The timely integration of crystal structure databases, such as CRYSTMET, ICSD etc., with quantum software, like VASP, OresteS, ElectrA
etc., allows ab initio cell and structure optimization on existing pure-phase compounds to be performed seamlessly with just a few mouse clicks. Application to the optimization of rough structure models, and possibly new atomic arrangements, is detailed. The ability to reproduce observed cell data can lead to an assessment of the intrinsic plausibility of a structure model, even without a competing model. The accuracy of optimized atom positions is analogous to that from routine powder studies. Recently, the ab initio symmetry-general least-squares extraction of the coefficients of the elastic tensor for pure-phase materials using data from corresponding entries in crystal structure databases was automated. A selection of highly encouraging results is presented, stressing the complementarity of simulation and experiment. Additional physical properties also appear to be computable using existing quantum software under the guidance of an automation scheme designed following the above automation for the elastic tensor. This possibility creates the exciting perspective of mining crystal structure databases for new materials with combinations of physical properties that were never measured before. Crystal structure databases can accordingly be expected to become the cornerstone of materials science research within a very few years, adding immense practical value to the archived structure data.
Data-driven learning and prediction of inorganic crystal structures
