Silicon is the most important material in the microelectronics industry. The diffusion of impurity atoms in silicon is critically influenced by intrinsic defects such as self-interstitials and vacancies, and it is therefore of great importance to improve our understanding of them. Unfortunately it has not been possible to detect self-interstitials directly, and the experimental situation regarding self-diffusion in silicon is still highly controversial. Indeed, experimental data have been used to support values of the diffusivity of the silicon self-interstitial that differ by ten orders of magnitude at the temperatures of around 800°C at which silicon is processed.

The consensus arising from density-functional theory (DFT) calculations is that the split-<110>, hexagonal, and tetrahedral self-interstitial defects are the lowest in energy. Another interesting suggestion is that self-diffusion could occur without point defects via exchange of neighbouring atoms in the perfect lattice, and Pandey (1986) proposed such a concerted exchange mechanism for self-diffusion in silicon.

Leung et al. performed diffusion quantum Monte Carlo (DMC) calculations and DFT calculations to determine the formation energies of self-interstitials in silicon. Within each method we found the split-<110> and hexagonal interstitials to be the most stable. The DMC formation energies are about 1 eV larger than the PW91-GGA values and 1.5 eV larger than the LDA values. Leung et al. used these DMC data to estimate a value for the activation energy for self-interstitial diffusion of about 5 eV, which is consistent with the value deduced from experiment of 4.84 eV. The activation energies predicted by the LDA and PW91-GGA density functionals are considerably lower than the experimental value, and do not provide a satisfactory explanation of self-diffusion in silicon.