Machine Learning Approaches for Predicting Crystal Systems: A Brief Review and a Case Study

Determining the crystal system and space group for a compound from its powder X-ray diffraction data represents the initial step of a crystal structure analysis. This task can constitute a bottleneck in the material science workflow and often requires manual interventions by the user. The fast development of Machine Learning algorithms has strongly impacted crystallographic data analysis. It offered new opportunities to develop novel strategies for accelerating the crystal structure discovery processes. This paper aims to provide an overview of approaches recently proposed for crystal system prediction, grouped according to the input features they use to construct the prediction model. It also presents the results obtained in predicting the crystal system of polycrystalline compounds, by using the lattice parameters to train some learning models.