Predicting the properties of ultrathin magnetic H_xCrS₂ from first principles and machine learning

An air stable H_xCrS₂ layered material has been synthesized by soft chemical methods, which can be exfoliated down to ultrathin layers, providing a promising path for the synthesis of two-dimensional (2D) magnets¹. Although a reliable synthesis method has been developed, the atomic structure is still unknown. Variables such as Cr vacancies and H impurities must be determined in order to understand the properties of this material for device applications. We used a combination of density functional theory (DFT), molecular dynamics (MD) and cluster expansion formalism to study the energetics as a function of Cr vacancies and H impurities. From here, we studied the stability, electronic and magnetic properties of these H_xCrS₂ structures and examined the effect of layering on these properties. In order to extend our calculations to a wider range of structures, we trained a machine learning algorithm with our DFT and MD calculated data to predict the properties of other H_xCrS₂ based materials outside our training set.

¹X. Song et. al, J. Am. Chem. Soc. 141, 15634 (2019)