Accurate structure and properties of experimental battery materials with AI-driven theoretical frameworks

The accurate structure of materials combined with theoretical methods provides insights into the atomic-scale mechanisms that are at play behind the bulk-scale properties generally measured with experiments. However, determining the atomic-scale structure including defects and interfaces in real materials is a significant challenge. Creating global optimization frameworks with machine learning interatomic potentials (MLIPs) offers significant means to determine the structure, properties, and dynamics of nanoscale materials relevant for energy storage and electronic applications. By leveraging MLIPs, we can navigate possible structures, assess their stability, and compare results with experimental measurements to incorporate realistic observations. In this talk, I will discuss the advancements of ML methods in materials modeling and examples of our methodology applied to battery cathodes and solid-state electrolytes. This approach offers an effective and powerful modality to "look" at these complex systems, complementing experimental techniques. This atomic-scale understanding is grounded in realistic experimental observations, thereby providing a comprehensive outlook on the system and allowing for accelerated materials design for energy storage applications.