A New Era of Inorganic Materials Discovery Powered by Data Science

Abstract

Advances in the synthesis, characterization, and high-throughput computation of inorganic compounds are rapidly proliferating the quantity and complexity of data available to scientists. By taking advantage of these extensive data sets, it is now possible to transition the field of solid-state chemistry toward guided materials discovery. Machine learning and associated methods in data science have for decades been used to identify correlations and patterns from large amounts of complex data. Recent applications of data science in materials chemistry have also shown its outstanding potential to investigate the composition–structure–property-processing relationships using similar data-centered techniques. In this chapter, a high-level overview discussing the relevance of data science in material chemistry is first presented, followed by a description of the typical workflow and necessary procedures for applying machine learning to inorganic materials chemistry. A discussion on some of the most commonly used algorithms and approaches is also provided. The current status of data-driven inorganic materials research is subsequently reviewed, with a specific focus on applications of machine learning in materials synthesis optimization, inorganic crystal structure prediction, and advancing physical property prediction. Finally, current challenges and opportunities related to machine learning in inorganic materials chemistry are considered.