Abstract: Data-driven approaches are becoming increasingly inseparable from materials design and synthesis processes. However, materials data is born in a variety of forms – including experiments, simulation, and literature – whose heterogeneity presents a challenge when seeking unified synthesis databases. Generating and connecting data sources using computation can aid discovery efforts, particularly within inorganic materials, where a variety of compositions, polymorphs, and synthesis routes aggravate the search for new structures. In this talk, I will describe how advances in high-throughput simulation, machine learning, and literature extraction enable the design of inorganic materials along with their synthesis routes. A case study on nanoporous materials will be used to illustrate this paradigm, where links between experiments and simulations explained over six decades of synthesis outcomes reported in the field and enabled realizing new catalysts from a database containing millions of data points. Finally, a special focus will be given to data sharing and how to accelerate materials synthesis using human-computer partnerships. Prepared by LLNL under Contract DE-AC52-07NA27344. IM: LLNL-ABS-837335
Bio: Daniel Schwalbe-Koda is a Lawrence Fellow at the Lawrence Livermore National Laboratory. He obtained his PhD in Materials Science and Engineering from MIT in 2022. His research interests include high-throughput simulations for predictive materials synthesis, literature extraction, and scientific machine learning.
