Dynamical mean-field theory applied to linear scaling density functional theory
-------------------------------------------------------------------------------
Cedric Weber
Abstract
--------
Phenomena that are connected to quantum mechanics, such as magnetism,
transport, and the effect of impurity atoms and disorder, and their relation to
material design and energy needs are important for almost every branch of the
industry. Density functional theory (DFT) was successful at making accurate
predictions for many materials, in particular compounds which have a metallic
behaviour. DFT combines high accuracy and moderate computational cost, but the
computational effort of performing calculations with conventional DFT
approaches is still non negligible and scales with the cube of the number of
atoms. A recent optimised implementation of DFT was however shown to scale
linearly with the number of atoms (ONETEP), and opened the route to large scale
DFT calculations.
Nonetheless, one bottleneck of DFT and ONETEP, is that it fails at describing
well some of the compounds where strong correlations are present, in particular
because the computational scheme has to capture both the band-like character of
the uncorrelated part of the compound and the Mott-like features emerging from
the local strongly correlated centres. A recent progress has been made in this
direction by the dynamical mean-field theory (DMFT), that allows to describe
the two limits (metal and insulator) in a remarkable precise way when combined
with DFT.
The ONETEP+DMFT implementation will be shortly discussed, and its applications
illustrated by two examples: i) the interplay of Mott and Anderson
localization within disordered Vanadium dioxide and ii) a typical biological
molecular system, iron porphyrin, which plays an important biological function
in human haemoglobin.