Exploring Charge Density Waves in two-dimensional NbSe2 with Machine Learning

Niobium diselenide (NbSe₂) has garnered attention due to the coexistence of superconductivity and charge density waves (CDWs) down to the monolayer limit. However, realistic modeling of CDWs-accounting for effects such as layer number, twist angle, and strain-remains challenging due to the prohibitive cost of first-principles methods. To address this, we develop machine learning interatomic potentials (MLIPs), based on the Allegro architecture-an E(3)-equivariant model -- specifically tailored to capture subtle CDW effects in NbSe₂. These MLIPs enable efficient exploration of commensurate and incommensurate CDW phases, as well as the dimensional dependence of the transition temperature, evaluated using the Stochastic Self-Consistent Harmonic Approximation (SSCHA). Our findings reveal a strong sensitivity of CDWs to stacking and layer number, and a slight suppression of the transition temperature with increasing thickness. This work opens new possibilities for studying and tuning CDWs in NbSe₂ and other 2D systems, with implications for electron-phonon coupling, superconductivity, and advanced materials design.