Adaptive Genetic Algorithm for Crystal Structure Prediction

We present a genetic algorithm (GA) for structural search that combines the speed of structure exploration by classical potentials with the accuracy of density functional theory (DFT) calculations in an adaptive and iterative way. This strategy increases the efficiency of the DFT-based GA by several orders of magnitude. This gain allows considerable increase in size and complexity of systems that can be studied by first principles. The method's performance is illustrated by successful structure identifications of complex binary and ternary inter-metallic compounds with 36 and 54 atoms per cell, respectively. The discovery of a multi-TPa Mg-silicate phase with unit cell containing up to 56 atoms is also reported. Such phase is likely to be an essential component of terrestrial exoplanetary mantles.