Smooth self-energy in the exact-diagonalization-based dynamical mean-field theory: Intermediate-representation filtering approach

TL;DR

This paper introduces a filtering method using the intermediate representation to obtain smooth, physically meaningful real-frequency self-energy spectra in DMFT-ED calculations, reducing discretization noise.

Contribution

The paper presents a novel IR-based filtering scheme that improves the quality of real-frequency self-energy in DMFT-ED, enabling more accurate spectral analysis without analytic continuation.

Findings

Effective noise filtering of self-energy using IR basis.

Enhanced smoothness and physical relevance of spectra.

Potential for improved spectral analysis in DMFT-ED.

Abstract

We propose a method for estimating smooth real-frequency self-energy in the dynamical mean-field theory with the finite-temperature exact diagonalization (DMFT-ED). One of the benefits of DMFT-ED calculations is that one can obtain real-frequency spectra without a numerical analytic continuation. However, these spectra are spiky and strongly depend on the way to discretize a continuous bath (e.g., the number of the bath sites). The present scheme is based on a recently proposed compact representation of imaginary-time Green's functions, the intermediate representation (IR). The projection onto the IR basis acts as a noise filter for the discretization errors in the self-energy. This enables to extract the physically relevant part from the noisy self-energy. We demonstrate the method for single-site DMFT calculations of the single-band Hubbard model. We also show the results can be further improved by numerical analytic continuation.