It had been reported, in a theoretical study, that the formation energy for a neutral Si self-interstitial was 4.9eV; using the diffusion quantum Monte Carlo method. This was 50% (30%) higher than the LDA (GGA) results of the so-called usual ab initio calculations. It was shown here that gap and finite-size corrections that were proposed in the past might be able to explain this difference, as long as the defect levels for point defects were well known. It was shown that B-interstitial systems were more difficult in that sense; thus questioning the generality of the approach.

Diffusion in Silicon and the Predictive Power of ab initio Calculations. W.Windl: Physica Status Solidi B, 2004, 241[10], 2313-8