Cross comparing ab-initio methods for estimating the Hubbard U: constrained random-phase approximation vs. linear response
by
OVGA/200
Accurate determination of the screened Coulomb interaction parameter (U) is critical for modeling strongly correlated quantum materials using DFT+U and DFT+DMFT. Constrained random-phase approximation (cRPA) and linear response theory (LRT) historically faced comparison challenges due to basis set differences. This work introduces maximally localized Wannier functions as unified projectors, enabling consistent U calculations across both methods.
For two benchmark systems: KCuF₃ and Sr₂FeO₄, the methods agree strongly when interacting and screening states are separable. However, cRPA becomes unreliable in systems with entangled subspaces (e.g., overlapping correlated/uncorrelated states), while LRT remains robust. The framework resolves methodological discrepancies by providing a common basis, leveraging Wannier functions' transferability.
This advancement bridges DFT+U and DFT+DMFT communities, enabling direct comparisons and fostering standardized U determination. Future applications could address lingering discrepancies in complex correlated systems, enhancing cross-method compatibility in electronic structure research.
Laboratory for Materials Simulations (LMS)