Guiding solvothermal synthesis

Abstract

Machine Learning Machine learning is rapidly revolutionizing computer-aided synthesis design, occasionally producing vivid use cases when the reaction parameters important for the synthesis are hidden in a complex chemical space. Xie et al. report a machine learning–assisted framework for the synthesis of metal-organic nanocapsules (MONCs), giant molecular building units potentially useful in different fields, based on predicting the crystallization propensity using experimental attempts as a training dataset. Machine-learning algorithms achieve prediction accuracies of more than 90%, considerably outperforming trained chemists, and the generated synthesis parameters direct solvothermal crystallization to new structures of MONCs. The proposed strategy for the discovery of new materials can be applied more broadly beyond MONCs. J. Am. Chem. Soc. 10.1021/jacs.9b11569 (2019).