

10 5 Binding Energy (eV)

Fig. 10. Photoemission spectra of 0-deficient TiO2(110) excited different photon energies, showing resonant enhancement in the intensity of the peak in the bandgap at around 1 eV binding energy when the photon energy is 47 eV. Adapted from ref. 149.

chemically shifted Ti core level features are observed to low binding energy. These observations are all broadly in line with the expected population of Ti 3d states at oxygen deficient surfaces. In particular the gap feature observed in UPS shows resonance enhancement at the Ti 3p core threshold [149] whilst the sensitivity of Ti 3p shallow core photoemission to defect states is enhanced by measuring spectra in coincidence with Ti M2,3VV Auger spectra [208]. Recent interest has raised the possibility of trying to establish in more detail the relationship between the observed spectral features and the local structure associated with oxygen deficiency. In principle oxygen deficiency may be accommodated as point defects by removal of surface bridging oxygen atoms or by removal of in-plane or subsurface oxygen atoms. However, as in the bulk, shear planes are expected to form in the near surface region if the level of oxygen deficiency becomes high. Finally there are possibilities for formation of new superstructures at oxygen deficient surfaces. We consider these possibilities in turn. The defects produced by argon ion bombardment probably involve a range of different and ill-defined local environments, but it has been argued that thermal annealing stoichiometric surfaces at elevated temperatures or exposure of them to low flux electron beams leads to preferential removal of bridging oxygen atoms: the maximum defect concentration produced by thermal

581

treatment alone corresponds to removal of about 33% of the bridging oxygen ions [202]. These defects expose underlying Ti ions and appear as bright features linking the original bright rows in STM [179]: the defects are quenched by exposure to oxygen or by oxygen transfer from the tip of the STM. Breaks in the bright rows [174] are now believed to correspond to subsurface oxygen vacancies [179]. In an early theoretical paper, Munnix and Schmeits [209] argued that isolated bridging oxygen vacancies do not give rise to occupied Ti 3d states, whereas subsurface oxygen vacancies do. This idea has been called into question by more recent theoretical work. Thus Ramamoorthy et al [157] using first principles density functional theory found that the bare (1x1) surface with all bridging oxygen ions removed was essentially metallic with an occupied band of Ti 3d states merging into the empty Ti 3d conduction band without a gap. By contrast Mackrodt et al. [210] concluded on the basis of unrestricted Hartree Fock calculations that the bare (110) surface should support a localised antiferromagnetic spin arrangement with the excess charge localised on surface Ti ions. Unfortunately neither of these theoretical approaches has dealt with isolated vacancies. Experimentally the nature of vacancy induced states has now been examined by angle resolved photoemission measurements on oxygen deficient TiO2(110) (1x1) surfaces prepared by prolonged annealing in UHV [211,212]. Although it was assumed that the (1x1) structure involved one vacancy per surface unit cell, it seems more likely that there was a lower concentration of randomly distributed vacancies. The weak dispersion of the vacancy induced states with parallel wavevector (fig. 11) was cited as evidence for the localisation of the states on specific lattice sites by polaronic lattice distortion. This conclusion is supported by rough estimates of the polaronic self trapping energy based on a very simple continuum dielectric model [197] or by rather more sophisticated calculations taking specific account of relaxation of oxygen ions around the centre trapping an electron via an effective local force constant [213]. Within the polaron model the strong excitations seen in EELS essentially corresponds to excitation of a self-trapped electron from a Ti^"^ site to an adjacent Ti"^^ site. Excitations of this sort are strongly dipole allowed and carry much greater oscillator strength than on-site 3d->3d transitions of the sort discussed in section 4.2. The polaron model is thus in accord with the high value of 0.1 for the oscillator strength of the EELS excitation inferred from analysis of the effects of defects on surface phonon frequencies in HREELS [201]. We note in addition that simple continuum models also account in a satisfactory fashion for differences in vertical ionisation energies and photoemission peak halfwidths between bandgap peaks associated with V"^^ dopant in Ti02 and Sn02 [148]. These differences arise from the large differences in dielectric constant between the two materials. Finally defective TiO2(110) surfaces have been studied by inverse photoemission spectroscopy. In addition to features observed leV and 4

582

-0.7 -0.8

>:-o.9 TiO.(llO) ^ , ^ „, hv"=22eV

II101

A

-PL

PC

Z-0.8i -0.9 -\

binding energy / eV

-1.0 -I

Fig. 11. Left hand panel: bandgap feature in photoemission (hv = 22eV) of thermally annealed TiOaCllO) as aftinctionof exposure to oxygen (a) OL (b) 10 L (c) 100 L (d) 2500L. Oxygen is seen to quench the gap state. Right hand panel: dispersion of the bandgap feature along rX and FY directions. Adapted from ref. 211.

eV above the Fermi level for perfect surfaces [165,214], new features appear at 3 eV for defective surfaces [215]. These are predicted by the calculations of Munnix and Schmeits discussed above. We turn next to consider shear planes at TiO2(110) surfaces. The earliest STM experiments on Ti02(l 10) surfaces that had been subject to very prolonged high temperature annealing in UHV revealed structures that w^ere attributed to shear planes [216,217], although this interpretation was later called into question [155]. More recently Norenberg et al [218] observed periodic steps on Ti02(l 10) that were consistent with formation of {112} shear planes. However, shear planes of this sort are not a feature of the bulk defect chemistry and the interpretation of these experiments was complicated by the simultaneous segregation of Ca to the (110) surface [219]. Most recently Bennet et al [220] have provided clearcut LEED and STM evidence for formation of the expected {132} series of shear planes at TiO2(110) surfaces following UHV annealing at 1223K for several hours. It was predicted many years ago that the 3d electron states associated with trapping of electrons at shear plane sites are lower in energy by about 0.3 eV as compared with simple self trapped electron states [221]. It would therefore be of interest to compare photoemission spectra from surfaces where shear planes provide the dominant mode of accommodation of non-stoichiometry with spectra from surfaces where point defects predominate. This experiment remains to be performed. Similarly there is considerable scope

583

for careful STS experiments which could in principle establish differences (if any) between the Ti 3d electronic levels associated with bridging and subsurface oxygen point vacancies and Ti 3d levels at shear plane sites. Early STS experiments on the effects of oxygen deficiency on the electronic structure of TiO2(110) surfaces were mostly performed under conditions where the topographic resolution was rather poor [222,223]. Nonetheless, oxygen deficiency was seen to produce clear signatures in lA'^ spectra. At stoichiometric surfaces, the spectra showed typical n-type semiconductor behaviour with significant empty state tunnelling currents at low positive sample bias, but essentially zero tunnelling current at negative sample bias. However upon reduction, filled state tunnelling was observed, the spectra from ion bombarded surface being almost metallic in appearance. Following these experiments little effort seems to have been devoted to STS under conditions where good topographic resolution is obtained on Ti02(l 10) and further work in this area is urgently needed. Finally we consider reconstructions induced by oxygen deficiency on TiO2(110). Moller and coworkers first characterised a (1x2) reconstruction on the (110) surface [224]. The initial model for TiO2(110) (1x2) was based on missing rows of bridging O ions, as discussed above in the context of Sn02(l 10) (1x2). Electron book-keeping dictates that if there are bridging oxygen vacancies, some surface or near surface Ti"*^ ions must be reduced to Ti^^. In agreement with these ideas strong bandgap intensity was found in photoemission from the (1x2) surface. Atomically resolved STM images appeared initially to provide support for the missing row model [175] but also gave evidence for significant lateral relaxations away from bulk terminated positions. However several problems were encountered with the model. Firstly, LDA calculations predicted that a (2x1) arrangement of bridging oxygen vacancies would be more stable than a (1x2) arrangement [157*, 225]. Secondly and most surprisingly Moller and coworkers in a second paper contradicted their own initial observations and found that a (1x2) surface could be prepared which essentially gave zero bandgap intensity in resonance photoemission [150]. Thirdly it was observed that the (1x2) reconstruction grew as an overlayer on top of the (1x1) surface when a reduced subsurface was annealed at elevated temperatures either in UHV or oxygen [170,173,178,226]. On the basis of dynamic STM experiments at high temperatures Onishi et ah suggested a radically different model for TiO2(110) (2x1) involving added Ti203 rows [170,173]. This model was shown to be consistent with measurements of O^ electron stimulated desorption ion angular distributions (ESDIAD), which revealed two parallel stripes rather than the single stripe expected for the missing row model [227]. Note however that in reference [157] the (1x2) reconstruction discussed here is called (2x1).

584

The added rows of the Onishi model are oxygen deficient with respect to a stoichiometric termination of the surface and simple considerations therefore suggest that 3d^ states associated with Ti^^ must be produced at such a surface. It is not therefore possible to reconcile this model with MoUer's observations [150]. The assertion that the (1x2) reconstruction supports no bandgap states has however itself been called into question by Patel et al [228], who found pronounced Ti 3d intensity in the gap in photoemission spectra excited with 47 eV photons. But this study in turn threw up the puzzling result that the bandgap intensity was not associated with chemically shifted Ti 3p structure in shallow core photoemission. More recently Thornton and coworkers have suggested a fiirther alternative model for the (1x2) reconstruction [229]. This involves added rows in which the Ti and O positions are close to those for a bulk terminated surface, but the added rows were presumed to terminate without bridging O ions. Thus we have another "reduced" termination, but unlike in the Onishi model it is possible to envisage a stoichiometric version of this reconstruction with bridging oxygen left in place. Thornton's model was later criticised by Tanner et al. [185], who carried out atomically resolved STM measurements that achieved better resolution of structure within the unit cell than in earlier work. They claimed that the Thornton model gave the wrong registry between (1x1) and (1x2) reconstructions. However, this claim has in itself provoked fiirther controversy [230,231]. The most recent dynamic elevated temperature STM measurements suggest that the two different added row (1x2) reconstructions may co-exist [232,233]. Further complexities in the surface chemistry of TiO2(110) have been uncovered by Diebold and coworkers who have made extensive and systematic studies of the "restructuring" of TiO2(110) surfaces by annealing bulk reduced (and therefore blue!) crystals in oxygen or UHV [234-238]. Restructuring arises by elimination of excess Ti present as either interstitials or in shear planes in the sub-surface region. Lightly reduced crystals only display (1x1) reconstructions after UHV annealing, whereas more heavily reduced crystals display areas of a (1x2) reconstruction that probably correspond to Onishi Ti203 added rows. The most heavily reduced crystals give rise to pseudo-hexagonal rosette structures upon annealing in oxygen. The rosette structure involves an incomplete Ti02 layer with both Ti and O vacancies. However the remaining ions occupy essentially bulk-like positions. The rosette overlayer is essentially stoichiometric and supports no occupied bandgap states. However, an LDA electronic structure calculation of the rosette overlayer predicted a significant widening of the bandgap as compared to the bulk value due to enhanced Ti-0 covalency within the rosettes [236]. In summary we conclude that despite the enormous effort devoted to the study of Ti02(l 10), the relationships between geometric and electronic structure

585

are not entirely clearcut. In particular there is still scope for definitive spatially resolved scanning tunnelling spectroscopic measurements that could in principle define the distinctive features of electronic structure associated with missing bridging oxygen atoms and subsurface oxygen vacancies on TiO2(110) (1x1); the differing (1x2) reconstructions; oxygen deficient shear planes in their intersection with the (110) surface; and the stoichiometric rosettes. It is probably impossible to prepare a surface where these different features do not co-exist to some extent, so that photoemission is too crude a tool to explore these subtleties of electronic structure. 4.43 TiO2(100) The TiO2(100) surface provides another example of the way in which understanding of the relationship between the geometric and electronic structure associated with defects at surfaces has evolved in the past 10 years. The earliest experiments on cleaved (1x1) TiO2(100) surfaces found very weak photoemission intensity in the bulk bandgap at hv=21.2 eV [153], presumably associated with defects generated during cleavage. In previous work it had been found that bombardment and annealing of TiO2(100) surfaces led to formation of superstructures, originally described as (1x3), (1x5) and (1x7) [239,240], the latter two associated with 3d electron states in the bulk bandgap. However, subsequent work showed that there are probably only two reconstructions, both of which can be produced by ion bombardment followed by well-defined annealing procedures. A stoichiometric (1x1) surface was obtained by annealing a bombarded surface in 10"^ mbar O2 at 870 K for 30 minutes, whereas a (1x3) surface was produced by annealing at the higher temperature of 1020-1070 K in UHV. Most surprisingly it was found that a small miscut which would have been expected to produce steps along the [001] direction led to LEED spot splitting in the [010] azimuth: the spot splitting may be misinterpreted in terms of higher order (Ixn) reconstructions [241]. Photoemission spectra measured at the resonance energy for the Ti 3p core threshold hv=47 eV (fig. 12) showed that the (1x1) surface supported no bandgap states, whereas for the (1x3) surface there was pronounced emission intensity in the gap [241]. The gap state also showed resonance enhancement at the Ti 2p core threshold [242]. Similar bandgap intensity is produced by Kdeposition [243,244]. A simple model for the (1x3) reconstruction was put forward, involving removal of every third row of on-top oxygen ions from an essentially bulk terminated (100) surface. The intensity of the bandgap emission was consistent with this model. However, grazing incidence X-ray diffi*action [245] led to development of a radically different model for the (1x3) surface

586

1 8

1

1

r

6 4 2 Binding Ener©^ (eV)

Fig. 12. Normal photoemission spectra of the (1x1) and (1x3) reconstructions of TiO2(100) excited at photon energies indicated. At 47 eV there is resonant enhancement of structure associated with Ti 3d states. The angle integrated spectrum at 47 eV is also shown. Adapted fromref 241.

based on {110} nanofacets*: the driving force for this reconstruction is simply the lower surface energy of rutile (110) as compared with rutile (100). Atomically resolved STM images of TiO2(100) (1x3) are consistent with the nanofacet model [247,248], which also provides a simple rational for the updown steps along [010] in terms of maximisation of the area of {110} facets [249]. The nanofacet model suggested on the basis of the XRD data involved an essentially stoichiometric termination and was not therefore consistent with the photoemission data. Thornton and co-workers therefore suggested that oxygen ions on the ridges of the nanofacets must be removed to give "bare" Ti ions at the top of the ridges [247,248]. Charge book-keeping then requires that some Ti ions in the near-surface region must be reduced. In agreement with the valence photoemission data, Ti 3p shallow core level photoemission spectra showed a chemically shifted component associated with Ti^^ for TiO2(100) (1x3). The * In the original literature the facets are referred to as microfacets. However, the length scales involved are in the nanometre range and we prefer the term nanofacet [246].

587

bare Ti ions on top of the ridge are the obvious candidates for reduction to Ti^"^. To explore this possibility, the azimuthal angular dependence of the intensity of the chemically shifted Ti 3p emission was measured and compared with photoelectron diffraction (PED) curves calculated using a multiple scattering formalism [250]. Qualitatively, computed PED curves with O vacancies and Ti^^ ions on the tops of the ridges gave better agreement with the experimental data than could be achieved with models assuming O vacancies on other possible sites on the nanofacets. R factor analysis confirmed this conclusion. More recent STM work has investigated surfaces where the transformation between the stoichiometric (1x1) reconstruction and the facetted (1x3) reconstruction is not complete. Two intermediate reconstructions with (1x3) periodicity have been identified and a plausible mechanism for their interconversion has been suggested [251]. These ideas have been further refined in the light of very recent non-contact atomic force microscopy (AFM) [252,253] and further surface XRD data [254] which has explored the magnitude of relaxations within the nanaofacet structure. On the basis of the discussion thus far, it appears that there is an excellent understanding of the relationship between the structure of TiO2(100) (1x3) surfaces and the nature of defect states. However two issues do remain to be resolved. First, in the light of the STS data for oxygen deficient SrTiOs alluded to in section 3.3 it might be anticipated that spatially resolved scanning tunnelling spectroscopy would provide evidence for occupied electronic states just below the Fermi energy on top of the ridges of the nanofacets. In fact the tunnelling spectra show that on the rows there is no filled state tunnelling and the I-V spectra are completely dominated by tunnelling into empty states. By contrast filled state tunnelling is found for sample biases negative of -1 V between the rows [247,248]. Second, a completely new model for the structure of Ti02 has been proposed on the basis of the most recent analysis of X-ray diffraction experiments [255]. An ab initio direct methods approach was used to re-analyse the original surface X-ray diffraction data set. The direct methods solution to the phase problem is widely used in determination of bulk structures from XRD data and leads to electron density maps without presupposition of a trial structure. The alternative model for the (100) surface derived in this way involves up-down topography with 3a periodicity. However the structure is quite distinct to that of the nanofacet model, which retains essentially the same linkage between TiOe octahedra as is found in the bulk structure. The new model involves extensive O deficiency in the near surface region and fusion of TiOe octahedra to eliminate O point defects in a fashion similar to that found in bulk shear planes. Clearly the new structure will be associated with bandgap states in photoemission but it remains to be seen if the model can be reconciled with the observed photoemission intensity, with the shallow core photoelectron diffraction data discussed above and with the extensive body of STM and AFM observations.

588

4.5W03andNaxW03 4,5.1 WOsiOOl) Tungsten trioxide (WO3) is a prototype d^ oxide based on a framework of comer sharing W06 octahedra. The formal oxidation state of tungsten is 6^, although due to strong covalency, the W ions do not carry the full 6^ charge. The neighbouring oxide ReOs has a simple cubic structure with an Re-O-Re angle of 180°, but in WO3 rotation and distortion of the WOe octahedra gives rise to a series of lower symmetry phases: there are at least five bulk phase transitions between 100 K and lOOOK. A monoclinic phase Y-WO3 is stable between 290K and 603K. It has cell parameters a = 7.297A, b = 7.539A, c = 7.688A and P = 90.91° and belongs to the space group P2i/n. The y-WOs structure represents a (2x2x2) superstructure based on an idealised cubic unit cell of dimension about 3.7A [256]. Crystals of WO3 cleave most easily to expose [001] surfaces, which because of the monoclinic symmetry are distinct from [100] and [010] surfaces. Stoichiometric monoclinic tungsten trioxide is an insulator at 300K with a band gap of about 2.6eV. However, WO3 readily becomes oxygen deficient to form WO3-X with variable oxygen composition parameter x. This oxygen deficiency greatly influences the bulk electronic transport properties by introducing donor electronic states into the upper half of the bulk bandgap: charge bookkeeping requires that each oxygen vacancy converts two W(VI) ions to W(V). For defect concentrations of x >10''^, point defects are eliminated by the formation of crystallographic shear planes belonging to the {ImO} series. These involve edge-sharing W06 octahedra [257]. In general the higher the degree of oxygen deficiency, the smaller m. Thus for very lightly reduced WO3, the shear planes are {120}, swinging round toward {010} for very highly reduced material. Electron spin resonance measurements on WO3.X (x>10"'^) in the low temperature 8 phase indicate that the 5d^ electrons associated with reduced W^^ centres are paired to give bipolarons [258]. The lattice distortion within the polaronic centres probably involves motion of the two W^^ ions toward each other. A simple rationale for the driving force for shear plane formation therefore involves bond formation between W^^ ions in edge-sharing pairs. A number of early studies by X-ray photoelectron spectroscopy (XPS) demonstrated that surface reduction of WO3 by thermal treatment in UHV or by exposure to electron and ion beams or UV photons could be monitored by changes in W 4f core level spectra [259-265]. Subsequently Bringans and coworkers used UV photoemission to follow changes in valence band spectra resulting from electron or ion bombardment of WO3(001) [266-267]. It was found that reduced surfaces were characterised by a broad and generally illdefined band of W 5d states extending from the valence band edge to the Fermi level: The overall spread of these defect induced states is much greater than for

589

TiOi. Annealing the highly oxygen deficient surfaces in a significant partial pressure of oxygen quenched the W 5d intensity, but annealing in UHV alone led eventually to a surface which was characterised by a sharp W 5d peak close to the Fermi energy. There was no attempt in any of this earlier work to relate the features observed in XPS or valence photoemission to specific atomic arrangements at the surface and indeed there was no real prospect of pursuing such ideas at that time. However it has recently been possible to combine STM and STS on WOsCOOl) surfaces with resonant photoemission [268]. The idealised cubic structure of WO3 may be described alternatively in terms of alternating layers of stoichiometry {0}-{W02}-{0}-{W02}- stacked along the [001] direction. These carry formal ionic charges {2-}-{2+}-{2-}-{2+}. To avoid termination of the [001] surface with a repeating normal dipole it is necessary to remove half the oxygen ions in the outermost oxygen ionic layer to give a sequence {Oo.5}-{W02}-{0}-{W02}. The charges can then be grouped into repeating quadrupolar units (fig. 13). Atomistic simulation show that the favoured configuration for the half monolayer of on-top oxygen involves an ordered c(2x2) (that is (V2xV2)R45°) arrangement [269]. Thus the "stoichiometric" [001] surface displays a c(2x2) reconstruction, which has been imaged by STM [269-271]. As with TiO2(110) (1x1) it was initially unclear whether the uppermost O ions or the bare metal cations in the WO2 plane appear as greyscale maxima at positive sample, although a simple consideration

.0.0

m •a ,-0

o---

••--C'-'o-

„o

O0.5

--0''

2+

o -o -O --

•O o

1-

:0 O

12+

.0 .0-

o-

::0 O -O"

1-

1-

} }

}

Fig. 13. Schematic representation of the structure of WO3 in terms of layers of ionic planes. A stoichiometric (001) surface cannot terminate in either a complete O or WO2 plane as this gives rise to a repeating dipole normal to the surface. The preferred termination with a half monolayer of on-top oxygen to give a c(2x2) reconstruction is shown in the left hand panel.

590

of the extent of W-O found in bandstructure calculations [272] suggested that there should be sufficient W 5d character in the states associated with the on-top oxygen for them to predominate. Very recently Altmann and co-workers have observed that alkoxide ions adsorb on the sites which appear as greyscale minima in STM images [273]. Chemically it only makes sense to assume that alkoxide species adsorb on bare cation sites: this in turn implies that the greyscale maxima must correspond to on-top oxygen ions. In contrast to TiO2(110) (1x1) it therefore appears that STM images are dominated by geometric rather than electronic structure effects. This difference must be attributed in part to the fact that the geometric height difference between the ontop O and the W ions is greater than for TiOaCllO). Additionally WO3 is more covalent than Ti02 so there is greater admixture of O 2p character into conduction band states. STM reveals that the oxygen deficiency produced by ion bombardment and/or UHV annealing of WO3(001) surfaces is initially associated with striking "trough defects" that dissect the c(2x2) terraces (fig. 14). The dominant photoemission intensity from these surfaces lies deep within the bulk bandgap with maximum intensity at about 1.5 eV binding energy [268]. This binding energy is similar to that associated with W-W a- bonding states in the oxide WO2, whose rutile-like structure is based on edge sharing WOe octahedra [274].

Fig. 14. 280 A X 280 A STM images of WOsCOOl) acquired at +2.0V sample bias and InA timnel current. Left hand panel is of a typical surface showing a c(2x2) reconstruction dissected by dark "defect troughs". Right hand panel shows a surface after prolonged annealing in UHV. The bright "raft" running diagonally across the image supports a p(lxl) reconstruction. Adapted from ref. 268.

591

Topographically resolved scanning tunnelling spectroscopy confirms that the occupied electronic states are associated with the trough like defects, which are therefore believed to correspond to shear-plane structures intersecting the surface (fig. 15). Metal-metal bonding between adjacent W ions within the shear planes pushes the states deep down into the gap [268,270,271]. Prolonged annealing of WOsCOOl) in UHV leads eventually to the growth of rafts on the c(2x2) terraces which support a new (1x1) reconstruction as shown in fig. 14 [275]. The simplest model for this reconstruction involves a termination in a bare WO2 layer which can be accommodated on top of the stoichiometric c(2x2) reconstruction provided that half a monolayer of oxygen ions are added to the O0.5 layer. Thus the overall stoichiometry of the raft structure relative to that of the stoichiometric surface is WO2.5 and the W ions are all formally 5d^ species in a +5 oxidation state. The raft structure appears to grow on the c(2x2) terraces by elimination of W-rich shear planes from the nearsurface region: this growth by net transport of cations from the bulk is analogous to the model proposed for growth of added Ti203 rows on Ti02(l 10) (1x2).

