a Perturbative Study of Correlated Electrons in Periodic Systems

In an attempt to obtain a better description than the local density approximation (LDA) for the electronic structure of narrow-band systems, we studied the dynamic effects of a local, Hubbard-type electron-electron interaction (U) by including fluctuations around the self-consistent, mean-field solution. We initially tested this perturbative approach on the periodic Anderson model. For a wide range of properties such as quasiparticle energies, spectral densities, and reduced hybridizations, we obtained a more reasonable overall description than the one given by LDA. Further, we were able to fix up a number of the outstanding weaknesses of the LDA, e.g. the energy dependence of the quasiparticle lifetime, and the appearance of satellite features. In the small U limit our method produced much better agreement for Fe, Co, and Ni than conventional LDA calculations for a number of experimental results, including effective masses, X-ray photoemission spectra, exchange splittings, and renormalized quasiparticle bands. We also obtain reasonable agreement for quasiparticle broadenings and the appearance of satellite features. For NiO we found that, despite the large value of U = 8eV, the results are dominated by the first-order term, as the gap strongly suppresses fluctuations. In the strong U limit we obtained good agreement for UPt_3, UIr_3, and UAu_3 with the experimental specific heat, and were able to understand the trends in the mass enhancement in terms of simple changes in the underlying band-structure. However, in contrast to the 3d-ferromagnets, we found that for these more strongly correlated, highly degenerate systems, our perturbative approach was unsatisfactory at high energies, because it does not capture the atomic nature of such systems.