Applicability of O(N)-like density functional study on the structural properties of nitrogen defects in SiC

Abstract

We investigate the quantitative applicability of a subspace optimization method formulated in terms of localized nonorthogonal orbitals. A Grassmann conjugate gradient algorithm is our subspace optimization method. We usethe substitution of a silicon atom by a nitrogen atom in cubic SiC with two different configurations as our test systems. Modifications of the existing method are made for the inclusion of a half-filled defect level. The approximation of localization for the orbitals gives linear scaling of the dominant parts of the algorithm thus allowing for the simulation of large systems. Numerical tests indicate that quite accurate quantitative results are obtained when using extended orbitals near the defect and a localization regions with radii of 7 Bohr for orbitals away from the defect.