Publication date: 10th April 2024
Solid-state ion conductors (SSICs) play an important role in electrochemical energy storage devices, such as all-solid-state batteries,[1] due to improved safety and high energy density. Considering the plethora of available SSIC candidate materials,[2] high-throughput experiments and simulations are essential in the search for suitable SSICs that offer high ionic conductivities. First-principles methods are the conventional approach to investigate the interplay between the anharmonic host lattice and mobile ions on an atomic level,[3,4] while being limited to few selected systems due to the large computational costs.
Here, we systematically explore the potential of machine-learning molecular dynamics (MLMD) for the prediction of the complex ion migration mechanisms including anharmonic vibrational properties in the host lattices.[5] We choose three classes of SSICs in which different mechanisms govern the migration of ions, and which all include vibrational anharmonicities: AgI, a strongly disordered Ag+ conductor; Li10GeP2S12, showing concerted Li+ migration; and tungsten-doped Na3SbS4, a Na+ vacancy conductor. These materials represent a variety of challenges for MLFFs including anharmonicities, dynamical potential energy surfaces as well as rare hopping events. Systematic comparison with ab initio molecular dynamics simulations reveals that MLMD captures accurately anharmonic vibrational properties in the host lattice and particularly is able to predict the underlying physical mechanism that governs ion migration. We will finally discuss the potential of MLMD simulations to complement experimental techniques in the investigation of finite-temperature vibrational properties and, thereby, to enable the rapid design of novel SSICs.
Funding provided by the Alexander von Humboldt–Foundation in the framework of the Sofja Kovalevskaja Award, endowed by the German Federal Ministry of Education and Research, by the Deutsche Forschungsgemeinschaft via Germany’s Excellence Strategy - EXC 2089/1-390776260, and by the TUM-Oerlikon Advanced Manufacturing Institute, are gratefully acknowledged. The authors further acknowledge the Gauss Centre for Supercomputing e.V. for funding this project by providing computing time through the John von Neumann Institute for Computing on the GCS Supercomputer JUWELS at Jülich Supercomputing Centre.