Liquid thermophysical properties of Ag-Si alloy based on deep learning potential

The knowledge of the thermophysical properties of liquid metals and alloys is essential for expanding the materials database and designing materials with good properties. In this work, we developed an interatomic potential using a deep neural network (DNN) algorithm for liquid Ag-Si alloys. Compared with ab initio molecular dynamics (AIMD) results, the DNN potential provided a good description of the information of energy, force, and structure features of the system in the simulated temperature range. Through this potential, we can obtain the thermophysical properties of different compositions of liquid alloys by simulation way. The computed thermophysical properties are in excellent agreement with the reported experimental data. The analysis of local structure indicates that the liquid ordering and stability strengthen upon cooling at the atomic level, eventually leading to an increase in thermophysical properties.