>

-n

1\

« 1

1 2 - 2 - 1 Sample bias / V

Fig. 15. Scanning tunnelling spectra (dl/dV versus sample bias) for WO3(001). Left hand panel: solid line, c(2x2) terraces; dashed line, defect troughs. Right hand panel: p(lxl) rafts. Partly adapted from ref 268.

592

Fig.l6. STM images of WO3(001) (1x1). Left hand panel: 30 A x 30 A empty state image taken at +2.0V, InA. Right hand panel: 25 A x 25 A filled state image taken at -0.4V, InA. Filled state imaging of this reconstruction is possible because of the metallic nature of the surface. Adapted from ref 275

0 - 1

3

binding energy / eV

Fig. 17. Resonant photoemission at 57.5 eV photon energy from WOsCOOl). Left hand panel: surface with defect troughs, giving defect states predominantly deep in the bandgap at 2 eV. Right hand panel: surface supporting large areas of the p(lxl) reconstruction. There is now a stronger peak at the Fermi energy with a metallic density of states. Adapted from ref 275.

593

Tunnelling spectra from the (1x1) rafts are characteristic of an essentially metallic material (fig. 15). In agreement with this observation it is possible to image the (1x1) reconstruction at both positive and negative sample bias, as shown in fig. 16. Photoemission spectra such as shown in fig. 17 from surfaces demonstrated by STM to support large areas of the (1x1) reconstruction are different to those where shear plane defects are dominant: the deep gap state observed for the latter is much attenuated and there is higher intensity near the Fermi energy. This suggests that the (1x1) surface is indeed metallic, in agreement with the conclusion from scanning tunnelling spectra, from (1x1) [275]. 4.5,2 Na^WOsiOOl) Introduction of sodium into empty 12-coordinate sites of the WO3 lattice results in a series of oxide bronzes. These have general formula NaxWOs, where 0.0<x<0.9. For low values of x the sodium tungsten bronzes retain the monoclinic structure of the parent WO3 and they are n-type semiconductors [276,277]. With increasing Na content the structure evolves through two distinct tetragonal phases and for x>0.43 the bronzes adopt an essentially cubic perovskite structure [278] closely related to that of Re03 (fig. 18). The Na 3s levels lie about 10 eV above the bottom of the W 5d bands and each added Na atom is therefore ionised to Na^, with donation of one electron into the W 5d band of local t2g symmetry [279]. For x values of less than 0.26, the 5d electrons are localised, probably by an interplay between polaronic effects, disorder =3.84A x=0.665

r*i O-

Fig. 18. Schematic representation of the structure of NaxWOs. WO2 and NaxO surface planes derived by bulk truncation are shown on the right.

594

induced (Anderson) localisation and on-site Coulomb repulsion [280,281]. However the cubic bronzes have metallic conductivities comparable with that of Cu. In this regime, the conduction electrons occupy a band which has an almost free-electron-like density of states, although the width of the band does not show an ^'^ variation with carrier concentration because Na intercalation leads to simultaneous narrowing of the W 5d conduction bands [279,282]. In addition the disorder and on-site Coulomb repulsion induce a minimum in the density of states or pseudogap at the Fermi energy [283,284] which may be seen in tunnelling spectra [285]. The gap becomes deeper as the metal to non-metal transition is approached and becomes apparent even at much lower resolution in photoemission for x values of around 0.4 [280,281]. In the metallic cubic phase the gap between the bottom of the occupied conduction band and the top of the valence band is much smaller than in the distorted monoclinic phase of WO3. Bandstructure calculations suggest that the bandgap narrowing could be attributed in part to structural effects [272, 286]. It should also be realised that bandgap shrinkage is a general phenomenon in passing from a non-metallic to a metallic state even when there is no structural change: the bandgap renormalisations arises from a Coulombic attraction associated with the donor ions and self-screening of the repulsion between valence and conduction electrons by the conduction electron gas [287]. Thus shrinkage of the intrinsic gap between valence and conduction bands has been observed in photoemission from systems such as CdO [288,289] and Sn02 [290] which may be doped to become metallic without structural change. In the present context, the sodium tungsten bronzes are of particular interest in view of the claim by Hochst et al that pronounced intrinsic surface states intensity is found in photoemission between the O 2p band and the metal conduction band at vacuum cleaved NaxWO3(001) surfaces [291]. It was suggested that the mutual Coulomb repulsion that pushes states out of the gap for insulating oxides [292] is screened out in a metallic oxide, thus allowing the emergence of gap states which appeared as strong features in He(I) photoemission spectra. The gap states showed pronounced dispersion with wavevector in the surface plane, as expected for a delocalised surface state. However it emerged that that most of the gap intensity in this work could be ascribed to a simple experimental artefact, namely satellite structure associated with He(I)p emission in the He lamp used to excite the photoemission spectra. On carefully vacuum annealed (001) surfaces the gap intensity in He(I) photoemission was found to very weak [293]. As with Na2/3WO3(110) [294] most of this intensity is attributable to a plasmon loss satellite of the conduction band peak. However, the experiments on the (001) surface [293] suffer from the ambiguity that the surface reconstruction was described as p(2x2). In the light of

595

the STM data discussed above this must correspond to the co-existence of c(2x2) and (2xl)+(lx2) reconstructions. STM and LEED demonstrates that NaxWOsCOOl) (with x values around 2/3) surfaces annealed to high temperatures above 600°C in UHV support a c(2x2) reconstruction analagous in most ways to the c(2x2) reconstruction of WO3(001). However, because the tungsten bronzes are metallic it is possible to image the reconstruction at both positive and negative sample bias, as shown in fig. 19 [295]. Very weak greyscale maxima appear in the centres of the squares that define the (V2xV2) R45° cell at positive sample bias: these are strongly enhanced at negative sample bias. The interpretation of these observations is that on-top oxygen atoms are imaged as the strong grey scale maxima and the bare W ions in the underlying WO2 plane appear as subsidiary maxima: the position of the Na ions is unclear from the STM images. The enhancement of the W ions at positive sample bias arises from an interesting aspect of electronic structure. The extent of mixing of O 2p states with W 5d states in the conduction band

-i

10

1

1

20

1

1-

30

Distance along [100] /Angstrom

Fig. 19. Right hand panels: STM images of the (V2xV2)R45° reconstruction on Nao.67W03(001) acquired at -0.6 V sample bias (upper) and +0.6 V sample bias (lower). Corresponding corrugation profiles are shown in the left hand panels, emphasising the fact that W ions in the centres of the squares of O ions which define the reconstruction are more pronounced in the filled state image. Adapted from ref 295.

596

increases in moving up through the t2g band: in the bulk inversion symmetry means that there is no mixing at all at T point where the bottom of the t2g band is found. Thus the W 5d states below the Fermi level that are interrogated at negative sample bias have less O 2p admixture than the empty states above the Fermi level that are involved in images taken with positive sample bias. This effect increases the contribution the W ions relative to on-top O in filled state images. Annealing (001) tungsten bronzes at temperatures below 600°C in UHV leads to a (1x2) reconstruction that has been studied by two groups by STM [296-298]. Jones et al. obtained well resolved images which were interpreted in terms of termination of the surface of a crystal with composition close to Nai/sWOs with a plane of stoichiometry NaxO [297,298]. The x2 periodicity apparently arises from dimerisation of oxygen ions in the surface plane to give "peroxide-like" surface species with an O-O separation less than twice the ionic radius of O^" (although the distance was still somewhat greater than in a true peroxide). The development of this model structure is shown alongside an STM image in fig. 20. A speculative model was proposed which involves reduction of

& Q^ 0^9 9^0 0

0 QfpQ 0 ^ ^ Q f ^ 0 ^ on3 On^ Q T ^ ^ ^ I F ^K-«iJli^ ^^'iS^

^^""V^

^^

0 Q^Q 0^0 ^Q 0 QQ QQ QQ

Q^

CIQ OQ ilQ

flO

^^^O ^ ^ ^ O ^^^O ^^'^^^ ^3f^^ ^^^p ^21^ l^^w

oa Q^ oa Q© ^^^^ fm^^^ f%^^^^ i"\^^^^

0@ 0 9 9® ^ ^ Fig. 20. Left hand panel: 23 A x 23 A STM image of the (2x1) reconstruction on Nao.67W03(100) taken at +0.4 V sample bias and 1 nA tunnel current. Top right hand panel: unrelaxed Nao.sO surface plane. The oxygen ions (large spheres) and sodium ions (small spheres) are assigned their conventional Shannon-Prewitt radii. Bottom right panel: schematic of the relaxed (2x1) reconstruction with "peroxide-like" oxygen ion dimers. Adapted from ref 298.

597

10

8

6

4

-2

binding energy / eV

Fig. 21. Photoemission spectra of Nao.67W03(001) (2x1) excited at the different photon energies indicated. The dashed line and arrow in the spectrum excited at 100 eV emphasise the feature attributed to the a bonding level in peroxide-like surface dimers. Adapted from ref 298.

charge in the surface layer by a mechanism different to that prevailing in the c(2x2) reconstruction. As two O^' ions move together, the bonding Gg and itu levels within the full-shell configuration fj^Ti^iZg^^ of the 02^^' unit will move down in energy. But the antibonding 7Cg and GU levels will move upward. If the Gu level crosses the Fermi energy, electron density will flow out of the localised orbital reducing the surface charge. A feature of this model is that it should be possible to observe the molecular-like orbitals associated with the dimer in a photoemission experiment. Using synchrotron radiation at around hv=100 eV in order to maximise the surface sensitivity and to be well off the W 5d resonance associated with the W 6p core threshold, a feature was observed (fig. 21) below the bottom of the valence in photoemission spectra at about 11 eV binding energy [298]. This was attributed to the ag bonding states. Structure is observed at a similar energy when K deposited on metals [299] or Ti02 [300] is exposed to oxygen and in this earlier work the feature is also ascribed to peroxide-like

598

surface species. However a problem with the model is that there was no firm indication of the GU antibonding levels in the vicinity of the Fermi level.

5. CONCLUDING REMARKS There has been very significant progress in the past decade in clarifying the nature of electronic states at oxides surfaces. Particular highlights include the clear-cut demonstration of the influence of reduced symmetry at surface sites in producing first order changes in the pattern of ligand field excited states in rocksalt transition metal monoxides (section 4.2); and verification of the importance of second order effects in inducing mixing between s and p states (which have opposite parity) at surface cation sites in MgO (section 4.1) and Sn02 (section 4.3). Second order effects do not appear to have a major influence on surface electronic structure for d^ oxides such as Ti02 and WO3, but for these materials the lowest empty nd and (n+l)s states have the same parity. The recent progress has depended in part on further refined application of established techniques such as photoemission and HREELS, but it is clear that STM and STS are now beginning to make a major impact. Work which attempts to correlate spatially resolved STS spectroscopic measurements with results from photoemission is in its infancy and there is undoubtedly scope for further work using STS on topographically well defined surfaces, especially when defects are present. The increased ability of theoretical models to capture the essential physics of surface electronic structure in a satisfactory fashion is a further significant development. However it is sobering to reflect on the fact that it is only in the past 5 years that first principles calculations have been able to account, for example, for surface states on SnO2(110). The chemical model of 5s-5p hybridisation at surface sites was first introduced in the early 1980s [130] based on ideas that had been prevalent in the chemical literature since the 1940s [301].

REFERENCES [1] [2] [3] [4] [5]

V.E. Henrich and P.A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, 1994 P.A. Cox, The Electronic Structure and Chemistry of Solids, Oxford University Press, Oxford, 1987. P.A. Cox, Transition Metal Oxides: An Introduction to their Electronic Structure and Properties, Clarendon Press, Oxford, 1992. W.A. Harrison, Electronic Structure and the Properties of Solids, W.A. Harrison, W.H. Freeman, San Francisco, 1980. J. Zaanen, G.A. Sawatzky and J.W. Allen, Phys. Rev. Lett., 55 (1985) 418.

599 [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40]

Y. Dou, R.G. Egdell, D.S.L. Law, N.M. Harrison and E.G. Searle, J. Phys.: Condens. Matter, 10(1998)8447. J.D. Levine and P. Mark, Phys. Rev., 144 (1966) 751. J.J. Yeh and I. Lindau, Atomic Data and Nuclear Data Tables, 32 (1985) 1. G.K. Wertheim, L.F. Mattheis, M. Campagna and T.P. Pearsall, Phy. Rev. Lett., 32 (1974)997. U. Diebold, H.S. Tao, N.D. Shinn and T.E. Madey, Phys. Rev. B, 50 (1994) 14474. D. Morris, Y. Dou, J. Rebane, C. E. J. Mitchell, R. G. Egdell, D. S. L. Law, A. Vittadini and M Casarin, Phys. Rev. B, 61 (2000) 13445. D. Morris, R. Dixon, F. H. Jones, R. G. Egdell, S. W. Downes and G. Beamson Phys. Rev B, 55 (1997) 16083. K.E. Smith and S.D. Kevan, Prog. Solid State Chem., 21 (1991) 49. K.E. Smith, Ann. Rep. Prog, Chem, C90 (1993) 115. K.E. Smith, K. Breuer, M. Greenblatt and W. McCarroll, Phys. Rev. Lett., 70 (1993) 3772. J.Y. Xue, L.C. Duda, A.V. Federov, P.D. Johnson, S.L. Hulbert, W. MacCarroll and M. Greenblatt, Phys. Rev. Lett., 83 (1999) 1235. K. Breuer, C. Stagerescu, K.E. Smith, M. Greenblatt and K. Ramanujachary, Phys. Rev. Lett., 76 (1996) 3172. G. Zwicker and K. Jacobi, Solid State Commun., 54 (1985) 701. R.T. Girard, O. Tjemberg, G. Chiaia, S. Soderholm, U.O. Karlsson, C. Wigren, H. Nylen and I. Lindau, Surf. Sci., 373 (1997) 409. N.B. Brookes, D.S.L. Law, H. Padmore, D.R. Warburton and G. Thornton, Solid State Commun., 57 (1986) 473. N.B. Brookes, D.S.L. Law, D.R. Warburton, P.L. Wincott and G. Thornton, J. Phys. Condensed Matter, 1 (1989) 4267. G. Thornton, N.B. Brookes, D.S.L. Law, D.R. Warburton and P.L. Wincott, Physica Scripta,41(1990)625. Z.X. Shen, C.K. Shih, O. Jepsen, W.E. Spicer, I. Lindau and J.W. Allen, Phys. Rev. Lett., 64 (1990) 2442. K.W. Goodman and V.E. Henrich, Phys. Rev. B, 49 (1994) 4827. K.W. Goodman and V.E. Henrich, Phys. Rev. B, 50 (1994) 10450. H. Ibach and D.L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations Academic Press, New York, 1982. J. Tersoff and D.R. Hamman, Phys. Rev. Lett., 50 (1983) 1998. H. Tanaka, T. Matsumoto, T. Kawai and S. Kawai, Jap. J. Appl. Phys., 32 (1993) 1405. H. Tanaka, T. Matsumoto, T. Kawai and S. Kawai, Surf Sci., 318 (1994) 29. H. Tanaka, T. Matsumoto, T. Kawai and S. Kawai, Surf Sci., 278 (1992) L153. A.L. Shluger, R.W. Grimes, C.R.A. Catlow and N. Itoh, J. Phys. Condensed Matter, 3 (1991) 8027. K.O. Legg, M. Prutton and C.G. Kinniburgh, J. Phys. C, 7 (1974) 4236. C.G. Kinniburgh, J. Phys. C, 9 (1976) 2695. J.B. Zhou, H.C. Lu, T. Gustafsson and P. Haberle, Surf. Sci., 302 (1994) 350. O. Robach, G. Renaud and A,. Barbier, Surf Sci., 401 (1998) 227. E.A. Colboum and W.C. Mackrodt, Surf Sci., 117 (1982) 571. C.A. Scamehom, N.M. Harrison and M.I. McCarthy, J. Chem. Phys., 101 (1994) 1547. M. Causa, R. Dovesi, C. Pisani and C. Roetti, Surf Sci., 175 (1986) 551. S. Pugh and M.J. Gillan, Surf Sci., 320 (1994) 331. L.N. Kantorovich, J.M. Holender and M.J. Gillan, Surf Sci., 343 (1995) 221.

600 [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81]

K. H. Rieder, Surf. Sci., 118 (1982) 52. A.B. Laponsky and N.R. Whetten, Phys. Rev., 120 (1960) 801. M. Liehr, P.A. Thiry, J.J. Pireaux and R. Caudano, Phys. Rev. B, 33 (1986) 5682. M.C. Wu, J.S. Comeille, C.A. Estrada, J.W. He and D.W. Goodman, Chem. Phys. Lett., 182 (1991) 472. M.C. Wu, C.A. Estrada and D.W. Goodman, Phys. Rev. Lett., 67 (1991) 2910. J.S. Comeille, J.W. He and D.W. Goodman, Surf. Sci., 306 (1994) 269. D.W. Goodman, J. Vac. Sci. Technol. A, 14 (1996) 1526. L.H. Tjeng, A.R. Vos and G.A. Sawatzl^, Surf. Sci., 235 (1990) 269. D. Ochs, W. Mausfriedrichs, M. Brause, J. Gunster, V. Kempter, V. Puchin, A. Shluger and L.N. Kantorovich, Surf Sci., 365 (1996) 557. L.N. Kantorovich, A.L. Shluger, P.V. Sushko, J Gunster, D.W. Goodman, P. Stracke and V. Kempter, Farad. Disc. 114 (1999) 173. L.N. Kantorovich, A.L. Shluger, P.V. Sushko and A.M. Stoneham, Surf. Sci., 444 (2000) 31. V.E. Henrich, G. Dresselhaus and H.J. Zeiger, Phys. Rev. Lett., 36 (1976) 158. V.E. Henrich, G. Dresselhaus and H.J. Zeiger, Phys. Rev. B, 22 (1980) 4764. P.A. Cox and A.A. Williams, Surf. Sci., 175 (1986) L782. M.C. Wu, C M . Truong and D.W. Goodman, Phys. Rev. B, 46 (1992) 12688 P.R. Underhill and T.E. Gallon, Solid State Commun., 43 (1982) 9. A.A. Williams, D. Phil. Thesis, Oxford, 1985. V.C. Lee and H.S. Wong, J. Phys. Soc. Japan, 45 (1978) 895. J.P. LaFemina and C.B. Duke, J. Vac. Sci. Technol.A, 9 (1991) 1847. C. Satoko, M. Tsukada and H, Adachi, J. Phys. Soc. Jap., 45 (1978) 1333. M. Tsukada, H. Adachi and C. Satoko, Prog. Surf Sci., 14 (1983) 113. A.M. Stoneham and M.J.L. Sangster, Phil. Mag. B, 43 (1981) 609. A.M. Stoneham, M.J.L. Sangster and P.W. Tasker, Phil. Mag. B, 44 (1981) 603. K. Saiki, W.R. Xu and A. Koma, Surf Sci., 287/288 (1993) 644. P.V. Sushko, A.L. Shluger and C.R.A. Catlow, Surf. Sci., 450 (2000) 153. J.W. He and P.J. Moller, Chem. Phys. Lett., 129 (1986) 13. R.R. Sharma and A.M. Stoneham, J. Chem. Soc. Faraday Trans. II, 72 (1976) 913. H.A. Kassim, J.A.D. Matthew and B. Green, Surf Sci., 74 (1978) 109. J.W. He and P.J. Moller, Surf Sci., 180 (1987) 411. C. Tegenkamp, H. Pfhur, W. Ernst, U. Malaske, J. Wollschlager, D. Peterka, K.M Schroder, V. Zielasek and M. Henzler, J. Phys. Condensed Matter, 11 (1999) 9943. M.C. Wu, C M . Truong and D.W. Goodman, Phys. Rev. B, 46 (1992) 12688. F. lUas and G. Pacchioni, J. Chem. Phys., 108 (1998) 7835. Y. Chen, R.T. Williams and W.A. Sibley, Phys. Rev., 182 (1969) 960. C Sousa, G. Pacchioni nad F. Illas, Surf. Sci., 429 (1999) 217. E. Castanier and C Noguera, Surf Sci., 364 (1996) 1. E. Castanier and C Noguera, Surf Sci., 364 (1996) 17. F. Fmocchi, J. Goniakowski and C Noguera, Phys. Rev. B, 59 (1999) 5178. W.C Mackrodt, N.M. Harrison, V.R. Saunders, N.L. Allan, M.D. Towler, E. Apra and R. Dovesi, Phil. Mag. A, 68 (1993) 653. M.D. Towler, N.L. Allan, N.M. Harrison, V.R. Saunders, W.C. Mackrodt and E. Apra, Phys. Rev. B, 50 (1994) 5041. G.A. Sawatzky and J.W. Allen, Phys. Rev. Lett., 53 (1984) 2339. J. Zaanen and G.A. Sawatzky, Prog. Theoret. Phys. SuppL, 101 (1990) 231.

601 [82] P.A. Cox and A.A. Williams, Surf. Sci., 152 (1985) 791. [83] A. Freitag, V. Staemmler, D. Cappus, C.A. Ventrice Jr., K. Al Shamery, H. Kuhlenbeck and H.J. Freund, Chem. Phys. Lett., 210 (1993) 10. [84] M. Pohlchen and V. Staemmler, J. Chem. Phys., 97 (1992) 2583. [85] A. Gorschluter and H. Merz, Phys. Review B, 49 (1994) 17293. [86] B. Fromme, Ch. Kock, R. Deussen and E. Kisker, Phys. Rev. Lett., 75 (1995) 693. [87] B. Fromme, M. Miiller, Th. Anschutz, C. Bethke and E. Kisker, Phys. Rev. Lett., 77(1996)1548. [88] W.C. Mackrodt and C. Noguera, Surf. Sci., 457 (2000) L 386. [89] C. Xu, W.S. Oh, Q. Guo and D.W. Goodman, J. Vac. Sci. Technol. A, 14 (1996) 1395 [90] Q. Guo, C. Xu and D.W. Goodman, Langmuir, 14 (1998) 1371. [91] M. Hassel, H. Kuhlenbeck, H.J. Freund, S. Shi, A. Freitag, V. Staemmler, S. Lutkoff and M. Neumann, Chem. Phys. Lett., 240 (1995) 205. [92] M.R. Castell, P.L. Wincott, N.G. Condon, C. Muggelberg, G. Thornton, S.L. Dudarev, A.P. Sutton and G.A.D. Briggs, Phys. Rev. B, 55 (1997) 7859. [93] M.R.Castell, S.L. Dudarev, P.L. Wincott, N.G. Condon, C. Muggelberg, G. Thornton, D.N. Manh, A.P. Sutton and G.A.D. Briggs, Surf Review and Lett., 4 (1997) 1003. [94] M.R. Castell, S.L. Dudarev, G.A.D. Briggs and A.P. Sutton, Physica B, 261 (1999) 717. [95] M.R. Castell, S.L. Dudarev, G.A.D. Briggs and A.P. Sutton, Phys. Rev. B, 59 (1999) 7342. [96] S.L. Dudarev, A.L Liechenstein, M.R. Castell, G.A.D. Briggs and A.P. Sutton, Phys. Rev. B, 56 (1997) 4900. [97] M. Marynowski, W. Franzen, M. El-Batanouny and V. Staemmler, Phys. Rev. B, 60 (1999) 6053 and references therein. [98] P.W. Palmberg, R.E. deWames and L.A. Vredevoe, Phys. Rev. Lett., 21 (1968) 682. [99] P.W. Pahnberg, R.E. deWames, L.A. Vredevoe and T. Wolfram, J. Appl. Phys., 40 (1969) 1158. [100] T. Wolfram, R.E. de Wames, W.F. Hall and P.W. Pahnberg, Surf. Sci., 28 (1971) 45. [101] P. Cantini, R. Taterek and G.P. Felcher, Phys. Rev. B, 19 (1979) 1161. [102] J.P. Toennies, G. Witte, A.M. Shikin and K.H. Rieder, J. Electron Spectrosc, 64/65 (1993) 677. [103] W.P. Brug, G. Chem, J. Duan, G.G. Bishop, S.A. Safron and J.G. Skofronick, J. Vac. Sci. Technol. A, 10 (1992) 2222. [104] H. Hayakawa, K. Namikawa and S. Miyake, J. Phys. Soc. Jap., 31 (1971) 1408. [105] K. Namikawa, T. Wolfram, H. Hayakawa and S. Miyake, J. Phys. Soc. Jap., 37 (1974) 733. [106] K. Namikawa, J. Phys. Soc. Jap., 44 (1978) 165. [107] A. Swan, M. Marynowski, W. Franzen, M. El-Batanouny and K.M. Martini, Phys. Rev. Lett., 71 (1993) 1250. [108] A. Swan, M. Marynowski, W. Franzen, M. El-Batanouny and K.M. Martini, J. Vac. Sci. Technol. A, 12 (1994) 2219. [109] M. Marynowski, A. Swan, W. Franzen, M. El-Batanouny and K.M. Martini, Surf Interface Anal, 23 (1995) 105. [110] J.J.M. Pothuizen, O. Cohen and G.A. Sawatzky, Materials Research Society Symposia Proceeding, 401 (1996) 501. [111] C. deGraaf, R. Broer and W.C. Nieuport, Chem. Phys. Lett., 271 (1997) 372. [112] M.El-Batanouny, G. Murthy, C.R. Willis, S. Kais and V. Staemmler, Phys. Rev. B, 58(1998)7391.

