Atomistic simulations on liquid Mg–Sr alloys assisted with deep learning potential

It has been demonstrated that Sr is a beneficial additive element to Mg alloys and could effectively upgrade the properties of Mg alloys. However, the local structure and thermophysical properties of Mg–Sr, which are important properties of the alloy, have rarely been reported. This work applied atomistic simulations, assisted by deep learning potential (DPL), conducted on the Mg–Sr liquid alloy to reveal its local structure and thermophysical properties. The reliability of the trained DPL model was thoroughly validated by the root-mean-square errors, energy and force comparison results, and local structure reproducibility. The radial distribution function and structural factor were adopted to assess the short-range and intermediate-range ordering of the Mg–Sr liquid alloy, whose component and temperature dependence were analyzed. Mg–Sr alloy is more compact in the Mg-rich system; all the ordering is temperature-negatively dependent. Densities, self-diffusion coefficients, and shear viscosities of Mg–Sr liquid alloys containing different Mg concentrations, covering the temperature range of 1100–1400 K, were predicted by deep learning molecular dynamics simulation, and the corresponding databases were established. The enthalpies of mixing and elemental activities of Mg–Sr liquid alloys at 1100 K were reliably evaluated. The mixing enthalpy is less than zero, and the dependence of the Mg and Sr activities on the Mg concentration presents a downward concave shape that deviates from Raoul's law, forming a negative deviation. Overall, this present study provides theoretical foundation and directional guidance for the development of Mg–Sr alloys.