Abstract:
We address the problem of a reliable description of strong electron correlations in metals. In particular we analyze the options one has at hand within the many-body perturbation theory to describes qualitatively correctly the transition between weak-coupling and strong-coupling regimes. We first discuss various levels of renormalizations one can introduce into the diagrammatic perturbation expansion. We show that a direct mass renormalization (one-particle self-consistency) is not as important as a charge renormalization introduced by a two-particle self-consistency. We demonstrate and explain the failure of simple approximations with only mass renormalization such as renormalized RPA, TMA and other FLEX-type approximations in the strong-coupling regime. We then show that an affordable improvement upon these approximations by a charge renormalization and a self-consistent determination of two-particle irreducible vertices within the parquet approach is capable to reproduce the Kondo asymptotics in the single-impurity Anderson model.