602 [113] P.M. Oliver, G.W. Watson, E.T. Kelsey and S.C. Parker, J. Mat. Chem., 7 (1997) 563. [114] P.W. Tasker, J. Phys. C, Solid State Phys., 12 (1979) 4977. [115] D.F. Cox, T.B. Fryberger and S. Semancik, Phys. Rev. B, 38 (1988) 2072. [116] D.F. Cox, T.B. Fryberger and S. Semancik, Surf. Sci., 224 (1989) 101 [117] D.F. Cox and T.B. Fryberger, Surf. Sci., 227 (1990) L105. [118]R.G. Egdell, p. 527 in Science of Ceramic Interfaces II, ed. J. Nowotny, Elsevier, Amsterdam 1994. [119] J.W. Erickson and S. Semancik, Surf Sci., 187 (1987) L658 [120] G.L. Shen, R. Casanova, G. Thornton and I. Colera, J. Phys. Condensed Matter, 3 (1991) S291 [121] F. H. Jones, R. Dixon, J. S. Foord, R. G. Egdell and J. B. Pethica, Surf Sci., 376 (1997) 367. [122] J.S.G. Irwin, A.F. Prime, P.J. Hardman, P.L. Wincott and G. Thornton, Surf Sci., 352 (1996) 480. [123] G. Thornton and coworkers, Phys. Rev.B, in press. [124] Semiconductors: Physics of Non-tetrahedrally Bonded Binary Compounds II, edited by O. Madelung, Landolt-Bomstein New Series, Group III, Vol 17, Part f. Springer Verlag, Berlin, 1984. [125] R. G. Egdell, S. Eriksen and W. R. Flavell, Solid State Commun., 60 (1986) 835. [126] J.M. Themlin, R. Sporken, J. Darville, R. Caudano, J.M. Gilles and R.L. Johnson, Phys. Rev. B, 42 (1990) 11914. [127] G.L. Shen, R. Casanova and G. Thornton, Vacuum, 43 (1992) 1129. [128] J. Szuber, Phys. Stat. Solidi B, 185 (1994) K9. [129] J. Szuber and G. Czempik, Vacuum, 48 (1997) 289. [130] P. A. Cox, R. G. Egdell, C. Harding, W. R. Patterson and P. Tavener, Surf. Sci., 123 (1982) 179. [131] J.M. Themlin, M. Chtaib, L. Henrard, P. Lambin, J. Darville and J.M. Gilles, Phys. Rev. B, 46 (1992) 2460. [132] C.S. Fang, F.M. Pan, W.S. Tse and S.R. Homg, Surf Sci., 211/212 (1989) 279. [133] C. S. Rastomjee, R. G. Egdell, M. J. Lee and T. J. Tate, Surf Sci., 259 (1991) L769. [134] C S Rastomjee, R G Egdell, G C Georgiadis, M J Lee and T J Tate Journal of Materials Chemistty, 2 (1992) 511 [135] R. Larciprete, E. Borsella, P. DePadova, M. Mangiantini, P. Perfetti and M. Fanfoni, J. Vac. Sci. Technol. A, 11 (1993) 336 [136] P. DePadova, R. Larciprete, C. Ottaviani, C. Quaresimam P. Perfetti, E. Borsella, C. Astaldi, C. Comicoli, C. Crotti, M. Matteucci, M. Zacchigna and K. Prince, Appl. Surf Sci., 104(1996)349. [137] R. G. Egdell, S. Eriksen and W. R. Flavell, Surf Sci., 192 (1987) 265. [138] J. Robertson, J. Phys. C , 12 (1979) 4767. [139] S. Munnix and R. Schmeits, Solid State Comm., 43 (1982) 867. [140] S. Munnix and R. Schmeits, Surf. Sci., 126 (1983) 20. [141] S. Munnix and R. Schmeits, Phys. Rev. B, 27 (1983) 7624. [142] T.J. Godin and J.P. Lafemina, Phys. Rev. B, 47 (1993) 6518. [143] I. Manassidis, J. Goniakowski, L.N. Kantorovich and M.J. Gillan, Surf. Sci., 339 (1995) 258. [144] T.T. Rantala, T.S. Rantala and V. Lantto, Surf Sci., 420 (1999) 103. [145] J. Goniakowski, J.M. Holender, L.N. Kantorovich and M.J. Gillan, Phys. Rev. B, 53 (1996) 957. [146] P. C. Hollamby, P. S. Aldridge, G. Morretti, R. G. Egdell and W. R. Flavell

603 Surf. Sci., 280 (1993) 393. [147 Semiconductors: Physics of Non-tetrahedrally Bonded Binary Compounds II, edited by O. Madelung, Landolt-Bomstein New Series, Group III, Vol 17, Part g. Springer Verlag, Berlin, 1984. [148 A. E. Tavemer, C. Rayden, S. Warren, A. Gulino and R. G. Egdell Phys. Rev. 6,51(1995) 6833. [149 Z.M. Zhang, S.P, Jeng and V.E. Henrich, Phys. Rev. B, 43 (1991) 12004 [150 J. Nerlov, Q.F. Ge and P.J. Moller, Surf. Sci., 348 (1996) 28. [151 R. Heise, R. Courths and S. Witzel, Solid State Commun., 84 (1992) 599 [152 R. Heise and R. Courths, Surf Sci., 287 (1993) 658 [153 V. E. Henrich and R.E. Kurtz, Phys. Rev. B, 23 (1981) 6280. [154 G. Charlton, P.B. Howes, C.L. Nicklin, P. Steadman, J.S.G. Taylor, C.A. Muryn, S.P. Harte, J. Mercer, R. McGrath, D. Norman, T.S. Turner and G. Thornton, Phys. Rev. Lett., 78 (1997) 495 [155 P. Reinhardt and B.A. Hess, Phys. Rev. B, 50 (1994) 12015. [156; D. Vogtenhuber, R. Podloucky, A. Neckel, S.G. Steinmann and A.J. Freeman, Phys. Rev. B, 49 (1994) 2099 [157 M. Ramamoorthy, R.D. King-Smith and D. Vanderbilt, Phys. Rev. B, 49 (1994) 7709. [158 M. Ramamoorthy, D. Vanderbilt and R.D. King-Smith, Phys. Rev. B, 49 (1994) 16723. [159 P.J.D. Lindan, N.M. Harrison, M.J. Gillan and J. A. White, Phys. Rev. B, 55 (1997) 15919. [160 S.P. Bates, G. Kresse and M.J. Gillan, Surf Sci., 385 (1997) 386. [161 N.M. Harrison, X.-G. Wang, J. Muscat and M. SchefHer, Faraday Discussions, 114 (1999)305. [162 J. Purton, D.W. BuUett, P.M. Oliver and S.C. Parker, Surf Sci., 336 (1995) 166. [163 G.N. Raikar, P.J. Hardman, C.A. Muryn, G. Vanderlaan, P.L. Wincott and G. Thornton, Solid State Commun., 80 (1991) 423 [164 P.J. Hardman, G.N. Raikar, C.A. Muryn, G. Vanderlaan, P.L. Wincott, G. Thornton, D.W. Bullet and P. Dale, Phys. Rev. B, 49 (1994) 7170. [165 A.K. See, M. Thayer and R.A. Bartynski, Phys. Rev. B, 47 (1993) 13722. [166 S. Munnix and R. Schmeits, Phys. Rev. B, 30 (1984) 2202. [167 M. Sander and T. Engel, Surf Sci., 302 (1994) L263. [168 A. Szabo and T. Engel, Surf Sci., 329 (1995) 241. [169 D. Novak, E. Garfimkel and T. Gustafsson, Phys. Rev. B, 50 (1994) 5000. [170 H. Onishi and Y. Iwasawa, Surface Sci., 313 (1994) L783. [171 H. Onishi and Y. Iwasawa, Chem. Phys. Lett., 226 (1994) 111. [172 Y. Yamaguchi, H. Onishi and Y. Iwasawa, J. Chem. Soc. Faraday Trans., 91 (1995) 1661. [173 H. Onishi, K. Fukui and Y. Iwasawa, Bull. Chem. Soc. Jap., 68 (1995) 2447. [174 S. Fischer, A.W. Munz, K.D. Schierbaum and W. Gopel, Surf Sci., 337 (1995) 17. [175 P.W. Murray, N.G. Condon and G. Thornton, Phys. Rev. B, 51 (1995) 10989. [176; S. Fischer, A.W. Munz, K.D. Schierbaum and W. Gopel, J. Vac. Sci. Technol. B, 14 (1996) 961. [177 H. Onishi and Y. Iwasawa, Phys. Rev. Lett., 76 (1996) 791. [178 U. Diebold, J.F. Anderson, K.O. Ng and D. Vanderbilt., Phys. Rev. Lett., 77 (1996) 1322.

604 [179] U. Diebold, J. Lehman, T. Mahmoud, M. Kuhn, G. Leonardelli, W. Hebenstreit, M. Schmid and P. Varga, Surf. Sci. 411 (1998) 137 . [180] L.P. Zhang, M. Li and U. Diebold, Surf. Sci., 412/413 (1998) 242. [181] K.O. Ng and D. Vanderbilt, Phys. Rev. B, 56 (1997) 10544. [182] O. Gulseren, R. James and D.W. Bullett, Surf. Sci., 377-379 (1997) 150. [183] H. Onishi and Y. Isasawa, Chem. Phys. Lett., 226 (1994) 111. [184] U. Diebold, W. Hebenstreit, G. Leomardelli, M. Schmid and P. Varga, Phys. Rev. Lett., 81 (1998) 405. [185] R.E. Tanner, M.R. Castell and G.A.D. Briggs, Surf Sci., 412/413 (1998) 672. [186] Q. Guo, I. Cocks and E.M. Williams, J. Phys. D, 31 (1998) 2231. [187] W. Gopel, J.A. Anderson, D. Frankel, M. Jaehnig, K. Phillips, J.A. Schafer and G. Rocker, Surf Sci., 139 (1984) 3427. [188] W. Gopel, G. Rocker and R. Feierabend, Phys. Rev. B, 28 (1983) 3427. [189] G. Rocker, J.A. Schaefer and W. Gopel, Phys. Rev. B, 30 (1984) 3704 [190] R. L. Kurtz, R. Stockbauer, T.E. Madey, E. Roman and J.L. de Segovia, Surf Sci., 218 (1989) 178. [191] J.M. Pan, B.L. Maschoff, U. Diebold and T.E. Madey, J. Vac. Sci. TechnoL, AlO (1992) 2470. [192] Y.W. Chung, W.J. Lo and G.A. Somorjai, Surf Sci., 64 (1977) 588. [193] V.E. Henrich, G. Dresselhaus and H.J. Zeiger, Phys. Rev. Lett., 36 (1976) 1335. [194] V.E. Henrich, G. Dresselhaus and H.J. Zeiger, Solid State Commun., 24 (1977) 623. [195] H.R. Sadeghi and V.E. Henrich, J. Catal, 109 (1988) 1. [196] K.E. Smith, J.L. Mackay and V.E. Henrich, Phys. Rev. B, 35 (1987) 5822. [197] R.G. Egdell, S. Eriksen and W.R. Flavell, Solid State Commun., 60 (1986) 835. [198] M.L. Knotek and P.J. Feibelman, Phys. Rev. Lett., 40 (1978) 964. [199] R.L. Kurtz, Surf Sci., 177 (1986) 526. [200] P.A. Cox, R.G. Egdell, S. Eriksen and R.G. Egdell, J. Electron Spectrosc, 39(1986)117. [201] S. Eriksen and R.G. Egdell, Surf Sci., 180 (1987) 263. [202] L-Q. Wang, D.R. Baer and M.H. Engelhard, Surf Sci., 320 (1994) 295. [203] V.M. Bermudez, J. Vac. Sci. TechnoL, 20 (1982) 51. [204] V.M. Bermudez, J. Vac. Sci. TechnoL, 20 (1982) 741. [205] G. Rocker and W. Gopel, Surf Sci., 181 (1987) 530. [206] J.T. Mayer, U. Diebold, T.E. Madey and E. Garfunkel, J. Electron Spectrosc, 73(1995)1. [207] A. Berko and E. Krivan, J. vac. Sci. TechnoL B, 15 (1997) 25. [208] A.K. See, W.K. Siu, R.A. Bartynski, A. Nangia, A.H. Weiss, S.L. Hulbert, X. Wu and C.C. Kao, Surf Sci., 383 (1997) L735. [209] S. Munnix and M. Schmeits, Phys. Rev. B, 31 (1985) 3369. [210] W.C. Mackrodt, E.A. Simsos and N.M. Harrison, Surf Sci., 384 (1997) 192. [211] A. Kodaira, Y. Sakisaka, T. Maruyama, Y. Haruyama, Y. Aiura and H. Katoh, Solid State Commun., 89 (1994) 9. [212] Y. Aiura, Y. Nishihara, Y. Haruyama, T. Komeda, S. Kodaira, Y. Sakisaka, T. Maruyama and H. Katoh, Physica B, 194 (1994) 1215. [213] A. Fujimori, A.E. Bocquet, K. Morikawa, K. Kobayashi, T. Saitoh, Y. Tokura,. Hase and M. Onada, J. Phys. Chem. Solids, 57 (1996) 1379. [214] A. E. Tavemer, P. C. Hollamby, P. S. Aldridge, R. G. Egdell and W. C. Mackrodt Surf ScL, 287/288 (1993) 653.

605 [215] A.K. See and R.A. Bartynski, J. Vac. Sci. Technol. A, 10 (1992) 2591. [216] G.S. Rohrer, V.E. Henrich and D.A. Bonnell, Science 250 (1990) 1239. [217] G.S. Rohrer, V.E. Henrich and D.A. Bonnell, Surf. Sci., 278 (1992) 146. [218] H. Norenberg, R.E. Tanner, K.D. Schierbaum, S. Fischer and G.A.D. Briggs, Surf. Sci., 396(1998)52. [219] L.P. Zhang, M. Li and U. Diebold, Surf Sci., 412/413 (1998) 242. [220] R.A. Bennett, S. Poulston, P. Stone and M. Bowker, Phys. Rev. B, 59 (1999) 10341. [221] C.R.A. Catlow and R. James, in M.W. Roberts and J.M. Thomas (eds), RSC Specialist Periodical Report on Chemical Physics of Solid Surfaces, The Royal Society of Chemistry, London, 1980, p. 108. [222] Z. Qian, J.M. Vohs and D.A. Bonnell, Surf Sci., 274 (1992) 35. [223] Z. Qian, J.M. Vohs and D.A. Bonnell, J. Am. Ceram. Soc, 76 (1993) 1137. [224] P.J. Moller and M.C. Wu, Surf Sci., 224 (1989) 265. [225] P.J.D. Lindan, N.M. Harrison, M.J. Gillan and J.A. White, Phys. Rev. B, 55 (1997) 15919. [226] A. Berko and F. Solymosi, Langmuir, 12 (1996) 1257. [227] Q. Guo, I. Cocks and E.M. Williams, Phys. Rev. Lett., 77 (1996) 1322. [228] R. Patel, Q. Guo, I. Cocks, E.M. Williams, E. Roman and J.L. deSegovia, J. Vac. Sci. Technol., 15(1997)2553. [229] C.L. Pang, S.A. Haycock, H. Raza, P.W. Murray, G. Thornton, O. Gulseren, R. James and D.W. Bullett, Phys. Rev. B, 58 (1998) 1586. [230] C.L. Pang, S.A. Haycock, H. Raza, P.W. Murray, G. Thornton, O. Gulseren, R. James and D.W. Bullett, Surf Sci., 437 (1999) 261. [231] R.E. Tanner, M.R. Castell and G.A.D. Briggs, Surf Sci., 437 (1999) 263. [232] R.A. Bennett, P. Stone, N.J. Price and M. Bowker, Phys. Rev. Lett. 82 (1999) 3831. [233] P Stone, R A Bennett and M Bowker, New. J. Phys. 1 (1999) 8 [234] M. Li, W. Hebenstreit and U. Diebold, Surf Sci. Letters, 414 (1998) L951. [235] M. Li, W. Hebenstreit, U. Diebold, M.A. Henderson, D.R. Jennison, P.A. Schultz and M.P. Sears, Surf Sci., 437 (1999) 173. [236] M. Li, W. Hebenstreit, U. Diebold, M.A. Henderson and D.R. Jennison, Faraday Discuss., 114(1999)245. [237] M. Li, W. Hebenstreit and U. Diebold, Phys. Rev. B, 61 (2000) 4926. [238] M. Li, W. Hebenstreit, U. Diebold, A.M. Tyryshkin, M.K. Bowman, G.G. Dunham and M.A. Henderson, J. Phys. Chem. B, 104 (2000) 4944. [239] Y.W. Chung, W.J. Lo and G.A. Somorjai, Surf Sci., 64 (1977) 588. [240] W.J. Lo, Y.W. Chung and G.A. Somorjai, Surf Sci., 71 (1978) 250. [241] C.A. Muryn, P.J. Hardman, J.J. Crouch, G.N. Raiker, G. Thornton and D.S.L. Law, Surf. Sci., 251/252 (1991) 747. [242] K.C. Prince, V.R. Dhanak, P. Finetti, J.F. Walsh, R. Davis, C.A. Muryn, H.S. Dhariwal, G. Thornton and G. vanderLaan, Phys. Rev. B, 55 (1997) 9520. [243] K. Prabhakaran, D. Purdie, R. Casanova, C.A. Muryn, P.J. Hardman, P.L. Wincott and G. Thornton, Phys. Rev. B, 45 (1992) 6969 [244]P.J. Hardman, R. Casanova, K. Prabhakaran, C.A. Muryn, P.L. Wincott and G. Thornton, Surf Sci., 270 (1992) 677. [245] P. Zschack, J.B. Cohen and Y.W.Chung, Surf Sci., 262 (1991) 395. [246] A. Howard, C.E.J. Mitchell, D. Morris, R.G. Egdell and S.C. Parker,

606 Surf. Sci., 448 (2000) 131. [247] P.W. Murray, P.M. Leibsle, H.J. Fisher, C.F.J. Flipse, C.A. Muryn and G. Thornton, Phys. Rev. B, 46 (1992) 12877. [248] P.W. Murray, F.M. Leibsle, C.A. Muryn, H.J. Fisher, C.F.J. Flipse and G. Thornton, Surf. Sci., 321 (1994) 217. [249] P.W. Murray, F.M. Leibsle, C.A. Muryn, H.J. Fisher, C.F.J. Flipse and G. Thornton, Phys. Rev. Lett., 72 (1994) 217. [250] P.J. Hardman, N.S. Prakash, C.A. Muryn, G.N. Raikar, A.G. Thomas, A.F. Prime and G. Thornton, Phys. Rev. B, 47 (1993) 16056. [251] H. Raza, C.L. Pang, S.A. Paycock and G. Thornton, Phys. Rev. Lett., 82 (1999) 5265. [252] H. Raza, C.L. Pang, S.A. Paycock and G. Thornton, Appl. Surf Sci., 140 (1999) 271. [253] C.L. Pang, H. Raza, S.A. Paycock and G. Thornton, Appl. Surf. Sci., 157 (2000) 233 [254] H. Zanjonz, H.L. Meyerheim, T. Gloege, W. Moritz and D. Wolf, Surf. Sci., 398 (1998) 369. [255] E. Landree, L.D. Marks, P. Zschack and C.J. Gihnore, Surf Sci., 408 (1998) 300. [256] A. Hamnett and J. B. Goodenough in : Semiconductors ed. O. Madelung, LandoltBomstein Numerical Data and Functional Realationships in Science and Technology. Group III, Volume 17g, Part III (Springer-Verlag, Berlin, 1984). [257] K. Kosuge, Chemistry of Non-Stoichiometric Compounds, Oxford University Press, Oxford, 1994. [258] O.F. Schirmer and E. Salje, J. Phys. C, 13 (1980) L1067. [259] G. Hollinger, T.M. Due and A. Deneuville, Phys. Rev. Letters, 37 (1976) 1564. [260] B.A. DeAngeHs and M. Schiavello, J. Solid State Chem., 21 (1977) 67. [261] R.J. Colton, A.M. Guzman and J.W. Rabalais, Accounts of Chemical Research, 11 (1978) 170. [262] R.J. Colton, A.M. Guzman and J.W. Rabalais, J. Appl. Phys., 49 (1978) 409. [263] E. Salje, A.F. Carley and M.W. Roberts, J. Solid State Chem., 29 (1979) 237. [264] T.H. Fleisch and G.J. Mains, J. Chem. Phys., 76 (1982) 780. [265] P.J.C. Chappell, M.H. Kibel and B.G. Baker, J. Catal., 110 (1988) 139. [266] R.D. Bringans, H. Hochst and H.R. Shanks, Vacuum, 31 (1981) 473. [267] R.D. Bringans, H. Hochst and H.R. Shanks, Phys. Rev. B, 24 (1981) 3481. [268] R.A. Dixon, J.J. Williams, D. Morris, J. Rebane, F.H. Jones, R.G. Egdell and S.W. Downes, Surf Sci., 399 (1998) 199. [269] P.M. Oliver, S.C. Parker, R.G. Egdell and F.H. Jones, J.Chem. Soc. Faraday Trans., 92 (1996) 2049. [270] F.H. Jones, K. Rawlings, J.S. Foord, P.A. Cox, R.G. Egdell, J.B. Pethica and B.M.R. Wanklyn, Phys. Rev. B, 52 (1995) R14392. [271] F.H. Jones, K. Rawlings, J.S. Foord, R.G. Egdell, J.B. Pethica, B.M.R. Wanklyn, S.C. Parker and P.M. Oliver, Surf Sci., 359 (1996) 107. [272] D.W. Bullett, J. Phys. C, 16 (1983) 2197. [273] R.E. Tanner, P. Meethunkij and E. I. Altmann, J.Am. Chem. Soc, submitted [274] A. Gulino, S. Parker, F.H. Jones and R.G. Egdell, J.Chem. Soc. Faraday Trans., 92 (1996) 2137. [275] F.H. Jones, R.A. Dixon and A. Brown, Surf. Sci., 369 (1996) 343. [276] P.G. Dickens and M.S. Whittingham, Chem. Soc. Quart. Rev., 22 (1968) 30. [277] P.A. Lightsey, D.A. Lillienfeld and D.F. Holcomb, Phys. Rev. B, 14 (1976) 4370.

