Advancing Materials Design Through Integrated AI and Physics-based Approaches

The accelerated discovery of novel materials is crucial for advancements in catalysis, energy storage, and electronics. Despite the recent progress in the computer-aided materials discovery, there are still significant challenges remaining including (1) reducing the computational cost for both describing electronic and atomic interactions within materials and exploring diverse chemical spaces, (2) expanding the capability of computational workflow in terms of the time and length scales to connect the computational insight with experimental observation, and (3) automating the execution of multi-step workflow to help reduce the manual effort.

In this presentation, we showcase our AI-based materials discovery workflows, including an ML-based force fields-accelerated discovery workflow for heterogeneous catalysts for CO2 reduction and the generative AI-accelerated workflow for battery material discovery. We will also describe how physics-based training data generation enhances the accuracy and efficiency of our AI-driven approaches, enabling faster and more reliable materials discovery.

‍