“…30 Because of its excellent performance and flexibility, deep learning interatomic potentials have enabled growing popularity in both chemistry and materials science, such as fuel oxidation, 31,32 phase change, 29,33 chemical catalysis 34 and material design. 35,36 In this work, a NN-based potential (NNP) model is developed to examine the melting behavior of boron nanoparticles with ab initio accuracy. We first validate the accuracy of the NNP comprehensively against DFT results via the prediction of atomic energy and force, crystallographic parameters, equation of state, elastic constant, phonon dispersion relationship, radial distribution function, as well as atomic thermal motion.…”