MSE 298 SEMINAR: Toward Predictive Materials Synthesis with Atomistic Simulations and Robust Machine Learning

Abstract: In the last decade, numerous advances in materials design were made using databases, high-throughput screening and machine learning (ML) approaches. While determining which structures should be synthesized is now a mature task, few theoretical approaches can predict how to synthesize them. In this talk, I will discuss how a unified view on entropy, simulations and ML can enable predictive models for the synthesis of inorganic materials. First, I will discuss how high-throughput simulations can optimize both the synthesis and performance of new nanoporous catalysts. Building on these successes, I will describe how information theory can be used for a range of tasks in atomistic simulations relevant to synthesis prediction, from improving the robustness of ML models to modeling synthesis mechanisms and phase transformations. Finally, I will describe how generative ML can be used to generate complex amorphous structures, from glasses to mesoporous materials, thus integrating advances in artificial intelligence, materials theory, and simulations.

Bio: Daniel Schwalbe-Koda is an assistant professor of materials science and engineering at UCLA. He was a Lawrence Fellow at the Physical and Life Sciences Directorate of the Lawrence Livermore National Laboratory, and received a doctorate in materials science and engineering from MIT in 2022. Some of his recent recognitions include the DOE Early Career Award (2024), a Scialog Fellow (2024), and the Forbes “30 Under 30” award in the ccience category (2023).