607 [278] Neutron diffraction shows that the structure has a very small tetragonal distortion and that the space group is I4/mmm. However for the present purposes we can describe the structure as cubic, see P. Wiseman and P. Dickens, J. Solid State Chem., 17 (1976) 91. [279] M. Kielwein, G. Saiki, G. Roth, J. Fink, G. Paasch and R.G. Egdell, Phys. Rev. B, 51 (1995) 10320. [280] R. G. Egdell and M. D. Hill, Chem. Phys. Lett., 85 (1982) 140. [281] M. D. Hill and R. G. Egdell, J. Phys. C, 16 (1983) 6205. [282] J.B. Goodenough, Progress in Solid State Chemistry, 5 (1971) 145. [283] J.H. Davies and J.R. Franz, Phys. Rev. Letters, 57 (1986) 475. [284] H. Ducker, W. von Niessen, T. Koslowski, M.A. Tusch and D.E. Logan, Phys. Rev. B, 59 (1999) 871. [285] A.K. Raychaudhuri, K.P. Rajeev, H. Srikanth and N. Gayathrin, Phys. Rev. B, 51 (1995)7421. [286] F. Cora, A. Patel, N.M. Harrison, R. Dovesi and C.R.A. Catlow, J. Am. Chem. Soc, 118(1996)12174. [287] I.Hamberg, K. F. Bergren, L. Engstrom, C. G. Granquist and B. E. Semelius, Phys. Rev. B, 30 (1984) 3240. [288] Y Dou, T. Fishlock, R. G. Egdell, D. S. L. Law and G. Beamson Phys. Rev. B, 55 (1997) R13381. [289] Y. Dou, R. G. Egdell, T. Walker, D. S. L. Law and G Beamson, Surf. Sci. 398 (1998) 241 [290] R.G. Egdell, J. Rebane, T. J. Walker and D.S.L. Law, Phys. Rev. B, 59 (1999) 1792. [291] H. Hochst, R.D. Bringans and H.R. Shanks, Phys. Rev. B, 26 (1982) 1702. [292] S. EUiatioglu and T. Wolfram, Phys. Rev. B, 18 (1978) 4509. [293] F.H. Potter and R.G. Egdell, Surf. Sci., 287/288 (1993) 649. [294] R.G. Egdell, H. Innes and M.D. Hill, Surf. Sci., 149 (1985) 33. [295] F.H. Jones, K. Rawlings, S. Parker, R.G. Egdell, J.S. Foord, P.A. Cox and J.B. Pethica, Surf. Sci., 336 (1995) 18L [296] G.S. Rohrer, W. Lu, M.L, Norton, M.A. Blake and C.L. Rohrer, J. Solid State Chem., 109 (1994) 359 [297] F.H. Jones, K. Rawlings, R.G. Egdell, J.S. Foord, P.A. Cox and J.B. Pethica, J. Chem. Soc. Chem. Commun., (1995) 2419. [298] F.H. Jones, K. Rawlings, R.A. Dixon, T.W. Fishlock and R.G. Egdell, Surf. Sci., 460 (2000) 277. [299] S.L. Qiu, C.L.Lin, J. Chen and M. Strongin, Phys. Rev. B, 41 (1990) 7467. [300] A.G. Thomas, P.J. Hardman, C.A. Muryn, H.S. Dhariwal, A.F. Prime, G. Thornton, J. Chem. Soc. Faraday Trans., 91 (1995) 3569. [301] J.D. Dunitz and L.E. Orgel, Advances in Inorganic Chemistry and Radiochemistry, 2 (1960) 1 and references therein.

Oxide Surfaces D.P. Woodruff, editor © 2001 Elsevier Science B. V. All rights reserved.

608

Chapter 15

Desorption Induced by Electronic Transitions, DIET, at Oxide Surfaces J. L. de Segovia^'* and E. M. Williams^ ^Institute de Ciencia de Materiales de Madrid. CSIC, E-28049 Cantoblanco Madrid, Spain ^ Surface Science Research Centre, The University of Liverpool, Liverpool L69 3BX, UK 1. HISTORICAL INTRODUCTION TO ESD AND PSD The electron stimulated desorption of ions, positive or negative, and neutrals induced by electronic transitions, ESD, has been widely studied for many years. In 1965 Lichtman and McQuistan reviewed several aspects of ESD [1], Madey and Yates in 1971 [2] and in the mid to late 80's the present authors reviewed the field of ESD and its application for surface adsorbate analysis [3,4]. An extensive review was published in 1991 by Ramsier and Yates [5] and in 1994 Madey reviewed the history of ESD [6]. The same phenomena can be observed when photons are the incident agents at the surface. In this case it is recognised as photon stimulated desorption, PSD. In addition to ions and neutrals, excitation leading to desorption as metastable species can also occur [7], as well as surface induced reactions. Thus, from 1982 the phenomenon was recognised in a more general form as Desorption Induced by Electronic Transitions, DIET [8, 9]. Apart from the general bibliography, a significant contribution to this field can be followed in the Proceedings of the Conference series on Desorption Induced by Electronic Transitions, DIET. In this introduction, and ahead of a discussion of specific oxide surfaces, it is of interest to consider some of the most relevant discoveries with the development of ESD. We shall also restrict the study to the desorption of positive ions from oxidised and oxide surfaces. The desorption of ions from adsorbed gases on surfaces was observed as early as 1918 by Dempster [10] using the then recently discovered mass spec* Author for correspondence:

[email protected]/[email protected]

609

trometer. The same approach was pursued later in 1950 by Plumlee and Smith [11] in 1961 using a magnetic deflection mass spectrometer. Most of the earlier studies of ESD were performed by detecting pressure changes within a vacuum system, such as in the pioneering work of Ishikawa in 1943 [12]. The first systematic studied of the role of impinging electrons in causing desorption of gases adsorbed on surfaces was performed by G. E. Moore in 1961 [13] using a magnetic deflection mass spectrometer. He observed desorption of O^ and CO"*" ions from CO adsorbed on molybdenum and tungsten surfaces. One of his most important findings was to show that ESD is a first order phenomenon, i.e. the number of desorbed ions is proportional to the number of impinging electrons. He also determined the dependence of the desorbed ion yield on the incident electron energy and found a maximum at about 70 eV. The minimum energy necessary to produce an O"^ was between 17 and 18.3 eV. The last years of the 50's and beginning of the 60's saw an increasing interest in the behaviour of ion currents in ionisation gauges when exposed to actives gases such us CO and O2. An anomaly arose in that gauges indicated pressures much higher than the true pressures in ultra high vacuum systems. The groups of Alpert, at the University of Illinois [14], Redhead [15] at National Research Council, Otawa, and Robins [16], independently discovered the role of the surface ions desorbed from the grids of gauges and mass spectrometers by impinging electrons. These ions contribute significantly to the apparent ion current, thus indicating values much higher than the true system pressure, even with the measuring instrument operated at low electron emission current. Fig. 1 shows a spectrum recorded by

Total pressure = 5 x lO'^torr

28 20 22119 44

Fig. 1. Mass spectrum recorded with a 90° magnetic deflexion mass spectrometer of the Davis Vanderslice type. Note the ESD surface ions at mass 19 (F"^), and the shoulder at mass 16 (O"^) with 5 eV kinetic energy. From Ref. [14]

610

Alpert's group with a 90° magnetic deflexion mass spectrometer of the Davis VandersHce type [17]. Note the peak at mass 19 assigned to ESD of adsorbed fluorine and the double peak at mass 16 corresponding to O^ from the gas phase and surface ions from ESD. The calculated kinetic energy of this surface ion at 5 eV was considerably in excess of the thermal range of energies normally associated with gas phase ionisation. The discovery of the threshold and energy distribution of ejected ions attracted much excitement amongst investigators since their form should reflected the character of the bond between the adsorbate and the surface. Redhead [18] measured the ion energy distribution of CO"^ ions from carbon monoxide at tungsten and found a distribution with two most probable kinetic energies at 1 and 7 eV. However, O"^ ions from O2 adsorbed on the same surface showed only a single peak at 7 eV. These findings led in 1964 to the presentation of the first theoretical interpretation of the ESD process independently by Redhead [19] and by Menzel and Gomer [20], later known as the MGR model. The systematic study of DIET demanded special instruments to study the phenomena, specifically to determine the desorption yield, the ion kinetic energy distribution and the energy threshold for the ejected ions. The focus upon ion detection, desorbed with a typical efficiency of around 10'^ ions/electron, naturally served as the most accessible monitor of the desorption process. Most early investigations of ion desorption parameters made use of the so-called electron desorption cell, with the exception of the contribution of Moore mentioned earlier. In its simplest form this consists of a sample situated in front of a filament, with a collector of either planar or hemispherical form, to the rear of the filament. In the hemispherical form the sample face will be entirely surrounded by the collector, with the sample at the centre of curvature. The arrangement can also incorporate further grids to prevent secondary electrons from the sample reaching the collector and to suppress photoelectron production at the collector due to the action of soft X-rays produced at the sample by the impinging electrons. This was the technique used by Redhead [21] to study ion energy distributions, ion yields and thresholds for ESD with oxygen on molybdenum and with carbon monoxide at tungsten and molybdenum. Degras and Lecante [22] used a similar device to study the adsorption of CO on tungsten, while Madey and Yates developed a similar electron desorption cell at NBS (now NIST) [23] to perform ESD studies of the reactivity of CO, NO and O2 with tungsten [24]. In an approach taken by Lopez-Sancho and de Segovia [25] an ionization gauge of the suppressor type [26] was adapted to study the reactivity of O2 on tungsten surfaces. The threshold for O^ desorption from O2 at W was determined at 24.2 eV with a value of most probable kinetic energy of 6.7 eV. Several devices based on the similar concept of a desorption cell have been used by other labo-

611

ratories; however, a common limitation was that the identity of desorbed ions could not be explicitly determined. Thus, to circumvent this limitation the next logical development was to couple an electron desorption cell to a mass spectrometer. Magnetic sector mass spectrometers had been utilised to investigate phenomenological aspects of ESD in the early 1960's as in the work of Young [27] and the systematic study of G. Moore [13] related with the adsorption of CO on W and Mo. One of the undesirable features, however, of magnetic deflection mass spectrometers is that mass separation is energy dependent. This was overcome with the advent of the quadruple mass spectrometer, QMS, which in addition offered a far more compact structure free from magnetic fields. Two of the first instruments to combine an electron desorption cell (of hemispherical form) with a quadrupole mass spectrometer were developed at NIST by Madey and Yates [28] and at Liverpool by Leek and Williams [29]. In both instruments the hemispherical collector contained an aperture just in front of the entrance to the QMS, so as to sample a fraction of the desorbed ion flux. While the hemispherical collector could detect the entire angular integrated ion yield, the QMS detected only those ions emitted in the solid angle subtended by the aperture. An instrument based on this principle was later developed in Madrid with the objective also to identify neutral desorbed species [30-32]. There has also been in addition a significant contribution to ESD using experimental methods based on changes in surface parameters as opposed to the direct detection of desorbed products. Menzel and Gomer [33] performed studies on the reaction of H2, CO, O2 and Ba adsorbed at W in a field emission electron microscope, monitoring changes in the work function of the tip. Dylla, King and Cardillo [34] using Auger electron spectroscopy, AES, and mass spectrometry techniques to study the adsorption of CO on plasma and oxygen-etched silicon surfaces, obtained data on total cross sections with both techniques which showed good mutual agreement. A combination of a double pass cylindrical mirror analyser with time of flight of ions was used for Traum and Woodruff [35] The ESD of CO adsorbed on Ir (111) has been studied by LEED using the intensity of the CO (VsxVs)/? 30°pattern (Shek, Withrow and Weinberg [36]); they obtained data on cross sections and rates of induced dissociation at the surface. Auger electron spectroscopy and mass spectrometer techniques have been applied to ESD of O2 on tungsten carbide for a broad range of incident electron energies, 80 eV to 2.5 keV, by Stori, Braun and Gomer [37], who found a maximum in ion yield at 300 eV. Using a similar instrumental approach, Priegge, Niehus and Baueri [38] investigated the ESD of O2 on W (111) and determinated a cross section of 2x10'^"^ cm^, much lower than the value 7x10'^^ cm^ reported by Lopez-Sancho and de Segovia at Madrid [25]. However, diffe-

612

rent results using different methods at the same sample can be found; Williams at Liverpool found cross sections of 1.3x10"^^ cm^ and 1.4x10'^^ cm^ for O2 on W using respectively a time of flight ion detection technique and a cylindrical mirror analyser, CMA, for Auger analysis [39]. The authors proposed that the explanation for the differences lay with the specificity of the ion response in ESD, dependent on the adsorption states contributing to the response. A review to 1989 on the use of ESD as tool for surface adsorption studies can be seen in Refs. [3,40,41]. In 1972 Madey at NIST [42] discovered that O^ ions desorbed from oxygen on tungsten had angular distributions with sharp maxima in the direction normal to the surface. Later, Madey, Czyzewski and Yates developed a sophisticated instrumentation incorporating microchannel plates [43] to improve the study and visualisation of the phenomenon. Patterns were observed with sharp structures, and so emerged the technique of Electron Stimulated Desorption Ion Angular Distribution, ESDI AD. The principle of the technique as identified from the work at NIST is that ions are initially emitted along the bond direction at the surface. This discovery stimulated theorists to refine the dynamics of the process taking into account the contribution of re-neutralisation processes of ions and of image and other short range forces. To a first approximation under typical desorption conditions, for angles up to around 40"" these opposing influences leave the distribution largely preserved [44-53]. Thus, by the early 80's ESDIAD was confirmed as a useful technique for characterisation of the bonding geometry of adsorbed molecules or surface atoms. In 1988 the ESDIAD technique was further improved by the addition of digital video data acquisition to obtain excellent patterns of surface-adsorbate structures [54]. As mentioned earlier desorption of ions, neutrals and metastables can also be caused by photons incident at the surface, in the process known as Photon Stimulated Desorption, PSD. In principle, there is an important difference between ESD and PSD. While in the first case, the energy of the primary electron interacting with the electronic structure of either the substrate or adsorbate atom can be partially transferred, in PSD the whole energy of the primary photon is transferred to the interacting electron. This raised the question as to whether this difference in primary excitation mechanism gives a different response with both processes. The first studies using photons of well-defined energies were performed by the Knotek group [55] at the Stanford storage ring. They found that ion yield curves are sharper in PSD than in ESD, although the onset of ion desorption is the same in both cases. Several laboratories were motivated to perform further experiments to compare ESD and PSD. Madey's group at NIST and Eastman's group at Wisconsin [56] in a joint experiment compared the anisotropy of ion emission with ESD and PSD, but found the angular distribution for O^ from W (HI) to be identical with both methods. PSD experiments by

613

Stockbauer, Madey and Hanson [57] using the SURF-II facility at NIST with C from Ti02 again showed similarity with BSD experiments. A review of ejection mechanism related to these different studies was presented by Feibelman in 1983 [58]. He analysed the areas where BSD can be useful as an analytical technique, and discussed experiments to elucidate the desorption pathways at the surface. At the end of this introduction, we can conclude that by the middle of the 1980's it was well established that electrons or photons of certain energy, generally higher than 5 eV, impinging at surface atoms or adsorbate produce electronic transitions, which yield ions, neutrals and metastables to the gas phase. So, once the principles of the BSD and PSD were established and that both primary excitation of the atoms yield similar results, the process was recognised by the general title of Desorption Induced by Blectronic Transitions, DIET. The main findings can be summarised as follows: (i) Exposure of surface atoms or adsorbate atoms to electron or photon fluxes above certain energy, gives rise to the desorption of ions, neutrals, fragments and metastables. (ii) The basic mechanism of ejection involves electronic excitation of the substrate atoms or adsorbed entities, (iii) Ions as well as neutrals are produced at specific threshold energies of the impinging electrons or photons, (iv) Ions and neutrals are produced with kinetic energies of a few electron volts; neutrals can also be produced in their ground or metastable states. (v) Cross sections for ion production are lower than those corresponding to gas phase ionisation, dependent on the adsorbate. Ion desorption cross sections are typically around 10"'' ions/electron with neutral desorption around 10"* neutrals/electron. (vi)

Ions are ejected in specific directions, which correspond to the metal atom-adatom bonding direction. Studies at this time related to several metal substrates with mainly small molecular gases (i) Hydrogen [30, 31, 59-65]. (ii) CO [2, 10, 24, 45,46, 66-73]. (iii) Oxygen [16, 23, 45, 74-79], (iv) H2O [45, 60, 61]. The latter two gases reflect a first step of enquiry into oxidation and corrosion problems at surfaces. Progress with alkaline atoms on several surfaces was pioneered by the Ageev group at St. Petersburg [80- 83]. Interest in DIET studies at oxides surfaces underwent an explosion in 1978, when Knotek and Feibelman at Sandia Laboratories [84] reported a new mechanism for BSD, the so-called "Auger decay of a core hole". The model explained well the thresholds and ion kinetic energy distributions for the ejection of O"*",

614

OH"^ and F"^ from transition metal oxides of maximal valence such as Ti02, V2O5 and WO3, for which, the MGR theory could offer no satisfactorily explanation. The authors subsequently published a reinterpretation of previous results from chemisorption systems to conclude that with ionic bonding the Auger mechanisms for desorption explained well the observed thresholds for ion production [85]. The new mechanism has become known familiarly as the KF model. In their survey Feibelman and Knotek entered into an in-depth review of thresholds in which they compared findings of energy loss spectroscopy, ELS, with thresholds as identified with changes in the slope of the ion yield as a function of electron energy. The significant finding was that singularities, inflexions and slope changes in the ion yield curve, could be unambiguously identified with core level excitation of the surface atom and/or electron levels of the adsorbate. For Ti02 and other full valence oxides they found that the O^ threshold began at about 22 eV, which corresponds to the ionisation of the O 2s level. A new increase of the O"*" ion yield at about 34 eV, could be correlated with ionisation of the Ti 3p level at 33.6 eV. Developments around the KF model gave rise to a wide interest in stimulated desorption processes and their role at oxide surfaces. Covalent systems were also found to reflect some features of the Auger decay model, which will be discussed in the next paragraph. 2. EJECTION MECHANISMS: THE FEIBELMAN-KNOTEK MODEL The basic principle of the Feibelman-Knotek model, KF model, as first published [84] for full valence oxides is based on an inter-atomic Auger neutralisation of a hole produced in the core level of the surface atom by the primary electron or photon. The model as related to Ti02 is sketched in Fig.2. The impinging electron (e) or photon (hv) creates a hole in the Ti 3p level (33.6 eV). Since in a full valence oxide the valence band is empty, one electron from the O 2p (13.5 eV) level decays to the 3p core level. In the situation two O 2p electrons take the excess energy and leave the atom; the initial Ti'^'^0^' configuration is thus transformed to a final state Ti"^"^ O"*". Here the O"^ is subject to a very high Coulomb repulsion force and, as a consequence, is ejected into vacuum. In a later communication Feibelman and Knotek [85] reviewed the general concepts of ESD over a range of chemisorbed systems and showed that many features can be easily understand using the Auger decay model instead of the MGR theory. For the Ti02 system the work by Knotek, Jones and Rehn [86] showed that the desorption by ESD and also PSD of H^, GH"^ and F"^ following water exposure (with residual fluorine) has major thresholds at the G 2s and F 2s core levels, consistent with the applicability of Auger decay model also to covalent che-

615 Auger electrons

Vacuum level

n

'(3d) Valence

r n Ti(3p)

rl r 1

'\^^^dy/)^/////^A

1

0^0 Ti02 bonding Final state yielding an ion by coulorabian repulsion

Fig. 2. Schematic of the Feibelman-Knotek model for the TiOa case. [3]

misorption systems. The Menzel group at Munchen [87, 88] studied the case of O2 and CO on W(IOO) and Ru(OOl), where the ion desorption yield shows thresholds associated with the ionisation of the C Is and O Is core levels (respectively to 270 eV and 510 eV). This was in addition to the usual valence band excitation (MGR). Water and fluorine adsorbed on several oxides have also shown ion desorption by Auger decay [89, 90]. Threshold for H"" ions were seen to be dominant with ESD of water at Al and W surfaces as found in joint experiments at Liverpool and Madrid [ 91,92] with thresholds associated with excitation within OH radicals. Thresholds of H"" with water at Nb were shown to be linked to the core Nb levels [93]. These findings, generally, supported the applicability of the KF model for desorption within some covalent systems. Knotek and Feibelman [94] examined the modification to a surface when exposed to ionising radiation and assesed the damage that can be produced. They addressed the stability of ionically bonded surfaces, where the KF mechanism applies, and concluded that Auger induced decomposition only occurs when the cation in the solid is ionised to relatively deep core levels. In the case of non-maximal oxides as with NiO, Freund's group [95] showed that whilst desorption of neutral NO and CO from NiO(lOO) and (111) surfaces has thresholds at the C Is, N Is and O Is core levels, it proceeds mainly on the basis of the MGR model, involving an excited state of the adsorbate. An overview of electronic desorption presented by Feibelman in 1983 [96] examined particularly the stability of the multiple-hole final state configuration leading to desorption. The presence of multiple holes , and associated hole-hole correlation

616

effects, is seen to lead to inherently longer iSnal state times life and so can facilitate desorption in covalent systems. These many body effects in desorption are also addressed in the work of Ramaker [97] and have their origin in the CiniSawatzky Configuration-Interaction theory of hole interaction. ESD induced by electrons extracted at the tip of a scanning tunnelling microscope (STM) has recently applied to the study of the H-Si and Cl-GaAs(l 10) systems by Shen and Avouris [98]. A feature of the STM is the capability to provide an intense electron density (10^^ A/cm^) of atomic dimensions directly at the surface atoms. Under these conditions, desorption can also proceed with multiple-vibrational excitation. A combination of ESD and STM holds promise as a tool for atomistic DIET, also as a means to asses the influence of electron irradiation at surfaces [99]. In ESDIAD an interesting aspect is the determination of the trajectory of the ion on leaving the surface, which will be discussed later in relation to the TiOi surface. Laser stimulated desorption has been also studied as in the work of Freund's group [100]. Low energy excitation and desorption dynamics from NO adsorbed on NiO (100) and (111) oxide surfaces is stimulated by a Ar-Fexcime laser with 6.4 eV energy. They found total cross sections of 6x10"^^ cm^ and explained their results on the basis of the MGR model. Induced ion desorption by resonant photon-stimulated desorption has been also observed. The yield of HT desorbed ions as a function of the incident photon energy from CeOi has been studied by the group at Sandia Laboratory. [101] They observed a sharp resonance in PSD of H^ using photon energies near the Ce 4d edge at 110 eV. The resonance is due to autoionisation from excited states from the interaction of the Ce 4d and 4f levels. The contribution of secondary electrons to DIET has also been studied. Jaeger and Stoehr proposed that secondary electrons are the predominant contribution to the H^ ion yield in the NHs/Ni system [102]. In a study of PSD ion yield of YC from the OH/YbO-Sm system, Schmidt-May, Senf and Voss explain the structure on the basis of the secondary electron contribution to the ion yield [103]. Hutson, Ramaker, Bermudez and Hoffbauer [104] found a significant contribution of secondary electrons to the ESD ion yield of OH^ from TiOi at incident electron energies higher than 50 eV. An approach taken by Madey's group was to study medium-energy backscattered electron diffraction at TiO2(110) [105]. A maximum in the backscattered electrons from the second layer of atoms was found at an incident electron angle of 45°, which corresponds to a close packed direction for the (110) surface. Under this conditions the incident electrons are focused onto the second layer atoms by the atoms in the first layer. In their study of ESD ion yield from a well characterised TiOi (110) surface, the Madrid group [106, 107] found that in addition to the threshold at the

617 200

2500

175 2000

150 <

O"*" ion current

125

• 1-4

00

« 100 ^

CIJ

(U

Secondary e emission

75 50

r? cd XJ

a o

25 200 300 e~ energy (eV)

400

0

500

Fig. 3. Ion yield from a nearly perfect Ti02 (110) surface and secondary electron emission. [107], both as a function of incident electron energy

Ti 3p core level, distinct increases in ion current also correlated with increases of the secondary electron emission as shown Fig. 3. As no core levels for O and Ti atoms exists at this energy it can be concluded that the sharp increases are due to an enhancement of the ion yield produced by secondary electrons from the Ti02 sublayer. It was also observed that these thresholds are very sensitive to the azimuth angle of the incident electron at the surface. It would appear that this phenomena could be related with the work of Maschhoff, Pan and Madey on the backscattered diffraction of electrons by the Ti02 sublayer [105]. 3. DIET AT OXIDE SURFACES In addition to the researches on oxygen adsorption at surfaces mentioned in the introduction, DIET has contributed significantly to the study of non-stoichiometric and stoichiometric oxides surfaces, as well as the study of the reactivity of oxygen with oxidised metals and semiconductors. However, we shall restrict the scope of this review to oxidised surfaces and oxides. Henrich and Cox have reviewed the oxide surfaces up to 1993, which includes a chapter on ESD and PSD [108].

