Accurate Excitations of the NV- Defect in Diamond via Embedding with Auxilliary-Field Quantum Monte Carlo

Predictive capability for excitations in defects would enable better understanding of the defects that occur in materials, and significantly improve our ability to engineer defects for applications, such as qubits. Dilute defect systems require large simulations and defects often host correlated states that require accurate and robust computations to treat. We present high-accuracy first-principles auxiliary-field quantum Monte Carlo (AFQMC) calculations applied to large systems via an embedding approach. This system contains a set of charged states with both closed-shell, weakly correlated states as well as open-shell, strongly correlated states, including the experimentally well-characterized triplet and singlet excitations. This provides a challenging realistic system to thoroughly test the accuracy of our approach. Utilizing the favorable scaling of AFQMC, we systematically study the convergence of our embedding approach by including up to hundreds of electrons in the embedded space. We determined the most effective procedure for converging excitations, and the degree of locality of the excitations, with possible implications to many other embedding approaches. The computed excitation energies are in very good agreement with experiment. We anticipate this method will become a general approach for first-principles characterization of correlated defects.