Recent advances and applications of machine learning in solid-state materials science

• DOI: 10.1038/s41524-019-0221-0
• Publication Date: 2019-08-08T10:20:54Z
• Authors: Jonathan Schmidt; Mário R. G. Marques; Silvana Botti; Miguel A. L. Marques
• Keywords: Interpretability; Toolbox

Abstract One of the most exciting tools that have entered material science toolbox in recent years is machine learning. This collection statistical methods has already proved to be capable considerably speeding up both fundamental and applied research. At present, we are witnessing an explosion works develop apply learning solid-state systems. We provide a comprehensive overview analysis research this topic. As starting point, introduce principles, algorithms, descriptors, databases materials science. continue with description different approaches for discovery stable prediction their crystal structure. Then discuss numerous quantitative structure–property relationships various replacement first-principle by review how active surrogate-based optimization can improve rational design process related examples applications. Two major questions always interpretability physical understanding gained from models. consider therefore facets importance Finally, propose solutions future paths challenges computational