Self-Energy Dispersion in the Hubbard Model

TL;DR

This paper introduces the concept of self-energy dispersion as an error measure for local theories in the Hubbard model, showing how adding a local potential can simplify the self-energy and improve the accuracy of local approximations.

Contribution

It proposes the self-energy dispersion as a new error bound, analyzes its behavior in the Hubbard model, and suggests a simple energy-based estimator for the local-to-nonlocal crossover.

Findings

Self-energy dispersion can be minimized by adding a local potential.

Local theories become more accurate with flattened self-energy.

A simple energy estimator predicts the transition from non-local to local self-energies.

Abstract

We introduce the concept of self-energy dispersion as an error bound on local theories and apply it to the two-dimensional Hubbard model on the square lattice at half-filling. Since the self-energy has no single-particle analog and is not directly measurable in experiments, its general behavior as a function of momentum is an open question. In this article we benchmark the momentum dependence with the Two-Particle Self-Consistent approach together with analytical and numerical considerations and we show that through the addition of a local single-particle potential to the Hubbard model the self-energy can be flattened, such that it is essentially described by only a frequency-dependent term. We use this observation to motivate that local theories, such as the dynamical mean-field theory, should be expected to give very accurate results in the presence of a potential of this kind. Finally, we propose a simple energy argument as an estimator for the crossover from non-local to local self-energies, which can be computed even by local theories such as dynamical mean-field theory.