Atomistic Calculation of the Melting Point of the High-Entropy Cantor Alloy CoCrFeMnNi

Abstract

The melting point of the high-entropy Cantor alloy CoCrFeMnNi was calculated by the classical molecular dynamics method. Interatomic potential as a set of artificial neural networks was used for simulation of this type for the first time. Neural network coefficients were optimized using machine learning technique with ab initio molecular dynamics data. Ab initio molecular dynamics simulation was carried out for a wide temperature range using the same initial crystalline state. The initial state for ab initio simulations was a special quasi-random structure optimized on pairs of the nearest neighbors. The two-phase method based on the movement of phase boundary in a crystal–melt system was used to calculate the melting point. It should be noted that, although the training set did not contain explicit two-phase configurations, the computed melting point proved to be in a satisfactory agreement with available experimental data. Thus, the melting point of the high-entropy CoCrFeMnNi alloy was calculated without the use of empirical data.