Fictive Impurity Models: an Alternative Formulation of the Cluster Dynamical Mean Field Method

TL;DR

This paper introduces Fictive Impurity Models as an alternative formulation of the cluster dynamical mean field method, interpreting clusters as computational tools for self-energy expansion rather than physical subunits, and proposes filtering techniques to address causality issues.

Contribution

It offers a novel perspective on cluster DMFT by framing clusters as algorithms for self-energy expansion and introduces filtering methods to improve causality.

Findings

Formalism matches perturbative calculations

Filtering methods mitigate causality problems

Provides a new interpretation of cluster methods

Abstract

"Cluster" extensions of the dynamical mean field method to include longer range correlations are discussed. It is argued that the clusters arising in these methods are naturally interpreted not as actual subunits of a physical lattice but as algorithms for computing coefficients in an orthogonal function expansion of the momentum dependence of the electronic self-energy. The difficulties with causality which have been found to plague cluster dynamical mean field methods are shown to be related to the "ringing" phenomenon familiar from Fourier analysis. The analogy is used to motivate proposals for simple filtering methods to circumvent them. The formalism is tested by comparison to low order perturbative calculations and self consistent solutions.