Google DeepMind study finds multi-lithium clusters in battery electrolyte simulations
Researchers ran molecular dynamics simulations of a 1 M LiPF6 solution in ethylene carbonate solvent, comparing a machine-learned force field (QRNN) against a conventional empirical force field (OPLS4). The two methods produced markedly different molecular structures around lithium ions, with the QRNN simulations revealing previously unidentified clusters containing multiple lithium ions bound together.