Surface Tension Calculations in Liquid Metals Using First Principles and Machine Learning

Surface tension plays a crucial role in various phenomena, from a steel wire ‘floating’ on water to materials processing. However, its computational prediction remains challenging due to limitations in empirical force fields, emphasizing the need for tailored approaches or force-field-free methods.

Our research aims to overcome this challenge by developing efficient computational methods for calculating the surface tension of liquids, focusing on elemental liquid metals. For this purpose, we use ab initio molecular dynamics (AIMD) and develop machine learning force fields (MLFF), both combining molecular dynamics and Kohn-Sham density functional theory. The two approaches potentially eliminate the dependence on empirical force fields, which can improve accuracy and transferability across different liquid systems.

Emphasizing computational efficiency, we present the results of our surface tension calculations using the exact method of Kirkwood and Buff (1949) and an approximation based on the work of cohesion. In addition, we evaluate their performance, revealing their relative strengths and limitations for liquid metal systems. Lastly, we compare our MLFF results with AIMD simulations and available experimental data, to validate our computational approaches.