Molecular crystal structure prediction (CSP) is challenging because it requires searching a high-dimensional configuration space with high accuracy. To this end, we combine first principles simulations with machine learning and genetic algorithm (GA) optimization. Our CSP workflow begins with a machine learned model for molecular solid form volume estimation [1]. Then, the random structure generator, Genarris [2] is used for pre-screening the configuration space to generate an initial population for the GAtor genetic algorithm [3]. GAs rely on the evolutionary principle of survival of the fittest to perform global optimization. GAtor's special features include a variety of crossover and mutation operators designed for molecular crystals, evolutionary niching to steer the GA to promising but under-explored regions, and a massive parallelization scheme. Once a crystal structure is found, Ogre [4] may be used to cleave surfaces from the bulk crystal structure, perform surface energy calculations, and calculate the Wulff shape to predict the crystal morphology.

Molecular crystal structure and shape prediction with Genarris, GAtor, and Ogre

ACS Spring Meeting 2021

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