The polarity in zinc-blende semiconductor compounds is calculated by extending Harrison’s bond-orbital theory to a cluster model and the result is compared to values from previous calculations. We find that polarities from the cluster calculations differ significantly from those predicted by the bond-orbital model and that this difference cannot be accurately calculated by including bond metallization in first-order perturbation theory in the extended bond-orbital model. Cluster polarities are found to be 15%–30% lower than polarities predicted by the bond-orbital model, and comparable to a Brillouin-zone integration of the corresponding tight-binding Hamiltonian. Since bond-antibond interaction corrections to the energy enter in second-order perturbation theory, metallization corrections to the bond energies are found to be comparable whether calculated from a cluster model or using second-order perturbation theory in the extended bond-orbital model.

• Received 26 June 1987

DOI:https://doi.org/10.1103/PhysRevB.36.7433