Moment-functional based spectral density-functional theory

TL;DR

The paper introduces MFbSDFT, a density-functional method that simultaneously computes ground state electron density and spectral functions using spectral moments, capturing correlation effects efficiently.

Contribution

It presents a novel spectral density-functional theory based on spectral moments, enabling efficient computation of spectral functions alongside ground state densities.

Findings

Captures correlation effects like valence-band satellites in Ni.

Reproduces Hubbard bands in SrVO₃.

Efficient algorithm for spectral function construction.

Abstract

We describe a density-functional method which aims at computing the ground state electron density and the spectral function at the same time. One basic ingredient of our method is the construction of the spectral function from the first four spectral moment matrices. The second basic ingredient is the construction of the spectral moment matrices from density functionals. We call our method moment-functional based spectral density-functional theory (MFbSDFT), because it is based on density-functionals for the spectral moments and because it allows us to compute the spectral function. If it is implemented in second variation our method consumes only a fraction more computer time than a standard DFT calculation with the PBE functional. We show that MFbSDFT captures correlation effects such as the valence-band satellite in Ni and the formation of lower and upper Hubbard bands in SrVO$_3$. For the purpose of constructing the spectral function from the first four $N\times N$ spectral moment matrices we describe an efficient algorithm based on the diagonalization of one hermitean $2N\times 2N$ matrix.