Machine-learned interatomic potentials for transition metal dichalcogenide Mo1−xWxS2−2ySe2y alloys

Abstract

Machine Learned Interatomic Potentials (MLIPs) combine the predictive power of Density Functional Theory (DFT) with the speed and scaling of interatomic potentials, enabling theoretical spectroscopy to be applied to larger and more complex systems than is possible with DFT. In this work, we train an MLIP for quaternary Transition Metal Dichalcogenide (TMD) alloy systems of the form Mo_1−xW_xS_2−2ySe_2y, using the equivariant Neural Network (NN) MACE. We demonstrate the ability of this potential to calculate vibrational properties of alloy TMDs including phonon spectra for pure monolayers, and Vibrational Density of States (VDOS) and first-order Raman spectra for alloys across the range of x and y. We show that we retain DFT level accuracy while greatly extending feasible system size and extent of sampling over alloy configurations. We are able to characterize the first-order Raman active modes across the whole range of concentration, particularly for the “disorder-induced” modes.