618

3.1. Oxidised Surfaces Bertel, Stockbauer and Madey investigated ESD with oxygen at oxidised Ti (0001) [109]. The O ion energy distribution for this case was found to be peaked at 5 eV with the threshold at 35 eV, consistent with the KF model. Oxidised chromiun with stoichiometry between CriOs and CrOi has been studied by PSD and photoemission [110, 111]. Ion emission of O^ is well correlated with 3p~>3d excitations. One of the most active groups in ESD studies of the adsorption of Ba, Li, Na, K and Ca at oxidised metals has been the group of Ageev at the Joffe Institute, St. Petersburg [112-116]. They made use of a surface ionisation detector to detect directly the desorbed alkali atoms. The atom kinetic energy distribution typically has multiple peaks, with energies over the range 0.2 Hi^^O+^^O is seen as being incorporated into the surface oxide. 3.2. Stoichiometric Oxides 3.2.1. The TiO2 Model Surface. Most of the work on stoichiometric oxides has been carried out on TiOi, especially samples with the (110) orientation which offers an interesting and contrasting character of possible adsorption sites. In addition this surface is the most stable and does not facet with thermal treatments. Fig.4 is the ball model of this surface [119]. Two kind of O"^ ions can be well differentiated at the surface: (i) ions located in bridging sites, which protrude out of the surface plane, and (ii) ions located at the in-plane sites. Ti atoms are six or five co-ordinated corresponding to positions associated with O"^ ions at type (i) or type (ii) sites, respectively. One of the first contributions of DIET to oxides was to correlate

619

THE STOICHIOMETRIC TiO, (110) SURFACE [110]

0 BRIDGING

[001]

0 IN-PLANE

Ti ATOMS

Fig. 4. The TiO2(110) ball model [158].

ESD witli the local geometry. In 1985 Kurtz, Stoc]<:bauer and Madey [120] studied the influence of the surface stoichiometry on the O"^ ion yield from nearly perfect Ti02 (110) and (001) surfaces as well as from defect surfaces produced by Ar"^ ion bombardment. This defect surface was shown to be deficient in O atoms. The unexpected result was that the sputtered surface (non-maximal valence) gave an ion yield which was three times that of the annealed (001), and seven times that of the annealed (110) surface, contrary to the predictions of the KF model. The conclusion was that the local environment of the Ti atoms influences the ion production, linked with the suppression of ion re-neutralisation processes. Later the authors studied ESDIAD at these surfaces, as shown in Fig. 5 where the patterns for the (001) and (110) surfaces at different stages of annealing of the initial sputtered surfaces are presented [121]. Note how the normal emission with the (001) surface increases with annealing temperature. The emission from the (110) surface shows a significant change in the pattern after annealing at 700 K. The study was extended by Kurtz [122] who showed that whilst the ion yield from both surfaces increases up to 700 K, at higher temperatures the ion yield from the (001) saturates, while that of the (110) surface decreases. He observed these changes to be well correlated with the form of the ion angle distribution. In the case of the (001) surface, the changes are lin]<:ed with the development of a faceted structure with ion emission from the edges of

620

ESDIAD from T i 0 2 (001)

(110)

sputtered

400 C annealed

800 C annealed

Fig. 5. ESDIAD patterns from Ti02 (001) and (110) surfaces after the indicated treatments [121].

the facets. The (110) emission shows a more complex behaviour in angular distribution with changing temperature. Ion emission was proposed to take place from step sites whose orientation changed with temperature. The final decrease in ion current at temperatures higher than 700 K was attributed to the estabHshment of flat surfaces with lower ion emission probability than the defect sites. Kurtz considered that the faceted and step edges are seen as bulk truncation containing reduced number of O coordinated Ti atoms, especially 4-fold coordinated sites, which would have a low probability of emission. Trajectory of ions on leaving the surface has been pursued mainly for Ti02. Clinton [123], and the Madey's groups [124, 125] initiated calculations based on

621

considerations of image and re-neutralisation forces on the ejected ion on its way to the vacuum. A more detailed model adopting charge densities linked to the (110) surface was later investigated by Walkup and Kurtz [126]. They pursued a theoretical calculation of the energies and trajectories of ion emission from TiO2(110) adopting a model ionic surface with potentials according to the superposition of Hartree-Fock charge densities of the individual ions of the solid. They predicted a lower probability of O^ desorption from the in-plane site as compared with bridging sites, with most probable kinetic energy of 4 eV from this location. For the bridging site the calculated kinetic energy was 7.5 eV with a pattern of ion emission concentrated in a central spot within lO"" of the surface normal. For the in-plane site some anisotropic features lead to an elongation of the pattern along the (110) direction, however, within an overal weak emission probability. The kinetic energy distribution of O"^ desorbed ions with BSD at Ti02 (HO) was investigated at both Liverpool and Madrid [127, 128] with findings of a bimodal energy distribution. Fig. 6 shows the kinetic energy distribution of O"^ ions for a range of primary electrons kinetic energy from 35 eV to 400 eV obtained at the Madrid laboratory [128,107]. Two most probable kinetic energies can be seen at 4 and 7 eV. It is interesting to observe that the intensity ratio of the 4 eV to 7 eV peak increases with incident electron energy. In addition, further studies have shown that when the surface is Ar"^ ion sputtered the shape of the ion kinetic energy distribution changes significandy with a decrease of the contribution at the high energy side. This suggests that since ion etching should remove mainly the O bridging oxygen, the high energy peak should be identified with O"^ ions desorbed form this bridging sites. This would be consistent with the calculated peak energy setting of 7.5 eV for the bridging oxygen site as 400

0

2

4

6

8

10

O"" KINETIC ENERGY (eV)

Fig. 6. Kinetic ion energy distribution of O"^ ions from Ti02(l 10) surface at the indicated incident electron energies [128].

622

found in the work of Walkup and Kurtz [126] leaving the 4 eV peak to be associated with the in-plane site. However, according to their calculations the probability of desorbing ions from the O-in plane oxygen is rather low, reflecting the action of repulsive forces mainly in the plane of the surface. An explanation may lie with the role of secondary electrons, as discussed earlier, which preferentially desorb ions from the in-plane site. This could arise from the reenforcement of back scattered electrons in the manner discussed by Maschhoff, Pan and Madey [105]. ESDIAD studies at Liverpool [133] at the clean TiO2(110)(lxl) surfaces reveal a single emission lobe, slightly elongated in the [110] direction as shown in Fig.7. It has been proposed that this character can be reconciled with emission due to binding oxygen ions, since there is a greater probability of re-neutralisation along the [001] direction linked with the proximity of oxygen atoms in the row. This would promote an elongation observed of the ESDIAD pattern in the direction perpendicular to the row as observed in the experiment. 3.2,2. The TiO2(100)(lx3) and (110) (1x2) Reconstructed Surfaces An interesting aspect of the study of TiOi single crystal surfaces has been the observations by ESDIAD of surface re-constructions induced by thermal treatments. At the TiOiClOO) surface, O^ ion emission along the surface normal at room temperature has been shown by the group at Liverpool [129] to transform into a two-lobe desorption feature after heating at 950 K. The accompanying LEED observation, carried out by reversing the electric potentials at the ESDIAD optic, demonstrated the development of a (1x3) surface. The off-nor-

Fig. 7. ESDIAD patterns from Ti02(l 10) surface: (a) (1x1) and (b) (1x2) reconstructed surface [132].

623

mal O"^ ion emission at around 40 ± 5 degrees is entirely consistent with desorption from (110) facets which feature in the so-called micro-facet model earlier proposed from X-ray diffraction and STM studies. Further annealing in oxygen was shown by ESDIAD to identify mixed phases leading to the eventual reestablishment of the (1x1) surface phase. ESDIAD findings for the reconstructed TiO2(110) surface from the Liverpool group in 1996, contributed significantly to resolve the nature of surface oxygen bonding upon transformation of the surface from (1x1) to (1x2). Two models had been proposed: Firstly, the missing-row model [130] , in which it is assumed that alternate rows of bridging rows of the (1x1) are removed to form the (1 X 2) structure (see Fig. 5). Secondly, a model in which the (1 x 2) surface consists of strings of Ti203 lying on top of an atomic arrangement similar to that of the original (1x1) surface, also yielding a (1x2) structure [131]. Fig. 7 shows the ESDIAD O"^ ion pattern from the (1x1) and the (1x2) surface obtained by annealing at 1200 K [132,133]. The change from one to two off-normal emissions with polar desorption angles of 18 ± 5 indicates a distinct change of bonding configuration for the surface oxygen; this can be reconciled with desorption from edge sites of oxygen atoms at the Ti203 strings but not with the missing row model. Subsequent collaboration in studies with photoelectron spectroscopy between the Liverpool and Madrid groups at the Daresbury Synchrotron [134] showed the absence of any Ti^"^ character upon thermal transformation. This further ruled out the missing row model of reconstruction, since Ti^"*" cations associated with the resulting four-fold co-ordination would be expected to contribute to the result. 3,2.3. Some further oxides surfaces studied by DIET. DIET has also been applied to the study of rare earth oxides, YBaCuO oxides as well as Si02 with significant contribution to the understanding of their structure. Loubriel, Knotek, Stulen, Koel and Parks studied the ESD and PSD of oxide films of Ce and Er and compared the results with those obtained by Soft X-ray absorption, SXA [135]. From the ESD experiments they concluded that the 5p level in the rare earth is involved in the electronic transitions for ion desorption from the surface. Schmidt-May, Sent, Voss, Kunz, Flodstrom, Nyholm and Stockbauer later studied the resonant electron and ion emission and the desorption mechanism from oxides of Sm, Eu and Yb [136]. They observed that the enhancement in photoelectron at the 4d edge is also found in the ion desorption yield. They concluded that the use of PSD as a proof of the specific desorption site is strongly limited because the initial excitation does not occur in the surface complex from which the detected ion is being desorbed. Rosenberg and Wen in 1988 observed the desorption of O2 from YBa2Cu307.x by excitation of the Cu 3p and Ba 4d core levels [137]. The ESD

624

desorption mechanism involves excitation and shake-up or correlation states while the PSD at the Ba 4d edge appears to involve excitation of the localised Ba^"^ states. Desorption of positive and negative ions from YBa2Cu307 was studied later by Hoffman, Moss, Paterson and Petravic [138].The origin of the O"^ ions is by excitation of the Ba 4s and Y 3p core levels with a threshold at 260 eV. Baragiola, Madey and Lazillotto have studied the ESD from Si02 surfaces [139]. They observed Si"^, Si"^^, Si"^^ multiple charged ions as well as O^ ions. The threshold occurs at the excitation of the Si 2p ar 110 eV but with some additional structure in the ion yield depending on the charge of the ejected ion. A threshold at the O Is level is also observed. A significant difference in the ion kinetic energy distribution profile of the O"*" ions with O"^ from other surfaces is that the energy tail extends up to 30 eV. Petravic [140] measured the threshold for O"^ and O' from Si02 at energies corresponding to excitation of the Si 2p and Si 2s core levels. Multiply charged Si ions, Si"^, Si^"^ and Si^"^, are found at excitation of the Si 2p level. Excitation of the O Is produced only Si"^^. As ejection mechanism the author proposes that Auger neutralisation of the core hole produce multiple-hole final states in the bonding orbital as applicable to the KF model. ESD of MgO(lOO) has been studied Gotoh, Takagi and Tominaga [141] using the time of flight technique. They found thresholds at 80 eV and 210 eV, which cannot easily be explained because these values are quite far from the Mg 2p core level excitation at 50.8 eV. The Madrid group also studied this MgO (100) surface [142] and Fig. 8 shows the ion kinetic energy with a bimodal structure having most probably kinetic energy at 6.1 eV and 9 eV. These -T

60 h

0

1

T"

— I — I — I —

MgO(1Q0) 500 eV

2

4 6 8 10 12 KINETIC ENERGY (eV)

14

Fig. 8. Ion kinetic energy distribution of desorbed O"^ from MgO(lOO) [142].

625

peaks are assigned to the role of defects on the surface. This interpretation does not, however, agree with the explanation for the lED reported by Gotoh, Takagi and Tominaga, who interpret the second peak as due to the existence of a physisorbed layer of O2, [141]. Recently a very interesting DIET study of ion desorption from mineral oxide surfaces of relevance to the origins of Na and K in the atmospheres of Mercury and the Moon has been performed by Madey and Yakshinskiy at Rutgers University [143]. Their objective was to understand the nature of the dominant processes for the generation of Na and K. On the basis of the alkali ion yields, desorption cross sections and ion energy distributions, they propose that Na and K adsorption at mineral surfaces is characterised by relatively strong chemisorption. The Na and K either diffuse up thorough the regolith or return to the surface from the atmosphere. The ESD desorption yields of neutral alkali atoms are significantly higher than the yields for ions. This work represents an interesting application of DIET to the understanding of the planetary formation. 4. ADSORPTION AND REACTIVITY AT OXIDES BY DIET Studies of adsorption and reactivity of gases at oxides surface using DIET have been mainly with Ti02 as a model surface, although other surfaces such as MgO have been also investigated. 4.1. Small molecules The site specificity of DIET processes with adsorbates at the Ti02 surfaces has been a topic of much interest. In 1980 Knotek [144] investigated the desorption of H^, O"^ and OH^ ions from hydrogen contaminated and after H2O exposed surfaces of Ti02 (001) and SrTi03 (111), either annealed or activated to point defect form. The TiO2(001) annealed surface contaminated with hydrogen from the residual vacuum showed H"^ and O"^ ions. The H"^ ions has thresholds at the ionisation potentials corresponding to the O 2s and Ti 3p core levels, while the O^ ions have a threshold at singly the Ti 3p core level; evidence that the H atoms are bonded to Ti and O surface atoms. The ion yield with water adsorbed at the Ti02 activated surface showed O"^, H"^ and OH"^ ions. H"^ ions have a major threshold at the O 2s core level with a minor onset at the Ti 3p core level. The threshold of the OH"^ ion is at the O 2s and Ti 3p core levels. Knotek thus concluded that the H^ is produced by the breakage of the OH bond. The Auger decay of an electron of the O 2s level can produce either a H"^ or an OH"^, depending of the molecular orbital from which the two electrons are removed. In the case of SrTiOs (111) exposed to water, desorption of H"^, O"^ and OH^ ions are also observed. The threshold for H"^ is at the Sr 4p core level, while that for a desorption of OH"^ is more closer to the value of the O 2s core level. A second

626

Fig. 9. Mass resolved TOF-ESDIAD images of highly O deficient T02(l 10) surface after ad sorption of 1 L NH3 at 160 K. Electron incident energy 200 eV. Inset shows the bond of the molecule to the surface [146].

threshold is observed at the Ti 3p. As in the case of Ti02, the threshold of the O"" is at the Ti 3p core level. Hydrogen adsorption on TiO2(110)-(lxl) surface has been studied using ESD and STM by Suzuki, Fukui, Onishi and Iwasawa [145]. H is adsorbed in two forms: (i) hydroxyl-type, 0-H and (ii) hydride-type, Ti-H. They used the STM to selectively depopulate the surface of H atoms by ESD with electrons at incident energy of 20 eV. At this energy O atoms are not desorbed. Diebold and IMadey studied the NH3-TiO2(110) system using ESDIAD with mass-resolved ion detection and for varying O defect concentrations at the surface induced by electron bombardment [146]. Fig.9 shows the mass resolved TOF-ESDIAD following adsorption of 1 L NH3 on a highly oxygen deficient surface. The H"" image shows the typical circular halo corresponding to the NH3 bonds illustrated in the inset. The N atom is bonded to the surface Ti atoms with the molecule in an upright position. The left part of the figure shows the image corresponding to O"^ ions from the substrate. 600

500

,,..

TiO2(110)/SO2

,

Saturated

400eV /

400

\

1

1

Ti02(110)/S02 AFTER HEATING AT R.T.

250 eV, /

200

250eV \

o 60eV

.^jL/^-^'^'TEi^

0

2

4

6

8

O'" KINETIC ENERGY (eV)

10

0

2

4

1

6

\ . |-~-T-

8

n

10

O" KINETIC ENERGY (eV)

Fig. 10. Left: Ion energy distribution of O"" from Ti02(l 10) surface after saturation with SO2 At 153 K. Right: after heating to room temperature [150].

627

The TiO2(110)-SO2 system at 153 K has been studied jointly by the Liverpool and Madrid groups using the ESD and ESDIAD techniques [147-150]. After SO2 saturation the O"" ion yield shows thresholds at 35 eV and 155 eV, which can be assigned to excitation of the respectively Ti 3p and S 2p core levels. A third threshold is observed at an energy of 240 eV corresponding to excitation of the S 2s core level. It is easily concluded that the origin of the O"" ions is at both the surface O atoms and in the SO2 molecule, thus implying sites with O surface atoms not covered by the SO2. Ion energy distributions were also determi-

[001]

/ ^ [ 1 T110]

[110]

[001]

Fig. 11. (a): ESDIAD pattern of Ti02( 110). (a) ESDIAD contour pattern measured with a sample bias of 150 V. (b)LEED pattern Ball model of the Ti02 surface. Black spheres, Ti atoms. White spheres bridging oxygen sites. Grey spheres, in plane oxygen atoms. An example of a sulphate SO4 species formed at the bridging oxygen is included in the model [148].

628

ned for both the clean and SO2 saturated surfaces, as shown in Fig. 10 for the range of electron incident energy iOOOOOOnvestigated. The shape of the lED at the clean surface is of the bimodal form earlier noted at this surface, see Fig.3. SO2 adsorption, however, produces a significant increase in the distribution at low energy with a most probable kinetic energy of 2 eV. The right part of Fig. 10 shows the influence on the ion kinetic energy distribution of heating the surface to room temperature. The enhancement of the low energy side shows that SOx radicals remains on the surface. The conclusion was that SO2 is molecularly adsorbed in agreement with NEXAFS experiments by Thornton et al. where it is proposed that SO2 is initially molecularly adsorbed on top of the five fold co-ordinated Ti atoms [151]. The ESDIAD pattern from the SO2 saturated surface as shown in the upper part of Fig. 11 reveals an elongation in the [iTo] direction as in the case of the clean surface. Without any new features being introduced, the ESDIAD findings in this case are inconclusive [148]. Nearly perfect surfaces of Ti02(l 10) are only weakly reactive at room temperature. However, O defect surfaces with or without deposited metals on the surface are more reactive as in the case of CO adsorption. The Madrid group has studied the reactivity of CO with O defective surfaces of TiO2(110) doped with Ta [152]. The O"^ ion yield for the clean surface and after CO adsorption follows the form shown in left part of Fig. 12. Thresholds for the clean surface occur at energies corresponding to excitations of the Ti 3p, Ta 4f, 4d and 4p core levels. Excitation of such a number of core levels yielding O"^ desorption has not previously been reported, moreover, their participation does not agree well with the expected lowering of the ionisation probability for inner core levels. There is a further singularity at about 120 eV which has been reported previously to occur with nearly perfect or defective surfaces, and probably related to seconda-

Ti02(110)+Ta CO Saturated

0

100

200

300

400

e- INCIDENT ENERGY (eV)

500

2

4

6

8

KINETIC ENERGY (eV)

Fig. 12. ESD of O^ of the O defective TiO2(110) Ta doped clean and CO adsorbed surfaces. Left: Ion yield. Right: Ion energy distribution [152].

629

ry electron emission as discussed earlier [106, 107]. From the measured threshold it can be concluded that in addition to the O atoms bonded to Ti atoms of the defective Ti02, there are additionally O atoms bonded on top of the evaporated Ta atoms. This implies that the Ta atoms migrate beneath the surface with the associated rearrangement of the O atoms. The O"^ ion yields changes dramatically when the Ta doped surface of TiO2(110) is dosed with CO, as shown by curve (b) on left part of Fig. 12. The thresholds at about 30 eV corresponding to the Ti 3p and Ta 4f core levels are not modified however. The threshold at the Ta 4d is not seen and a new threshold at 285 eV corresponding to excitation of the C Is is observed. This implies that CO molecules remain un-dissociated on the surface and bonded to the Ti atoms rather than to the Ta atoms. The O"^ ion kinetic energy distribution of the clean surface show the typical bimodal structure as commented upon before and shown in the right part of Fig 12 (b). The analysis of the ion profile upon CO adsorption shows that in addition to the two peaks at 4 eV and 6.3 eV with the clean surface, there is a peak at 3.2 eV related with the C-0 bond. The adsorption of CO on O defective TiOiCllO) was also investigated [153] with similar results to the Ta doped surface. In this study it is noticeable that thresholds at electron incident energies lower than the value corresponding to excitation of the Ti 3p core level are not observed. However, at an incident electron energy of 285 eV corresponding to the C Is core level a threshold is observed, but which gives rise to a far smaller increase of the ion current as compared to the Ti doped surface. The interesting conclusions are that the main desorption mechanism involves excitation of core levels, and that the Ta atoms makes the surface more reactive. The adsorption of water on TiOi (001) has been studied by Bermudez and Hoffbauer [154]. The neutral yield for OH desorption shows thresholds at 11.6, 21 and 32 eV incident electron energies, with the greatest emission at 11.8 eV incident electron energies. This unexpected value has not been reported before and is some 10 eV lower than the threshold for OH"^ at 22-24 eV. It would be interesting to extend the availability of data relating to the desorption of neutrals, however, there are understandably problems with detection sensitivity. Menzel has compared the adsorption of H2O and D2O on the Ti02 defective surface. Whilst the dynamics of the response reflected the isotopic influence, adsorption proceed at the same places [155]. The ESD and electron damage of adsorbed deuterated water on MgO(OOl) has been also studied by the Madrid group [156,157]. The adsorption of D2O produces a significant increase of the D"^ desorbed ion while the intensity of the O^ decreases. There is also a slight increase of OD^ desorbed ions. Desorbed H^ ions are also observed due to some contamination from residual hydrogen in the system. The left part of Fig. 13 shows the ion yield for the D^, H^, O"^ and OD^ mm desorbed ions obtained at 150 eV electron incident energy. The threshold for desorption oftheD"^ and H^ ions occurs at incident electron energy of 30

630 - , — . — I — . — , — . — I — . — I — . — I — I — I — , — , — ,

MgO(100) / l O U D g O )

-

MgO(100)

..M^.

