High-throughput dynamical mean-field theory computations for 3d-perovskite oxides

One of the major challenges in high-throughput computations of strongly correlated materials, such as 3d perovskite oxides, is the non-transferability of the screened Coulomb interaction U, even among compounds within the same family. The U values obtained from either constrained DFT (cDFT) or constrained random phase approximation (cRPA) often exhibit a non-linear dependence on increasing d-electron valence at the B site, as commonly observed in transition-metal ABO₃ perovskites. This typically necessitates empirical fine-tuning or costly first-principles evaluations. Here, we employ the recently proposed constrained dynamical mean-field theory (cDMFT), which self-consistently incorporates essential vertex corrections, to investigate perovskite oxides ABO₃ (A = Ca, Sr, La; B = V, Cr, Mn, Fe, Co, Ni) using structures from the Crystallography Open Database. We perform direct, parameter-tuning-free comparisons between the spectral properties computed within the embedded DMFT (eDMFT) framework and experimental photoemission spectra. The excellent agreement achieved without empirical adjustment of highlights the predictive capability of cDMFT and opens a pathway toward high-throughput DMFT computation of correlated materials.