The second-electron affinity of the oxygen atom is often required in modelling the behaviour of ceramics. It is usually taken to be about 8 eV, independent of the structure of the oxide being modelled. We show that this is not so. Quantum-mechanical calculations show that the true second-electron affinity of an in-crystal O2- ion depends on both the chemical composition and the nuclear geometry. Moreover, for a binary oxide at its equilibrium geometry, this true second-electron affinity is about 40% greater than 8 eV. This discrepancy between the true second-electron affinity and the value of 8 eV derived from semi-empirical potentials determined by fitting to experimental crystal data arises from the structural dependence of the true value. In a semi-empirical treatment, this dependence becomes absorbed into the quantity regarded as the short-range cation-oxide interaction with the consequence that the resulting second-electron affinity is smaller than the true value. A further consequence is that these short-range cation-oxide interactions are dependent on the crystal structure. Thus, for example, a semi-empirical potential determined from a binary oxide with the rock salt structure cannot be transferred to a polymorph with the eightfold-coordinated CsCl structure.
