Beyond MX2: Transition metal chalcogenides discovered via selective random structure search

Abstract

The canonical MX2 formulation in the class of transition metal chalcogenides (TMCs) has received intensive attention due to its promising potential in various electronic applications. However, TMCs can adopt a remarkably diverse array of stoichiometries, elemental combinations, and dimensional architectures, which have not been investigated thoroughly. In this work, we present a Selective Random Structure Search framework for exploring TMCs beyond conventional MX2 stoichiometries and dimensionalities. By combining random structure generation with machine-learning-based pre-screening, we systematically identify thermodynamically stable phases across the MoxSy system. Apart from reproducing the existing structures, we report a family of previously unknown MoxSy nanotubes and generalize them via elemental substitution to multiple compositions, predicting several promising low-dimensional nanostructures for experimental synthesis.