Abstract:

Large-scale atomic simulations can capture ample phenomena that are difficult to solve experimentally or based on quantum mechanics in materials science. Accurate interatomic potentials are a factor in the validity of atomic calculations. It has been a challenge to develop efficient potential functions to resolve complex phenomena in complex materials. The machine learning potentials based on first principles are currently widely studied. Here, we developed a highly accurate potential for Nickel which is one of the most important metals in the aerospace industry using the Deep Potential (DP) approach. The generated DP correctly describes the structural, energetic, and mechanical property-related properties of Ni in the fcc and hcp structures. The developed potential simultaneously exhibits excellent predictability, such as the ability to accurately predict dislocations, grain boundaries, idea strength, and phonon spectrums. Therefore, the as-developed potential expects to provide strong support for further solving mechanical problems in nickel-based alloys.