Machine-learning accelerated identification of exfoliable two-dimensional materials

Abstract Two-dimensional (2D) materials have been a central focus of recent research because they host variety properties, making them attractive both for fundamental science and applications. It is thus crucial to be able identify accurately efficiently if bulk three-dimensional (3D) are formed by layers held together weak binding energy that, thus, can potentially exfoliated into 2D materials. In this work, we develop machine-learning (ML) approach combined with fast preliminary geometrical screening, exfoliable Starting from combination descriptors crystal structures, work out subset that accurate predictions. Our final ML model, based on random forest classifier, has very high recall 98%. Using SHapely Additive exPlanations analysis, also provide an intuitive explanation the five most important variables model. Finally, compare performance our best model deep neural network architecture using same descriptors. To make algorithms models easily accessible, publish online tool Materials Cloud portal only requires 3D structure as input. provides practical yet straightforward assess whether any compound layers.