The electronic structures of the core and valence ion states of metal oxides

Abstract

Abstract SCF-Xα SW MO calculations on metal core ion hole states and X-ray emission (XES) and X-ray photoelectron (XPS) transition states of the non- transition metal oxidic clusters MgO 6 10− , AlO 4 5− and SiO 4 4− show relative valence orbital energies to be virtually unaffected by the creation of valence orbital or metal core orbital holes. Accordingly, valence orbital energies derived from XPS and XES are directly comparable and may be correlated to generate empirical MO diagrams. In addition, charge relaxation about the metal core hole is small and valence orbital compositions are little changed in the core hole state. On the other hand, for the transition metal oxidic clusters FeO 6 10− , CrO 6 9− and TiO 6 8− relative valence orbital energies are sharply changed by a metal core orbital or crystal field orbital hole, the energy lowering of an orbital increasing with its degree of metal character. Consequently O 2p nonbonding → M 3d-O 2p antibonding (crystal field) energies are reduced, while M 3d bonding → O 2p nonbonding and M 3d-O 2p antibonding → M 4s,p-O 2p antibonding (conduction band) energies increase. Charge relaxation about the core hole is virtually complete in the transition metal oxides and substantial changes are observed in the composition of those valence orbitals with appreciable M 3d character. This change in composition is greater for e g than for t 2g orbitals and increases as the separation of the e g crystal field (CF) orbitals and the O 2p nonbonding orbital set decreases. Based on the hole state MO diagrams the higher energy XPS satellite in TiO 2 (at about 13 eV) is assigned to a valence → conduction band transition. The UV PES satellites at 8.2 eV in Cr 2 O 3 and 9.3 eV in FeO are tentatively assigned to similar transitions to conduction band orbitals, although the closeness in energy of the crystal field and O 2p nonbonding orbitals in the valence orbital hole state prevents a definite assignment on energy criteria alone. However the calculations do clearly show that charge transfer transitions of the e g bonding → e g crystal field orbital type would generally occur at lower energy than is consistent with observed satellite structure. A core electron hole has little effect upon relative orbital energies and is only slightly neutralized by valence electron redistribution for MgO and SiO 2 . For the transition metal oxides a core hole lowers the relative energies of M3d containing orbitals by large amounts, reducing O → M charge transfer and increasing M 3d crystal field → conduction band energies. Large and sometimes overcomplete neutralization of the core hole is observed, increasing from CrO 6 9− to FeO 6 10− to TiO 6 8− . as the O → M charge transfer energy declines. High energy XPS satellites in TiO 2 may be assigned to O 2p nonbonding → conduction band transitions while lower energy UV PES satellites in FeO and Cr 2 O 3 arise from crystal field or O 2p nonbonding → conduction band excitations. Our “shake-up” assignment for FeO 6 10− , CrO 6 9− and TiO 6 8− are less than definitive because no procedure has yet been developed to calculate “shake-up” intensities resulting from transitions of the type described. However the results do allow a critical evaluation of earlier qualitative predictions of core and valence hole effects. First, we find that the comparison of hole or valence state ionic systems with equilibrium distance systems of higher nuclear and/or cation charge (e.g. the comparison of the FeO 6 10− Fe 2p core hole state to Co 3 O 4 ) is dangerous. For example, larger MO distances in the ion states substantially reduce crystal field splittings. Second, core and CF orbital holes sharply reduce O → M charge transfer energies, giving 2e g → 3e g energy separations which are generally too small to match observed satellite energies. Third, highest occupied CF-conduction band energies are only about 4–5 eV in the ground states, but increase to about 7–11 eV in the core and valence hole states of the transition metal oxides studied. The energetic arguments presented thus support the idea of CF and/or O 2p nonbonding → conduction band excitations as assignments for “shake-up” satellites, at least in oxides of metals near the beginning of the transition series.