Melting simulations of high-entropy carbonitrides by deep learning potentials.

The melting temperature is a crucial property of materials that determines their potential applications in different industrial fields. In this study, we used a deep neural network potential to describe the structure of high-entropy (TiZrTaHfNb)CN carbonitrides (HECN) in both solid and liquid states. This approach allows us to predict heating and cooling temperatures depending on the nitrogen content to determine the melting temperature and analyze structure changes from atomistic point of view. A steady increase in nitrogen content leads to increasing melting temperature, with a maximum approaching for 25% of nitrogen in the HECN. A careful analysis of pair correlations, together with calculations of entropy in the considered liquid phases of HECNs allows us to explain the origin of the nonlinear enhancement of the melting temperature with increasing nitrogen content. The maximum melting temperature of 3580 ± 30 K belongs to (TiZrTaHfNb)CN composition. The improved melting behavior of high-entropy compounds by the addition of nitrogen provides a promising way towards modification of thermal properties of functional and constructional materials.