Systematic dynamical mean-field theory study of 3d perovskite oxides with uniform Coulomb interactions

TL;DR

This study develops a high-throughput, parameter-free eDMFT framework for 3d perovskite oxides, enabling accurate, large-scale electronic structure predictions without the need for material-specific Coulomb interaction parameters.

Contribution

The paper introduces a robust, parameter-tuning-free high-throughput eDMFT approach using fixed Coulomb interactions, validated against experimental spectra for 3d perovskite oxides.

Findings

Spectral properties are mainly governed by dynamical self-energy.

Self-consistently screened Coulomb interactions fall within narrow ranges.

High-throughput eDMFT shows excellent agreement with photoemission data.

Abstract

Strongly correlated transition-metal perovskite oxides pose a fundamental challenge for electronic-structure theory and for large-scale, data-driven materials discovery. While DFT+DMFT provides a quantitatively accurate description of such systems, its high-throughput application is hindered by the need to determine material-specific Coulomb interaction parameters ($U$). First-principles approaches such as the cRPA predict a highly nonlinear and non-transferable evolution of the interaction strength across chemically similar ABO$_3$ perovskites. Here we show that this paradigm does not extend to the large-energy-window eDMFT, which employs highly localized orbitals and treats electronic correlations and screening self-consistently within the same many-body framework. As a result, spectral properties are governed primarily by the dynamical self-energy rather than by static interaction-induced energy shifts. Recent constrained-eDMFT calculations demonstrated that, for broad classes of $3d$ transition-metal oxides, the self-consistently screened Coulomb interactions naturally fall within relatively narrow ranges for correlated metals and insulators. Motivated by these findings, we implement a high-throughput eDMFT framework employing physically derived interaction values of $U=6$ eV for metals and $U=10$ eV for insulators together with $exact$ double counting. We test this framework using systematic high-throughput eDMFT calculations for ABO$_3$ compounds (A = Ca, Sr, La; B = V--Ni) and benchmark the resulting spectral functions against photoemission experiments, where we find overall excellent agreement. Our results establish that charge self-consistent eDMFT enables robust, parameter-tuning-free high-throughput many-body calculations for correlated oxides, opening a practical pathway toward predictive electronic-structure databases for strongly correlated materials.