Abstract
We show that a pair-potential model [Phys. Rev. B 40, 10 351 (1989)] for covalent bonding can be used to describe microcluster configurations in semiconductors. The model regards bonding as due to electrons in covalent bonds. In its original form, there was no bond-bond term (or bond-charge term) to account for interaction between electrons in different bonds, and it took the form of a pair potential truncated after the fourth bond. In this paper, we introduce a bond-bond force in the form of another pair potential acting between bonds. The relative strength of this force distinguishes between the different cluster configurations observed in carbon and silicon. The configurations of various silicon clusters found with this model are related neither to the close-packed metallic microclusters nor to the open diamond structure of bulk silicon. This is in good agreement with electronic-structure calculations but differs from results obtained previously with empirical potentials derived from bulk structures. The additional term does not affect the excellent reproduction of the stability of various bulk lattices found previously, and the potential is therefore transferable between the two regimes.