GW method for 3d and 4f systems.

Aryasetiawan F.

Lund Univ Dept Theoret Phys Solvegatan 14a S-23362 Lund Sweden
Max Planck Inst Festkorperforsch D-70569 Stuttgart Germany

Conventional methods for calculating the electronic self-energy of a crystal within the GW approximation are mostly based on a plane-wave approach, While these methods have been successful in treating s-p systems, applications to systems with localized states such as 3d and 4f systems have been hampered by the large size of the computation. We present a method for calculating the self-energy of 3d and 4f systems within the GW approximation. Unlike, conventional plane-wave methods, the method uses products of LMTO's. Due to the small number of basis functions, the method allows full calculations of the dielectric matrix and makes it feasible to treat systems which would otherwise be beyond the present computational capability. Applications have been made to transition-metal systems (Ni, NiO) and more recently to an f-system (Gd) and to the 3d semicore states in ZnSe, GaAs and Ge with encouraging results. The band structure of Ni is greatly improved in particular the overestimated LDA bandwidth is narrowed by similar to 1 eV bringing it much closer to the experimental value, In NiO, the LDA gap, which is much too small (0.2 eV in LDA and 4.0 eV experimentally), is widened to similar to 5 eV. Since the initial LDA band gap is quite different from the final one, we find that it is important to perform the calculation self-consistently. Application to the semicore states in ZnSe, GaAs has also yielded good agreement with experiment. We derived a simple formula to describe the increasing error from ZnSe to Ge in the LDA eigenvalues. Finally, application to an f-system Gd has revealed some new features: a very small quasiparticle weight (0.3) and the presence of a low-energy satellite structure in the spectral function.

A reprint of this paper can be obtained at the cond-mat in Germany or in the US .

Physica B, 237 321-323, 1997.

Max-Planck Institut für Festkörperforschung;
Postfach 80 06 65   D-70506 Stuttgart