Hubbard U parameters for transition metals from first principles

Using the linear response-based constrained local density approximation (cLDA) approach we systematically computed the Hubbard U parameters for series of 3d, 4d, and 5d transition metals. We compare the results with estimations by the constrained random phase approximation (cRPA) method and discuss the performance of the self-consistent density functional theory + U (DFT + U) method for prediction of lattice parameters, work functions, d-bandwidths and d-band centers. Interestingly, we found that blindly applied the standard, fully localized limit (FLL) version of the DFT + U approach heavily overestimates the positions of d-band centers with respect to the Fermi level, but much better agreement with experiment is obtained when applying a more realistic, Wannier-type representation of d orbitals for projection of d states occupancies. We present another, independent estimate of the Hubbard U parameter based on the comparison of Hartree-Fock and DFT eigenvalues, and positions of d-band centers. The so-derived estimates are surprisingly well consistent with the ones derived from the above-mentioned first principles approaches, and allow for validation of cRPA or cLDA results for the disputed cases, including Cu, Ag, and Au for which large U parameters are obtained from the cLDA method.

Automated summary

In this study, the Hubbard U parameters for transition metals were systematically calculated using different methods, and it was found that using a more realistic projection method produced results that were more in line with experimental data and consistent with other approaches.