Realistic many-body approach to materials with strong nonlocal correlations
Our project aims at substantial progress in the development of a realistic dynamical
mean-field theory. This is achieved by advancing our present-day impurity solvers and
by improving the interface to band theory. We employ optimized quantum Monte-Carlo techniques for
the numerical evaluation of the recently developed dual-fermion approach which is
extended for a general impurity model with several orbitals and sites.
Our goal is to include non-local correlations beyond the single-site DMFT, to include
long wave-length modes and to calculate dynamical two-particle correlation functions.
These finite-temperature (T>0) methods are checked against a T=0 solver that will be developed by
using a recent reformulation of the (dynamical) density-matrix renormalization group
in terms of matrix-product states. The prime goal is to provide and apply a T=0
multi-orbital impurity solver without a sign problem and for real frequencies.
Vital methodical progress is
envisaged at the interface of DMFT to effective single-particle methods: This
comprises tailored basis sets for an efficient representation of the correlated
subspace, global charge self-consistency as well as access to phase diagrams and atomic
forces due to a reliable scheme to compute the free energy. Correlated transition
metals and compounds as well as correlation-induced insulators and frustrated lattice
systems are considered.