* •

O* lED (500 eV) , 0

20

40

CLEAN 3 L (DjO)

60

1

-'.v'-A^v^^,.^^ H" Oh

--^WwOt

_ , ^^:^^^ ^ - ^ -

OD^ -J

100

200

300

400

Electron Energy (eV)

500

2

i-^_J

I

L_

4 6 8 10 12 Ion kinetic energy (eV)

Fig. 13. ESD of D2O adsorbed on MgO(lOO) at room temperature. Left: ion yield of BSD D^, H^, O^ and OH^ ions. Right: Ion kinetic energy distribution of O^ [142].

eV, while for O"*^ and DO^ the onset is at 55 eV. The ion energy distribution of the O"^ surface ions shows a bimodal peak with most probable kinetic energies at 5.6 eV and 9.1 eV, as presented in the right part of Fig. 13. The ion kinetic energy distribution of the desorbed D^ has a peak at 3.5 eV and a shoulder at about 7 eV. The ion energy distribution of HT shows two maxima at 3.3 eV and 6.7 eV in good agreement with the values for the D"^ ions. From these results it is possible to propose an adsorption model: the O^ ions have their origin through excitation the Mg 2p in the Mg-0 bond. The D^ and H^ ions have their origin by excitation of the O 2s core level from the 0-D or 0-H bonds. The OD^ ions are desorbed by excitation of the Mg 2p from the Mg-O-D bond. No ion yield was found for D"^ at an ionisation energy of 55 eV, so no total dissociation of the D2O molecule is foreseen. However heavy electron irradiation of the D2O layer shows only D^ and O"^ ions, which means that irradiation produces full dissociation of the adsorbed species. It is interesting to note that irradiation of the clean MgO(lOO) surface produces a dramatic decrease in the intensity ratio of the 9.1 eV to the 6.1 eV. peaks. This finding supports a hypothesis that the higher energy peak could be produced at surface defect sites, which are neutrahsed by the electron irradiation. An other interesting aspect of ESD is the transmission of ions through matter. Yakshinskiy, Akbulut and Madey have studied the transmission of 0^ ions through Au films deposited on Ti02 (HO). When the Au covers 5% of the surface they observe strong attenuation of the ESD O"^ ion intensity, due to the coverage of steps and defect sites of the TiOi by the Au atoms. At coverage between 5 and 30 % at the TiOi surface, O"^ attenuation is due to geometrical blocking of the substrate by growing clusters of Au atoms in a three dimensional form. The Madrid group in co-operation with Shierbaum at the University of Tii-

631

60

Pt 4f

/

TiOgCllO)^ 5

"Ti 3p /

.

"7 "" 100

40

•5 .

200

Pt/TiOgCllO) 0.3 ML

300

400-

20

^~

500

ELECTRON ENERGY [eV]

KINETIC ENERGY E^ [eV]

Fig. 14. BSD of O^ from clean Ti02 (110) surface and after evaporation of 0.2 ML of Pt. Left: O"^ ion yield. Right: O"^ ion energy distribution [158].

bingen [158] have studied the ESD of the adsorption of Pt on the Ti02(l 10) surface. The left side of Fig. 14 shows the O"^ ion yield for the clean and after desorption of 0.3 IVIL of platinum. Thresholds are seen at the Ti 3p and Ti 3p and Pt 4f for the clean and Pt dosed respectively. Platinum produces a significant decrease of the O"*" intensity due to the blocking of the Ti atoms by the Pt atoms. The right part of Fig. 14 shows the ion energy distribution of the clean and Au doped surface. The clean surface shown the typical bimodal profile characteristic of the nearly perfect clean surface with maxima at 4 eV and 7 eV. When the surface is covered by 0.3 IVIL of Pt the intensity of the peak at 4 eV decreases significantly. The intensity of the most probably kinetic energy at about 7 eV, however, decreases only slightly. As pointed out before, the intensity of the peak at 4 eV is possibly limited to emission from the in-plane O atoms by the reinforcement action of secondary or backscattered electrons. On this basis, as ion emission from the O at 7 eV, is not affected significantly, it could taken to indicate that Pt atoms adsorbs mainly on top of the Ti atoms. 4.2. Large molecules Recently the Liverpool Group pioneered studies with ESD and ESDIAD of molecules of the carboxylic acid family, viz formic, acetic and benzoic acid, at Ti02 (110) surfaces. The adsorption proceeds with the loss of hydrogen leading to the establishment of a carboxylate. In the case of acetic acid [159] [160] the reaction is: CH3-COOH = CH3-C00(ads) + H,l(ads)ESD shows consistently the displacement of O"^ ions of the substrate by H"^ ion species of the surface carboxylate, although to maintain stability of adsorbate

632

Fig. 15. ESDI AD of acetate adsorbed on Ti02(l 10) . (a) Angular distribution of O"^ ions from the clean surface, (b) H"^ ions with a (2x1) acetate overlayer . (c) H"^ ions from a (2x1) acetate overlayer after annealing to 450 K. (d) H"^ ion pattern as (b) but after exposure to 350 eV lectrons with a total dose of 3x10"^ C/cm^ . (e) H^ ions from a (2x1) overlayer of formate. [159]

ion current particular care is necessary to maintain the incident electron flux at low levels (typically < 100 nA cm"^). The simultaneous LEED observations performed at these low levels of electron density showed the surface to be reconstructed to a (2x1) pattern, consistent with 0.5 ML coverage of carboxylate OOat saturation coverage. Fig. 15 provides a comparison of the O"^ ESDIAD images from the clean surface and the H"^ species at saturation coverage of ace-

633

H

H

H^H

A ?• ? Ti 100 200 300 Azimuth angle (arb.unit)

Ti

bridging form

T

R ^/° Ti

(b) Bidentate form

Fig. 16. Left: A one-dimensional intensity profile along the line AB of the image shown in Fig. 15 (b). Lower part of the figure is the result of the curve-fitting using Gaussians. Right: The adsorption model illustrating the bidentate and bridging forms of the acetate. [159].

acetate. The angular distribution of the H"^ is more widely spaced, indicating greater angles of desorption. The profile of the intensity along the line AB is shown in the left part of Fig. 16 where it can be seen to be resolved into three Gaussian forms. The central part corresponds to emission normal to the surface and the two peaks at each side to off-normal emissions. Since the off-normal emission has circular symmetry, the image can be viewed as a superposition of a central lobe and a doughnut shaped halo. The central lobe is attributed to H"^ from surface hydroxyls. Simulation of the ion trajectories yields a polar angle of 35 degrees for the off-normal emissions. The halo of off-normal emissions is identified with the free rotation of the methyl (CH3) group in a model with the C-C axis perpendicular to the surface. Either bridging or bi-dentate forms of bonding of the carboxylate may be envisaged, as illustrated in the right part of Fig. 16 . The bridging form is favoured since all the five-fold co-ordinated surface Ti"^^ cations are converted back to the preferred six-fold co-ordination of the bulk, moreover, the adsorbate induced (2x1) reconstruction can be associated with the adjustment of the Ti"^"^ - Ti"^^ distance to accommodate the smaller 0 - 0 distance within acetate. ESDIAD findings with the larger benzoic acid species (C6H6COOH) are illustrated in Fig. 17 (a). Again the observation of a central peak intensity originating from H+ ions emitted along the surface normal, however, findings of relatively weak off-normal H-i- features. These exhibit, moreover, an azimuthal

634 a)

b)

o Hydrogen -60

-30

0

o

[110]

Oxygen in benzoate

30

Polar Angle (degrees)

Lattice oxygen

Titanium

Carbon

Fig. 17. (a) H"" ion ESDIAD image from 0.5 ML of benzoate on Ti02(l 10) ( Ep=300 eV, target bias 90 V) and (b) an intensity profile of the ESDIAD pattern taken along the [110] direction (soUd Une). The polar angle corresponds to thefield-freelaunch angle of the ion from the surface. Dashed line: schematic depiction of the normal and off-normal contributions. Right: Ball model illustrating the orientation of adsorbed benzoate and the relative positions of the aromatic rings of a benzoate dimmer. The location of a hydroxyl species is also shown [160].

anisotropy, with emissions preferentially along the directions of the major crystal axes. These are interpreted in association with simultaneous STIVl studies to reflect a "T"- type linking between adjacent aromatic rings of the benzoate. A ball model of the surface is shown in the right part of Fig. 17(b); the weak emissions arise from hydrogen on either side of the apex of the upright molecule, while hydrogen atoms on either side at the base of the ring, through their interaction with the bridging oxygen sites, contribute to the organisation of dimerised benzoate rows along [001]. Deliberate electron beam irradiation was seen to lead to the decomposition of these carboxylates at the surface, leading as in the case of acetate to the desorption of neutral species ketene (CH2CO) and CH3/CH4. The electron induced decomposition of methanol adsorbed on vacuum annealed TiO2(110) surfaces has been investigated over a broad temperature range by Henderson et al [161]. The Liverpool and IMadrid groups have studied the stability of glycine layers on Ti02(l 10) in the energy range 20-120 eV using synchrotron radiation [162]. As

635 Table 1 Summary of the main DIET parameters of oxidised metal surfaces or stoichiometric oxides (See ref. 3 for more parameters until 1985) Surfaceoxide

Atom/MoEjected leculespecie Surface Configuration W-0 0" Up-right 0 0"

02AV(poly) oxidised 02AV(poly) oxidised 02AV(poly) W-0 oxidised O2AV(110) oxidised O2AV(100) oxiddised 02/Mo(poly) State 1

TiO2(110) |TiO2(110) 02/Si(lll) Oxidised H2/Ti02

1(001)

24

163

0

6.0x10-^^

164

0"

2.0x!0-^^

165

State 2 Mo-0

Ti-0

0" 0^+0 0^

W-0 V-0 Ti-O-H

0" 0^ OH^

Ti-F

F" 0"

TiO2(001)

24

17.5 7

21.8 33.6 35.5 37.6 21.8 33,6 31.9

0^ Ti-0 Si-O-H Si-F Ti-O-H Ti-H/Ti-O

Most Desorption probable cross section REF. cm^ kinetic energy eV 7.0x10-^" 6.7 25 8.8x10-^^ 4.5x10-^^ 20 3.4x10'^"

0 0" 0 0"

O2AV(110) Ti02 single crystal WO3 single crystal V2O5 polycrystalline Ti02/0H TiO2(001)

0^

Threshold ev

0" 0" H" F" H" H^ 0^

35 38 24 38.5 21 32

0 Ti W V 0 Ti F

2s 3p 4f 3p 2s 3p 2s

Not measurable 1.3x10-^^ 2.6x10-2^ 166 6x10"^^ 2x10-^^ 84

2 3.75 4.75(after annealing) 4 7.5 (Ti 3p) 2.5, 4, 6.5

122

( 0 2s) (Ti 3p) ( 0 2s) (Ti3p)

58

127

167

1

636 Table 1 (Continued) Atom/MoSurfaceleculeoxid e Surface Configuration -H Ti - N - H -H

Ejected specie

Threshold ev

Most probable kinetic energy eV

dissociation Ti-0 0 Ti-S0 Ti-0 C0/Ti02(ll Ti-0(from dissociated 0) ( 0 deficient) CO) Ti-0 Ta-0 CO/Ti02 Ti-CO (110)+Ta Ta-CO Mg-0 Mg-O-D Mg-O-D p20/MgO Mg-H (from (100) residual) Ti-0 Ti02+Au (slighdy 0 After 5A Au deficient NO/NiO(100 02/Ce

02/Er

Ce-0

Er-0

1.5x10'^^ 35

0"

155

S 2p

2.0

0" 0"

35 35

Ti 3p Ti 3p

4,7 3.4,4.3, 7.4

153

0" 0" 0" 0" 0" OD"^ D" H^

Ti Ta C C 55 55 30 30

4,6.3

152

3p 4f, 4d, 4p Is Is Mg 2p Mg 2p 0 2s 0 2s

4,7 0 0" 3.44, 7.20 Strong atenuation NO

0"

H" 0"

150

3.2,4,6.3 6.1,9.2 3.0 3.5,6.5 3.3,6.7

3.6x10'"

157

6.3x10-'° 1.1x10-^^ 168

6x10-^'^

95 135

0"

H2"

Yb-H Mg-0

4,7.2

0"

OH^ 02/Yb MgO(lOO)

REF.

1.0x10-^' (stoichiometric surface) 146 0.9x10"^^ (defective surface)

NH3/Ti02 (110)

SO2/ TiO2(110)

Desorption cross section cm^

18.5 Ce5p/0 2s 33-37 Er5p 25 0 2s 25 0 2s 28 31 F2s 178 Yb 4d 531.6 0 2s 531.6 0 2s

103 169

637

Tabla 1 (Continued) Ejected Atom/Mospecie leculeSurface Configuration O2 Yba2Cu307-x CU-O2 Ba-0 0" Yba2Cu307 Surfaceoxide

0-

Si02

Si02

films

films

0" SiO^ Si^"

Si^'^

Threshold

Ev

REF.

137 138

139

1.8-4.7 140

100 Si2p 150 Si 2s n=l 170-190 n>l 4 4 2

02^

1.7x10-'' 6.2x10-'^ 5.4x10-'''

CH3OH CH3O

0" 0"

Desorption cross section cm^

79 Cu 3p 102 Ba4d 260±5 Ba4s+Y3p +Y3s 100±5 Cu 3p +Ba 4d+Cu3s+Y 3d+Ba4p Si2p 0 1s 4.6

02/Si(lll) CD30D/Ti02 (110) NH2CH2CO H /TiO2(110) TiO2(001) Defective Annealecd

Most probable kinetic energy eV

38

Ti3p 1.7/3.8 4.75

161 162

122

an assessment technique, use was made of the photoelectron spectroscopy of the valence band, the 2s band and the Ti 3p level. Photons induced desorption of glycine multilayer produces first orders desorption of the zwitterionic adsorbed species. The zwiterionic form corresponds to the neutral glycine form with the H from the carboxyl group rearranged to join the amino group, i.e. NHs'^-CHiCOO". When the multilayer was subject to photon irradiation to reduce the coverage to less than two monolayers, the remaining surface species are C and NHx with x=l,2. The photon cross section for desorption is significantly large at 5.4x10"^^ cml

638

5. SUMMARY OF RELEVANT KINETICS PARAMETERS The table contains data on the configuration and identity of desorbed ions as well as parameters of ESD. The extent of the data reinforces the significance of the contribution of DIET to our knowledge of oxide surfaces and their reactivity with gas uptake. 6. CONCLUSIONS We have presented a review of the main results obtained by DIET, which have contributed significantly to knowledge of the characterisation of oxides and oxidised surfaces, in both non-stoichiometric and stoichiometric form, as well as elucidating features of the interaction of gases at these surfaces. Following early experiments centred around the MGR model, the explosion of interest with oxides came with the establishment of the KF model. The main emphasis in the field has been placed on the TiOi as a prototypical model oxide. In addition to contributions with adsorption-desorption kinetics, a highlight of DIET has been very elegant way in which the local bonding has been derived from studies of thresholds and ion kinetic energy distributions. The powerful contribution of ESDIAD in obtaining the orientation of molecules and radicals at the surface as well as features of surface geometry is also remarkable. No surface technique in isolation can completely encompass a phenomenon, but DIET serves as an excellent complementary technique in many spheres of surfaces studies. In the future, processes of DIET will remain at the forefront of Surface Science; in particular, the possibility to combine DIET and STM is very exciting for DIET studies at an atomistic level. Also the work with larger molecules could be further extended to biological surfaces, building on the significance of titanium as a biocompatible material. ACKNOWLEDGEMENTS We thank T..E. Madey for useful comments on the manuscript. We also acknowledge the contributions from our colleagues and collaborators G. Thornton, E. Roman, S. Bennett, C. Greenwood, Q. Guo, M.C. Torquemada, I. Colera, E. Soria and I. Cooks. [1] [2] [3]

D. Lichtman and R. B. MacQuistan, in: Progress in Nuclear Energy Series IX, Vol.4, Pergamon Press, 1965, p. 95 T.E. Madey and J.T. Yates, J. Vac. Sci. Technol. 8 (1971) 525 J.L. de Segovia, in: Trends in Quantum Electronics. A.M. Prokhorov and I.Ursu (Eds.), Springer Verlag, Berlin Heidelbelberg, 1986, p. 217; Rev. Roum.Phys. Tome 23, num. 4-6 (1988) 925

639 [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

[34] [35] [36] [37] [38] [39] [40] [41]

E.M. Williams and J.L. de Segovia, Vacuum, 39 (1989) 633 R.D. Ramsier and J.T. Yates, Jr., Surf. Sci. Rep. 12 (1991) T.E. Madey, Surf. Sci., 299/300 (1994) 824 I.G. Newsham and D.R. Sandstrom, J. Vac. Sci. Techno!., 9 (1972) 596; J. Vac. Sci. TechnoL, 10 (1973) 39 Desorption Induced by Electronic Transitions, DIET VII, Eds. E. M. Williams and R.E. Palmer, Surf. Sci. 390 (1997) Desorption Induced by Electronic Transitions, DIET VIII, Eds. T. E. Madey, F. M. Zimmermann and R.A. Bartynski, Surf. Sci., 451 (2000) A.J. Dempster, Phys. Rev., 11 (1918) 316 R.H. Plumlee and L.P. Smith, J. Appl. Phys., 21 (1950) 811 Y. Ishikawa, Rev. Phys, Chem. Japan, 15 (1942) 83; 117. Proc. Imp. Acad. (Tokyo)I9 (1943) 380 G.E. IVIoore, J. Appl. Phys., 32 (1961) 1241 D. Alpert, Physics Today, 16 (1963) 22 P.A. Redhead, Vacuum, 13 (1963) 253 J. Robins, Can. J. Phys., 41 (1963) 1385 W. Davis and T. Vandershce, in 1960 Vacuum Symposium Transactions , Pergamon Press, 1961, p. 417 P.A. Redhead, J. Vac. Sci. TechnoL, 4 (1967) 57 P.A. Redhead, Can. J. Phys., 42 (1964) 886 D. IVIenzel and R. Gomer, J. Chem. Phys., 41 (1964) 3311 P.A. Redhead, J. Vac. Sci. TechnoL, 4 (1967) 57 ; Nuovo Cimento, Suppl. V (1967) 587 D.A. Degras, J. Lecante, Nuovo Cimento, Suppl. V, (1967) 598 T.E. Madey, J. T. Yates, Jr., Surf. Sci. 11 (1968) 327 J.T. Yates, Jr. T.E. Madey , J.K. Payn, Nuovo Cimento, Suppl. V (1967) 559 J.M. Lopez-Sancho, J.L. de Segovia, Surface Sci., 30 (1972) 419 W.C. Schueman, Rev. Sci. Instrum., 34 (1963) 700 J.R. Young, J. Appl. Phys., 31 (1960) 921 T.E. Madey, J.T. Yates, Jr. J. Vac. Sci. TechnoL, 8 (1971) 39 K.W. Asskcroft, J.H. Leek, D.R. Sandstrom, B.P. Stimptom and E.M. Williams, J. Phys. E. 5 (1972) 1106 F. Gonzalez, Ph.D. Thesis, Universidad Complutense, Facultad de Ciencias Fisicas, Madrid, 1982 F. Gonzalez and J.L. de Segovia, Phys. Scr., T4 (1983) 132 J.P. Adrados and J.L. de Segovia, Vacuum 34 (1984) 737 D. Menzel, R. Gomer, J.Chem. Phys. 41 (1964); in: Desorption Induced by Electronic Transitions, DIET I, N.H. Talk, M.M. Traum. J.C. Tully and T.E. Madey (Eds.), Springer Series in Chem. Phys., Vol. 24, Springer, Berlin, 1983, p. 44 H.F. Dylla, G. King, M.J. Cardillo, Surf. Sci., 74 (1978) 141 M.M. Traum and D.P. Woodruff, J. Vac. Sci. TechnoL, 17 (1980)1202 M.L. Shek, S.P. Withrow, W.H. Weimberg, Surf. Sci., 75 (1978) 635 H. Stori, P. Braun and R. Gomer, Surf. Sci., 141 (1984) 654 S. Priegge, H. Niehus, E. Baueri, Surf. Sci., 75 (1978) 635. S.W. Bellard and E.M. WiUiams, Surf. Sci., 80 (1979) 450 D. Menzel, J. Vac. Sci. TechnoL, 20 (1982) 538 E.M. Williams and J.L. de Segovia, Vacuum, 7/8 (1989) 633

640 [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53]

[54]

[55] [56] [57] [58]

[59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78]

T.E. .Madey, Surf. Sci., 33 (1972) 355 J.J. Czyzewski, T.E. Madey, J.T. Yates, Jr., Phys Rev. Lett., 32 (1974) 777 J. Gersten, R. Janow and N. Tzoar, Phys. Rev. Lett., 36 (1976) 610 R. Janow and N. Tzoar, 69 (1977) 253 T.E. Madey, in: Springer Series in Chemical Physics, Vol. 17, Springer, Berlin, 1981, p. 80 Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci., 141 (1984) 285 Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci., 169 (1986) 405 J. Gersten, R. Janowand, N. Tzoar, Phys. Rev. Lett., 36 (1976) 610; J. I. Gersten, R. J. Janow and N. Tozar, Phys. Rev. Lett. B 11 (1975) 1267 W.L. CHnton, Surf. Sci., 112 (1981) L791 Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci., 141 (1984) 285 Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci., 169 (1986) 405 C.Z. Dong, P. Nordlander and T.E. Madey, in: Desorption Induced by Electronic Transitions, DIET IV, G. Bertz and P. Varga (Eds.), Springer Series in Surface Science, Vol.19, Springer -Verlag, Berhn-Heidelberg, 1990, p. 34 A.L. Johnson, R. Stockbauer, D. Barak and T.E. Madey, in: Desorption Induced by Electronic Transitions, DIET III, R.H. Stulen and M.L. Knotek (Eds.), Springer Series in Surface Science, Vol. 13, Springer, Beriin, 1988, p. 130 M.I. Knotek, V.O. Jones and V. Rehn, Phys. Rev. Lett., 43 (1979) 300 T.E. Madey, R.L. Stockbauer, D.E. Eastman and J.F. Van der Veen, Phys. Rev Lett., 45 (1980) 187 D.M. Hanson, R.L. Stock bauer and T.E. Madey, Phys. Rev., B 24 (1981) 5513 P.J. Feibelman, in: Desorption Induced by Electronic Transitions, DIET I, N.H. Tolks, M. M. Traumm, J.C. TuUy and T.E. Madey, Springer Series in Chemical Physics, Vol. 24, Springer, Berlin, 1983, p. 63 D. Menzel, R. Gomer, J. Chem Phys., 41 (1964) 3311 T.E. Madey, Surf. Sci., 36 (1973) 281 F. Gonzalez, Ph.D. Thesis, Universidad Complutense, Facultad de Ciencias Fisicas, Madrid 1982 F. Gonzalez and J.L. de Segovia, Phys. Scr., T4 (1983) 132 M. Nishijima and F.M. Propst, Phys. Rev., B 2 (1970) 2368 D. Lichtman, F. N. Simon and T.R. Kirst, Surf. Sci., 9 (1968) D. Menzel and R. Gomer, J. Chem Phys., 41 (1964) 3320 P.A. Redhead, Appl. Phys. Lett., 4 (1964) 166 D. Menzel and R. Gomer, J. Chem. Phys., 41 (1964) 3320 D. A. Degras and J. Lecante, Nuovo Cimento, Suppl. V, (1967) 598 D. Lichtman, Y. Shapira, in: Chemistry and Physics of Solid Surfaces, R. Vanselov (Ed.), CRC, Boca Raton, 1978, p. 397 D. Menzel, Ver. Bunsenges Phys. Chem., 73 (1968) 591 R.M. Lambert and C M . Comrie, Surf. Sci., 59 (1976) 33 M.L. Shek, S.P. Withrow and W.H. Weinberg, Surf. Sci., 72 (1978) 678 T.E. Madey, Surf. Sci., 79 (1979) 575 M. Nishijima and T. Murotani, Surf. Sci., 32 (1972) 459 S. Priegge, H. Niehus, E. Bauer, Surf. Sci., 75 (1978) 635 R.J. Musket, Surf. Sci., 74 (1978) 423 S.W. Bellard and E.M. Williams, Surf. Sci., 80 (1979) 450 F.P. Netzer and T.E. Madey, Surf. Sci., 27 (1983) L102

