Accurate diffusion data in oxide melts is critical for modeling high-temperature processes in metallurgy and geosciences, yet available measurements remain scarce and inconsistent. This study presents a high-throughput molecular dynamics (MD) investigation of self-diffusion in CaO–Al 2 O 3 –SiO 2 melts. A total of 119 simulations were performed across 40 compositions within the liquid phase between 1400 C and 1600 C, producing diffusivities for Ca, Al, Si, and O. The trends were analyzed in relation to melt structure and fitted using the Thibodeau–Jung model to obtain two parameterizations: an MD-only fit and a hybrid fit combining MD and selected experimental data. The MD-only fit showed excellent internal consistency and reproduced a physically meaningful diffusivity ranking. The hybrid fit improved agreement with experiments, particularly for Al, Si, and O. Both fits were compared to Kwon’s fully experimental parameterization and validated against diffusion couple experiments and phase-field simulations of Al 2 O 3 and SiO 2 particle dissolution. Unlike Kwon’s parameterization, the MD-based parameterizations capture structural trends and reproduce the experimentally observed uphill diffusion. This work establishes a validated, transferable framework for self-diffusion in oxide melts and demonstrates the potential of MD-informed strategies for predictive transport modeling in data-scarce multicomponent oxide systems. • 119 MD diffusivity calculations across 40 compositions and 5 temperatures • Validated parameterizations link atomistic MD with mesoscale transport modeling • Hybrid parameterization improves agreement with experimental diffusivities • MD-based fits reproduce diffusion couples and capture uphill diffusion • Correct timescale predictions for and particles dissolution
Verbeeck et al. (Sun,) studied this question.