First-principles study on the lithiation process of amorphous SiO anode for Li-ion batteries with Bayesian optimization.

Amorphous silicon monoxide (a-SiO), which contains Si atoms with various valence states, has attracted much attention as a high-performance anode material for lithium (Li) ion batteries (LIBs). Although current experiments have provided some information during charge/discharge cycles, further investigation of structural changes at the atomic scale is needed. To investigate the lithiation process of a-SiO using first-principles simulations and machine learning techniques, we developed a computational code employing Bayesian optimization to efficiently identify stable sites for Li insertion in the large search-space of amorphous models to reproduce the actual lithiation process and compared this approach to the conventional random scheme by applying it to an a-SiO model previously generated with neural network potentials.