Predicting the electronic and structural properties of two-dimensional materials using machine learning

Abstract

Machine-learning (ML) models are novel and robust tools to establish structure-to-property connection on the basis of computationally expensive ab-initio datasets. For advanced technologies, predicting novel materials and identifying their specification are critical issues. Two-dimensional (2D) materials are currently a rapidly growing class which show highly desirable properties for diverse advanced technologies. In this work, our objective is to search for desirable properties, such as the electronic band gap and total energy, among others, for which the accelerated prediction is highly appealing, prior to conducting accurate theoretical and experimental investigations. Among all available componential methods, gradient-boosted (GB) ML algorithms are known to provide highly accurate predictions and have shown great potential to predict material properties based on the importance of features. In this work, we applied the GB algorithm to a dataset of electronic and structural properties of 2D materials in order to predict the specification with high accuracy. Conducted statistical analysis of the selected features identifies design guidelines for the discovery of novel 2D materials with desired properties.