Alloying Process at the Interface of Au-Li Studied Using Neural Network Potential

Abstract

In this paper, we report construction of neural network potentials (NNPs) of Au-Li binary systems based on density functional theory (DFT) calculations and analyses of alloying properties. To accelerate construction of NNPs, we proposed an efficient method of structural dataset generation using the symmetry function-based principal component analysis. We investigated the mixing energy of Au_1－xLi_x with fine composition grids, which were achieved owing to the lower computational cost of NNPs. The obtained results agree well with the DFT values, where we found previously unreported stable compositions. In addition, we examined the alloying process starting from the phase separated structure to the complete mixing phase using Au/Li superlattice structures. We found that when multiple adjacent Au atoms dissolved into Li, the alloying of the entire Au/Li interface started from the dissolved region. These results demonstrate the applicability of NNPs toward miscible phase and provides the understanding of the alloying mechanism.