Developing a machine learning Ni-Co-Al ternary interatomic potential for temperature and strain-rate dependent mechanical behaviors of Ni-based single crystal superalloys

Ni-based single crystal (SX) superalloys, with excellent mechanical properties in high temperature, are widely used as the preferred materials in the aerospace field. Atomistic simulations provide an effective means to understand the underlying physics of microstructure-property relationship. However, the reliability of atomistic simulation results highly depends on the precision of potential. In this work, a machine learning potential (MLP) for the ternary Ni-Co-Al system is developed using neuroevolution potential (NEP) framework which maintains a high degree of fidelity and computational efficiency from ab initio molecular dynamics simulations (AIMD). It is demonstrated that the Ni-Co-Al MLP can be used to investigate the temperature and strain-rate dependent mechanical behaviors of Ni-based superalloys with the effect of [Figure presented] precipitation. Specifically, the following case studies about γ/γ^′ system are conducted: (1) the temperature and strain rate effect on the γ/γ^′ phase boundary characteristics and uniaxial tensile deformation behaviors; (2) the shock-induced phase transformation and spallation behaviors at different strain rates. The atomistic findings promote the fundamental understanding of the mechanical responses of Ni-based SX superalloys with [Figure presented] precipitation at different temperature and strain rate conditions. Moreover, the MLP development process proposed in this work provides a feasible framework to develop diverse multicomponent interatomic potentials.