Deep-learning potential method to simulate shear viscosity of liquid aluminum at high temperature and high pressure by molecular dynamics

Abstract

The shear viscosity of matter and efficient simulating methods in a wide range of temperatures and densities are desirable. In this study, we present the deep-learning many-body potential (the deep potential) method to reduce the computational cost of simulations for the viscosity of liquid aluminum at high temperature and high pressure with accurate results. Viscosities for densities of 2.35 g/cm3, 2.7 g/cm3, 3.5 g/cm3, and 4.27 g/cm3 and temperatures from melting points to about 50 000 K are calculated. The results agree well with the experiment data at a pressure near 1 bar and are consistent with the simulation of first-principles at high pressure and high temperature. We reveal the behavior of the shear viscosity of liquid Al at a range where the current experimental results do not exist. Based on the available experimental data and newly generated simulation data, we propose a modified Enskog-Dymond theory, which can analytically calculate the viscosity of Al at this range. This research is helpful for numerous potential applications.