Predicting the minimum energy pathway of 1H to 1T phase transition of 2D monolayer ScS2 using machine learning

Abstract

The minimum energy pathways and transition energy barriers during the 1H to 1T phase transitions of two-dimensional transition metal dichalcogenide (TMD) ScS₂ is calculated via density functional theory calculations (DFT) and is predicted using machine learning (ML) approach The resulting transition barrier is 0.19 eV. Among the tested machine learning models, Gaussian Process Regression achieved the highest accuracy with an absolute mean error of 2.50×10^−3. Evaluating the runtime efficiency of DFT and ML reveals that the appropriately trained machine learning models are significantly faster than DFT calculations. The ML models used in the study aside from Linear Regression are non-parametric. The atomic numbers of ScS₂, its bond lengths, and bond angles were utilized to train the ML models. This research contributes to the understanding of TMD phase transitions and highlights the potential of machine learning in materials science.