641 [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95]

[96]

[97] [98] [99] [100]

[101] [102] [103]

[104]

M.Q. Ding, E.M. Williams, J.P. Adrados and J. L. de Segovia, Surf. Sci., 140 (1984) L264 V.N. Ageev, Yu.A. Kuznetsov, and B.V. Yakshinskii, Fiz. Tverd. Tela, (Leningrad) 24 (1982) 349 [Sov. Phys. Solid State, 24 (1982) 199 V.N. Ageev, O.P. Burmistrova and Yu.A. Kuznetsov, Fiz. Tverd. Tela, (Leningrad) 29 (1987) 1740 [Sov. Phys. Solid State, 29 (1987) 1000 V.N. Ageev, O.P. Burmistrova, and Yu.A. Kuznetsov, Fiz. Tverd. Tela, (Leningrad) 31(a) (1989) 91 [Sov. Phys. Solid State 31 (1989) 1526 V.N. Ageev, Yu.A. Kuznetsov, Pisma Zh.Tekh. Fiz., 16(5) (1990) 38 [Sov. Tech. Phys. Lett., 16 (1990) 179 M.L. Knotek and P.J. Feibelman, Phys. Rev. Lett., 40 (1978) 964 P.J. Feibelman and M.L. Knotek, Phys. Rev., B 18 (1978) 6531 M.L. Knotek, V.O. Jones and V. Rehn, Phys. Rev. Lett., 43 (1979) 300 P. Feulner, R. Treichler and D. Menzel, Phys. Rev. B 24 (1981 7427 D. Menzel, J. Vac. Sci. TechnoL, 20(3) (1982) 538 I. W. Owen, N. H. Brookes, C.H. Richardson, D.R. Warburton, F.M. Quinn, D. Norman and G. Thornton, Surf. Sci. 178 (1986) 897 M.L. Knotek, R.H. Stulen, G. M. Loubriel, V. Rehn, R.A. Rosemberg and C.C. Parks, Surf. Sci., 133 (1983) 291 M.Q. Ding, E.M. Williams, J.P. Adrados and J. L. de Segovia, Surf. Sci., 140 (1984) L264 M.Q. Ding and E.M. Williams, Surf. Sci., 160 (1985) 189 S. Rey and J.L. de Segovia, J. Vac. Sci. Technol.,A5 (1987) 562 M.L. Knotek and P.J. Feibelman, Surf. Sci., 90 (1979) 78 M. Menges,B. Baumeister, K. Al-Shamery, B. Adaam, Th. Mull, H.-J.Freund, C. Fisher, D. Weide and P. Andersen, in: Desorption Induced by Electronic Transitions, DIET V, A.R. Burns, E.B. Stechel and D.R. Jennison, Springer Series in Surface Sciences, Vol. 31, Springer, Berlin, 1993, p. 275 P. J. Feibelman, in: Desorption Induced by Electronic Transitions, DIET I, N.H. Talk, M.M. Traum, J.C. Tully and T.E. Madey, Springer Series in Chem. Phys., Vol. 24, Springer, Berlin, 1983, p. 62 D.E. Ramaker, J. Chem. Phys., 78 (1983) 2998 T.C. Shen and P. Avouris, Surf, Sci., 390 (1997) 35 I.D. Cooks, Q. Guo and E.M. WilUams, Surf. Sci., 390 (1997) 119 M. Menges, B. Baumeister, K. Al-Shamery,B. Adam, Th. Mull, H.-J Freund, C. Fischer, D. Weider, and P. Andresen, in: Desorption Induced by Electronic Transitions, DIET V, A.R. Burns, E.B. Stechel and D.R. Jennison (Eds.), Springer Series in Surface Sciences, Vol. 31, Springer, Berlin, 1993, p. 275 B.E. Koel, G.M. Loubriel and M.L. Loubriel, M.L. Knotek, R.H. Stulen,R.A. Rosenberg and C.C. Parks, Phys. Rev., B 25 (1982) 5551 R. Jaeger and J. Stoehr, Phys. Rev., B 28 (1983) 1145; R. Jaeger, J. Stoehr and T. Kendelewicz, Surf. Sci., 134 (1983) 547 J. Schmidt-May, F. Senf, J. Voss, C. Kunz, A. Flodstroem, R. Nyholm and R. Stockbauer, in: Desorption Induced by Electronic Transitions, DIET II, W. Brenig and D. Menzel, Springer Series in Surface Science, Vol. 4, Springer, Berlin, 1984, p. 94; Surf. Sci., 163 (1985) 303 F.L. Hutson, D.E. Ramaker, V.M. Bermudez and M.A. Hoffbauer, J. Vac. Sci. TechnoL, A3(3) (1985) 1657

642 [105] [106] [107] [108] [109] [110] [111]

[112] [113] [114] [115] [116] [117] [118] [119] [120]

[121] [122] [123] [124] [125]

[126]

[127] [128] [129] [130] [131] [132] [133] [134]

B.L. Maschhoff, J-M. Pan and T.E. Madey, Surf. Sci., 259 (1991) 190 M.C. Torquemada, J. L. de Segovia and E. Roman, Surf. Sci., 337 (1995) 31 J.L. de Segovia, Vacuum, 47 (1996) 333 Y. E. Henrich and P. A. Cox in: The Surface Science of Metal Oxides, Cambridge University Press, 1994, p. 369. E. Bertel, R. Stockbauer and T.E. Madey, Surf. Sci., 141 (1984) 355 E. Bertel, R. Stockbauer, R. L. Kurtz, D.E. Ramaker and T. E. Madey, Phys. Rev. B 31(1985)5580 E. Bertel, R. Stockbauer, R.L. Kurtz, T.E. Madey and R.E. Ramaker, in: Desorption Induced by Elnectronic Transitions, DIET II, Springer Series isn Surface Science, Vol. 4, Eds. W. Brenig and D. Menzel, Springer, Berlin, 1985, p. 84 V.N Ageev, Yu.A. Kuznetsov and N.D. Potekhina, Phys. Solid State, 36(2) (1994) 790; Surf. Sci., 367(1996)113 V.N Ageev, Yu.A. Kuznetsov and N.D. Potekhina, Phys. Solid State, 38(2) (1996) 335 V.N Ageev and Yu.A. Kuznetsov, Phys. Solid State, 39(4) (1997) 671 V.N Ageev, Yu.A. Kuznetsov and T.E. Madey, Surf. Sci., 390 (1997) 146 V.N Ageev, Yu.A. Kuznetsov and T.E. Madey, Phys. Rev., B 38 (1998) 2248 M. Akbulut, N.J. Sack and T.E. Madey, Surf. Sci., 351 (1996) 209 R.L. Kurtz, R. Stockbauer, T.E. Madey, E. Roman and J.L. de Segovia, Surf. Sci., 218(1989) 178 K.D. Schierbaum, S. Fischer, M. C. Torquemada, J. L. de Segovia, E. Roman and J.A. Martin-Gago, Surf. Sci. 345 (1996) 261 R. L. Kurtz, R. Stockbauer and T.E. Madey, in: Desorption Induced by Electronic Transitions, DIET II, Springer Series in Surface Science,vol. 4, W. Brenig and D. Menzel, Springer, Berlin, 1985, p. 85. R.L. Kurtz, R. Stockbauer and T.E. Madey, Nuclear Instr. and Methods in Phys. Res., B13 (1986) 518; J. Vac. Sci. Technol., A4(3) (1986) 1248 R.L. Kurtz, Surf. Sci., 177 (1986) 526 W.L. Clinton, Surf. Sci., 112 (1981) L791 Z. Miskovic, J. Vukanic and T.E. Madey, Surf. Sci., 169 (1986) 405 C.Z. Dong, P.Nordlander, and T.E. Madey, in: Desorption Induced by Electronic Transitions, DIET IV, G. Betz and P. Varga (Eds.), Springer Series in Surface Sciences, Vol. 19, Springer, Berlin, 1990, p. 34 R.E. Walkup and R. L. Kurtz, in: Desorption Induced by Electronic Transitions, DIET III, Springer Series on Surface Science, Vol. 13, R. H. Stulen and M.L. Knotek(Eds.), Springer, Berlin, 1988, p. 160 C.L. Greenwood, E. M. Williams, G. Thornton, S. L. Bennet, E. Roman, J. L. de Segovia and M. C. Torquemada, Surf. Sci., 287/288 (1993) 386 M.C. Torquemada and J. L. de Segovia, J. Vac. Sci. Technol., A12(4) (1994) 2318 Q. Guo, I. Cocks and E.M. Williams, Surf. Sci., 366 (1996) 99 P.J. Moller and M.C. Wu, Surf. Sci., 224 (1989) 265 H. Onishi and Y. Iwasawa, Surf. Sci., 313 (1994) L783 Q. Guo, I. Cocks and E.M. WiUiams, Phys. Rev. Lett., 77 (1996) 3851 I.D. Cocks, Q. Guo and E.M. Williams, Surf. Sci., 390 (1997) 119 R. Patel, Q. Guo, I. Cocks, E.M. Williams, E. Roman and J.L. de Segovia, J. Vac. Sci. Technol., A15(5) (1997) 2553

643 [135]

G. Loubriel, M.L. Knotek, R.H. Stulen, B.E. Koel and C.C. Parks, J. Vac. Sci. Techno!., Al(2) (1983) 1145. [136] J. Schmidt-May, F. Sent, J. Voss, C. Kunz, A. Flodstrom, R. Nyholm and R. Stockbauer. Surf. Sci., 163 (1985) 303 [137] R.A. Rosengerg and C.-R. Wen, Phys. Rev., B37 (1988) 9852 [138] A. Hoffman, S. D. Moss, J.K. Paterson and M. Petravic, J. Appl. Phys., 78 (1995) 6858 [139] R.A. Baragiola, T.E. Madey and Ann-Marie Lanzillotto, Phys. Rev., B 41 (1990) 9541 [140] M. Petravic, Phys. Rev., B 48 (1993) 2627 [141] T. Gotoh, S. Takagi and G. Tominaga, Vacuum, 41 (1990) 213; in: Desorption Induced by Electronic Transitions, DIET IV. Springer Series in Surface Science, vol. 19. Eds. G. Betz and P. Varga, Springer-Verlag, Berlin, 1990, p. 327. [142] I. Colera, E., J.L. de Segovia, E.L. Roman and R. Gonzalez, Vacuum, 48 (1997) 647 [143] T.E. Madey, B.V. Yakshinskiy, V.N. Ageev and R.E. Johnson, J. Geophys. Res. 103 (1998)5873 [144] M.L. Knotek, Surf. Sci., 101 (1980) 334; Surf. Sci., 91 (1980) L17 [145] S. Suzuki, Ken-ichi Fukui, H. Onishi and Y. Iwasawa, Phys. Rev. Lett. 84 (2000) 2156 [146] U. Diebold and T.E. Madey, J. Vac. Sci. Techno!., A 10(4) (1992) 2327; U. Diebold and T.E. Madey, in: Desorption Induced by Electronic Transitions, DIET V, A.R. Burns, E.B. Stechel and D.R. Jennison (Eds.), Springer Series in Surface Science, Vol.31, Springer, Berlin, 1993, p. 284 [147] M.C. Torquemada, J.L. de Segovia, E. Roman, G. Thornton, E.M Williams and S.L. Bennett, in: Desorption Induced by Electronic Transitions, DIET V, A.R. Burns, E.B. Stechel and D. R. Jennison, Springer Series in Surface Science, Vol. 31, Springer, Berlin, 1993, p. 289 [148] C.L. Greenwood, E.M. Williams, G. Thornton, E. Roman, J.L. de Segovia and M.C. Torquemada, in: Desorption Induced by Electronic Transitions, DIET V, A.R. Burns, E.B. Stechel and D.R. Jennison (Eds.), Springer Series in Surface Science, Vol. 31, Springer, Berlin, 1993, p.289 [149] J.L. de Segovia, M.C. Torquemada and E. Roman, J. Phys. Condens. Matter., 5 (1993) A139 [150] M.C. Torquemada and J.L. de Segovia, J. Vac. Sci. Techno!., A12 (1994) 2318 [151] C.A. Muryn, P.J. Hardman, J. J. Grouch, G. N. Raiker and G. Thornton, Surf. Sci., 251/252(1991)747 [152] M.C. Torquemada and J.L. de Segovia, Surf. Sci., 331-333 (1995) 219 [153] M.C. Torquemada, J.L. de Segovia and E. Roman, Surf.Sci., 337 (1995) 31 [154] V.M. Bermudez and M. A. Hoffbauer, Phys. Rev., B 30 (1984) 1125. [155] D. Menzel, in: Interactions on Metals Surfaces, Ed. R. Gomer, Springer-Verlag, New York, 1975, p. 101 [156] I. Colera, E. Soria, J.L. de Segovia, E.L. Roman and R. Gonzalez, J. Vac. Sci. Techno!., A 15 (1997) 1698 [157] E. Soria, J.L. de Segovia, I. Colera and R. Gonzalez, Surf. Sci., 390 (1997) 140 [158] K.D. Schierbaum, S. Fischer, M. C. Torquemada, J.L. de Segovia, E. Roman and J.A. Martin-Gago, Surf. Sci., 345 (1996) 261 [159] Q. Guo,I. Cocks and E.M. Williams, J. Chem. Phys., 106 (1997) [160] Q. Guo, I. Cocks and E.M Williams, Surf. Sci., 393 (1997) 1

644 [161] [162] [163] [164]

[165] [166] [167] [168] [169]

M.A. Henderson, S. Otero-Tapia and M.E. Castro, Surf. Sci., 4127413 (1998) 252 E. Soria, E Roman, E.M. Williams and J.L. de Segovia, Surf. Sci., 433-435 (1999) 543 T.E. Madey, J. T. Yates Jr., D.A. King, C.J. Uhlaner, J. Chem. Phys., 52 (1970) 5215 R. Gomer, in: Desorption Induced by Electronic Transitions, DIET I, . N.H. Tolk, M. M. Traum, J.C. Tully and T.E. Madey (Eds.), Springer Series in Chemical Physics, Vol. 24, Springer, Berlin, 1983, p. 40 S. Priegge, H. Niehus and E, Bauer, Surf. Sci., 75 (1978)635 C. Leung, Ch. Steinbruchel and R. Gomer, Appl. Phys., 14 (1977) 79 M.L. Knotek, Surf. Sci., 334-340 (1980) 334 B. Yakshinskiy, M. Akbulut and T.E. Madey, Surf. Sci., 390 (1997) 132 R. L. Kurtz, R. Stockbauer, R. Nyholm, S.A. Flodstrom and F. Senf, phys. Rev., B35 (1987) 7794

645

Index

a-AbOs, 292. See AI2O3 y-like AbOsCOOOiydi-tert-butyl nitroxide, 234

ab initio Hartree-Fock methods, 38 ab initio model potential, 98 absorption band, 523 acetaldehyde, 428, 436 acetate, 223, 422 aceticacid, 223, 488, 631 acetylene, 423 acid rain, 31 acid-base reaction, 22, 65, 137 acoustic mode, 520 acrylic acid, 423 activation energy, 420 adsorbate structures on oxide surfaces, 199 adsorption, 25, 257, 326 adsorption energy, 334 AFM, 418, 443, 451, 502, 587 Al(l 11)70, 50 AI2O3, 15,49,292, 550 AI2O3 surface energies, 492 AI2O3 thin films, 393 Al2O3(0001), 49, 60, 79, 80, 490, 499 Al2O3(0001), reconstruction, 43, 266 Al2O3(0001), structure, 266, 342 Al2O3(0001)/Cu, 233 Al2O3(0001)/di-tert-butyl nitroxide, 234 Al2O3(0001)/H2O,351 Al2O3(100), 292 Al2O3(110),292 Al2O3(lll)/NiAl(110),353 Al203(lll)/N02,352 AI2O3, surface structures, 233 AI2O3/ di-tert-butyl nitroxide, 356 Al203/doxyl-stearic acid, 358 Al203/NiAl, 292 AI2O3/OH, 539 Al203/Pd, 540 Al2O3/Re(0001), 393 Al203/Rh, 535 aldol, 428

aldoHsation, 428 aliphatic alcohols, 432 alkali chemisorption, 26 alkyne decomposition, 488 alumina, 15 anatase, 435, 444 anatase (001), 478 anatase (101), 476 anatase surfaces, 476 Anderson localisation, 594 anharmonicity, 516 anion vacancies, 110 anisotropic surface properties, 485 antiferromagnetism, 47, 281, 568, 571 anti-fluorite M2O structures, 45 antiphase boundaries, 316 atmospheric aerosol particles, 30 atomic relaxation, 70 Au(110)/CO, 390 Au(lll)/N02,352 augmented spherical wave method (ASW), 143 autocompensation, 10, 79, 268, 312, 447 B band gap, 550 band structure, 552 band theory, 5 BaO, 45 BaTiOs, 24 BaTiO3(110),79 BaTi03(lll),56,79 benzoate, 224, 423 benzoic acid, 633 bipolarons, 588 Blyholder model, 389 bond orbital model, 63 bond transfer model, 63, 80, 82 Bragg scattering, 259 Brewster angle, 522 Bronsted acid sites, 168 brucite, 110 bulk defects, 445 bulk termination, 448

646

CaO,45, 108 CaO/C02, 109 CaO/Na, 126 CaO/NiO(100), 396 CaO/NiO(100), NO adsorption, 397 carbonate, 227 carboxylate, 223, 631 carboxylic acids, 422 catalysts, oxide-supported, 23 catalytic oxidation, 165 catalytic reactions, 137, 420 cation vacancies, 121 CdO, 550, 553, 594 Ce02,45, 616 CeO2(lll),60 C-H bond activation, 189 C-H bond breaking, 127 charge compensation, 80 charge distribution, 70, 111 charge neutrality, 3, 10 charge transfer, 448 charge transfer energy, 4 charge transfer excitation, 330 charge transfer insulators, 552 charged defects, 96 charging, 23, 559 chemical activity, 373 chemisorption, 22 CI, 413 classical methods, 36 classical potentials, 94 cleavage, 12 cluster calculations, 44, 96, 138 cluster diameter, influence on activity, 388 CO adsorption, 27, 326, 332 CO binding energies at MgO(lOO), 106 CO oxidation activity, 388 C-0 stretching frequency, 105 Co304(lll),283 combination band, 528 conduction band minimum, 452 configuration interaction, 39, 97, 616 constant initial state spectra, 555 CoO, 45, 553, 555, 566 CoO(100),46,567 CoO(lOO), charge density, 569 CoO(l 11), preparation, 282

CoO(lll), structure, 281 CoO-MgO soHd solutions, 124 core level photoemission, 580, 586 core level shifts, surface, 16 corundum, 9, 11,49,266 corundum structure oxides, 341 corundum(OOOl), 341 corundum(10-12), 14 Coulomb interaction, screened, 42 Coulomb repulsion energy, 3 Coulomb self-interaction, 41 CO valency, 552 covalent bonding, 20 Cr(l 10) oxidation, 293 Cr203, 49, 293 Cr203 surface structures, 235 Cr2O3(0001),67,343,531 Cr2O3(0001), structure, 288, 342 Cr2O3(0001)/C2H4, 352 Cr2O3(0001)/CO, 347 Cr2O3(0001)/CO2, 235, 345, 351 Cr2O3(0001)/Cr(110),344 Cr2O3(0001)/H2O, 350 Cr2O3(0001)/NO, 236 Cr2O3(0001)/O/CO2, 351 Cr2O3(0001)/O2, 346 Cr2O3/Al2O3(0001), 309 Cr203/Fe203 superlattice, 314 Cr2O3/Pt(lll),310 Cr304, 309 cristobalite, 57 Cr02, 360 CrO2/TiO2(100), 309 CrO2/TiO2(110), 309 Cr03, 309 crotonaldehyde, 428 CRYSTAL computer program, 99 crystal shape, 11 crystal symmetry mismatch, 316 crystal truncation rods, 260 crystallographic shear planes, 138, 168, 461,504 CS2O, 45 CTR. See crystal truncation rods CU2O, 490 CuO, 530 cyclotrimerisation, 423

647 D d—> d excitations, 35, 46, 68 dangling bonds, 312 decarboxlyation, 423 defect concentrations, 96 defects at MgO surfaces, 94 dehydration, 412, 421 dehydrogenation, 412, 421, 488 dehydroxylation processes, 110 density functional theory. See DFT density of states. See DOS desorption induced by electronic transitions. See DIET DFT, 39, 94, 95, 451 DIET, 608, 613, 638 dioxymethylene, 431 dipole coupling, 527 dipole moment, 75 dipole moment, total, 44 dipole scattering, 564 dipole selection rule, 521 direct methods in surface crystallography, 262 dissociative adsorption, 25 di-tert-butyl nitroxide, 234, 355 divacancies, 122 d-orbitals, 5 DOS, 67, 82 DOS, V2O3 bulk, 147 DOS, V2O5 bulk, 144 DOS, VO2 monoclinic and rutile, 146 doxyl-stearic acid, 358 dynamic dipole coupling, 527 dynamic mean field theory (DMFT), 43

edge dislocation, 24 EELS, 50, 55, 68, 118, 330, 344, 381, 556 electron acceptors, 26 electron correlation, 3, 7, 36, 100 electron counting rules, 79 electron distribution, 62 electron donors, 26 electron energy loss spectroscopy. See EELS electron paramagnetic resonance. See EPR electron screening, 8 electron stimulated desorption. See ESD

electron stimulated desorption ion angular distribution. See ESDIAD electron transfer, 63 electron-density difference function, 261 electronegativity, 38, 65 electronic density of states, layer by layer, 16 electronic properties, 2 electronic structure, 16, 326, 373, 550 electrostatic field, 77 electrostatic interactions, 552 electrostatic potentials, 38 EMBED computer program, 99 embedded clusters, 44, 98 epitaxial films, 21, 200, 289, 373, 478, 518 epitaxial films, crystal symmetry mismatch, 316 epitaxial films, growth, 302 epitaxial films, lattice mismatch, 313 epitaxial films, residual strain, 313 epitaxial films, structure, 301 EPR,48, 106, 111, 113, 120, 126 equilibrium shape, 491 ESD, 458, 608, 613, 623, 638 ESDIAD, 223, 224, 422, 449, 465, 583, 612, 620, 638 ESR, 97, 234, 353, 356, 358 ethanol, 432 etherification, 415 ethylene epoxidation, 392 exchange energy, 100 exchange interactions, 38 excitons, 558, 562

Fcentres, 98, 381,563 facetting, 48, 103, 276, 414, 426, 471, 541 Fe203, 49, 293, 497 Fe2O3(0001), 15,79,341,499 Fe2O3(0001), structure, 305, 342 Fe2O3(0001), surface terminations, 50 Fe2O3(0001)/H2O, 350 Fe203 (0112), 49 Fe203 (1120), 49 Fe203/Al203, 303 Fe203/Cr203 superlattice, 314 Fe2O3/Pt(lll),303

