Neuroevolution potential enabled accurate prediction of thermophysical properties for high-temperature Ga2O3 melt

Reliable thermophysical parameters of Ga 2 O 3 melt are essential for the reliable numerical simulation and process optimization of β -Ga 2 O 3 crystal growth. However, experimental measurements of melt properties near the melting point remain extremely challenging due to the ultrahigh melting temperature of Ga 2 O 3 . Herein, we develop a neuroevolution potential (NEP) trained on density functional theory (DFT) datasets to perform large-scale molecular dynamics (MD) simulations. Using this approach, we provided the first estimates of thermal conductivity and surface-tension temperature coefficient of Ga 2 O 3 melt, while systematically evaluating its density, isobaric thermal expansion coefficient and viscosity, with all predictions compared against available experimental data. The simulated results provide a comprehensive set of thermophysical data for numerical simulation of β -Ga 2 O 3 melt growth and demonstrate the capability of machine learning potentials for predicting material properties under extreme conditions.