The prediction of both stable and metastable crystal structures by atomistic modeling has proven to be a challenging yet important task. Traditionally the phase diagrams of a given chemical model have been generated by comparing the free energy of experimentally known crystal structures. Ideally one should be able to predict new stable or metastable crystal structures without any apriori knowledge of the structure should look like. In practice given the large number of possible atomic configurations at the molecular level simple algorithms are not sufficient for this purpose. In this work we present approaches based on genetic algorithms and machine learning to idenfify both stable and meta-stable crystal structures across. These algorithms are capable of learning on the fly in order to find and correctly predict stable structures for given thermodynamic conditions.

Predicting phase diagrams via free energy based genetical algorithms

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.