First-principles total-energy calculations of interstitial Fe were performed. Particular attention was paid to the matrix elements of hyperfine interactions with the Fe nucleus and with Si ligand nuclei. The total energy calculations were performed within the general framework of density functional theory. Many-particle effects were treated using the local spin-density approximation. The Dyson equation approach was used to solve the Kohn-Sham equation of density functional theory and to calculate the Green’s function, using linear muffin-tin orbital theory and the atomic spheres approximation (although the latter approximation did not permit the inclusion of lattice relaxation effects). The total-energy calculations led to a model, for the electronic structure, which was dominated by the covalent hybridization of Fe d-states with Si p-states. The results explained reactions of the electronic system which occurred when passing from Fei0 to Fei+. By using a modified Ludwig-Woodbury model, the exchange splitting of the Fe d-states was first considered. The latter states were then further split by the crystal field. The hyperfine matrix elements for Si:Fei0 were calculated directly, and exhibited excellent agreement with the results of electron-nuclear double resonance experiments. In the case of Si:Fei+, account also had to be taken of the spin-orbit interaction of Fe d-orbitals. Because these total-energy calculations ignored spin-orbit interactions, a method was also used which exploited symmetry-adapted many-particle wave functions. These functions were used to calculate the matrix elements for hyperfine interaction with the Fe nucleus and several Si ligand nuclei. Localization properties of the magnetization density were used to determine

the many-particle wave function and to obtain data on hyperfine interactions. It was shown that covalent hybridization of the Fe d-states with Fe p-states and Si orbitals could not alone explain the discrepancy between calculated and measured data on hyperfine interactions with the Fe nucleus. However, the incorporation of a dynamic Jahn-Teller effect led to consistent results. Calculated data for ligand hyperfine interactions permitted ligand shells to be precisely related to measured electron nuclear double resonance data.

H.Weihrich, H.Overhof: Physical Review B, 1996, 54[7], 4680-95