Combining ab initio accuracy and large-scale Molecular Dynamics with Machine Learning: an application on Transition Metal Dichalcogenides.*

Amidst the exciting landscape of materials science, Transition Metal Dichalcogenides (TMDs) have emerged as exciting subjects of exploration. These versatile low-dimensional materials, which include monolayer TMDs like MoS2 and MoSe2, offer a rich tapestry of optical and dynamical properties with applications across various scientific domains. In our pursuit of understanding TMDs, we've employed a molecular dynamics method powered by neural network potentials. This innovative approach enables us to delve into these materials with remarkable precision, and we have determined phonon dispersion curves and frequency-dependent dielectric constants at finite temperatures for monolayer systems. Importantly, our approach is adaptable and can readily incorporate any new findings.