Charge transfer in multicomponent oxides

Abstract

The transfer of charge between different ions in an oxide plays an essential role in the stability of these compounds. Since small variations in charge can introduce large changes in the total energy, a correct description of this phenomenon is critical. In this work, we show that the ionic charge in oxides can strongly depend on its atomic environment. A model to assign point charges to atoms as a function of their atomic environment has recently been proposed for binary alloys [C. Wolverton, A. Zunger, S. Froyen, and S.-H. Wei, Phys. Rev. B 54, 7843 (1996)] and proven to be very successful in screened solids such as semiconductors and metals. Here, we extend this formalism to multicomponent oxides and we assess its applicability. The simple point-charge model predicts a linear relation between the charge on an atom and the number of unlike neighbors, and between the net value of the charge and the Coulomb field at a given site. The applicability of this approach is tested in a large-supercell self-consistent tight-binding calculation for a random Zr-Ca-O alloy. The observed fluctuations of the ionic charge about the average linear behavior (as a function of the number of unlike neighbors) was larger than 0.25 electrons even when many shells of atomic neighbors were considered in the fit. This variation is significant since it can introduce large errors in the electrostatic energy. On the other hand, for small absolute values of the charge, the ionic charge varied linearly with the Coulomb field, in agreement with previous findings. However, for large Coulomb fields, this function saturates at the formal chemical charge.