Speaker
Description
Accurate total energy calculations within the combined Density Functional Theory and Dynamical Mean-Field Theory (DFT+DMFT) framework are essential for predictive studies of correlated materials, but remain computationally demanding when using numerically exact impurity solvers such as continuous-time quantum Monte Carlo (CT-QMC). This limitation is particularly severe for problems involving structural degrees of freedom such as volume optimization or lattice distortions in correlated materials where repeated total energy evaluations are required.
In this work, we explore the use of approximate impurity solvers to significantly accelerate many-body total energy calculations. In particular, we employ an exact diagonalization (ED) solver based on the EDIpack library and assess its performance against CT-QMC benchmarks. Using SrVO₃ as a prototypical correlated metal, we perform a systematic comparison of total energies and including calculations as a function of unit cell volume.
We show that, despite its approximate nature due to bath discretization, the ED solver can in particular cases achieve close agreement with CT-QMC results when an appropriate bath parametrization is employed, while computational costs are reduced by orders of magnitude. This efficiency makes ED-based approaches particularly well suited for structural optimization and the study of correlation-driven lattice effects, including Jahn-Teller distortions in perovskites.
Our results demonstrate that approximate solvers, when carefully implemented, provide a reliable and efficient route for many-body total energy calculations within DFT+DMFT, enabling scalable studies of structurally complex correlated systems.