Three common sp-based orthogonal tight-binding models (Goodwin-Skinner-Pettifor, Kwon et al., Sawada) were compared with regard to their suitability for use in atomic-scale semiconductor simulations. It was found that an increased complexity of the functional form did not necessarily lead to improved results. For most of the static and dynamic properties which were studied, none of the methods was clearly superior to the others. However, there was a distinct correlation between the properties which were used in the parameterisation and the advantages of a given model. Thus, the Kwon et al. model was particularly effective in predicting defect properties. A first comprehensive review was presented of tight-binding predictions of the nature of the structure around a vacancy. Both the Goodwin-Skinner-Pettifor and the Kwon et al. tight-binding models predicted 7 vacancy structures having comparable formation energies (except for the tetrahedral form in the case of Goodwin-Skinner-Pettifor, and C2v in the case of Kwon et al.). The Kwon et al. model predicted that the split vacancy would be the most stable vacancy structure, although it also predicted a low-energy tetragonal structure.

A Comparison of Orthogonal sp-Based Tight-Binding Models for Silicon. P.D.Godwin, A.K.Roberts, D.Yu, K.Dean, P.Clancy: Computational Materials Science, 2001, 21[1], 135-48