Safe and efficient batteries is one of the key technologies for electrification of transport and sustainable energy storage and thus enabling the green transition. The intercalation-type Li-ion battery is by far the most studied and commercially successful battery type. Electrodes in these batteries have traditionally been ordered crystalline materials, but improvements in these materials’ capacity and stability are needed. Recent studies suggest that such improvements can be achieved by the use of electrode materials with different kinds of disorder, for example materials undergoing order-disorder transitions during charge/discharge cycling.

In this project, I propose to use topological data analysis and machine learning methods to enable the computational design of such disordered electrode materials with improved performance. To this end, I have divided the project into four tasks. First, atomic structures of the selected systems will be generated. This will be done using molecular dynamics simulations as well as based on experimental x-ray/neutron scattering data that are analyzed using reverse Monte Carlo, genetic algorithm, or particle swarm optimization algorithms. Second, topological features of these atomic structures will be identified using topological data analysis. When these data are combined with a classification-based machine learning algorithm, it will be possible to construct topological metrics that are correlated to the materials’ propensity to possess large tunnels that enable Li ion motion. Third, models for predicting the dynamics of the conducting Li ions will be constructed using graph neural networks.

Based on this analysis, the relative importance of the various structural environments surrounding the Li ions on their dynamics can be quantified. Fourth, the insights gained in the previous two tasks will be used to design new improved electrode materials based on high-throughput molecular dynamics simulations and machine learning regression models.

Taken as a whole, the proposed research will enable battery scientists to find “order in disorder” in a promising new family of electrode materials, which in turn will enable future development of novel batteries. Two experts in machine learning applications, disordered materials, and topological data analysis will supervise the project.