Quantum Monte Carlo assessment of embedding for a strongly-correlated defect: interplay between mean-field and interactions (2505.00845v1)

Abstract: Point defects are of interest for many applications, from quantum sensing to modifying bulk properties of materials. Because of their localized orbitals, the electronic states are often strongly correlated, which has led to a proliferation of quantum embedding techniques to treat this correlation. In these techniques, a weakly correlated reference such as density functional theory is used to treat most of the one-particle states, while certain states are singled out as an active space to be treated with an effective interaction. We assess these techniques in the context of an iron defect in aluminum nitride by referencing to a fully correlated quantum Monte Carlo description. This comparison allows us to have access to detailed information about the many-body wave functions, which are not available experimentally. We find that errors in the underlying density functional theory calculation, and thus choice of the active space, lead to qualitatively incorrect excited states from the embedded model. These errors are extremely difficult to recover from by adding corrections such as double counting or many-body perturbation theory.