Photoemission satellites “on” and “off” resonance in transition metals and their oxides

Resonant photoemission has attracted considerable interest since 1977, when an enhancement of the valence band satellite in Ni at photon energies close to the 3p core level binding energy was first observed. Interpretations for the presence of this satellite, and similar satellites found in other transition metals, have included autoionization, M_2,3VV Auger processes, interband transitions, shake-up excitations and various combinations of these. We explain the energies of these photoemission satellites using a simple model, corrected for the final state effect of hole-hole Coulomb interaction and relaxation by using U_eff determined empirically from our Auger results. “Off” resonance the satellites observed are interpreted in terms of a shake-up process and at resonance as merely the M_2,3VV Auger feature occurring at the energy of the shake-up satellite. We show that the recent conclusions of Chandesris et al. [Phys. Rev. B27 (1983) 2630], which state that the correlation energy between two 3d holes is constant for Cr through to Ni, are incorrect and that the second satellite, observed at resonance in Ni is merely the manifestation of a peak observed in the M_2,3VV Auger spectra. Further we conclude that the 3 eV increase in the satellite to main photoemission peak separation in going from Ni to NiO, is unlikely to be due to an increase in U_eff as tentatively assigned by Thuler et al. [Phys. Rev. B27 (1983) 2082]. We suggest that for “off” resonance it is a shake-up process involving empty 4sp states at the bottom of the conduction band, whilst at resonance it is due simply to changes in the energy of the M_2,3VV Auger transition.