648 Fe203/S02,31 Fe304, 57, 76, 293, 497 Fe3O4(100), 79 Fe304(l 11), 15,499 Fe304(lll)/H5C6CHCH2, 237 Fe304/Al203, 303 Fe3O4/MgO(001),317 Fe3O4/Pt(lll),303 Feibelman-Knotek model, 614, 619 FeO, 45, 46, 497 FeO(l 11)/H5C6CHCH2, 237 FeO(lll)/Pt(lll), 304 ferroelectric field effect transistors, 1, 24 ferromagnetic planes, 281 field-effect transistors, 24 fluorine desorption, 609 fluorite MO2 structures, 45 formaldehyde, 415, 487 formate, 221, 414, 538 formic acid, 120, 412 FP-LAPW, 143, 147 free-electron band theory, 5 F5 centres, 70, 102, 112 Fs centres, stabiHty and formation energies atMgO, 115 Fuchs-Khewer modes, 515, 567 full-potential linear augmented plane wave method. See FP-LAPW

generalised gradient approximation. See GGA GGA, 40, 578 GIXS, 257, 328, 473, 585 glycine, 424, 634 gold clusters, 399 grain boundary diffusion, NiO, 117 grazing incidence X-rays, 258 Green's function, 42 growth of oxide films, 302 GW method, 41 H H2O dissociation, 257 Hartree-Fock, 36, 95, 451, 564, 581 HAS, 207 HCOOH, 412 He atom diffraction, 559, 571

helium atom scattering. See HAS hematite, 15 hematite/S02, 31 hexagold phosphine-stabilised cluster, 399 high pressure STM studies, 399 high-resolution electron energy loss spectroscopy. See HREELS Hohenberg-Kohn theorem, 100 HOMO, 39, 65, 177 homogeneous electron gas (HEG), 39 HREELS, 514, 516, 556, 559, 598. See also EELS Hubbard U, 3, 42, 552 hybridisation, 20 hydrogen abstraction, 126, 138 hydrogen adsorption energies, 163 hydrophilicity, 506 hydroxyl, 109,416,634 hydroxylation processes, 110 hyperfine interactions, 113 I lABS, 514 impact scattering, 519 impurity atoms, 124 inelastic atom beam spectroscopy. See lABS infra-red reflection-absorption spectroscopy. See IRAS infra-red spectroscopy. See IR spectroscopy intercalation, 594 interfacial structure, 392 inverse photoemission, 7, 554, 556, 574, 581 ion scattering spectroscopy. See ISS ionic charge, 62 ionicity, 2 iono-covalent materials, 78 IR spectroscopy, 104, 344 Ir(lll)/C0,611 IRAS, 104, 346, 349, 351, 389. See also RAIRS iron oxides, surface structures, 236 isopropanol, 435 ISS, 209, 395, 396, 572

649

Janak's theorem, 65 K ketene, 423, 634 ketonisation, 415 kinematical approximation, 259 kinks, 70, 105 Kohn-Sham equations, 40, 100 Koopman's theorem, 65

LaCrOs, 25 LaFeOs, 25 lanthanide rare earths, 6 lattice mismatch, 313 Laue condition, 260 layered structure oxides, 363, 555 LCAO, 143, 561 LDA, 39, 115, 563, 578 LDOS, 67 LDOS, surface, 84 LEED, 200, 203, 208, 213, 268, 328, 343, 361,410,449,559 Lewis acids, 26 Lio.9Mo60i7, 555 Li20, 45 Li02, 558 local charge neutrality, 3 local density approximation. See LDA local density of states. See LDOS local spin density approximation. See LSDA Lorentz oscillator model, 525 low energy electron diffraction. See LEED LSDA, 40 LUMO, 39, 65 LYP correlation functional, 100 M Madelung constant, 105, 560 Madelung field, 97, 115 Madelung potential, 52, 61, 64, 104, 161, 245,551,560 Magneh phases, 6, 139, 497, 499 magnetic properties, 317 magnetite, 76, 317

maleic anhydride, 423 Mayer bond orders, 156 MBE, 302 McMurray reaction, 415 metal catalysts, oxide-supported, 23 metal vacancies, 563, 570 metal-insulator transition, 7 metallic termination, 84 metastable He atom diffraction, 571 metastable impact electron spectroscopy. See MIES methane oxidation, 487 methanol, 424 methanol oxidation, 425 Mezel-Gomer-Redhead model. See MGR model Mg(0H)2, 110 MgO, 10, 12, 45, 374, 412, 517, 550, 553 MgO smoke, 104 MgO surface defects, optical spectra, 118 MgO surface defects, summary, 101 MgO surface energies, 493 MgO surface point defects, theory, 94 MgO surface, computed F centre properties, 114 MgO(lOO), 13, 24, 46, 60, 67, 70, 71, 101, 413,436,492,530,558 MgO(lOO), band structure, 73 MgO(lOO), electronic structure, 375 MgO(100),F5 centres, 112 MgO(100),HREELS, 561 MgO(lOO), preparation, 264 MgO(lOO), structure, 263, 328 MgO(lOO), surface defects, 381, 562 MgO(lOO), surface vacancies, 103 MgO(100),V centres, 122 MgO(lOO), vacancy structure, 47 MgO(100)/Ag, structure, 210 MgO(100)/C2H2, structure, 206 MgO(100)/C6H6, 374 MgO(100)/Ca, structure, 209 MgO(100)/CH3OH, 378, 384, 518 MgO(100)/CO, 104, 333, 336 MgO(100)/CO, charge density, 125 MgO(100)/CO, structure, 207 MgO(100)/CO2, structure, 208 MgO(100)/D2O, 378, 629 MgO(100)/defect energy levels, 116

650 MgO(100)/Fe, structure, 213 MgO(100)/H2O, 338 MgO(100)/H2O, structure, 202 MgO(100)/Mo(100), 374 MgO(100)/N2O, 383 MgO(100)/Ni, structure, 212 MgO(100)/Ni/CO, 125 MgO(100)/NO, 336, 383 MgO(100)/O, 624 MgO(100)/OH,205 MgO(100)/Pd, structure, 212 MgO(lll), 81, 128, 276, 329, 413, 492 MgO, energy levels, 551 MgO, impurity atoms, 124 MgO, S doping, 126 MgO, Se doping, 126 MgO, surface divacancies, 122 MgO, surface structures, 202 MgO/CDsCN, 106 MgO/CH4, 120 MgO/C02, 108 MgO/H2, 124 MgO/HCOOH, 120 MgO/Li, 126 MgO/NiO layers, 69 MgO/NiO(100), 395 MgO/NiO(100), NO adsorption, 397 MgO/OH, 109 MgO/S02, 107 MGR model, 610 micro facets, 128, 471 MIES, 374, 379, 382, 560, 571 mixed valency, 137 mixed valency vanadium oxides, 140 mixed-metal oxides, 392 Mn304(lll),283 MnO, 45 MnO(lOO), 565 MnO(l 11), preparation, 282 MnO(lll), structure, 281 Mo(100)/MgO(100), 561 molecular adsorption, 25 molecular beam epitaxy, 302 molecular dynamics simulation, 340 molecular orbitals, 96 molybdenum oxides, 136, 168 molybdenum oxides, adsorption, 186

molybdenum oxides, bulk electronic structure, 172 molybdenum oxides, bulk structure, 170 molybdenum oxides, surface oxygen vacancies, 182 molybdenum oxides, surface structure, 172 molydenum oxide, gas-phase clusters, 169 momentum conservation, 258, 516, 555 M0O2, 168,171,360 MoO2(011), 180 MoO2(011), electronic structure, 181 MoO2(011), structure, 174 M0O3, 57, 137, 168, 170,486 MoOsCOlO), 58, 502 MoOsCOlO), electronic structure, 175 MoOaCOlO), structure, 173 MoOsCOlO), surface oxygen vacancies, 183 MoO3(010), surface reduction, 180 MoOsCOlOyCHsOH, 501 MoO3(010)/CH4, 189 MoO3(010)/CO, 189 MoO3(010)/H, 186 MoOsCOlOyOH, 186 MoOsClOO), 58 MoOsOOO), structure, 174 MoO3(100)/CH3, 189 M0O3/H2O, 504 Mo03/methanol, 188 morphology of surfaces, 490 Mott transition, 146 Mott-Hubbard insulator, 552 Mott-Hubbard metal, 147 Mulliken charges, 72, 153 Mulliken electronegativity, 65 multi-phonon structure, 519 N N2O4 dimer, 354 Nao.9Mo60i7, 555 Na2/3WO3(001), 597 Na20, 45 nanofacets, 586 nanosized Au clusters, 402 NaxW03, 550 NaxWO3(001), 593 near-edge X-ray absorption fine structure. See NEXAFS

651 Neel temperature, 282, 568 neutron diffraction, 206 NEXAFS, 200, 208, 214, 216, 218, 223, 224, 225, 227, 231, 234, 237, 238, 240, 246, 349 Ni/CO, 332 Ni304, 284 Ni3O4(lll),280 NiAl(lOO) oxidation, 292 NiAl(llO) oxidation, 292 NiAl(110)/Al2O3/Pd/CO, 540 NiAl(110)/Al2O3/Rh,536 NiO, 45, 550, 553, 555 NiO(lOO), 46, 60, 67, 565 NiO(lOO), charge density, 569 NiO(lOO), electronic structure, 330 NiO(lOO), structure, 328 NiO(100)/CO, 334 NiO(100)/CO, structure, 216 NiO(100)/H2O, 339 NiO(100)/H2S, structure, 217 NiO(100)/MgO(100), 546 NiO(100)/NO,331,334 NiO(100)/NO, structure, 214 NiO(110),66 NiO(lll), 80, 128 NiO(l 11), p(2x2) reconstructions, 276 NiO(l 11), p(2x2) thin film structure, 289 NiO(l 11), preparation, 277 NiO(lll), structure, 329 NiO(l 11), surface reduction, 279 NiO(lll)/Au(lll),289,329 NiO(lll)/CO, structure, 219 NiO(lll)/Ni(lll),289,329 NiO(lll)/NO, structure, 218 NiO(lll)/OH, 128 NiO, surface structure, 213 NiO/CaO(100),396 NiO/CO,332,615 NiO/MgO layers, 69 NiO/MgO(100), 395 NiO/NO,615, 616 NiO-MgO solid solutions, 124 NIXSW, 204 non-polar surface structures, 59 non-stoichiometric surfaces, 69 normal incidence X-ray standing waves. ^e^ NIXSW

O octahedral structure, 8 OH surface groups, 109 optical excitation, 557 optical phonons, 515 optical phonons, surface, 517 optical spectra, 118 organic reactions, 409 overtones, 519 oxidation chemistry, 303 oxide film superlattices, 314 oxide films, 301 oxide superconductors, 2 oxide supported catalysts, 23 oxide surface states, 553 oxide surface/oxygen thermodynamic equilibrium, 74 oxide surfaces, adsorption geometries, 199 oxide surfaces, history, 1 oxide-gas interfaces, 21 oxide-liquid interfaces, 21 oxide-oxide interfaces, 23 oxide-solid interfaces, 21 oxide-water interfaces, 22 oxygen deficiency, 580, 590 oxygen diffusion, 161 oxygen pressure - thermodynamic equilibrium, 74 oxygen vacancies, 17, 98, 457, 564, 579 oxygen vacancies, spectroscopic signature, 72 oxygen-deficient surfaces, 36 oxygen-plasma-assisted growth, 304

paramagnetic defects, 122 paramagnetic F"^ centre, 112 paramagnetic O anions, 126 paramagnetic phase, 7 particulate catalysts, 486 Patterson function, 261, 278, 473 Pd particles, 541 Pd/CO, 540 Penning de-excitation, 560 perovskite, 489 perovskite structure, 54 perovskite surfaces, 76 PhD, 214

652 phonon overtones, 519 phonons, 515 photocatalysis, 30, 437 photoelectron diffraction, 472, 587 photoelectron diffraction (scanned energy mode). See PhD photoelectron diffraction. X-ray. See XPD photoemission, 7, 554, 574, 578, 582, 585, 597 photoemission satellites, 8 photoemission, core levels, 8, 311 photoemission, resonant, 19 photoemission, valence levels, 8, 365 photon stimulated desorption. See PSD photoreactions, 424, 437 physisorption, 22, 26, 108 PIRS, 207, 208 pits at surfaces, 504 plasmons, 556 point charge arrays, 98 point defects, 20 point defects, MgO surfaces, 94 Poisson distribution, 519 polar surfaces, 11, 36, 75 polar surfaces, electronic structure, 78 polarisation infrared spectroscopy. See PIRS polarity healing, 80 polarons, 581,593 pressure gap, 21 propanoate, 224 propionate, 423 PSD, 612, 613, 623 pyridine, 240

quantum cluster, 99 quartz, 57 quasi-particle spectrum, 35, 41, 67 R RAIRS, 222, 230, 415, 514, 521. See also IRAS Rayleigh mode, 520 reaction order, 420 reactions, organic, 409 redox reactions, 22, 26, 137 reduced Ti02, 19

reflection absorption infrared spectroscopy. See RAIRS reflection high energy electron diffraction. See RHEED relaxation, atomic, 70 ReOs, 550 resonant photoemission, 19, 577, 589, 592 RHEED, 226, 314 rhenium trioxide, 9 rocksalt oxides, structure, 327 rocksalt structures, 12, 45, 59 rocksalt(100),45 rocksalt(lll), 128 rocksalt(l 11) structure, 48 rocksalt(l 11), octopolar reconstruction, 83 Ru(001)/CO,615 Ru(001)/O2,615 Ru02, 360, 550 RuO2(110)/CO,361 RuO2(110)/Ru(0001),361 rutile, 9, 444 rutile structure, 51 rutile structure oxides, 359 rutile structure oxides, structure, 360 rutile(llO), 16

sample preparation, 262 sapphire, 266 sapphire(OOOl), 60 scanning tunnelling microscopy. See STM scanning tunnelling spectroscopy. See STS Schrodinger equation solution, 99 screened Coulomb interaction, 42 screw dislocations, 504 secondary ion mass spectrometry. See SIMS segregation of impurities, 263 self-interaction correction method, 40 self-trapping, 581 semi-empirical methods, 36 SEWS, 514 SEXAFS, 200, 210, 217, 232, 233, 244 Shannon-Prewitt radii, 596 Si(lOO), 527 SIMS, 426, 461, 467 Si02, 57, 97, 436, 530, 558, 624 Si02/Pd/C0, 541

653 SnO(lOl), 575 Sn02, 51,360, 550, 553, 572 SnO2(110),53, 572 SnO2(110), electronic structure, 574 SnO2(110), reconstructions, 573 SnO2(110), structure, 430 SnO2(110)/CH3OH,429 SNS junctions, 24 SO2 oxidation, 31 solid-vapour equilibrium, 497 solution-deposited clusters, 402 spin distribution, 70 spin polarised HREELS, 566 spin valves, 2 Sr2Ti04, 489 SrCu02, 489 SrO, 45, 489 SrTiOs, 54, 489, 555 SrTiOs surface orientations, 495 SrTiOsCOOl), 478 SrTiOsCOOl), structure, 286 SrTiOsClOO), 55, 70, 75, 77, 80, 558 SrTiOsCl 10), 56,79 SrTi03(lll),56,79 SrTi03(lll)/H20,625 stacking faults, 308 step edges, 70 STM, 201, 319, 328, 341, 360, 361, 386, 393, 401, 403, 410, 418, 443, 451, 500, 557, 568, 570, 575, 579, 590, 592, 638 STM at high pressure, 399 STM tip functionalistion, 454 stoichiometric surfaces, 36 stoichiometric surfaces, structures, 59 stoichiometric Ti02, 18 stoichiometry, 496, 550 stoichiometry and oxygen pressure, 498 strain fields, 313 structural anisotropy, 485 structure, 8 structure of clean surfaces, 262 STS, 386, 387, 472, 558, 583, 587, 598 styrene, 237 supercell approach, 96 superconductors, 2, 550 superexchange coupling, 47, 571 superlattice films, 25, 314 superoxide anion, 121

surface acoustic mode, 520 surface charging, 200 surface core level shifts, 16 surface crystallography, 262 surface defects, 381, 455 surface defects, summary for MgO, 101 surface diffraction, 260 surface diffusion, 71, 117 surface electromagnetic wave spectroscopy. See SEWS surface electronic structure, 16 surface energy, 11, 61, 342, 491 surface exciton, 562 surface extended X-ray absorption fine structure. See SEXAFS surfaced centres, 70, 102, 563 surface F centres, reactivity, 120 surface Fuchs-Kliewer modes, 531 surface hydroxyl groups, 109 surface magnetism, 568 surface metallisation, 75 surface morphology, 490 surface optical phonons, 517 surface oxygen vacancies, 157, 457 surface oxygen vacancy energies, 158 surface phonons, 343 surface pitting, 504 surface polarity, criterion, 76 surface preparation, 469 surface reactions, 108 surface reconstruction, 256, 463 surface relaxation, 59, 64, 265 surface rumpling, 265, 559 surface segregation, 263 surface selection rule, 521 surface stability, electrostatic condition, 78 surface states, 553, 576, 585, 598 surface steps, 11, 13, 105 surface stoichiometry, 496, 572 surface structure, 256 surface structure, non-polar stoichiometric surfaces, 59 surface structure, quantitative methods, 199 surface structure, VO, 148 surface vacancies, 415 surface vibrational modes, 343

654 surface X-ray diffraction. See SXRD and GIXS surfaces, non-stoichiometric, 69 SXRD, 200, 209, 210, 212, 450, 578

TDS, 106, 334, 339, 347, 350. See also TPD tetrahedral structure, 8 theory of oxide surfaces, 35 thermal desorption spectroscopy. See TDS thermistors, 7 thin film stoichiometry, 304 thin films, 289, 301, 373, 478, 489, 531 Ti(0001)/O,618 Ti203, 49, 419, 446, 464, 623 TisOs, 446 tight-binding, 36, 575 TiO, 45 Ti-0 phase diagram, 446 Ti02, 51, 359, 385, 409, 550, 577, 614 TiOz bulk defects, 445 Ti02 bulk structures, 444 Ti02 powders, 428 Ti02 surface structure, 443 Ti02 vicinal surfaces, 475 TiO2(001), 52, 410, 474, 478, 488 TiO2(001), faceting, 427 TiO2(001)/C2H5OH,433 TiO2(001)/CH3OH,427 TiO2(001)/H2O, 625 TiO2(100),52,585 TiO2(100) structure, 470 TiO2(100), reconstruction, 275 TiO2(100), structure, 273 TiO2(101),476 Ti02(l 10), 16, 51, 60, 63, 70, 360, 410, 499,578,616,619 Ti02(l 10) structure, theory, 450 TiO2(110),AFM,455 TiO2(110), electronic structure, 582 Ti02(l 10), oxygen vacancies, 457 TiO2(110), reconstruction, 622 Ti02(l 10), reduced, 19 TiO2(110), STM,452 Ti02(l 10), stoichiometric, 18 Ti02(l 10), structure, 219, 273, 446 TiO2(n0), surface defects, 455

Ti02(l 10), surface preparation, 469 TiO2(110), surface reconstructions, 463 Ti02(l 10), surface relaxation, 449 Ti02(l 10), surface steps, 455 TiO2(110)/(NH4)6Mo7O24.4H2O, 228 TiO2(110)/{Rh(CO)2Cl}2, 230 TiO2(110)/Au,385 TiO2(110)/Au/CO,388 TiO2(110)/Au/CO:O2,387 TiO2(110)/Au/O2,400 TiO2(110)/bi-isonicotinic acid, 225 TiO2(110)/C2H5OH,433 TiO2(110)/Ca,460 TiO2(110)/CH3COO,422 TiO2(110)/CH3OH,425 TiO2(110)/CO,628 TiO2(110)/glycine, 424 TiO2(110)/H,458 TiO2(110)/H2O,362 TiO2(110)/H3CCOOH, 223 TiO2(110)/H5C2COOH, 224 TiO2(110)/H5C6COOH, 224, 633 TiO2(110)/HCOO, 363, 417, 632 TiO2(110)/HCOOH,536 TiO2(n0)/HCOOH, structure, 221 TiO2(110)/K,232 TiO2(110)/Mo(CO)6, 529 TiO2(110)/Na,226 TiO2(110)/Na/CO2,227 TiO2(110)/NH3,626 TiO2(110)/O,621 TiO2(110)/O2,466 TiO2(110)/OH,416 TiO2(110)/Pd/CO, 541,545 TiO2(110)/Pt, 631 TiO2(110)/Rh(CO)2, 533 TiO2(110)/Rh(CO)4Cl2, 533 TiO2(110)/Rh2CO4Cl2,528 TiO2(110)/SO2,231,626 TiO2(110)/Ta,628 Ti02, colour, 468 Ti02, interstitial Ti, 462 Ti02, photoreactions, 437 Ti02/Ag, 508 Ti02/Au, CO oxidation, 386 Ti02/CH30H photoreaction, 438 Ti02/Fe, 555 Ti02/Rh(CO)2, 544

655 Tischetiko reaction, 436 TPD, 373, 379, 426, 428 transition metal oxides, 136, 552 transmission band, 523 tridymite, 57 troposphere, 30 tunnelling, 23 U ultra-violet photoelectron spectroscopy. SeeUFS UO2, 45 UO2/CH3OH, 432 UPS, 18, 375,410, 554 UPS, ¥205(010), 144

V centres, 121 V203,49, 139 ¥203(0001), 26, 156 ¥203(0001), structure, 149 ¥2O3(0001)/Au(lll), 343, 365 ¥2O3(0001)/CO, 27, 350 ¥205,57,139,363,488 ¥205(001), 365 ¥205(001)/Na, 364 ¥205(010), 58, 144 ¥205(010), electronic structure, 154 ¥205(010), structure, 151 ¥2O5(010)/H, 162 ¥205(010)/H2 dissociation, 164 ¥2O5(010)/H2O, 162, 168 ¥2O5(010)/NH3, 168 ¥205(010)702, 166 ¥2O5(010)/OH, 162 ¥205, band structure, 144 ¥205, electronic properties, 147 ¥60i3, 500 vacancy diffusion, 71 vacancy diffusion, at MgO surfaces, 117 valency, 5 van der Waals bonding, 12 van der Waals gaps, 501 vanadia, 136 vanadium oxide surfaces, 148 vanadium oxides, 136, 139, 157, 303 vanadium oxides, adsorption, 162

vanadium oxides, bulk electronic properties, 142 vanadium oxides, bulk structure, 140 vanadium oxides, clean surface properties, 152 vibrational spectroscopy, 514, 557 vicinal surfaces, 475 vinyl, 423 VO, 45, 139 VO surface structures, 148 VO, bulk electronic properties, 148 VO2, 51, 139,360 VO2(011), structure, 150 VO2(110)/TiO2(110), 361, 365 VO2, electronic properties, 145 VO2, monoclinic and rutile, 146 W W(100)/CO,615 W(100)/H2O,618 W(100)/O2,615 W(l 11)702,611 W70, 610 weakly polar surfaces, 84 WO2, 360 WO3, 550 WO3(001), 588 WO3(001), STS, 591 Wulff construction, 491 wurtzite, 238 wurtzite structure, 53 X XAFS,312 XPD, 221, 306, 307, 449 XPS, 311, 312, 410, 426, 554, 588 X-ray absorption fine structure. See SEXAFS and NEXAFS X-ray diffraction, 52, 258 X-ray photoelectron diffraction. See XPD. See XPD X-ray photoelectron spectroscopy. See XPS X-ray scattering, 256

YBa2Cu307, 489 YBa2Cu307-x, 15, 263, 623

656

zinc oxide surface structures, 238 Zn(0001)/CO, 246 Zn(0001)/CO2, 246 ZnO, 10, 53, 54, 412, 555 ZnO(OOOl), 21, 54, 82, 284, 413, 488 ZnO(0001),419,488 ZnOCOOOiyCsHsN, 248 ZnO (0001)/C5H5N, 244 ZnOCOOOiyCeHsOH, 248 ZnO(0001)/C6H6, 247 ZnOCOOOiyCHsOH, 247 ZnO (0001)/CO, 242 ZnO (0001)/CO2, 243 ZnO(0001)/Cu/CO, 546 ZnO(0001)/HCOOH, 247 ZnO (0001)/HCOOH, 243 ZnO (0001)/K, 244 ZnO(1010), 284 ZnO (1010), surface structure, 239 ZnO(1010)/C6H6,239 ZnO(1010)/C6N3Hh,241 ZnO(1010)/C7N2H6,241 ZnO(1010)/C7N3H7,241 ZnO(1010)/CO2,238 ZnO (1010)/HCOOH, 238 ZnO/CsHsN, 240 Zr02, 45 ZrO2(100),431 ZrO2(100)/CH3OH,430 ZrO2(110),431