Predicting Kinetic Reactivity of Solid-State Battery Interfaces

In this work, we combine ab initio simulation with machine learning to build the first model for predicting the kinetics of interfacial reactivity in solid-state Li metal batteries. We first use density functional theory molecular dynamics (DFT-MD) to simulate the time evolution of several solid-state electrolyte (SSE) candidate materials interfaced with Li metal. Unlike state-of-the-art grand potential convex hull calculations for predicting interfacial stability, simulation with DFT-MD explicitly considers kinetic reaction barriers between two phases. We use the resulting data from these simulations to build a data-driven model that uses computationally inexpensive structural and thermodynamic features to predict the DFT-MD kinetic reactivity of these interfaces. This first-of-its-kind model enables more targeted and effective selection of solid electrolytes for solid-state Li metal batteries. Solid-state batteries offer safer and more energy-dense alternatives for energy storage systems. Like current liquid electrolyte batteries, the stability of interfaces between different battery components remains a critical requirement for longevity and device performance. With over 20,000 Li containing inorganic solids in the Materials Project database, the current high throughput technique to assess the chemical compatibility of candidate interfaces between SSEs and battery electrodes consists of a thermodynamic static parameter, the driving force for chemical mixing, with an arbitrary global assumption of a kinetic stabilization barrier. The current version of this kinetic reactivity model outperforms the purely thermodynamic parameter with a significantly lower misclassification of reactive/unreactive interfaces and predicts over 1,000 additional stable candidate interfaces that could be used for metric-specific material searches.