On-the-fly machine learning force field study of liquid-Al/α-Al2O3 interface

The atomic structure of the liquid Al/α-Al2O3 interface was investigated using on-the-fly machine learning force field based on first-principles calculations. We found that three different terminal surfaces of the liquid Al/α-Al2O3{0 0 0 1} systems exhibited the same interfacial characteristics. Liquid Al atoms at the interface formed an Al-rich layer and several ordering layers. Analysis of mean square displacement and vibrational density of states indicated that the characteristics of Al-rich layer at the interface is close to solid, while liquid ordering layers possess a confined liquid nature. Furthermore, we simulated a larger system with more than 1000 atoms, which showed that periodic boundary conditions caused size effects that affected the ordering of the liquid atoms at the interface. Hence, the present study has the potential to generate substantial interest and inspire further research into the utilization of machine-learned force fields for simulating various liquid–solid interfaces.