Density functional theory calculations of defect energies using supercells

Reliable calculations of defect properties may be obtained with density functional theory (DFT) using the supercell approximation. We systematically review the known sources of error and suggest how to perform calculations of defect properties in order to minimize errors. We argue that any analytical error-correction scheme relying on electrostatic considerations alone is not appropriate to derive reliable defect formation energies, certainly not for relaxed geometries. Instead we propose finite size scaling of the calculated defect formation energies, and compare the application of this with both fully converged and 'Gamma' (Gamma) point only k-point integration. We provide a recipe for practical DFT calculations which will help to obtain reliable defect formation energies and demonstrate it using examples from III-V semiconductors.