High-throughput Wannierizations and screening of novel materials

Maximally localized Wannier functions (MLWFs) are widely used in computational condensed matter physics and materials science. We recently developed some very effective automated Wannierization algorithms: projectability-disentangled Wannier functions (PDWFs), to construct atom-centered Wannier functions that span both occupied and unoccupied states, and manifold-remixed Wannier functions (MRWFs), that mix these into target manifolds. These methods make it possible to reliably Wannierize both composite bands (e.g., the valence bands of insulators), or entangled bands (of metals, or the valence plus conduction bands of insulators), as exemplified here for over 20,000 crystal structures taken from the Materials Cloud three-dimensional crystals database (MC3D). We showcase this MLWF database and its accuracy and efficiency in calculating many physical properties with several applications geared toward the discovery of high-performance materials, including thermoelectrics, polar crystals, and nonlinear Hall materials.