Understanding Charge Density Wave Phase Transition in 1T-TaS2 via Machine Learning Force Field

Abstract

Recently, the charge density wave (CDW) phases and their properties in various condensed matter systems have been widely studied. In condensed matter, the formation of CDWs and periodic lattice distortions leads to changes in the electronic transport properties. Among these materials, 1T-type tantalum disulfide (1T-TaS2) is a layered material, and its CDW phases indicate a two-step phase transition depending on thickness and temperature. Experimental studies have shown that its physical properties depend on the stacking order of its layers, which affects the CDW phase transitions. However, due to its strong correlation between electronic structure and complex atomic geometry, understanding the CDW phase transition in 1T-TaS2 is challenging from a theoretical perspective. In this study, we investigated the tendency for structural change from molecular dynamics (MD) with the machine learning force field (MLFF), to simulate multiscale dynamics. We extracted a MLFF for a 1T-TaS2 monolayer based on Ab Initio Molecular Dynamics (AIMD) simulation training data on an Angstrom scale. Using this MLFF in MD simulations, we performed large-scale (∼< 100 nm^2) and long-time (∼< ns) simulations of temperature-dependent dynamics. Through these simulations, we discovered CDW phase transitions and domain wall formations at various temperatures and times, and analyzed their atomic composition. Our results not only theoretically predicted the temperature-dependent bulk CDW phase transition of 1T-TaS2, but also observed microscopic dynamics. This suggests that it will provide explanations consistent with previous experimental results. In future studies, our MLFF approach methodology could be applied to investigate bulk 1T-TaS2. Moreover, our findings may contribute to future studies that analyze light-induced hidden phases in 1T-TaS2 in detail, and could be extended to understand CDW phase transitions in other materials.