Machine Learning-Based Recommendation of Optimal Crystallization Conditions for Organic Small Molecules

Crystallization is an important process in a broad range of industries, though studies on this topic remain complicated. Recently, machine learning has been applied to resolve complex issues in chemistry and material science. Here we present a machine learning model to propose crystallization experiments for organic small molecules. This model has been integrated into a robotic platform that performs experiments automatically. To improve applicability and accuracy, the model was trained on both simulated and experimental data. In comparative case studies, polymorph screening experiments by the platform yielded a high rate of solid products, and the number of forms obtained by the platform equaled those obtained by human researchers. The model provides a data-based perspective of the promoting and inhibiting influences to crystallization from molecular and interaction features. This work demonstrates the feasibility of applying machine learning techniques to solid-state studies to boost efficiency and deepen understanding.