Self-consistent dynamical Hubbard functional for correlated solids

Abstract

Many-body functionals of the Green's function can provide fundamental advances in electronic-structure calculations, due to their ability to accurately predict both spectral and thermodynamic properties, such as angle-resolved photoemission spectroscopy (ARPES) experiments and total energies of materials. However, fully first-principles, self-consistent calculations with these dynamical functionals remain a major challenge, ultimately limiting their application to thermodynamic quantities, and restricting spectral predictions to one-shot calculations. In this paper, we present a fully self-consistent treatment of the electronic structure of solids using a dynamical functional. Our approach leverages the so-called dynamical Hubbard functional, which generalizes the DFT+$U$ correction by incorporating frequency-dependent screening, augmenting the static density functional to accurately describe both spectral and thermodynamic properties of materials with $d$- or $f$-localized orbitals near the Fermi level. To enable this, we employ the algorithmic-inversion method based on a sum-over-poles representation, resulting in a numerically accurate self-consistent scheme for frequency-integrated properties, while keeping real-axis spectral resolution for dynamically-resolved quantities. Using this framework, we study the paradigmatic correlated solid SrVO$_3$, accurately reproducing its spectral features, essentially confirming previous one-shot predictions, and improving the description of its equilibrium properties, such as the equilibrium volume and bulk modulus, bringing these significantly closer to experimental measurements.